Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This paper builds on previous structural studies from the Springer lab, which show that the I-domain of the malaria sporozoite surface protein TRAP can switch between two conformations ("open" and "closed"), similar to what happens with the I domain of integrins. In the current study the authors explore whether the open and closed conformations of TRAP are important for sporozoite migration and infectivity by generating parasites expressing each form locked in place using introduced disulfides. Locking TRAP into either form inhibits sporozoite gliding, invasion of the mosquito salivary glands and transmission from mosquito to mouse. The addition of low levels of a reducing agent can partially rescue these effects with the open form, which the authors suggest points to a requirement for TRAP to undergo a switch between its open and closed states to support these key processes in the parasite's life cycle.

The paper is well written and the data and methods are clearly presented.

Major Comments:

Lines 109-115: The entire paper relies on the premise that the introduced disulfides lock the I domain of TRAP in either the open or closed conformation. In the absence of experimental proof that this is the case, it would be helpful to the reader to have more detail on how this can be confidently inferred from the previous work on integrins -- perhaps as a supplementary figure.

The partial rescue of motility in the S210C/Q216C parasites by 50-100mM DTT is a VERY indirect approach to testing whether the conformational switch between the open and closed states is functionally important. The authors are appropriately reserved in their conclusions, but wildtype parasites are ultimately a better (less confounded) comparator; the authors may want to stress this more.

The implications of the current results for the authors' previous stick-and-slip model of sporozoite motility should be discussed.

Minor Comments:

The authors are experienced in the use reflection interference contrast microscopy to visualize attachment of the parasite to the substrate. Did they do any RICM on the mutants? Although not necessary for the paper, this would be a good way to support the interpretation of some of their results (eg, lines 185-190).

Line 127: Figure 2A does not show "no growth difference to wild type mice"

Line 154: "higher, but not significantly higher" - if the data don't meet the significance threshold being used, they cannot be called higher

Line 186: "showed nearly no floating parasites compared to the controls S210C and cFluo" - cFluo and S210C/Q216C show the same amount of floaters in Fig 5B. Also, S210C/Q216C HLS show similar levels of floaters to both controls in 2A.

Fig S2 defines unproductive gliding to include waving, but Figs. 5 and S3 scores waving as a separate category

Fig S3B and Fig 5D,E appear to be the same data presented twice

Line 304: "the inability of sporozoites with the TRAP I domain to migrate" (?)

Line 226: please explain what the + > ++++ qualitative descriptors signify in the tables and how they were scored

Lines 282, 312: the authors should mention here the extensive work that has been done on efficacy of viral vaccines directed against a particular conformational state of the immunogen

Fig 2B: The labels above the lanes are incomplete and therefore confusing; suggest sticking to the same nomenclature as in 2A

Typographical errors on lines: 58 (space missing), 80-81 (parentheses), 155 (Figure misspelled), 188 (expressing the closed (S210C/F224C)), 306 (comma after both)

Referees cross-commenting

Since all three reviewers questioned whether the introduced disulfides would have the assumed effects on I domain structure, the authors should provide a stronger rationale for this assumption or -- as suggested by reviewer 2 -- some actual data to support it. Reviewer 2's comment about the extracellular ligands available for the parasite to bind to in the gliding assay and whether this could influence the outcome (and relevance to what occurs in vivo) also deserves consideration.

Significance

This is an elegant study that contributes important new information to our understanding of apicomplexan parasite motility and the function of the TRAP protein. The results will be of significant interest to those who study parasite motility, and likely also to those who study the role of integrins in cellular adhesion and signaling. The data nicely connect what was previously known about conformational changes in integrins with parasite adhesion to the substrate and motility.

I have expertise in the area of parasite motility.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

This work seeks to investigate the effects on sporozoite motility of hypothesized conformational changes in TRAP's integrin domain. Previous structural studies suggested that the Integrin-like domain of TRAP assumes 'open' and 'closed' conformations induced by ligand binding. Here the authors hypothesize that TRAP's ability to dynamically switch between these states is crucial for sporozoite motility. The hypothesis is interesting and the authors elected to test it by 'fixing' TRAP in constitutively 'open' and 'closed' conformations by introducing Cys residues that are presumed to form disulphide bonds resulting in these states.

Major Comments

The approach is creative. The data that are presented are of high quality with adequate reproducibility. However, as written the manuscript does not provide a rationale for the specific substitutions they chose, how these substitutions are expected to lead to 'open' and 'closed' states, and how they mimic the natural route of TRAP transitioning between the 'open' and 'closed' conformations.

Furthermore, there is no biochemical, biophysical or modeling data demonstrating that the introduced mutations impact the folding/unfolding of TRAP's I domain in the manner hypothesized. Therefore, it is difficult to interpret subsequent phenotypic data from the two mutant lines. While mutant parasites display defects in gliding motility, these defects are unexpected, perhaps pointing to alternative explanations - such as aberrant inter- or intra-molecular disulphide bonding in TRAP's extracellular domain.

About 10% of mutant sporozoites assumed to be in a constitutively 'closed' conformation display gliding motility and they move faster that WT. Yet this mutant did not cause patency in mice when introduced either via mosquito bite or IV injection. In contrast, the mutant assumed to be in an 'open' conformation displayed no motility in vitro but was able to infect mice (albeit at significantly reduced compared to controls). Presumably these data suggest that the in vitro gliding motility assays used are insufficient for testing the effect of these mutations on motility that is relevant in vivo. The model is that TRAP's interaction with extracellular ligands stabilizes its 'open' position. This suggests that motility assays conducted with extracellular matrix eg Matrigel are a more appropriate test of motility.

Experiments with DTT are difficult to interpret since there is once again no biochemical evidence that this treatment leads to a change in conformation of TRAP. DTT's effect on motility of the mutant could be non-specific. Overall, conclusions that are presented need to be supported by more data.

Significance

Sporozoite motility is a prerequisite for infection by Plasmodium of the liver. A better mechanistic understanding of this process is significant for our understanding of the first step of malaria infection. TRAP is the major adhesin on the sporozoite surface and its loss abrogates sporozoite motility. The authors are to be commended for undertaking a challenging study.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Plasmodium sporozoites rely on gliding motility to invade the salivary glands of the mosquito vector where they reside until being inoculated into the vertebrate host. Once in the vertebrate host, sporozoites again have to glide to pass through liver cells and reach the final host cell before exo-erythrocytic replication occurs. In sporozoites, thrombospondin-related anonymous protein (TRAP) is the key adhesin, linking the molecular motor to the substrate. TRAP and TRAP-like proteins contain on the extracellular side an inserted (I) domain that is responsible for substrate binding. The I domain can be in an open or closed state which can be fixed by mutating specific amino acids. Baumann et al. then assessed the effect of these states on gliding motility and infection. Whilst no effect was seen for both mutants with closed or open states until the production of haemolymph sporozoites, the number of salivary gland sporozoites were highly reduced in both mutants. The closed mutant was not able to be transmitted from mosquitoes to mice whilst the open mutant is severely impacted in transmission. Gliding of both mutant is highly impaired. Some of these phenotypes could be reverse by adding a reducing agent.

Major comments:

This is an elegant study with exhaustive experiments to address the research questions. One of the things I wondered about is the effect of exchanging amino acids. As far as I understood, structural information of integrins from other organisms informed the authors about positions that should be mutated and about the impact (or loss of it) on the 3D structure of the domain. Would it not be useful to run a structure prediction programme such as Alphafold on all the mutants to at least confirm stability of the domain structure upon mutation in silico?

Some of the concentrations of reducing agents tested are very high. Can the authors be sure that the parasites are not dead?

Minor comments:

• Please make sure that punctuation is correct (e.g. missing commas) and that there are no other typing errors.
• Line 50: This sentence may imply mechanical rupture of oocysts rather than active egress of the sporozoite. Please re-phrase.
• Line 82: Do you mean a complexed Mg2+ ion?
• The introduction stops rather abruptly. Maybe a sentence of the significance of this study could be added.
• Line 361: 20 million parasites each? Please re-phrase to make it clearer.
• Line 363: Field of view needs to defined: What is the magnification used to observe exflagellation?
• Materials and Methods: Please write out reagent names for the first time.

Figures:

• Figure 1A is very small. The label font of the whole figure 1 is often too small.
• Figure 1D: Please list the components in the figure legend, e.g. blue: substrate etc.
• Figure 3: The figure title should be changed as there is no complete failure of sporozoites with fixed I-domains to invade salivary glands.
• Figure 5: for clarity, it would be easier if Figure C would not be squeezed into the legends of A and B.

Significance

This study addresses in detail the mechanism of Plasmodium TRAP I domain in gliding motility and transmission. It builds on work from many years and groups including their own and contributes to deepen our understanding of TRAP-family proteins, gliding motility in Plasmodium sporozoites and maybe even in other Plasmodium stages or other Apicomplexan parasites.

This work is of interest to researchers in the field of Plasmodium mosquito stages and transmission as well as scientist who work on apicomplexan gliding motility and transmission.

I have previously worked on Plasmodium mosquito stages, but currently work on Toxoplasma gondii.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

The Kim et al. paper titled "PRMT5 links lipid metabolism to contractile function of skeletal muscles" reports how the arginine methyltransferase PRMT5 affects lipid metabolism in myofibers by stabilizing the mSREBP1 protein and repressing the expression of the PNPLA2 gene. The genetic deletion of PRMT5 in muscle results in the loss of lipid droplets in myofibers and a loss in muscle strength. Additionally, there is a change in muscle fiber types, moving from an oxidative state to a more glycolytic one. While the authors present compelling data on PRMT5's role in muscle metabolism, there are some concerns on the mouse model used and the sequencing data.

Major concerns:

1. The mouse model used in this study is PRMT5fl/fl , Myl1cre in order to genetically delete PRMT5 in skeletal muscle. While there is no issue with the KO mice, the WT mice are PRMT5fl/fl , Myl1+ which is not an acceptable control. It is known that Cre itself can have a phenotype, and additionally Myl1 is very highly expressed. Thereby, there is a large amount of Cre in the KO mice, but none in the WT which may contribute to the differences seen between WT and KO mice. The appropriate WT control is PRMT5+/+ Myl1cre and the experiments would need to be repeated using this mouse genotype as the WT.

2. We appreciated the comment and have analyzed many mice from the various control groups (WT, Myl1Cre, Prmt5f/f, and Myl1Cre/Prmt5f/+). Long story short, we have maintained and used Myl1Cre for multiple projects and in several previous publications (PMID: 25794679; PMID: 27644105), and never observed a phenotype of the Myl1Cre In the current study during the development of the Myl1Cre-Prmt5KO mouse model, we had to first breed Myl1Cre with Prmt5f/f to generate the Myl1Cre/Prmt5f/+ mice, and then breed Myl1Cre/Prmt5f/+ with Prmt5f/fmice to generate the Myl1Cre/Prmt5f/f mice. During the breeding we had generated many Myl1Cre/Prmt5f/+ mice (at least 10). We only observed phenotypes in the Myl1Cre/Prmt5f/f mice but not in the Myl1Cre/Prmt5f/+ (heterozygous KO) mice. In line with our observation, no phenotypes were described in the original report on the generation of the Myl1Cre mice by Steve J. Burden and his colleagues (Bothe, Genesis, 2000, PMID: 10686620). Also, in consistent to our choice of the floxed mice as control, Pereira et al (EMBO Molecular Medicine, 2020, PMC7005622 ) used the Ndufsf/+/Myl1Cre-/- or Ndufsf/f/Myl1Cre-/- as control for their Ndufs3f/f/Myl1Cre+/- KO mice.

3. Given the situation, we trust that the reviewer will agree that the Prmt5f/f mice are appropriate controls and repeating all the experiments with a new control model will not only require years of work but also violate IACUC’s and NIH’s 3R policy in reducing unnecessary use of animals. We added a sentence to state that the Prmt5f/f are phenotypically identical to Myl1Cre/Prmt5f/+ mice in the revised manuscript.

In the scRNA-Seq the authors claim that PRMT5 is not expressed in quiescent muscle stem cells. However, the data set that is used only has approximately 250 muscle stem cells, which would not provide much coverage. It would be necessary to validate this claim by using other data sets, such as Tabula Muris or publicly available bulk RNA-Seq.

• As suggested, we queried Tabula Muris on Prmt5 expression in skeletal muscles based on scRNA-seq. The results showed that Prmt5 is expressed at very low levels in various mononuclear cell populations in the skeletal muscle. Specifically, only 8% of satellite cells had detectable levels of Prmt5, while 92% satellite cells had no detectable levels of Prmt5 (Table 1). We included the results in the revised Supplemental Table S1.

The ChIP-Seq data shown was performed on 3T3-L1 cells and is not appropriate for a muscle paper. The ChIP-Seq must be performed on muscle cells in order to confirm their conclusion.

• We trust that the reviewer understand that ChIP-seq only represents a discovery tool that needs to be experimentally validated. Although the ChIP-Seq data and identification of the PRMT5 binding peak at the Pnpla2 gene was based on 3T3-L1 cells, we have validated enrichment of PRMT5 on the potential binding region through ChIP-qPCR experiments. Repeating the ChIP-seq on muscle cells will not add additional support to the conclusion.

The authors claim that loss of PRMT5 leads to a gradual loss of muscle fiber size but has no effect on myogenesis. The evidence to support that claim is shallow, being based solely on CSA and total number of myofibers, along with a loss of lean body mass. To confirm this statement, it would be best to quantify the CSA and # of myofibers in EDL and TA at P7 and P21. Further, a regeneration assay would also demonstrate if myogenesis is compromised or not.

• Thank you for this suggestion. As Myl1Cre is only expressed in post-differentiation myocytes and myofibers (Bi et al, 2016, eLife, PMID: 27644105), we do not expect the Myl1Cre/Prmt5f/f to impact muscle development. Nevertheless, we now provide data on analysis of muscles at P7 and P21, as well postnatal muscle regeneration. The results are included in Supplementary Fig S2.

The data presented shows that there is fiber type switch from oxidative to glycolytic, along with a decrease in muscle strength in the PRMT5 KO mice. This seems counterintuitive to what is known in the field as glycolytic fibers are viewed as being capable of generating more force than oxidative, while having less endurance. The authors should clarify this point and elaborate more on their conclusion that the loss of strength is due to an altered metabolism.

• We agree with the reviewer that an increased abundance of glycolytic myofibers should increase muscle force under normal conditions. However, the increased glycolytic fibers in the Prmt5 mKO mice is associated with metabolic deficiencies. These fibers are atrophic (smaller than control fibers of the same myosin type), devoid of lipid droplets and had less force production. We added the following sentences to the discussion in this matter. Minor concerns:

• The term myocyte is not the most accurate word for describing skeletal muscle. Myofibers would be best for fully differentiated muscle, myotubes for in vitro differentiation and myocytes should be reserved for differentiated cells that have not fused yet (1-2 nuclei/cell). I would recommend changing all mentions of myocyte to myofiber and changing "myocyte specific" with regards to the mouse models as "skeletal muscle specific"

• Thank you for your suggestion. We updated the nomenclature accordingly.

Figure 1 Panel F and G would be clearer if they were labelled as 2 months old mice

We edited the Figure 1F, and 1G to include the mouse age. 3. Figure 2 Panel E, G and H the control line appears to be missing error bars

Error bars were in the original panels, but the variation was very small, making it hard to visualize the error bars. 4. Figure 2 legend should specify that these mice are 2 months old for the sake of clarity

We added mouse age to the Figure 2 legend. 5. Figure 3 Panel G and H are not informative and are misleading. The supplemental panels show that when normalized to body mass there is no difference in O2 consumption or CO2 production. These should be replaced with the supplemental panels

We reformatted the figure as suggested.

Figure 4 Panel E and F, are these separate experiments? The values do not match between the 2 panels. If these are separate it should be made more clear. a. Error bars appear to be missing in Panel E

Error bars were included (as it was obvious after FCCP) but the error bars are small for the rest of the time points. b. As this is an experiment where the state of the cell is incredibly important (metabolism of myoblast is much different to myotube), the authors must demonstrate that there is no defect in in vitro differentiation by showing the fusion index assay for these cells and a representative image of the myotubes.

The differentiation of the KO cells was normal as Myl1cre is only expressed after differentiation. we isolated myoblasts from WT and Prmt5MKO mice and differentiated for 3 days to stain MyoG and MF20 to measure fusion index. However, we did not see any change in fusion index from differentiated myotubes (Supplementary Fig. S4A)

c.The authors should mention in the legend that these are 3 DM myotubes

Figure legends are updated.

1. Figure 5 Panel G, a qPCR to confirm the O/E of the PRMT5 and SREBP1a in the test samples would be necessary.

We added the PRMT5 and SREBP1 O/E data (Supplementary Fig S5C, D). a. The double O/E cells are marked as significantly different, but it is unclear to which group they are significantly different to.

We edited the figure to highlight the comparison. 8. Figure 7 Panel B, body weight would be more informative in g rather than percentage.

We reformatted the figure as suggested.

It would be interesting to test whether the KO of Pnpla2 also rescues the fiber type switch.

We added Fiber type staining in dKO muscles. Based on quantification in Sol and EDL muscles, increased Type IIB (glycolytic myofibers) in Prmt5MKO mice was significantly decreased in Prmt5/Pnpla2MKO mice (Supplementary Fig S6D-F).

Reviewer #1 (Significance (Required)):

Overall the manuscript provides insight into the role of PRMT5 in regulating the metabolism of skeletal muscle. While the paper provides some interesting data, it is severely hampered by the improper WT controls lacking the cre alleles. In order for the data to be reliable, all of the WT samples will need to be replaced with PRMT5+/+ Myl1 cre mice. There are other papers looking at the role of PRMT5 in skeletal muscle, however these are more focused on muscle stem cells than the myofibers. Therefore this paper does fill in a gap with regards to the metabolic role of PRMT5 and how this can affect skeletal muscle function. This paper would most likely be of interest to a specialized audience, mostly those in the skeletal muscle field.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In the manuscript "PRMT5 links lipid metabolism to contractile function of skeletal muscles", Kim et al., describe the role for the arginine methyl transferase PRMT5 in maintaining skeletal muscle homeostasis. The authors showed that myocyte-specific deletion of PRTM5 results in the loss of muscle mass, reduced motor-performance, a fiber type switch from oxidative to glycolytic fibers along with reduced lipid content in the myofibers. The authors reasoned that absence of PRMT5 results in reduced methylation and stability of mSREBP1 and an increase in the expression of adipose triglyceride lipase (ATGL).

Major Comments

1. Experiments looking at the interaction between PRMT5 and mSREBP1a using overexpressed proteins are used to conclude that PRMT5 methylates SREBP1a. While the results are consistent with these conclusions, the finding remains correlative. Methylation assays using purified PRMT5 and SREBP1a would be required to make a definite conclusion that PRMT5 methylates SREBP1a. These should be performed. In the absence of such data, the authors would need to adjust their conclusions to say that in the presence of PRMT5, SREBP1a becomes methylated, and that it remains to be determined if this is directly mediated by PRMT5.

[response] We thank the reviewer for this comment and wished we could perform in vitro methylation assay to address whether SREBP1a is directly methylated by PRMT5. However, our co-author (Dr. Changdeng Hu) who is an expert of PRMT5 biochemistry unfortunately died recently, hampering the validation of SREBP1 as PRMT5 substrate. We would also like to mention that several other studies have reported SREBP1 as substrate of PRMT5 in cancer cells (Liu et al, 2016, Cancer Research, https://doi.org/10.1158/0008-5472.CAN-15-1766 ).We cited and discussed the paper in the manuscript. Our new results using enzymatic inhibitor of PRMT5 (BLL3.3) further supports that PRMT5 mediates methylation of SREBP1 (Supplementary Fig. S5A).

Stability studies in figure 5E have been performed in HEK293 cells using over-expressed proteins. While these results show that PRMT5 protects SREBP1a from degradation, the significance of this in muscle is less clear. Western blots of these proteins in C2C12 cells show that the over expression of PRMT5 does not stabilize SREBP1a (Flag) despite the appearance of increased methylation. To solidify the concept that PRMT5-dependent methylation of SREBP1 leads to its stabilization, cycloheximide experiments should be performed in myotubes generated in culture from the PRMT5mko mice. The stability of the endogenous SREBP1a gene could be monitored using antibodies.

[response] We performed that study as suggested. Consistent with the notion that PRMT5 stabilizes SREBP1, we found extremely low levels of SREBP1 proteins (nearly undetectable) in Prmt5 KO myotubes, in contrast to the robust expression of SREBP1 in WT myotubes (Supplementary Fig. S5B). This extremely low levels of SREBP1 precludes us from examining degradation after cycloheximide treatment

The measurement of methylation levels of SREBP1a are complicated by the fact that the protein levels are destabilized in the absence of PRMT5. The authors should use a PRMT5 inhibitor experiments in complement to these overexpression studies to measure the relative methylation of mSREBP1a (SMY10/mSREBP1).

[response] We used PRMT5 inhibitor, BLL 3.3, to support that mSREBP1a methylation is mediated by PRMT5 (Supplementary Fig. S5A).

Minor Comments 4. In figure 1H, there are more nuclei surrounding the myofibers. The authors should document the number of PAX7 cells per myofiber in the Control and Prmt5MKO mouse strains as it helps understand the additive effect of change in PAX7 cells during muscle atrophy.

[response] We quantified the number of Pax7+ cells in muscles and myofibers of WT and KO mice (Supplementary S2D,E).

In figure 5, the legend title mentioned ATGL. However the stability or methylation results of ATGL are not presented anywhere in the manuscript. Only in figure 6 did the authors show the differential expression of ATGL in the presence and absence of PRMT5. This title for Figure 5 should corrected.

[response] Thanks for clarifying. We fixed title of Figure 5 in the manuscript.

For better representation, the authors should consider moving the western blot panel (S3A) and lipid droplet staining data (S3E) to main figures and some of the data on force generation and body weights to supplementary. [response] We updated the figures as suggested.

Reviewer #2 (Significance (Required)):

This is an interesting manuscript that provides novel insight into the role for PRMT5 in muscle homeostasis. While previous studies have looked at the PRMT5 from a transcriptional standpoint during muscle differentiation, this work shows a role for PRMT5 in controlling the metabolic state of myofibers through transcriptional and potentially non-transcriptional mechanisms. Identification of the transcriptional regulation of Pnpla2 gene by PRMT5 is confirmed by mouse rescue experiments with the double KO mice. However the role for non-transcriptional control of SREBP1a stability through methylation by PRMT5 is not as clearly established. To strengthen this aspect of the manuscript, additional experiments are needed.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The lipid droplets represent an energy store and central hub of lipid metabolism in cells. In the skeletal muscle, abundant lipid droplets are present in myofibers (especially in oxidative Type 1 and IIA myofibers) and thought to play a role in supplying energy through fatty acid oxidation (FAO) to power muscle contraction. Increased abundance of lipid droplets in myofibers is associated with poor muscle function and insulin resistance in patients of type 2 diabetes. Paradoxically, myofibers of trained athletes also contain higher than normal levels of lipid droplets but with better contractile function and insulin sensitivity. The molecular mechanism underlying this "Athlete's Paradox" has been unclear, due to the lack of understanding of what controls biogenesis and metabolism of lipid droplets in the myofiber.

In this manuscript, Kuang and colleague provide compelling evidence to support a key role of PRMT5 in maintaining lipid droplets in the myofiber. They generated a conditional knockout mouse model to disrupt the Prmt5 gene in the myofibers. This leads to an astonishing depletion of lipid droplets in myofibers. The Prmt5 null myofibers also exhibited poor contractile function and classical signs of atrophy. To determine if the depletion of lipid droplets drives muscle atrophy or if muscle atrophy drives depletion of lipid droplets, the authors introduced a secondary lesion (ATGL-KO) in the Prmt5 null myofibers that would preserve the lipid droplets by preventing ATGL-mediated lipolysis. This restored lipid droplet content and largely rescued muscle contractile function, demonstrating that depletion of lipid droplets drives muscle atrophy and impairs muscle contractile function. The authors also performed a series of biochemical and molecular biology assays to show that PRMT5 stabilizes SREBP1a through dimethylation, enhancing the activity of this master transcriptional regulator of de novo lipogenesis. PRMT5 also methylates H4R3, and the methylated H4R3 represses the transcription of Pnpla2 (ATGL - which controls lipolysis), therefore inhibiting degradation of lipid droplets. These results together illustrate the dual role of PRMT5 in promoting lipid biogenesis and inhibiting lipid droplet degradation.

Reviewer #3 (Significance (Required)):

Collectively, this study identify PRMT5 as a key regulator of lipid metabolism in the muscle and establish a causal relationship between lipid droplet and muscle contractile function, and point to scarce lipid droplets as a driver of muscle atrophy. PRMT5 has previously been reported to regulate early myogenesis but its role in post-fusion myofibers has never been reported, therefore the conceptual novelty of this study is high. There are couple of minor points authors should consider: 1) While muscle-specific Prmt5-KO mice show reduced muscle mass, it is not clear whether this is a developmental effect or muscle atrophy. Authors should measure a few markers of muscle atrophy such as Atrogin1 and MuRF1 and overall levels of ubiquitination. Alternatively, authors can also subject the mice to conditions of muscle atrophy such as starvation and measure how various markers of atrophy are affected.

[response] Thank you for this suggestion. As the Myl1Cre is only expressed in post-differentiation myocytes and myotubes/myofibers, we did see any developmental defects during postnatal myogenesis (Supplementary Fig S2A-C). We also checked Atrogin-1 and MuRF1 in WT and KO muscle tissues based on your advice, but we did not find any significant difference (Supplementary Fig S1F). These two genes are muscle-specific E3 ubiquitin ligases involved in protein degradation by the ubiquitin proteasome pathway. This finding is consistent with the idea that there are multiple mechanisms that can lead to muscle atrophy, and our study clearly elaborates that dysregulation of lipid catabolism by PRMT5 is the main pathway associated with muscle wasting.

2) Please show representative EcoMRI images for body composition analysis.

[response] Thank you for your comments, but the EcoMRI equipment does not provide images of body composition. It only provides data of lean mass, fat mass and water in gram. We presented those data in the manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The lipid droplets represent an energy store and central hub of lipid metabolism in cells. In the skeletal muscle, abundant lipid droplets are present in myofibers (especially in oxidative Type 1 and IIA myofibers) and thought to play a role in supplying energy through fatty acid oxidation (FAO) to power muscle contraction. Increased abundance of lipid droplets in myofibers is associated with poor muscle function and insulin resistance in patients of type 2 diabetes. Paradoxically, myofibers of trained athletes also contain higher than normal levels of lipid droplets but with better contractile function and insulin sensitivity. The molecular mechanism underlying this "Athlete's Paradox" has been unclear, due to the lack of understanding of what controls biogenesis and metabolism of lipid droplets in the myofiber.

In this manuscript, Kuang and colleague provide compelling evidence to support a key role of PRMT5 in maintaining lipid droplets in the myofiber. They generated a conditional knockout mouse model to disrupt the Prmt5 gene in the myofibers. This leads to an astonishing depletion of lipid droplets in myofibers. The Prmt5 null myofibers also exhibited poor contractile function and classical signs of atrophy. To determine if the depletion of lipid droplets drives muscle atrophy or if muscle atrophy drives depletion of lipid droplets, the authors introduced a secondary lesion (ATGL-KO) in the Prmt5 null myofibers that would preserve the lipid droplets by preventing ATGL-mediated lipolysis. This restored lipid droplet content and largely rescued muscle contractile function, demonstrating that depletion of lipid droplets drives muscle atrophy and impairs muscle contractile function. The authors also performed a series of biochemical and molecular biology assays to show that PRMT5 stabilizes SREBP1a through dimethylation, enhancing the activity of this master transcriptional regulator of de novo lipogenesis. PRMT5 also methylates H4R3, and the methylated H4R3 represses the transcription of Pnpla2 (ATGL - which controls lipolysis), therefore inhibiting degradation of lipid droplets. These results together illustrate the dual role of PRMT5 in promoting lipid biogenesis and inhibiting lipid droplet degradation.

Significance

Collectively, this study identify PRMT5 as a key regulator of lipid metabolism in the muscle and establish a causal relationship between lipid droplet and muscle contractile function, and point to scarce lipid droplets as a driver of muscle atrophy. PRMT5 has previously been reported to regulate early myogenesis but its role in post-fusion myofibers has never been reported, therefore the conceptual novelty of this study is high.

There are couple of minor points authors should consider:

1. While muscle-specific Prmt5-KO mice show reduced muscle mass, it is not clear whether this is a developmental effect or muscle atrophy. Authors should measure a few markers of muscle atrophy such as Atrogin1 and MuRF1 and overall levels of ubiquitination. Alternatively, authors can also subject the mice to conditions of muscle atrophy such as starvation and measure how various markers of atrophy are affected.
2. Please show representative EcoMRI images for body composition analysis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In the manuscript "PRMT5 links lipid metabolism to contractile function of skeletal muscles", Kim et al., describe the role for the arginine methyl transferase PRMT5 in maintaining skeletal muscle homeostasis. The authors showed that myocyte-specific deletion of PRTM5 results in the loss of muscle mass, reduced motor-performance, a fiber type switch from oxidative to glycolytic fibers along with reduced lipid content in the myofibers. The authors reasoned that absence of PRMT5 results in reduced methylation and stability of mSREBP1 and an increase in the expression of adipose triglyceride lipase (ATGL).

Major Comments

1. Experiments looking at the interaction between PRMT5 and mSREBP1a using overexpressed proteins are used to conclude that PRMT5 methylates SREBP1a. While the results are consistent with these conclusions, the finding remains correlative. Methylation assays using purified PRMT5 and SREBP1a would be required to make a definite conclusion that PRMT5 methylates SREBP1a. These should be performed. In the absence of such data, the authors would need to adjust their conclusions to say that in the presence of PRMT5, SREBP1a becomes methylated, and that it remains to be determined if this is directly mediated by PRMT5.
2. Stability studies in figure 5E have been performed in HEK293 cells using over-expressed proteins. While these results show that PRMT5 protects SREBP1a from degradation, the significance of this in muscle is less clear. Western blots of these proteins in C2C12 cells show that the over expression of PRMT5 does not stabilize SREBP1a (Flag) despite the appearance of increased methylation. To solidify the concept that PRMT5-dependent methylation of SREBP1 leads to its stabilization, cycloheximide experiments should be performed in myotubes generated in culture from the PRMT5mko mice. The stability of the endogenous SREBP1a gene could be monitored using antibodies.
3. The measurement of methylation levels of SREBP1a are complicated by the fact that the protein levels are destabilized in the absence of PRMT5. The authors should use a PRMT5 inhibitor experiments in complement to these overexpression studies to measure the relative methylation of mSREBP1a (SMY10/mSREBP1).

Minor Comments

1. In figure 1H, there are more nuclei surrounding the myofibers. The authors should document the number of PAX7 cells per myofiber in the Control and Prmt5MKO mouse strains as it helps understand the additive effect of change in PAX7 cells during muscle atrophy.
2. In figure 5, the legend title mentioned ATGL. However the stability or methylation results of ATGL are not presented anywhere in the manuscript. Only in figure 6 did the authors show the differential expression of ATGL in the presence and absence of PRMT5. This title for Figure 5 should corrected.
3. For better representation, the authors should consider moving the western blot panel (S3A) and lipid droplet staining data (S3E) to main figures and some of the data on force generation and body weights to supplementary.

Significance

This is an interesting manuscript that provides novel insight into the role for PRMT5 in muscle homeostasis. While previous studies have looked at the PRMT5 from a transcriptional standpoint during muscle differentiation, this work shows a role for PRMT5 in controlling the metabolic state of myofibers through transcriptional and potentially non-transcriptional mechanisms. Identification of the transcriptional regulation of Pnpla2 gene by PRMT5 is confirmed by mouse rescue experiments with the double KO mice. However the role for non-transcriptional control of SREBP1a stability through methylation by PRMT5 is not as clearly established. To strengthen this aspect of the manuscript, additional experiments are needed.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The Kim et al. paper titled "PRMT5 links lipid metabolism to contractile function of skeletal muscles" reports how the arginine methyltransferase PRMT5 affects lipid metabolism in myofibers by stabilizing the mSREBP1 protein and repressing the expression of the PNPLA2 gene. The genetic deletion of PRMT5 in muscle results in the loss of lipid droplets in myofibers and a loss in muscle strength. Additionally, there is a change in muscle fiber types, moving from an oxidative state to a more glycolytic one. While the authors present compelling data on PRMT5's role in muscle metabolism, there are some concerns on the mouse model used and the sequencing data.

Major concerns:

1. The mouse model used in this study is PRMT5fl/fl , Myl1cre in order to genetically delete PRMT5 in skeletal muscle. While there is no issue with the KO mice, the WT mice are PRMT5fl/fl , Myl1+ which is not an acceptable control. It is known that Cre itself can have a phenotype, and additionally Myl1 is very highly expressed. Thereby, there is a large amount of Cre in the KO mice, but none in the WT which may contribute to the differences seen between WT and KO mice. The appropriate WT control is PRMT5+/+ Myl1cre and the experiments would need to be repeated using this mouse genotype as the WT.
2. In the scRNA-Seq the authors claim that PRMT5 is not expressed in quiescent muscle stem cells. However, the data set that is used only has approximately 250 muscle stem cells, which would not provide much coverage. It would be necessary to validate this claim by using other data sets, such as Tabula Muris or publicly available bulk RNA-Seq.
3. The ChIP-Seq data shown was performed on 3T3-L1 cells and is not appropriate for a muscle paper. The ChIP-Seq must be performed on muscle cells in order to confirm their conclusion.
4. The authors claim that loss of PRMT5 leads to a gradual loss of muscle fiber size but has no effect on myogenesis. The evidence to support that claim is shallow, being based solely on CSA and total number of myofibers, along with a loss of lean body mass. To confirm this statement, it would be best to quantify the CSA and # of myofibers in EDL and TA at P7 and P21. Further, a regeneration assay would also demonstrate if myogenesis is compromised or not.
5. The data presented shows that there is fiber type switch from oxidative to glycolytic, along with a decrease in muscle strength in the PRMT5 KO mice. This seems counterintuitive to what is known in the field as glycolytic fibers are viewed as being capable of generating more force than oxidative, while having less endurance. The authors should clarify this point and elaborate more on their conclusion that the loss of strength is due to an altered metabolism.

Minor concerns:

1. The term myocyte is not the most accurate word for describing skeletal muscle. Myofibers would be best for fully differentiated muscle, myotubes for in vitro differentiation and myocytes should be reserved for differentiated cells that have not fused yet (1-2 nuclei/cell). I would recommend changing all mentions of myocyte to myofiber and changing "myocyte specific" with regards to the mouse models as "skeletal muscle specific"
2. Figure 1 Panel F and G would be clearer if they were labelled as 2 months old mice
3. Figure 2 Panel E, G and H the control line appears to be missing error bars
4. Figure 2 legend should specify that these mice are 2 months old for the sake of clarity
5. Figure 3 Panel G and H are not informative and are misleading. The supplemental panels show that when normalized to body mass there is no difference in O2 consumption or CO2 production. These should be replaced with the supplemental panels
6. Figure 4 Panel E and F, are these separate experiments? The values do not match between the 2 panels. If these are separate it should be made more clear.
   • a. Error bars appear to be missing in Panel E
   • b. As this is an experiment where the state of the cell is incredibly important (metabolism of myoblast is much different to myotube), the authors must demonstrate that there is no defect in in vitro differentiation by showing the fusion index assay for these cells and a representative image of the myotubes.
   • c. The authors should mention in the legend that these are 3 DM myotubes
7. Figure 5 Panel G, a qPCR to confirm the O/E of the PRMT5 and SREBP1a in the test samples would be necessary.
   • a. The double O/E cells are marked as significantly different, but it is unclear to which group they are significantly different to.
8. Figure 7 Panel B, body weight would be more informative in g rather than percentage.
9. It would be interesting to test whether the KO of Pnpla2 also rescues the fiber type switch.

Significance

Overall the manuscript provides insight into the role of PRMT5 in regulating the metabolism of skeletal muscle. While the paper provides some interesting data, it is severely hampered by the improper WT controls lacking the cre alleles. In order for the data to be reliable, all of the WT samples will need to be replaced with PRMT5+/+ Myl1 cre mice. There are other papers looking at the role of PRMT5 in skeletal muscle, however these are more focused on muscle stem cells than the myofibers. Therefore this paper does fill in a gap with regards to the metabolic role of PRMT5 and how this can affect skeletal muscle function. This paper would most likely be of interest to a specialized audience, mostly those in the skeletal muscle field.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements [optional]

We thank the reviewers for their comments and very helpful suggestions to improve the manuscript. All the reviewers address that further confirmation of the causality of activity-induced AMPK activation and AMPK-induced mitochondrial fission and mitophagy regulating dendritic outgrowth in immature neurons would strengthen the significance of this study. We believe that this is the first study demonstrating that AMPK mediates activity-dependent dendritic outgrowth of immature neurons, and that regulation of mitophagy is critical for dendrite development.

We can perform most of the experimentations and corrections requested by the reviewers. We have already made several revisions and are currently working on additional experiments. All experiments will be finished in several weeks and we expect to submit a full revision by the due date.

2. Description of the planned revisions

Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

Reviewer #1-1.- MMP alone is not a good indicator of mitochondrial health. For instance, ATPase inhibitor causes increase in MMP and complex I inhibition diminish MMP and in both cases mitochondrial function is impaired. On the other hand, authors use increased flickering and mitochondrial ROS production as an indicator of enhanced respiration but they could also be used as indicators of mitochondrial dysfunction. Other assays, such as oxygen consumption, are needed to assess the mitochondrial function.”

*Related comments by Reviewer #2-C. In figure 6 it is unclear what is the significance of the TMRM "flickering" parameter quantified and the difference between the control and knockdown condition is small on average. *

Increase in TMRM flickering and mitochondrial ROS production, which we used as indicators of enhanced mitochondrial respiration, can certainly also be caused by mitochondrial dysfunction. We think it difficult to adopt an oxygen consumption assay in our system, as the transfection efficiency in the primary hippocampal culture is low (~10%). Instead, we plan to assess the mitochondrial function in control and AMPK deficient cells by using an ATP FRET sensor targeted to mitochondria (Mito-ATeam, Imamura et al., PNAS, 2009; Yoshida et al., Methods Mol Biol 2017). Mito-ATeam will be transfected in neurons to compare mitochondrial ATP synthesis in control and AMPK deficient neurons.

*Reviewer #1-2.- It would be interesting to show a better characterization of the mitophagy flux and to test whether pharmacological or genetic stimulation of mitophagy could revert the effect of AMPK KD on dendritic outgrowth, ultimately linking AMPK, mitophagy and dendritic outgrowth. The latter experiments may be challenging but not impossible, for example see (PMID: 27760312). *

We understand that it is important to demonstrate more strongly the correlation of the AMPK-induced peripheral fission and subsequent mitophagy of fragmented mitochondria with dendritic outgrowth. We will attempt the suggested experiment to see if induction of autophagy could revert the dendritic hypoplasia by AMPK KD. However, because AMPK deficiency generates elongated mitochondria defective in fission rather than fragmented mitochondria that are failed to undergo mitophagy, we doubt that activating mitophagy will properly remove damaged mitochondria.

In parallel to the above experiments, we currently analyze if inhibition of mitochondrial fission or mitophagy would phenocopy the hypoplastic dendrites of AMPK-deficient neurons, and if the activation of fission would rescue the phenotypes of AMPK KD, to strengthen the causality of AMPK-dependent fission, autophagy and dendrite outgrowth. So far we have observed that inhibition of mitochondrial fission by MFF knockdown or inhibition of autophagy by bafilomycin treatment strongly suppress dendrite outgrowth. MFF knockdown also leads to the elongation of mitochondria with decreased association of p62-puncta, strikingly reminiscent of AMPK-deficient neurons. Please see attached figures. Completed analyses will be included in the full revision.

*Reviewer #1-4.- Results clearly indicate that AMPK enhances mitochondrial fission, and that AMPK is necessary for proper dendritic outgrowth. However, as indicated, the role of AMPK-dependent mitochondrial fission in promoting dendritic growth is not well demonstrated. A possible, and not very difficult experiment, would be the expression of non-phosphorylable MFF S155/172 mutant (perhaps is also needed to knock down the endogenous MFF). Use of this mutant would abolish AMPK-dependent mitochondrial fission while preserving its other functions. *

Related comments by Reviewer #3-3. The authors could further confirm the claim by examining how mutations in Mff and ULK2 which cannot be phosphorylated by AMPK can rescue defects in mitochondrial fission and spine density.

We will examine if the expression of non-phosphorylable MFF S155/172 mutant would cause defective autophagy and dendritic arbor growth similarly to AMPK KD neurons. In addition, we will test whether MFF S155/172 mutant would inhibit activity-induced mitochondrial fission to strengthen the link between activity-AMPK-MFF-autophagy axis and dendritic outgrowth.

*Reviewer #1-Minor 2.- It is intriguing that as shown in Fig. 2A, rather than an increase in pAMPK/AMPK at DIV5 seems there is less phosphorylation despite FRET analysis indicate more AMPK activation. On the other hand, most of the blots in Fig. 6 seem to be overexposed. *

The exposure time of WB in Fig. 2A was adjusted so that all lanes can be compared. We will fix the exposure time.

Reviewer#2-A. Most of the evidence on the role of AMPKa2 relies on a shRNA-based strategy. The authors have performed this approach with the best practice, including selecting 2 shRNA plasmids for each gene, and performing a rescue experiment with shRNA -resistant cDNA. Yet, it is critical to provide stronger evidence with all the tools available to demonstrate the role of AMPKa2 in dendritic development. This is especially important because the effect reported by the authors is a transient effect: indeed, dendritic development appears abnormal in very young neurons (P5) but largely normal afterwards (P10). Hence one cannot discard a non specific effect on cell viability or sampling effect. The number of neurons counted is fairly low (about 30 neurons per condition) and it is not clear if they come from several independent cultures. It is known that plasmid preparation can impact cell viability and performing the experiment with only one batch of plasmid prep could explain why one plasmid would produce a short-lived effect on cell morphology. Two shRNA constructs are presented in figure S2A but only one is used for morphological experiments quantified in S2D-E with again a very low N number. The specific experiments I would recommend would be to increase the N: at least 25-30 neurons counted per culture, 3 independent cultures, and presenting the results of the two shRNA plasmids for both AMPKa1 and AMPKa2. Furthermore, the immunofluorescence validation of knockdown provided in figure S2B is not really convincing, a nuclear marker is lacking to determine where cells are (it seems that many cells are present in the image, maybe some of them with low AMPKa2 expression as well). A quantification should be provided as well as evidence for shRNA #1 and #2. *

*

We thank the reviewer for valuable suggestions to improve our manuscript. All the knockdown analyses were done from three independent experiments using different mouse litters and multiple batches of plasmid prep. N number was low because of a low transfection efficiency in the primary culture. We will repeat experiments and increase the sample number. We will also present results of the two shRNA constructs. We will redo the immunofluorescence for validation of shRNA knockdown and replace Extended Data Fig. 2B which was pointed out as not being clear.

*Reviewer #2-C. The observation, in vivo, that dendritic development is normal at P10 is intriguing but this reconciles the observation of altered dendritic development with previous studies demonstrating that AMPK knockout has little effect on brain development, as well as previous studies (Mairet-Coello et al. Neuron 2013, Lee et al. Nat commun 2022) targeting AMPKa2 in the hippocampus of AD mouse models by in utero electroporation. This is a critical aspect of the paper and as stated in the discussion, the previous studies only looked at the end product (neuronal morphology appears normal after development) but not the process of neuronal development and maturation. The in vitro experiment offer the possibility to study dendritic development over time in the same population of neurons, either through selected time points, or through time lapse imaging. This would strengthen one of the most original aspect of this work. *

We thank the reviewer for an important suggestion. We will analyze if dendritic morphology and mitochondria would recover in later stages in culture. However, the dendritic growth defects in AMPK KD neurons are apparently more severe in culture and our preliminary results have shown that dendritic growth defects and mitochondrial elongation persist until 10DIV. We anticipate that AMPK deficiency is complemented by certain compensation mechanisms in vivo that are not present in culture, such as chemical signals or synaptic inputs from correct afferents. We will confirm the recovery of dendritic outgrowth in vivo using an AMPK alpha2 knockout mouse. We will include the results in vivo and in vitro in the revised manuscript.

* The authors use a FRET probe to witness AMPK activity, and this part raises a lot of questions. A lot of the signal matches the regularly spaced activity peaks suggesting that FRET response is a coincidence detector of calcium waves. Hence, is the FRET signal influenced by intracellular calcium concentration, or changes in pH? To address this question, the proper control would be to use a FRET biosensor with a mutated AMPK phosphorylation site and demonstrating the absence of response to calcium waves. *

We think it unlikely that the FRET probe detects calcium concentration or pH change, as its kinetics and timing are different from calcium spikes. For confirmation, we will examine a FRET probe lacking phosphorylation sites to negate that calcium waves directly activate the FRET probe.

* Also, the parameter used for quantification is a so-called "number of FRET peaks over 3 minutes" for which the biological significance is unknown. On average there are 1-2 such "peaks" in control conditions (figure 4). These peaks have low amplitude, sometimes around 0.05-0.1 of the YFP/CFP ratio, which is about what is expected even in AMPKa2 knockdown cells (figure S4C). Are there changes in the baseline of FRET signal? *

We monitor FRET at 3-5 sec intervals and is set to 3 minutes due to gradual photobleaching. Although the event frequency is 0-4 times per 3 minutes observation, it is nearly absent in AMPK KD (1 small peak in 3 cells out of 40 cells) or activity deprivation, which we consider a significant difference. We have replaced Figure 4B, 4D, 4I, 4J andExtended Data Fig.4E. The basal FRET signal is lowered in AMPK KD cells, but also varies depending on the expression level of the probe. For comparision of the results shown in Figure 4 and Extended Data Fig.4, we have changed the y-axis to the normalized FRET signal {FRET/FRETbaseline} and jRGECO signals (DF/F0) in Fig. 4F, Extended Data Fig.4C, 4D.

*

*

*Finally, given that calcium peaks and AMPK activity peaks overlap, one key observation is the continued presence of calcium peaks upon AMPKa2 knockdown in figure S4D. Yet, the scale for jRGECO1 intensity in figure S4D differs from the scale in figure 4, making it difficult to interpret. It seems that on average the delta (peak-baseline) is 2000 in wild-type cells (figure 4), compared to 500 in AMPKa2 knockdown cells, which suggests a strong reduction in calcium signal amplitude upon knockdown of AMPK. This should be clarified to demonstrate that the FRET probe peaks are really due to AMPK activity. Also, the effect of STO-609 should be added to this figure. *

We think that the presence of calcium transients in AMPK KD cells supports our conclusion that AMPK is downstream of calcium signaling. The amplitude of calcium spikes was actually lowered in AMPK KD cells. We think it is due to the reduction of the cell size and complexity in KD cells. To negate that AMPK inhibition affects calcium influx, we will examine if acute inhibition by an AMPK inhibitor will suppress only FRET signals but not calcium waves. In addition, we will monitor calcium waves and FRET signals in neurons treated with STO-609 or AICAR. STO-609 and AICAR should decrease and increase FRET signals without affecting calcium influx.

• Other comments by Reviewer #2*
• Similarly, the number of events in figure 5F-G is really low. Is a difference between 0.02 in the control group and 0.01 in the knockdown group physiologically relevant?*

Since p62 puncta contact only a small mitochondrial region, the overlap area of mitochondria with p62 in the total mitochondrial area is small. We will analyze the number of p62 puncta associated with mitochondria per unit dendritic area.

1. • Lines 339-350, the authors discuss about a putative regulatory loop involving AMPK dephosphorylation. Since this part of the discussion is based on the FRET signal, the authors should consider if an alternative explanation could be the kinetics of the biosensor dephosphorylation.* We will revise Discussion to argue about alternative possibilities of dynamic oscillation of the FRET signal when we get data from the above experiments.

*In terms of significance, I would have two major criticisms. The first is that it appears that many of the findings by the authors are redundant with observations of the roles of CAMKK2-AMPK-MFF-ULK1 in AD model mice, see for example the work by Polleux (Mairet-Coello et al. Neuron 2013, Lee et al., Nat commun 2022). As said above, my opinion is that the paper should put more emphasis on the transient effect of AMPK, which would be a novel observation and, as the authors rightfully discuss, a phenotype potentially overlooked in previous studies of AMPK KO mice. The second is that many points in the discussion seem to be over reached and are not entirely supported by the data. As an example lines 298-299 "leading to mitochondrial dysfunction with low respiratory activity" (not addressed in this manuscript), lines 312-313 "multiple signatures of mitochondrial dysfunction such as reduced delta-Psi-m and ROS production" (biological significance of these parameters?), lines 332-334 "AMPK phosphorylation dynamically oscillates in dendrites, depending on Ca2+ influx and CAMKK2 activity, while it is independent of LKB1" (the authors don't study AMPK phosphorylation, and the experimental data has many limitations that need addressing), etc. *

We thank reviewer’s guidance. We think this is the first study showing AMPK function in dendritic arbor growth in immature neurons before synaptogenesis. We will rewrite the manuscript to emphasize that neuronal activity in immature neurons regulates dendrite formation via AMPK in a short time window during brain development. Discussion will be revised according to the data of the ongoing additional experiments.

Reviewer#3-1. All these studies are done in invitro neuronal culture modal with transfection of ShRNAs to Knockdown AMPK. An alternative possibility is that authors could use an AMPK Conditional Knockout mouse models Conditional deletion of (AMPKα1/α2 (AMPKα1−/−; AMPKα2F/F; Emx1-Cre) derived neurons for this study.

We showed the effect of AMPK knockdown in hippocampal neurons in culture and in vivo (Fig. 2). For validation, we also examined CRISPR interference (Extended data Fig.2). We will examine in vivo phenotypes in pyramidal neurons in AMPK alpha2 knockout mice to further validate our observation.

Description of the revisions that have already been incorporated in the transferred manuscript

Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. If no revisions have been carried out yet, please leave this section empty.

*Reviewer #1-1.- Authors use mitochondrial membrane potential (MMP), MMP flickering and mitochondrial ROS production as indicators of mitochondrial function, but this is not convincing. To analyze MMP, authors use TMRM fluorescence normalized by mitochondria area. This is not correct, using this strategy would mean that a symmetric fission would instantly double MMP and fusion would half MMP. The analysis must be made by tracing ROIs of the same surface in different mitochondria and determining TMRM fluorescence in these ROIs. *

We have reanalyzed TMRM fluorescence using the method indicated. As a result, TMRM fluorescence show a slight but significant decrease (p=0.0071) in AMPK KD cells. Extended Data Fig. 5C has been replaced accordingly. We thank the Reviewer for kind guidance!

Reviewer #1-3.- The authors treat neurons with glutamate to support the view that synaptic activity activates AMPK and promotes mitochondrial fission. However, the concentration used (100 mM) may be excitotoxic. Synaptic activity can be induced by electric field stimulation, although this require equipment that may not be available in the authors' lab. Another alternative is network disinhibition with bicuculine or to use lower concentrations of glutamate. In any case, since neurons are immature and may respond differently from mature neurons, it would be worth to verify synaptic activity by analyzing Ca2+ transients.

*Reviewer #1-Minor 3.- It is necessary more explanation about spontaneous Ca2+ transients in immature cultures. What percentage of neuros experience it? Is it synchronized? *

*Related comments by Reviewer#2-D. It is well established and thus not surprising that AMPK activity increases in response to synaptic activity. It is more surprising to witness such an effect of activity in very immature neurons, where presumably synapses are sparse and not well developed. For example dendritic segments in Figure 1E and 3A don't have dendritic spines. Western-blot and/or immunofluorescence of synaptic markers with comparison to fully mature neurons would complete figure 1 and make the case whether the reported effects are marginal or a strong driver for dendritic development and AMPK regulation. *

We thank the reviewers’ point that we failed to emphasize in the original manuscript.

We focus on AMPK function during activity-dependent dendritic outgrowth in immature neurons before the onset of synaptogenesis. It has been shown that synaptogenesis occurs in dissociated hippocampal cultures between 7-12 DIV (eg, Renger et al., Neuron 2001) and that developing dendrites at 5 DIV are activated by ambient glutamate which is spontaneously released from nearby immature axon terminals and undergo spontaneous Ca2+ transients, and this non-synaptic activity is important for dendritic outgrowth (Andreae and Burrone, Cell Rep 2015). We have observed that Ca2+ transients in individual neurons are variable in frequency and magnitude and are not synchronized in consistent with previous studies. We have performed immunofluorescence with a synaptic marker PSD95 and confirmed that dendritic spines are not yet differentiated and PSD95 is sparsely distributed along the dendritic shaft in DIV5 hippocampal neurons. We describe the nature of Ca2+ transients in the Results more clearly and provide high magnified images and immunofluorescence with a synaptic marker PSD95 of the neurons at DIV5 and DIV13 as a new Fig. 1A. We believe that this is the first indication of AMPK function in non-synaptic neuronal activity during dendritogenesis.

We have observed induction of mitochondrial fission in neurons treated with 1 µM glutamate. Extended Data Fig. 1E has been replaced accordingly. Since GABA is known to induce depolarization in immature neurons (Soriano et al., PNAS 2008), we would like to exclude bicuculine treatment from this analysis.

*Reviewer #1-5.- The statistical analysis seems appropriate, but it is confusing that sometimes non-parametric and sometimes parametric tests are used. It is not indicated which test is used to determine normality since the methods section lacks a statistical analysis section.

*

We have revised Methods and have described statistical analysis in detail.

*Reviewer #1-Minor 1.- Authors should double check the analysis shown in Fig. 1A. As it is shown, Ca2+ transients are 2-3% higher than basal, when the video shown in video 1 seem to indicate much more. *

Thank you for pointing this out. In the original version, the percentile change was erroneously measured across the entire visual field, including areas without neurons. We have replaced Fig. 1B (original Fig. 1A) with reanalyzed data in the proximal region of the apical dendrite.

*Reviewer #1-Minor 4.- It is interesting that AMPK KD in vivo impairs dendritic architecture at P5, however at P10 the defect seem to be somehow compensated. This result apparently detracts from the relevance of the findings, however last year was published a paper in which in an animal model of Huntington's disease dendritic architecture is delayed during the first week but normalizes thereafter. Despite later normalization in dendritic architecture, this early defect in maturation has effects in adulthood as pharmacological restoration of arborization during the neonatal period suppresses some phenotypes observed in adulthood (PMID: 36137051). I believe that discussing this paper would help the reader to recognize the potential relevance of the findings. *

*Related comments by Reviewer #2: Nonetheless let aside the technical concern, if their findings hold true, this is an intriguing mechanism. There are interesting parallels to be made with observations of altered morphology and excitability of neurons in Huntington's disease model mice during the first postnatal week. These changes spontaneously reverts and are undetectable in the second week (Braz et al. Science 2022). Thus, precedent suggests that indeed dendritic development can take a slow course, and this study also suggests that this is important later since normalization of abnormal excitability during the first week in HTT mice prevents some of the phenotypes later in life. Here again, an interesting parallel could be made with the known role of AMPK in synaptic loss in AD models. *

We thank the reviewers for the supportive comments. We will refer this paper and discuss about potential significance of the transient defects in early dendrite morphology in AMPK deficient neurons.

*Reviewer#2-B. The Crispr method lacks validation which should be provided somehow. The drug-based experiment relies on compound C, a notoriously non specific AMPK inhibitor (see for example Bain et al. Biochem J 2007, or Vogt et al. Cell Signal 2011). Data obtained with Compound C is hard to interpret given the number of kinases that are affected by the drug and should be removed from the manuscript. *

We have added immunofluorescence images for validation of AMPK deletion by CRISPRi (Extended Data Fig. 2F).

We think the results of Compound C treatment support our conclusion in combination with KD and CRISPRi, but will delete the results in accordance with this comment.

• Other comments by Reviewer#2*
• Figure 5A-C relies on the quantification of fission events that appear very rare (0.4 event per 20 minutes). The difference between the two groups is between 0.1 and 0.2 events on average. Since this was quantified on a fairly low number of cells (N=14), it is hard to appreciate exactly how many events have been observed and the actual physiological relevance. Furthermore individual datapoints should be added to the figure to estimate variability.*

The number of fission events was counted in mitochondria in a unit length of dendrite of similar diameter, and normalized by the number of mitochondria. The values were thus small as they represent average number of events in one mitochondrion in 20 minutes. We have replaced the Fig. 1K, 3F, 5B and 5C to show the number of fission events in mitochondria included in a unit length of dendrites of similar diameter. Individual data points have been included.

Reviewer #3-4. Authors showed activity-dependent calcium signaling controls mitochondrial homeostasis and dendritic outgrowth via AMP-activated protein kinase (AMPK) in developing hippocampal neurons do the cortical mitochondria respond the same way as the hippocampal neurons?

Thank you for the comment. As pyramidal neurons in the cerebral cortex and hippocampus are basically the same origin, it is likely that they share the same signaling. We use hippocampal neurons in this study to perform quantitative analysis of dendritic morphology in the same type of neurons. Primary cultures of cortical neurons contain multiple different cell types, making it difficult to analyze the same cell type.

4. Description of analyses that authors prefer not to carry out

Please include a point-by-point response explaining why some of the requested data or additional analyses might not be necessary or cannot be provided within the scope of a revision. This can be due to time or resource limitations or in case of disagreement about the necessity of such additional data given the scope of the study. Please leave empty if not applicable.

*Reviewer #1: If activity is observed in only a portion of the neurons, taking advantage of the stablished long-term live imaging protocol in the authors' lab, it would be interesting to study in the same culture whether neurons that experience spontaneous activity develop more than those that do not. *

We prefer not to carry out this analysis, as activity-dependent dendritic growth has already been well described in previous papers. It will take considerable time to observe the number of neurons for analysis of correlation between Ca2+ transients and dendrite morphology. We would like to focus our effort to demonstrate AMPK signaling during activity-dependent dendritic growth.

Reviewer#3-2. Another technical issue here, most of the experiments are carried out on Neurobasal media, which has a lot of glucose plus substitution of glutamax might be not the perfect conditions for AMPK. Authors could not obtain evidence supporting the regulation of mitochondria biogenesis by PGC1α phosphorylation and expression. This surprise me, if you could reduce the glucose concentration if might change.

We observed little or no changes in phosphorylation of PGC1alpha by enhancing or suppressing neuronal activity or AMPK activity. As mitochondrial biogenesis is very active in growing neurons, we surmise that PGC1alpha and mitochondrial biogenesis is regulated by multiple mechanisms during neuronal differentiation and AMPK activation/inhibition might not induce visible changes. We agree the reviewer that there is room to seek the conditions under which changes in PGC1alpha can be detected, but we do not see why Neurobasal plus glutamax is not suitable for this study. Multiple papers studying AMPK function in cultured neurons use similar culture media (Sample et al., Mol. Cell. Biol., 2015; Muraleedharan et al., Cell. Rep., 2020; Lee et al., Nat. Commun., 2022). We might see PGC1 phosphorylation by glucose deprivation, as it decreases glycolysis-derived ATP and thereby activates AMPK. Since we focus on AMPK activation by calcium signals, we are afraid that it would be difficult to distinguish AMPK activation by ATP deficiency or calcium signaling in glucose deficient condition. In addition, glucose deprivation would affect neuronal activity (which consumes large amount of ATP) and neuronal differentiation including dendritic outgrowth.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Hatsuda and colleagues builds on previous work from their group and others, to further investigate the activity-dependent dendritic arbor development and AMPK-dependent mitochondrial quality control. It recently been illustrated that over-activation of the CAMKK2-AMPK kinase dyad mediates synaptic loss through coordinated phosphorylation of MFF-dependent mitochondrial fission and ULK2-dependent mitophagy (Aβ42 oligomers trigger synaptic loss through CAMKK2-AMPK-dependent effectors coordinating mitochondrial fission and mitophagy). Though this is progress is modest, present study showed neuronal activity induces activation of two downstream effectors of AMPK, MFF and ULK1, which are the key regulators of mitochondrial fission and mitophagy.

Major Concerns:

1. All these studies are done in invitro neuronal culture modal with transfection of ShRNAs to Knockdown AMPK. An alternative possibility is that authors could use an AMPK Conditional Knockout mouse models Conditional deletion of (AMPKα1/α2 (AMPKα1−/−; AMPKα2F/F; Emx1-Cre) derived neurons for this study.
2. Another technical issue here, most of the experiments are carried out on Neurobasal media, which has a lot of glucose plus substitution of glutamax might be not the perfect conditions for AMPK. Authors could not obtain evidence supporting the regulation of mitochondria biogenesis by PGC1α phosphorylation and expression. This surprise me, if you could reduce the glucose concentration if might change.
3. The authors could further confirm the claim by examining how mutations in Mff and ULK2 which cannot be phosphorylated by AMPK can rescue defects in mitochondrial fission and spine density.
4. Authors showed activity-dependent calcium signaling controls mitochondrial homeostasis and dendritic outgrowth via AMP-activated protein kinase (AMPK) in developing hippocampal neurons do the cortical mitochondria respond the same way as the hippocampal neurons?

Significance

It recently been illustrated that over-activation of the CAMKK2-AMPK kinase dyad mediates synaptic loss through coordinated phosphorylation of MFF-dependent mitochondrial fission and ULK2-dependent mitophagy (Aβ42 oligomers trigger synaptic loss through CAMKK2-AMPK-dependent effectors coordinating mitochondrial fission and mitophagy). Though this is progress is modest, present study showed neuronal activity induces activation of two downstream effectors of AMPK, MFF and ULK1, which are the key regulators of mitochondrial fission and mitophagy.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript 'Calcium signals tune AMPK activity and mitochondrial homeostasis in dendrites of developing neurons' by Hatsuda and collaborators aims at studying the interrelationship between neuronal activity, mitochondria dynamics (ie. fusion and fission mechanisms) and dendritic development. The authors provide evidence linking activity-dependent activation of a CAMKK2-AMPK pathway and the regulation of mitochondria fission and autophagy. Based on the literature, they focus on the roles of the mitochondria fission factor MFF and the autophagy regulator ULK1, both previously known targets of AMPK.

This work parallels previous observations in the context of Alzheimer's disease and as such the discovery of a molecular link between CAMKK2-AMPK, MFF/ULK1 and the regulation of dendritic mitochondria is not surprising. The change of biological context raises interesting question although the relevance of these observations is not addressed in this manuscript.

As a general comment, the work is well structured, reads easily. Iconography and figures organization are good. Major criticisms would concern the tools used to study AMPK and challenge some of the observations, as such I believe these are essential to address to validate the findings.

Major comments

• A. Most of the evidence on the role of AMPKa2 relies on a shRNA-based strategy. The authors have performed this approach with the best practice, including selecting 2 shRNA plasmids for each gene, and performing a rescue experiment with shRNA -resistant cDNA. Yet, it is critical to provide stronger evidence with all the tools available to demonstrate the role of AMPKa2 in dendritic development. This is especially important because the effect reported by the authors is a transient effect: indeed, dendritic development appears abnormal in very young neurons (P5) but largely normal afterwards (P10). Hence one cannot discard a non specific effect on cell viability or sampling effect. The number of neurons counted is fairly low (about 30 neurons per condition) and it is not clear if they come from several independent cultures. It is known that plasmid preparation can impact cell viability and performing the experiment with only one batch of plasmid prep could explain why one plasmid would produce a short-lived effect on cell morphology. Two shRNA constructs are presented in figure S2A but only one is used for morphological experiments quantified in S2D-E with again a very low N number. The specific experiments I would recommend would be to increase the N: at least 25-30 neurons counted per culture, 3 independent cultures, and presenting the results of the two shRNA plasmids for both AMPKa1 and AMPKa2. Furthermore, the immunofluorescence validation of knockdown provided in figure S2B is not really convincing, a nuclear marker is lacking to determine where cells are (it seems that many cells are present in the image, maybe some of them with low AMPKa2 expression as well). A quantification should be provided as well as evidence for shRNA #1 and #2.
• B. To complete the shRNA-based experiments, the authors use a single cell Crispr approach, as well as a pharmacological approach. The Crispr method lacks validation which should be provided somehow. The drug-based experiment relies on compound C, a notoriously non specific AMPK inhibitor (see for example Bain et al. Biochem J 2007, or Vogt et al. Cell Signal 2011). Data obtained with Compound C is hard to interpret given the number of kinases that are affected by the drug and should be removed from the manuscript.
• C. The observation, in vivo, that dendritic development is normal at P10 is intriguing but this reconciles the observation of altered dendritic development with previous studies demonstrating that AMPK knockout has little effect on brain development, as well as previous studies (Mairet-Coello et al. Neuron 2013, Lee et al. Nat commun 2022) targeting AMPKa2 in the hippocampus of AD mouse models by in utero electroporation. This is a critical aspect of the paper and as stated in the discussion, the previous studies only looked at the end product (neuronal morphology appears normal after development) but not the process of neuronal development and maturation. The in vitro experiment offer the possibility to study dendritic development over time in the same population of neurons, either through selected time points, or through time lapse imaging. This would strengthen one of the most original aspect of this work.
• D. It is well established and thus not surprising that AMPK activity increases in response to synaptic activity. It is more surprising to witness such an effect of activity in very immature neurons, where presumably synapses are sparse and not well developed. For example dendritic segments in Figure 1E and 3A don't have dendritic spines. Western-blot and/or immunofluorescence of synaptic markers with comparison to fully mature neurons would complete figure 1 and make the case whether the reported effects are marginal or a strong driver for dendritic development and AMPK regulation. Furthermore the authors use a FRET probe to witness AMPK activity, and this part raises a lot of questions. A lot of the signal matches the regularly spaced activity peaks suggesting that FRET response is a coincidence detector of calcium waves. Hence, is the FRET signal influenced by intracellular calcium concentration, or changes in pH? To address this question, the proper control would be to use a FRET biosensor with a mutated AMPK phosphorylation site and demonstrating the absence of response to calcium waves. Also, the parameter used for quantification is a so-called "number of FRET peaks over 3 minutes" for which the biological significance is unknown. On average there are 1-2 such "peaks" in control conditions (figure 4). These peaks have low amplitude, sometimes around 0.05-0.1 of the YFP/CFP ratio, which is about what is expected even in AMPKa2 knockdown cells (figure S4C). Are there changes in the baseline of FRET signal? Finally, given that calcium peaks and AMPK activity peaks overlap, one key observation is the continued presence of calcium peaks upon AMPKa2 knockdown in figure S4D. Yet, the scale for jRGECO1 intensity in figure S4D differs from the scale in figure 4, making it difficult to interpret. It seems that on average the delta (peak-baseline) is 2000 in wild-type cells (figure 4), compared to 500 in AMPKa2 knockdown cells, which suggests a strong reduction in calcium signal amplitude upon knockdown of AMPK. This should be clarified to demonstrate that the FRET probe peaks are really due to AMPK activity. Also, the effect of STO-609 should be added to this figure.

Other comments

• A. Figure 5A-C relies on the quantification of fission events that appear very rare (0.4 event per 20 minutes). The difference between the two groups is between 0.1 and 0.2 events on average. Since this was quantified on a fairly low number of cells (N=14), it is hard to appreciate exactly how many events have been observed and the actual physiological relevance. Furthermore individual datapoints should be added to the figure to estimate variability.
• B. Similarly, the number of events in figure 5F-G is really low. Is a difference between 0.02 in the control group and 0.01 in the knockdown group physiologically relevant?
• C. In figure 6 it is unclear what is the significance of the TMRM "flickering" parameter quantified and the difference between the control and knockdown condition is small on average. Rather, the data presented in figure S5 would suggest that there is no difference in TMRM signal.
• D. Lines 339-350, the authors discuss about a putative regulatory loop involving AMPK dephosphorylation. Since this part of the discussion is based on the FRET signal, the authors should consider if an alternative explanation could be the kinetics of the biosensor dephosphorylation.

Significance

The manuscript by Hatsuda and collaborator studies the roles of neuronal AMPK in the development of hippocampal neurons, specifically the authors describe a transient effect on dendritic development. To this reviewer's opinion, this is the major findings of this paper. Although the physiological implications of this finding are unknown, this is beyond the scope of this paper.

Yet in terms of significance, I would have two major criticisms. The first is that it appears that many of the findings by the authors are redundant with observations of the roles of CAMKK2-AMPK-MFF-ULK1 in AD model mice, see for example the work by Polleux (Mairet-Coello et al. Neuron 2013, Lee et al., Nat commun 2022). As said above, my opinion is that the paper should put more emphasis on the transient effect of AMPK, which would be a novel observation and, as the authors rightfully discuss, a phenotype potentially overlooked in previous studies of AMPK KO mice. The second is that many points in the discussion seem to be over reached and are not entirely supported by the data. As an example lines 298-299 "leading to mitochondrial dysfunction with low respiratory activity" (not addressed in this manuscript), lines 312-313 "multiple signatures of mitochondrial dysfunction such as reduced delta-Psi-m and ROS production" (biological significance of these parameters?), lines 332-334 "AMPK phosphorylation dynamically oscillates in dendrites, depending on Ca2+ influx and CAMKK2 activity, while it is independent of LKB1" (the authors don't study AMPK phosphorylation, and the experimental data has many limitations that need addressing), etc.

Nonetheless let aside the technical concern, if their findings hold true, this is an intriguing mechanism. There are interesting parallels to be made with observations of altered morphology and excitability of neurons in Huntington's disease model mice during the first postnatal week. These changes spontaneously reverts and are undetectable in the second week (Braz et al. Science 2022). Thus, precedent suggests that indeed dendritic development can take a slow course, and this study also suggests that this is important later since normalization of abnormal excitability during the first week in HTT mice prevents some of the phenotypes later in life. Here again, an interesting parallel could be made with the known role of AMPK in synaptic loss in AD models.

Reviewer expertise: I have expertise in neuronal metabolism

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Hatsuda et al study the role of synaptic activity-dependent Ca2+ signals in activating AMPK and regulating mitochondrial homeostasis and dendritic outgrowth in developing neurons. Using cultures of immature primary hippocampal neurons, the authors found that Ca2+ transients activate AMPK, which regulates mitochondrial fission and dendritic growth. AMPK KD and pharmacological manipulation of AMPK activity confirmed a role of AMPK in mitochondrial morphology that correlated with impaired mitophagy and mitochondrial function. These findings led the authors to conclude that activity-dependent activation of AMPK promotes mitochondrial fission, which facilitates removal of dysfunctional mitochondria and thus contributes to maintaining a healthy mitochondrial pool that is necessary in the intense energetic effort that requires dendritic outgrowth.

Major comments:

Meanwhile the activation of AMPK and its role in mitochondrial fission and dendritic outgrowth is in general well demonstrated, the conclusion that AMPK is necessary for proper dendritic outgrowth because by promoting asymmetric mitochondrial fission facilitates mitophagy of dysfunctional mitochondria and thus ensures adequate generation of energy for dendritic outgrowth still seems preliminary.

1. Authors use mitochondrial membrane potential (MMP), MMP flickering and mitochondrial ROS production as indicators of mitochondrial function, but this is not convincing. To analyze MMP, authors use TMRM fluorescence normalized by mitochondria area. This is not correct, using this strategy would mean that a symmetric fission would instantly double MMP and fusion would half MMP. The analysis must be made by tracing ROIs of the same surface in different mitochondria and determining TMRM fluorescence in these ROIs. But even in the case that there were changes in MMP, that it does not seem to be the case, MMP alone is not a good indicator of mitochondrial health. For instance, ATPase inhibitor causes increase in MMP and complex I inhibition diminish MMP and in both cases mitochondrial function is impaired. On the other hand, authors use increased flickering and mitochondrial ROS production as an indicator of enhanced respiration but they could also be used as indicators of mitochondrial dysfunction. Other assays, such as oxygen consumption, are needed to assess the mitochondrial function.
2. It would be interesting to show a better characterization of the mitophagy flux and to test whether pharmacological or genetic stimulation of mitophagy could revert the effect of AMPK KD on dendritic outgrowth, ultimately linking AMPK, mitophagy and dendritic outgrowth. The latter experiments may be challenging but not impossible, for example see (PMID: 27760312).
3. The authors treat neurons with glutamate to support the view that synaptic activity activates AMPK and promotes mitochondrial fission. However, the concentration used (100 M) may be excitotoxic. Synaptic activity can be induced by electric field stimulation, although this require equipment that may not be available in the authors' lab. Another alternative is network disinhibition with bicuculine or to use lower concentrations of glutamate. In any case, since neurons are immature and may respond differently from mature neurons, it would be worth to verify synaptic activity by analyzing Ca2+ transients.
4. Results clearly indicate that AMPK enhances mitochondrial fission, as previously reported, and that AMPK is necessary for proper dendritic outgrowth. However, as indicated, the role of AMPK-dependent mitochondrial fission in promoting dendritic growth is not well demonstrated. For example, AMPK could regulate dendritic outgrowth through its role on cytoskeletal dynamics. A possible, and not very difficult experiment, would be the expression of non-phosphorylable MFF S155/172 mutant (perhaps is also needed to knock down the endogenous MFF). Use of this mutant would abolish AMPK-dependent mitochondrial fission while preserving its other functions.
5. The statistical analysis seems appropriate, but it is confusing that sometimes non-parametric and sometimes parametric tests are used. It is not indicated which test is used to determine normality since the methods section lacks a statistical analysis section.

Minor comments:

1. Authors should double check the analysis shown in Fig. 1A. As it is shown, Ca2+ transients are 2-3% higher than basal, when the video shown in video 1 seem to indicate much more.
2. It is intriguing that as shown in Fig. 2A, rather than an increase in pAMPK/AMPK at DIV5 seems there is less phosphorylation despite FRET analysis indicate more AMPK activation. On the other hand, most of the blots in Fig. 6 seem to be overexposed.
3. It is necessary more explanation about spontaneous Ca2+ transients in immature cultures. What percentage of neuros experience it? Is it synchronized? If activity is observed in only a portion of the neurons, taking advantage of the stablished long-term live imaging protocol in the authors' lab, it would be interesting to study in the same culture whether neurons that experience spontaneous activity develop more than those that do not.
4. It is interesting that AMPK KD in vivo impairs dendritic architecture at P5, however at P10 the defect seem to be somehow compensated. This result apparently detracts from the relevance of the findings, however last year was published a paper in which in an animal model of Huntington's disease dendritic architecture is delayed during the first week but normalizes thereafter. Despite later normalization in dendritic architecture, this early defect in maturation has effects in adulthood as pharmacological restoration of arborization during the neonatal period suppresses some phenotypes observed in adulthood (PMID: 36137051). I believe that discussing this paper, and others with similar message (if they exist, I do not know), would help the reader to recognize the potential relevance of the findings.

I believe that all the proposed experiments would strongly help to support the claims of the paper and are perfectly feasible during the time given for a revision and economically affordable.

Significance

As authors already cite in their work, several groups have shown that mitochondrial fission is necessary for proper dendritic growth. Here, the authors have proposed that synaptic activity in immature neurons induce mitochondrial fission via AMPK activation and subsequent MFF phosphorylation. Many of the findings here are confirmations of previous work, for instance activity-dependent activation of AMPK (PMID: 25698741, PMID: 27012879) and induction of mitochondrial fission by AMPK via MFF phosphorylation (PMID: 26816379). The novelty of the work is in studying these processes in immature neurons and how this affects dendritic growth, which is of interest for cellular neuroscientist, but I do not think it represents a conceptual breakthrough. A more detailed understanding of the mechanism proposed, i.e. mitochondrial fission facilities removal of dysfunctional mitochondria to maintain the high energy demands of growing dendrites, would greatly enhance the significance of the study.

This reviewer specializes in neuronal cell biology, specifically the study of the mechanisms by which mitochondrial function regulates aspects of neuronal physiology, including neuritic outgrowth.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Point-by-point response to reviewer comments

General statement

Several studies have previously demonstrated functional links between the death receptors (DR) TRAIL-R1/2 and the Unfolded protein response (UPR). In this manuscript, we describe the previously unrecognized IRE1-dependent dual regulation of the expression of another DR, CD95, and CD95L-induced cell death. Our work therefore adds to the current knowledge on the functional links existing between UPR and DR signaling and provides novel mechanistic insights on a dual regulation involving both transcriptional and post-transcriptional control of the expression of CD95 mRNA expression by IRE1. To demonstrate this, we have used both genetic (overexpression of XBP1s or dominant-negative forms of IRE1) and pharmacologic (IRE1 RNase inhibitor) approaches and cellular models of glioblastoma (GB) and triple-negative breast cancer (TNBC). We show that IRE1 RNase activity promotes CD95 expression and CD95-mediated cell death via the transcription factor XBP1s whilst IRE1 RNase limits CD95 expression and cell death via its ability to cleave RNAs (through RIDD, for Regulated IRE1-dependent decay of RNAs, activity). Furthermore, we report that IRE1-mediated control of CD95 expression is active in vivo, using a model of CD95-mediated fulminant hepatitis in mice. Lastly, we correlate these results to the pathology by showing that CD95 expression is decreased in RIDD high or XBP1s low human GB and TNBC tumors.

We thank the reviewers for their fair assessment of our manuscript and for their insightful comments. Below, we describe the experiments we plan to carry out to address the reviewers’ comments.

Reviewer #1 (Evidence, reproducibility and clarity (Required)): Summary: Here the authors argue that IRE1 activation has opposite effects on Fas/CD95 expression/stability in a number of contexts, via either RIDD-dependent degradation of Fas mRNA or XBP-1-mediated induction of Fas expression, which led to either increased or decreased sensitivity to Fas-induced apoptosis in a number of settings. Major issues: The study is somewhat preliminary and inconclusive in that it is not clear why the RIDD function of XBP-1 appears to predominate in vitro in the cell lines examined, leading to modest increases in Fas expression levels (Figure 1) when IRE1 DN versus IRE1 WT constructs are overexpressed, which is at odds with the latter part of the paper which suggests that inhibition of RIDD led to reduced Fas expression levels. However, this could be due to supraphysiological levels of IRE1 being expressed under overexpression conditions, leading to confounding results. Similarly, when XBP-1s is overexpressed in vitro (Figure 5) the modest increases in CD95/Fas expression and sensitization to Fas-induced cell death may not be fully representative of what would be observed at physiological levels of XBP-1s activation. The in vivo results obtained using an IRE1 RNase inhibitor (MKC8866) contradict the earlier part of the study (as Fas levels decreased and there was protection from Fas-induced liver toxicity) and this could be due to a multitude of reasons. There is no doubt that impacting on IRE1 activity has interesting effects on CD95/Fas expression, which can be up- or -down-regulated, with consequences for cell death induced via engagement of the latter receptor, however, the manuscript does not offer a lot of clarity on which outcome is the predominant one in the context of engagement of the UPR. I have the following suggestions for improvement.

We thank the reviewer for this overall positive assessment.

1. The authors should induce ER stress using Thaps, Brefeldin A and Tunicamycin, and explore the effects of doing this on Fas expression levels in the context of silencing endogenous IRE1, XBP-1 and PERK.

We do agree with this reviewer that the proposed experiments might further highlight which of the IRE1-dependent control of CD95 expression dominates upon ER stress induction. Therefore, we will perform the requested experiment in the various cell lines already used in the manuscript.

We propose to evaluate the expression of CD95 (at the mRNA and total protein levels) under ER stress induction (by different ER stressors) upon knock-down of IRE1 or XBP1. Other DRs (TRAIL-R1 and 2) have been shown to be induced by PERK activation and it is also demonstrated that PERK and IRE1 signaling pathways coregulate each other. As such, we also propose to assess whether PERK could also control CD95 expression in this setting.

1. The authors should explore the effects of silencing of IRE1, XBP-1 and PERK on constitutive Fas expression and the outcome of Fas/CD95-induced apoptosis in cells not experiencing an overt activation of the UPR (i.e. in the absence of Thaps, Brefeldin A or other UPR inducer).

We thank the reviewer for their suggestion and will perform the requested experiments as proposed.

1. The specificity of MKC8866 at the concentration used (30 uM) is unclear. What effect does MKCC have on sensitivity towards Fas-induced apoptosis, similar to the type of experimental set up presented in Figure 5A, 5B?

Regarding the specificity of MKC8866, this drug has been optimized and refined from a family of IRE1-specific endoribonuclease inhibitors initially obtained from a chemical library screen [1-3]. This salicylaldehyde analog has already shown to be effective in multiple cancer models including breast [4, 5] and prostate [2] cancers. We have recently demonstrated its efficacy in a GB mouse model [6]. It is therefore a widely used IRE1 inhibitor, including in the dose range 10-30 mM used in this study (e.g [4, 5]). We therefore do not think it is in the scope of this manuscript to re-assess it specificity. However, we will aim at testing an additional IRE1 inhibitor to assess whether similar effects can be observed on CD95 expression in cells. To do so, we propose to use a novel IRE1 kinase inhibitor developed in the laboratory (DOI: 10.26434/chemrxiv-2022-2ld35 – Accepted iScience) and shown to efficiently blunt IRE1 activity in GB. As also suggested by the reviewer, we will assess whether the use of MKC-8866 can affect CD95L-induced cell death in cell lines.

1. Similarly, what effects does MKC8866 (at 30 uM) have on key Fas pathway determinants, such as Fas, FLIPL, FLIPs, Caspase-8, FADD, RIPK1, A20, CYLD, cIAP-1, cIAP-2 and Bid? There are many points at which MKC8866 could influence the outcome of Fas receptor engagement beyond the receptor itself.

In the present manuscript, we have shown that MKC-8866 reduces CD95 expression in mouse liver (IHC depicted in Figure 4B and S3B) in vivo and that, when used at 30 mM in vitro, it prevents the loss of CD95 expression induced by tunicamycin or thapsigargin in U87 cells (Fig 1C-F). We do agree with the reviewer that IRE1 may impact CD95-induced cell death beyond modulating CD95 expression, as also already discussed in the present manuscript. Therefore, and as suggested, we will assess whether MKC-8866, used at 30 mM, also impacts on the basal cellular expression of the various components of CD95 signaling mentioned by this reviewer.

Minor issues:

1. For the Fas mRNA cleavage experiments presented in Figure 2, there are no irrelevant control mRNAs to allow the reader to judge whether the effects presented are specific to Fas mRNA or are commonly observed for many mRNAs at these amounts of IRE1 (1 ug, 0.5 ug, which appear high).

The expression of Fas mRNA was already normalized to GAPDH (which does not seem to vary upon incubation with IRE1). We nevertheless will test the expression of additional “irrelevant” RNAs as suggested by the reviewer.

Reviewer #1 (Significance (Required)): General assessment: this is an interesting study, as there is little knowledge currently concerning how the UPR influences Fas expression or Fas-dependent outcomes. However, the impact of this work is limited by the overexpression approaches used, which could produce artifactual results, as well as the contradictory message of the study.

Although we think that the message of the manuscript is indeed complex, the work presented herein does not rely exclusively on overexpression approaches as our genetic-based results are also comforted by the use of pharmacologic inhibitors of IRE1.

Advance: the advance reported here is relatively modest and limited in scope due to the inconclusive nature of the data presented.

Audience: this study will be of interests to specialists in the UPR and cell death communities.

We thank the reviewer for acknowledging the overall novelty of our work. We do hope that the experiments proposed will address her/his remaining concerns.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The authors address here for the first time the connection between CD95, which is known as Fas, and ER stress. The role of another DR, TRAIL-R2 has been already reported, but this is the first study uncovering the link between Cd95 system and ER stress. The study is performed on the high level and supported by all necessary controls. They find the connection between IRE1 and CD95 and show that it might play a role in Cd95 signaling and attenuate CD95-mediated cell death.

Further, the correlation between CD95 expression and IRE is found in tumors. Importantly the authors find out the connection between XBP1 and CD95 expression, which was not reported to date. Hence, it is a very important and highly essential piece of research.

We thank the reviewer for these very positive comments and the acknowledging of the novelty and importance of our study.

However, I would like to clarify the several issues:

1: Figure 1. Tunicamycin obviously leads to deglycosylation of CD95, which is indicated by the appearance of 35 Kda band. This should be highlighted and commented.

We agree, this will be commented on in the text.

1. Figure 2c, d. The piece of mRNA structure, which is synthesized, might have the different secondary structure and might be not cleaved by IRE, accordingly. More detailed comments have to be provided in this regard.

The model depicted in Figure 2B is a predicted computational secondary structure of CD95 mRNA. In the experiments performed in Figure 2A, C and D the mRNA was extracted from U87 cells prior to incubation with recombinant IRE1 and the resulting products analyzed using RT-qPCR with primers flanking different portions of the CD95 mRNA sequence. For Figures 2C and D, the primers used flank the two sites which were predicted to be cleaved by IRE1 based on previous work from our lab [7]. Even though we cannot exclude that additional sites can be targeted beyond these two, the fact that the amplification of CD95 sequence is reduced in samples pre-incubated with recombinant IRE1 strongly suggests that IRE1 is indeed able to cleave CD95 mRNA in these regions in vitro. We will modify the main text to further explain this point.

1. Figure 3. Caspase-8-3 western blots show beautiful effects but did authors see some effects further downstream, eg on PARP1 cleavage? Was cell death (not viability) measured as well? Can you comment on this?

This is absolutely right, we will test PARP-1 cleavage in this setting as suggested. Given the morphology of the cells we observed in the viability experiments, we would expect a similar trend using cell death assays. However, we do agree with the reviewer that this should be proven experimentally, so we will perform these experiments again using cell death assays as a read out.

1. Did the authors looked at the DISC assembly? Did they detect some differences there?

No, we did not. We would expect some difference given the impact we have observed on CD95 expression, caspase-8 activation and cell death of expressing dominant negative forms of IRE1, but this of course needs to be actually tested. We are in the process of optimizing CD95 DISC experiments in our lab and we therefore hope to be able to address this reviewer’s comment in a revised version of the manuscript.

Reviewer #2 (Significance (Required)):

This is an excellent study. The authors address here for the first time the connection between CD95, which is known as Fas, and ER stress. The role of another DR, TRAIL-R2 has been already reported, but this is the first study uncovering the link between Cd95 system and ER stress. The study is performed on the high level and supported by all necessary controls. This is an important advance for the death receptor field.

Thank you again for these very positive comments and your insightful appreciation of our work.

References

1. Volkmann, K., Lucas, J. L., Vuga, D., Wang, X., Brumm, D., Stiles, C., Kriebel, D., Der-Sarkissian, A., Krishnan, K., Schweitzer, C., Liu, Z., Malyankar, U. M., Chiovitti, D., Canny, M., Durocher, D., Sicheri, F. & Patterson, J. B. (2011) Potent and selective inhibitors of the inositol-requiring enzyme 1 endoribonuclease, J Biol Chem. 286, 12743-55.
2. Sheng, X., Nenseth, H. Z., Qu, S., Kuzu, O. F., Frahnow, T., Simon, L., Greene, S., Zeng, Q., Fazli, L., Rennie, P. S., Mills, I. G., Danielsen, H., Theis, F., Patterson, J. B., Jin, Y. & Saatcioglu, F. (2019) IRE1α-XBP1s pathway promotes prostate cancer by activating c-MYC signaling, Nat Commun. 10, 323.
3. Langlais, T., Pelizzari-Raymundo, D., Mahdizadeh, S. J., Gouault, N., Carreaux, F., Chevet, E., Eriksson, L. A. & Guillory, X. (2021) Structural and molecular bases to IRE1 activity modulation, Biochem J. 478, 2953-2975.
4. Logue, S. E., McGrath, E. P., Cleary, P., Greene, S., Mnich, K., Almanza, A., Chevet, E., Dwyer, R. M., Oommen, A., Legembre, P., Godey, F., Madden, E. C., Leuzzi, B., Obacz, J., Zeng, Q., Patterson, J. B., Jager, R., Gorman, A. M. & Samali, A. (2018) Inhibition of IRE1 RNase activity modulates the tumor cell secretome and enhances response to chemotherapy, Nat Commun. 9, 3267.
5. Almanza, A., Mnich, K., Blomme, A., Robinson, C. M., Rodriguez-Blanco, G., Kierszniowska, S., McGrath, E. P., Le Gallo, M., Pilalis, E., Swinnen, J. V., Chatziioannou, A., Chevet, E., Gorman, A. M. & Samali, A. (2022) Regulated IRE1α-dependent decay (RIDD)-mediated reprograming of lipid metabolism in cancer, Nat Commun. 13, 2493.
6. Le Reste, P. J., Pineau, R., Voutetakis, K., Samal, J., Jégou, G., Lhomond, S., Gorman, A. M., Samali, A., Patterson, J. B., Zeng, Q., Pandit, A., Aubry, M., Soriano, N., Etcheverry, A., Chatziioannou, A., Mosser, J., Avril, T. & Chevet, E. (2020) Local intracerebral Inhibition of IRE1 by MKC8866 sensitizes glioblastoma to irradiation/chemotherapy in vivo, 841296.
7. Voutetakis, K. D., D.; Vlachavas, E-I., Leonidas, DD.; Chevet, E.; Chatzioannou, A. (In preparation) RNA sequence motif and structure in IRE1-mediated cleavage.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors address here for the first time the connection between CD95, which is known as Fas, and ER stress. The role of another DR, TRAIL-R2 has been already reported, but this is the first study uncovering the link between Cd95 system and ER stress.The study is performed on the high level and supported by all necessary controls. They find the connection between IRE1 and CD95 and show that it might play a role in Cd95 signaling and attenuate CD95-mediated cell death.

Further, the correlation between CD95 expression and IRE is found in tumors. Importantly the authors find out the connection between XBP1 and CD95 expression, which was not reported to date. Hence, it is a very important and highly essential piece of research.

However, I would like to clarify the several issues:

1: Figure 1. Tunicamycin obviously leads to deglycosylation of CD95, which is indicated by the appearance of 35 Kda band. This should be highlighted and commented. 2.Figure 2c, d. The piece of mRNA structure, which is synthesized, might have the different secondary structure and might be not cleaved by IRE, accordingly. More detailed comments have to be provided in this regard. 3. Figure 3. Caspase-8-3 western blots show beautiful effects but did authors see some effects further downstream, eg on PARP1 cleavage? Was cell death ( not viability) measured as well? Can you comment on this? 4. Did the authors looked at the DISC assembly? Did they detect some differences there?

Significance

This is an excellent study. The authors address here for the first time the connection between CD95, which is known as Fas, and ER stress. The role of another DR, TRAIL-R2 has been already reported, but this is the first study uncovering the link between Cd95 system and ER stress.The study is performed on the high level and supported by all necessary controls. This is an important advance for the death receptor field.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Here the authors argue that IRE1 activation has opposite effects on Fas/CD95 expression/stability in a number of contexts, via either RIDD-dependent degradation of Fas mRNA or XBP-1-mediated induction of Fas expression, which led to either increased or decreased sensitivity to Fas-induced apoptosis in a number of settings.

Major issues:

The study is somewhat preliminary and inconclusive in that it is not clear why the RIDD function of XBP-1 appears to predominate in vitro in the cell lines examined, leading to modest increases in Fas expression levels (Figure 1) when IRE1 DN versus IRE1 WT constructs are overexpressed, which is at odds with the latter part of the paper which suggests that inhibition of RIDD led to reduced Fas expression levels. However, this could be due to supraphysiological levels of IRE1 being expressed under overexpression conditions, leading to confounding results. Similarly, when XBP-1s is overexpressed in vitro (Figure 5) the modest increases in CD95/Fas expression and sensitization to Fas-induced cell death may not be fully representative of what would be observed at physiological levels of XBP-1s activation. The in vivo results obtained using an IRE1 RNase inhibitor (MKC8866) contradict the earlier part of the study (as Fas levels decreased and there was protection from Fas-induced liver toxicity) and this could be due to a multitude of reasons. There is no doubt that impacting on IRE1 activity has interesting effects on CD95/Fas expression, which can be up- or -down-regulated, with consequences for cell death induced via engagement of the latter receptor, however, the manuscript does not offer a lot of clarity on which outcome is the predominant one in the context of engagement of the UPR. I have the following suggestions for improvement.

1. The authors should induce ER stress using Thaps, Brefeldin A and Tunicamycin, and explore the effects of doing this on Fas expression levels in the context of silencing endogenous IRE1, XBP-1 and PERK.
2. The authors should explore the effects of silencing of IRE1, XBP-1 and PERK on constitutive Fas expression and the outcome of Fas/CD95-induced apoptosis in cells not experiencing an overt activation of the UPR (i.e. in the absence of Thaps, Brefeldin A or other UPR inducer).
3. The specificity of MKC8866 at the concentration used (30 uM) is unclear. What effect does MKCC have on sensitivity towards Fas-induced apoptosis, similar to the type of experimental set up presented in Figure 5A, 5B?
4. Similarly, what effects does MKC8866 (at 30 uM) have on key Fas pathway determinants, such as Fas, FLIPL, FLIPs, Caspase-8, FADD, RIPK1, A20, CYLD, cIAP-1, cIAP-2 and Bid? There are many points at which MKC8866 could influence the outcome of Fas receptor engagement beyond the receptor itself

Minor issues:

1. For the Fas mRNA cleavage experiments presented in Figure 2, there are no irrelevant control mRNAs to allow the reader to judge whether the effects presented are specific to Fas mRNA or are commonly observed for many mRNAs at these amounts of IRE1 (1 ug, 0.5 ug, which appear high).

Significance

General assessment: this is an interesting study, as there is little knowledge currently concerning how the UPR influences Fas expression or Fas-dependent outcomes. However, the impact of this work is limited by the overexpression approaches used, which could produce artifactual results, as well as the contradictory message of the study.

Advance: the advance reported here is relatively modest and limited in scope due to the inconclusive nature of the data presented.

Audience: this study will be of interests to specialists in the UPR and cell death communities.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

The authors do not wish to provide a response at this time.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Overview:

This manuscript addresses the emerging nexus linking the machinery associated with clathrin endocytosis (clathrin-coated pits; CCPs), flat clathrin lattices (FCLs) and the recently discovered Reticular Adhesions (RAs). This is timely work, reflecting recent foci on the relationship between these structures and systems. Initially clearly identifying reductions in FCL and RA formation on fibronectin, the authors sought to clarify the mechanisms that suppress or prevent FCL / RA formation on this matrix. Knock-down of integrin avb5 (core RA component) suppressed both RA and FCL formation, suggesting a dependence of FCLs on this integrin. This was supported by acute avb5 inhibition via cilengitide (avb5 and avb3 inhibitor) which caused disassembly of existing RAs and FCLs.

Notably, the inverse relationship also appears true, with suppression of core clathrin endocytic machinery (AP2 complex components) being sufficient to greatly deplete RA formation. Supporting this finding, overexpression of a dominant negative-acting protein fragment (AP180 c-terminal fragment) blocked both AP2 localisation to the plasma membrane and RA formation.

To unmix this bi-directional dependence further, the authors used acute cilengitide treatment followed by washout and post-washout incubation to first deplete RAs (cillengitide) and then allow monitoring of en masse RA formation after cilengitide washout. This is an effective experiment, however, the analysis would benefit from greater depth, particularly relating to the order of events. Analysis of this aspect seems central to the thrust of the paper, and some statistical analysis of either static co-occurrence or dynamic ordering in large numbers of FCL / RA structures (i.e. hundreds) would be of value.

The authors next focused on the observation that fibronectin suppressed both FCL and RA structures, by assessing the role of fibronectin-receptor integrin b1. Acute antibody mediated integrin b1 inhibition (mab13) and integrin b1 knock down both confirmed that in cells on fibronectin, suppression of integrin b1 is sufficient to permit massive upregulation of both FCL and RA formation. This is surprising and very interesting. It raises questions about the actual ECM requirements for avb5-mediated reticular adhesion formation. It would seem that fibronectin per se can support very efficient RA / FCL formation, but that normally concurrent integrin b1 activities would suppress this. Given this implication, it would be especially important to clarify the purity of FN ECM coating (as explored in questions 1-3 below) at the time of imaging. The discussion addresses a number of topics, and proposes a mechanistic model to explain the results presented. I don't find the mechanism very convincing, as the directionality of the dependence between FCLs and RAs is not clearly delineated by the experiments presented, in my opinion. That there is co-dependence is convincingly shown, but whether there is directionality, and what order of events underpins FCL then RA or RA then FCL formation, is unclear. Nonetheless, the evidence presented does generally support the idea of a shift in the way we consider the role of endocytic machinery in adhesion regulation, from a disassembly only related function to additional functions associated with adhesion formation and maintenance. Ideally, the mechanisms around this new assembly / maintenance function would be further delineated here, but regardless, this work does point in the direction of important new questions in this area. Further discussion about the potential role of this interdependent regulatory process in, for example, mitosis, seems unwarranted and should probably be removed.

Questions:

1. A technical question on the replating experiments onto specific matrix proteins; after coating surfaces with the purified ECM components or controls, what media were the cells replated in? Ideally this should be serum-free media, to ensure that the ECM components of FBS / FCS are not immediately added to the purified ECM components. This should be clarified in the methods.
2. Related to above, I cannot see how long cells were plated onto the different ECM conditions. This would be relevant to know and should be clarified because cells will secrete ECM over time and thus the purity of the ECM components addressed is dependent on the length of time cells are incubated and imaged for after attachment.
3. Similarly, it is noted that cells plated on 'glass' support RA formation. It should be clarified what ECM component is then actually responsible for cell adhesion and adhesion complex (RA, FA or other) formation
   • since this requires an ECM component of some type. Presumably, 'on glass' means whatever ECM is either derived from the media used during cell attachment / incubation (if that media contains serum, which is vitronectin rich), or whatever ECM is secreted by the cells themselves over the attachment / incubation period prior to imaging.
4. In the cilengitide washout experiments, the evidence shown in figures seems to suggest that AP2-positive FCLs form in locations where avb5 (probably RAs) are already present, whereas avb5 positive structures do not form from AP2-only structures. Statistical analysis of this pattern (i.e. which protein is present first) would be valuable to address directionality. Notably, in Figure 3G, it appears that avb5 is present and increasing prior to the subsequent arrival of AP2.
   • a. I would suggest that the cilengitide experimental results(3E-G) be shown in a separate figure from the endocytosis inhibition results (3A-D).
5. The integrin b1 inhibition and knock down results are clear and interesting. Clarifying the ECM components present during these experiments would be valuable to interpretation of the paper.

Significance

see above

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

Hakanpää and colleagues report on the relationship between flat clathrin lattices (FCLs) and reticular adhesions, with FCLs being preposed to nucleate reticular adhesions. Overall the experimental work is high quality and the data are generally well presented.

Major comments:

The Introduction is very brief and doesn't cover the information required to understand the paper. There are three cellular structures to be understood: focal adhesions, reticular adhesions and FCLs. The intro jumps straight into the FCLs (the paper is written from a FCL point of view) but there is no information about the other two structures really particularly the differences between them. Furthermore, there is nothing in the intro about cellular adhesion or why this is even worth studying. The authors should fully revise this Intro, there is much room for improvement!

Fig 4D I could not understand this plot and the legend did not describe it properly. What are the units of FCL frequency? I guess it is FCLs per some distance (10µm?), the images need a scale bar. OK, I read the description in the methods and now I see it is the proportion of total CCSs that are FCLs; so frequency is the wrong term. The legend says that there were 32 videos and there are 32 points on the plot but what we need to know is: where n = 1 cell, what did the FCL frequency do over time? A line is drawn on the graph, no info on what the line is and the fit is poor (R2 < 0.5). The authors should take their series of data points from individual cells and fit curves to each and describe the summary statistics of the parameters of these fits OR average the data and fit to that, using the 1 SD of the data for weighting the fit. Probably more data is required to do any meaningful fitting here. To my eye it looks like FCL frequency goes from 0.3 to a plateau of 0.5 at 10 min and that more timepoints between 0 and 10 min would have been useful.

Fig 3F/G is a nice expt. It looks as though the ITGB5 signal is already creeping up when the AP2 arrives. I agree that they accelerate together, but the prior accumulation of integrin is at odds with the conclusion that AP2/clathrin nucleates the adhesion. This experiment is missing two controls: what is the behaviour of ITGB5 in AP2 negative regions? What happens to both signals in the continued presence of cilengitide? These controls are needed to conclude that AP2 is nucleating the adhesion.

Minor comments:

Fig 6 - several typos - "adaptor proteins" "engagement" "containing"

Significance

Previously, endocytosis (clathrin-mediated) was thought to decrease cellular adhesion by removal of integrins. This paper suggests that the same machinery can be used to build adhesions. This is a surprising conclusion that will be of interest to many cell biologists; the topic of clathrin and adhesion is being actively explored by several labs either from the adhesion or the endocytosis sides. I have been following this topic from a distance and don't know the details of all the published papers, but this paper does seem to add something new over the recent work from Taraska, Sonnenberg, Montagnac, Strömblad.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, Hakanpää et al explored the connection between flat clathrin lattices and reticular adhesions and the regulatory mechanism underlying the formation of these two structures in the U2OS cells. The author provided evidence that the composition of the extracellular matrix plays critical roles in their formation and concluded that fibronectin and its receptor, β1 integrin, inhibit the assembly of FCLs and RAs. The author depleted several components of the clathrin mediated endocytosis machinery and could show that it blocked the formation of reticular adhesions.

Major comments

1. In Fig. 1, the author plated U2OS cells on surfaces coated with different ECM components and then measured the frequency of FCLs. How long were the cells allowed to attach to the surfaces before they were imaged? Were the cells serum-starved before seeding? Given the fact that cells attached well even on BSA-coated dishes, I guess the cells were allow to attach and spread for at least overnight. In this case, the ECM components (vitronectin is very abundant in the serum in a concentration of 200-400 ug/mL, fibronectin is another one) in the culture medium would coat the glass surface and this has profound effects on the adhesion status of the cells. Thus, more details of this experiment need to be included and more attention should be pay regarding the data interpretation. In Fig. 1C, it seems like the mere glass surface induced the most FCL formation. However, if the cells are grown on the glass overnight or for days, the major component of the surface would actually be vitronectin and fibronectin (maybe more) rather than glass, thus it is not accurate to say that 'VTN reduced FCL frequency to some extent compared to glass' (Line 67).
2. Line 91-93. It is not accurate to claim that the reticular adhesions are the only type of cellular adhesion maintained during mitosis. In several studies, active integrin β1 are found along the retraction fibers (Dix et al, Dev Cell, 2018; Chen et al, NCB, 2022). The importance of αVβ5 integrin in the spatial memory during mitosis was only shown in the in vitro cell culture. In fact, mice lacking αVβ5 integrin or its ligand vitronectin are both viable and show no major defects during embryonic development (Zheng et al, PNAS, 1995; Huang et al, Mol Cell Biol, 2000), suggesting reticular adhesions are dispensable in cell division in vivo. I advise to change it into'RAs are composed of αVβ5 integrin and are maintained during mitosis in culture'.
3. It has been shown that fibronectin and laminin coating inhibit formation of reticular adhesions (Lock et al, NCB, 2018, Fig. S7). This study should be cited in Fig. 1. I suggest to also include laminin in Fig. 1C to make the list of ECM components more complete.
4. Fig. 5C, the fluorescent intensity of αVβ5 integrin is increased dramatically when integrin β1 was depleted compared to the control shRNA. Although the images were collected in the TIRF mode, it is important to measure αVβ5 and β1 integrin level by immunoblot to confirm the knockdown efficiency of β1 integrin and exclude the possibility that the increase of RA formation is not due to the compensation by upregulation of αVβ5.
5. Antibody-blocking or depletion of β1 integrin both lead to accumulation of FCL and RA formation, indicating that the activation of β1 integrin is critical in the inhibition of FCL and RA. Activation of β1 integrin depends on talin and kindlin, which bind β5 integrin with a much lower affinity. Would depletion of talin or kindlin cause FCL and RA formation similar to inhibition of β1 integrin?

Minor comments

1. In most of the quantifications, only the number of the cells measured were mentioned in the legend. The number of replicate experiments is missing. It should be included in the legend as well.
2. A recent study (https://doi.org/10.1242/jcs.259465) demonstrated the molecular mechanism underlying the localization of αVβ5 integrin in flat clathrin lattices. It should be mentioned in the introduction.

Significance

Although it is not novel that FCLs and RAs share same localization and might actually be the different parts of the same structure (Zuidema et al, JCS, 2022), the observation that inhibiting β1 integrin stimulates FCL and RA assembly is interesting as it indicates the counter-balance between the αVβ5 and α5β1 integrins. It is a pity that the author did not dig deeper into the mechanism underlying this interesting finding, which should greatly increase the impact of this study.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer 1:

We would like to thank you for taking the time to review our manuscript. Your thoughtful and insightful comments have greatly improved the quality of our work. We appreciate your thoroughness in evaluating our study and providing valuable feedback.

Your constructive criticism and suggestions have helped us identify areas that needed further clarification and improvement, and we are grateful for your efforts in guiding us towards a stronger manuscript.

Thank you again for your time and expertise in reviewing our work. We hope that you find our revisions satisfactory and look forward to hearing your thoughts on the revised manuscript.

Reviewer #1 (Evidence, reproducibility and clarity (Required)): *

In this manuscript by Sharma and colleagues, the authors investigate the transcriptional regulation of the TAL1 isoforms - that derive from differential promoter usage and/or alternative splicing - and the contribution of TAL1 long and TAL1 short protein isoforms in normal haematopoietic development and disease.

The study suggests that TAL1 transcript isoforms are fine-tuned regulated. By using CRISPR/Cas9 techniques, the authors show that the enhancer -8 (MuTE) and enhancer -60 differentially regulate the TAL1 isoforms. Whether the remaining enhancers at the TAL1 locus (see Zhou Y et al, Blood 2013) also differentially regulate TAL1 transcription remains to be elucidated.

The authors found that TAL1 short isoform interacts strongly with T-cell specific transcription factors such as TCF3 and TCF12, as compared to TAL1 long isoform. TAL1 short shows an apoptotic transcription signature and it fails in rescuing cell growth as compared to TAL1 long in T-ALL. In addition, TAL1 short promotes erythropoiesis.

Lastly, the authors suggest that altering TAL1 long and TAL1 short protein isoforms ratio could have a potential therapeutic application in disease, but further studies are needed. *

We would like to thank you for your time and effort in reviewing our manuscript. Your constructive feedback and insightful comments have been immensely valuable in improving the quality of our work. Your expertise in the field has undoubtedly contributed to the credibility and accuracy of this research. In addition, your dedication and attention to detail have been instrumental in shaping the final version of the manuscript.

* I have a number of comments: Figure 1 It was not mentioned that MOLT4 cells also have MuTE. Do Jurkat and MOLT4 share a similar profile in terms of TAL1 transcript isoforms? It would have been very interesting to see whether the TAL1 transcript isoforms are similar in SIL-TAL1+ cells (e.g RPMI-8402). In these cells, TAL1 activation results from a deletion that fuses the 5' non-coding region of SIL with TAL1. *

Thank you for your comment. We apologize for the confusion regarding the MOLT4 cells in our analysis. We have now updated the manuscript to explicitly mention the presence of MuTE in MOLT4 cells (Line 127). Additionally, we agree that it would be interesting to investigate whether the TAL1 transcript isoforms are similar in SIL-TAL1+ cells, such as RPMI-8402. To address this point, we have included the CCRF-CEM cell line that harbors the SIL-TAL1 recombination in our analysis. We have updated the manuscript with these new findings (Fig. 1C&D and S1A&B). Thank you for bringing this to our attention.

Figure 2 * It is not very clear how the expression of the short isoform delta exon 3 is quantified. Detailed information and a schematic of the primer location could be helpful. *

Thank you for your comment. We apologize for any confusion regarding the quantification of the expression of the short isoform (delta exon 3). The detailed information and schematic of the primer location can be found in Supplementary Figure 2B. We have included the location of each primer used in real-time PCR analysis for the quantification of all TAL1 isoforms. We hope this additional information will address your concerns.

* The results on Figure 2 derive from complex Cas9/CRISPR experiments. A schematic representation showing the location of the following elements is missing: CTCF sites, CTCF gRNA target region, dCas9-p300 gRNA target region and -60 enhancer. *

We agree that providing a schematic representation of the Cas9/CRISPR experiments would be helpful for better understanding the data in Figure 2. We have now included a detailed schematic of the location of the CTCF sites, CTCF gRNA target region, dCas9-p300 gRNA target region and -60 enhancer in Supplementary Figure 2E. We believe this new figure will provide a clearer overview of the experiments performed and will aid in the interpretation of the results.

* Are the levels of dCas9-p300 WT and dCas9-p300 MUT comparable in transfected HEK 293 cells? Were those possibly measured by qPCR or Western Blot? Why the authors chose to use 293T cells for the CTCF del as the enhancer usage around the locus must be so different from haematopoietic cells. *

Thank you for your question. We have added Western Blot analysis to compare the levels of dCas9-p300 WT and dCas9-p300 MUT in transfected HEK293T cells, as suggested. The results are presented in Supp. Fig. S2H.

Regarding the choice of HEK293T cells for the CTCF deletion experiment, we selected this cell line for its low expression of TAL1, which contributes to a high dynamic range when tethering p300 core to a closed chromatin region. We have added a clarification of our rationale for using HEK293T cells in the revised manuscript (Lines 177-8). Thank you for your valuable feedback.

* Is CPT - camptothecin? A control gene that is sensitive to CPT treatment would ensure the inhibitor is working. *

Thank you for your comment. Indeed, CPT stands for camptothecin, and this information is already included in the methods section. We have also added this information to the results section (Line 221) to make it clearer.

Regarding the suggestion to use a control gene sensitive to CPT treatment, we agree that this could be a useful addition to our experimental design. To address this, we have quantified the amount of TAL1 transcript to an endogenous control which is not transcribed by RNA Polymerase II (RNAPII) (18s rRNA). As a positive control, we compared Cyclo A, our endogenous control, to 18s rRNA and observed a reduction (Supp. Fig. S2K). This allows us to confidently conclude that the inhibitor is working as intended.

Thank you for bringing up this point, and we hope that our response addresses your concern.

*

In supplementary Figure 2D, the reduction in expression in Jurkat Del-12 is restricted to TSS2. There is no reduction in TAL1 TSS1 and TAL1 TSS4 (this is not clear from the result description section). As seen, these isoforms are upregulated and that could suggest a compensatory mechanism mediated by alternative promoter activation. The fact that Jurkat Del-12 express TAL1 from MSCV-TAL1 could also suggest that TSS1 and TSS4 are upregulated by TAL1 or indirectly, by other members of the TAL/LMO complex (see Sanda T et al, Cancer Cell 2012) *

Certainly, we appreciate your feedback. Supplementary Figure 2D indeed shows that the MuTE enhancer has a differential effect on the promoters, and we have now included this in the text of the manuscript. Regarding the TAL1-long isoform, while MSCV-TAL1 in the Jurkat Del-12 cell line does give rise to this isoform, our results from Figure 3A did not find TAL1-long to have a differential effect on TAL1 promoters. It is important to note that the experiment conducted was an exogenous construct in HEK293T cells, which has its limitations. Thus, the speculation that TAL1-long drives the result in supplementary Figure 2D is possible, and we have added this to the text. Thank you for bringing up this important point (Lines 167-9).

Figure 3 * A. Are the levels of TAL1 short cDNA and TAL1 long cDNA comparable in the co-transfection luciferase experiments? The overexpression of the isoforms does not reflect the endogenous expression levels in cell lines where one of the isoforms is more predominantly expressed (e.g Jurkat cells express low levels of TAL1 short). *

Thank you for your comment. To address your concern, we have added real time (Supp. Fig. S3A) as well as Western blot in a new figure (Supp. Fig. S3B) to show that the levels of TAL1-short and TAL1-long cDNA are comparable in the co-transfection luciferase experiments. Additionally, we observed a very low amount of endogenous TAL1 isoforms in the cell line (Supp. Fig. S3A&B), which was below detection using these methods. This suggests that the effect of the endogenous TAL1 in this cell line is low. We appreciate your feedback, and we hope this additional information addresses your concern.

* Figure 4 Are the levels of flag-TAL1 long and flag-TAL1 short comparable? The levels of expression could explain the low intensity signal for TAL1 long. *

Thank you for your insightful comment. Indeed, the issue of isoform quantification is critical in understanding the functional differences between TAL1-short and TAL1-long. To address this concern, we performed careful quantification of the isoforms and made sure that the amount was equal or slightly in favor of TAL1-long before conducting the experiments in this manuscript. We have also added a Western blot in Supp. Fig. S3A and real time in Supp. Fig. S3B showing the similar amount of the two isoforms. Furthermore, in Figure 4A, we provided the amount of each isoform in the input section, showing a higher amount of TAL1-long. This strengthens our result, which shows that TAL1-short binds stronger to TCF-3 and 12. Protein levels for ChIP-seq experiment (Fig. 4B-H) is now in Supp. Fig. S4B. We thank you for bringing up this important point, and we hope that our additional data and clarifications have addressed your concern

*Is there any reason for not performing a depletion of endogenous TAL1 prior to the ChIP seq flag experiment? *

Thank you for your comment. In our experience, infecting Jurkat cells with shRNA or an expressing vector systems can induce some cellular stress, and we did not want to add additional stress to the cells by depleting endogenous TAL1. Since we immunoprecipitated using a Flag-tagged protein, we did not see a need to deplete the endogenous TAL1 protein. However, in our RNA-seq experiment, depletion of endogenous TAL1 was critical, and we have added this additional step in this experiment.

* Could the authors speculate about MAF motif enrichment in both isoforms and not in TAL1-total? *

Thank you for bringing up this interesting point. It is worth noting that while all ChIP-seq experiments were performed in Jurkat cells, not all of them were conducted by us. In particular, ChIP-seq of TAL1 total was performed by Sanda et al., 2012, using an endogenous antibody against both isoforms, whereas we conducted ChIP-seq for TAL1-short and TAL1-long using a FLAG tag antibody in cells expressing each of the isoforms. Therefore, the different conditions of these experiments may have contributed to the observed MAF motif enrichment in both isoforms and not in TAL1-total. While we cannot provide a definitive explanation, we speculate that the overexpression of the isoforms or the presence of the FLAG tag may have facilitated the detection of the MAF motif. We have added this discussion to the manuscript to acknowledge and address this interesting observation (Lines: 307-8).

1. Sanda et al., Core transcriptional regulatory circuit controlled by the TAL1 complex in human T cell acute lymphoblastic leukemia. Cancer Cell 22, 209-221 (2012).

   * Do TAL1 long and TAL1 short recognise the same DNA motif? *

This is indeed a very interesting question but a difficult one to answer since TAL1 does not bind to the DNA alone but in a complex. In this situation, the ChIP-seq de-novo binding results suggest motifs that could be recognized by TAL1 or any of its complex partners. Using previous data, TAL1’s binding motif is CAGNTG (Hsu et al., 1994), while this motif was not identified in our analysis of the TAL1-total or FLAG-TAL1-long ChIP-seq results, we did, however, identify this sequence in FLAG-TAL1-short ChIP-seq results (p value=1e-93). We predict that this discrepancy is due to the complex nature of transcription factors binding and the fact that the ChIP-seq results were not all done in the same way. We have now added this to the discussion (Lines: 419-25).

1. L. Hsu et al., Preferred sequences for DNA recognition by the TAL1 helix-loop-helix proteins. Mol Cell Biol 14, 1256-1265 (1994).

* Figure 6 In A and B, are the levels of flag-TAL1 long and flag-TAL1 short in transduced K562 comparable? In C and D, are the TAL1 levels reduced at the protein level?*

Thank you for your question. To answer your question, we added Western Blot analysis to show the comparable levels of flag-TAL1-long and flag-TAL1-short in transduced K562 cells (Supp. Fig. S6C). In Figure 6C and D, we also added Western Blot analysis to show the reduction in TAL1 protein levels upon shRNA-mediated knockdown(Supp. Fig. S6B).

* Minor points: Figure 1 A. Include a scale bar *

To address this, we included coordinates of the components of the gene marked in the figure.

* C. Loading control such as GAPDH is missing in the Western Blot. Are CUTLL cells the same as CUTTL-1? *

We added loading controls as requested now supplementary Fig. 1C, S2C, S3A, S4B, S6B&C. Yes, CUTLL is the same as CUTLL-1 we have now fixed this in the text (Line 120).

D. Adjust scale of the CHIP seq tracks in K562 cells in order to see the peak summit. *Include genome build *

Thank you for your comment. We have adjusted the scale of the ChIP-seq tracks in K562 cells as suggested to improve the visualization of the peak summit. However, one of the peaks still had a much higher signal and the summit is still missing from this particular peak. To address this, we have added a new figure in the supp. Fig. S1C materials where we adjusted the peak to show the summit. Please note that in this track, the chromatin structure at the enhancers is missing, and therefore, we did not include it in the main figure. Thank you for bringing this to our attention.

We have added a genome build hg19 to the figure legend.

* In supplementary Figure 1B, the symbol scheme is not clear *

Thank you for this note, we have replaced the figure and added text to make it clearer.

* Figure 2 A & C. Remove 'amount' from the Y axis. Is the total mRNA amount calculated as % of the reference genes? It could be specified on the y axis or figure legend. *

We have removed the word "amount" from the Y axis as requested. Total mRNA amount is normalized relative to the reference genes (∆∆Cq) by Bio-Rad's CFX Maestro software (version 2.3) according to the formula:

where:

• RQ = Relative Quantity of a sample
• Ref = Reference target in a run that includes one or more reference targets in each sample
• GOI = Gene of interest (one target)

* In supplementary Figure 2C, a loading control is missing.*

We have added alpha-tubulin to this figure.

* Figures 4, 5 and 6 Size of the figures should be increased. *

We have increased the figure size as suggested. *

Reviewer #1 (Significance (Required)): The study from Sharma and colleagues is novel and it extends the knowledge on TAL1 regulation and the role of TAL1 in development and disease. Although the study suggests that there is a correlation between enhancers, chromatin mark deposition at exons and regulation of alternative splicing, the mechanistic link is not fully elucidated.*

To further elucidate the mechanistic link between the MuTE enhancer, broad H3K4me3 modification spanning 7.5 Kbp from TAL1 promoter 1 to promoter 5 (as shown in Fig. 1D), and alternative splicing, we conducted experiments where we manipulated KMT2B, a component of the SET1/COMPASS complexes responsible for methylating H3K4. Our findings indicate that silencing KMT2B in Jurkat cells led to a significant 30% increase in TAL1-∆Ex3 (Fig. 2H and Supp. Fig. S2I&J). These results contribute to a more comprehensive understanding of the molecular mechanisms underlying TAL1 alternative splicing regulation.

The findings on TAL1 short protein are interesting but the data on TAL1 long lacks some refinement so then robust conclusions can be drawn. * The experimental data lacks a few controls. The text is clear and prior studies could be better referenced. *

We have made an effort to better reference out manuscript.

* As TAL1 is a very crucial transcription factor oncogene in T-ALL, the study is important as it addresses a very relevant question in the field that is the regulation of the transcription of TAL1 and the functional relevance of both TAL1 short and TAL1 long isoforms. *

Reviewer 2: *

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary: Sharma et al. thoroughly characterized the regulation of TAL1 by mapping the use of its five promoters and enhancers, which together transcribe five transcripts, coding for two protein isoforms. For that purpose the authors used few cell lines: Jurkat as a T-ALL cell line, chronic myeloid leukemia (CML) cell line K562 and HEK293T with low TAL1 expression, as well as CutLL and MOLT4. They profiled the chromatin marks H3K27ac and H3K4me3 at the TAL1 locus, and show that when a the -8 enhancer is compromised tha chromatin marks change, and not only the expression level of TAL1 is reduced, the level of exon 3 skipping is increased. When the -60 enhancr was activated, TAL1 expression increased, and exon 3 skipping was reduced. Those findings indicate that in tal1, transcription and alternative splicing are co-regulated, independent of RNAPII. The authors also show that as an autoregulator, TAL1-short has a preference to TSS1-3 of TAL1, which is not shared by TAL1-long, and that each of the 5' UTR affect Tal1 expression differently. TAL1-short binds E-proteins more strongly than TAL1-long, binds many more sites than TAL1-long and stronger, and each isoform has unique set of targets. Finally, the authors set to identify the different functions of the TAL1 isoforms, and showed that Tal1-short slows cell growth and leads to TAL1-short but not TAL1-long leads to exhaustion of hematopoietic stem cells and promotes differentiation into erythroids. This paper used for the first time TAL1 isoform specific ChIP-seq, which enable accurate definition of isoform-specific targets in Jurkat cells. They demonstrated an interaction between choice of TSS and alternative splicing, and isoform specific functions. Given the clinical importance of TAL1 and the meticulous work performed to characterize its isoform specific regulation and function, I find this manuscript of interest, and only have minor suggestions to improve readability. *

Thank you for taking the time to carefully review our manuscript on the regulation and function of TAL1 isoforms. We appreciate your positive feedback on our comprehensive characterization of TAL1 regulation using chromatin profiling and isoform-specific ChIP-seq. We are glad that you found our findings on the co-regulation of transcription and alternative splicing, as well as the isoform-specific functions of TAL1, to be of interest.

We also appreciate your suggestions to improve the readability of the manuscript and have made the necessary revisions accordingly. Your feedback has been invaluable in strengthening the quality of our work, and we are grateful for your contribution to the scientific community.

* Minor comments: Add explicitly the motivation for choosing the cell line in each part. *

We have added motivation (Lines: 157-8, 177-8, 192-194, 235-6 text that was on the previous version: 192-194, 379-80).

* Figure 1 - Consider marking the promoter numbers and the enhancers names in the same names as in text (-8,-60 etc.), to make it easier for the readers to understand which enhancers is being discussed. *

This in a very important point. We have added the numbering to Figure 1D and Supp. Fig. S2A, B & E.

*P5, P18 - ProtParam is only a prediction tool, and does not supply an experimental measurement, as may be assumed from text. Please rephrase accordingly. *

The words “prediction tool” were added in the indicated paragraphs (Lines 115 and 427).

* Figure 2B/D - y axis label unclear, not explained in text. In accordance, unclear if the change is in the amount of RNA, or the ratio between the long and short variants. *

Thank you for this comment. We greatly appreciate your feedback and suggestions. To make our calculations, which are the norm in the splicing field, clearer, we have now added text to Figure 4 and provided more detailed explanations in lines 670-73. We hope that these modifications will improve the clarity and comprehensiveness of our manuscript.

*Consider removing the bars and increasing the dots, to make the graphs cleaner. *

We removed the bars throughout the manuscript for a cleaner look.

* P8 - The term '5C' may require more explanation, depending on target audience. *

We have added text to explain the technique (Lines 179-81).

* Figure 3 - the trend is that TAL1-short promotes transcription from all five TSSs. However, only in TSS1-3 is the difference significant, but the difference between the long and short forms is not significant. It is unclear if "The mean of three independent experiments done with three replicates" means overall there are three replicates per condition or nine. Please rephrase to clarify. *

Thank you for your comment. To clarify, we want to state that each biological experiment was done in three technical replicates, resulting in a total of nine replicates for each condition. We apologize for any confusion and have now rephrased to: The mean was calculated from three independent biological experiments, each performed with three technical replicates (Lines: 696 and 699).

*Fig 4 A - it seems that many of the sites bound by Tal1 total are not bind by either Tal1-short or Tal1-long. Indeed very little overlap between Tal1-short and Tal-1-total is seen in Fig 4I as well. It seems Tal1-long has very few peaks. Consider adding a discussion of possible reasons. *

We agree that these findings are noteworthy and warrant further discussion. We added text to the discussion section to explore potential reasons for these observations (Lines 416-25).

* Fig 4c - it is hard to distinguish the different lines. Consider a more clear visualization. Also, some text is in a font size too small to read. *

We have changed the format of the figure and took out the input data from the main figure to help the visualization. The input data appear in the Supp. Fig. S4C.

* Fig 4 D-H - will be useful to see the numbers, not just the % divided by %. *

A table with the specific numbers can be found in Supp Figure 4F-J.

* Fig 4 legend - 'I&L' possibly means 'I-L'. P14 - refer to where the results of the 'validation using real-time PCR' are shown. P16 - symbol replaced by an empty rectangle 20 􀀀M *

Thank you for these valuable comments, we have fixed/added these in the manuscript.

* Figure 6D - Y axis value seem strange (fold change relative to day 0 should be 1 at day 0). Consider different Y axis label for C and D to clarify. *

Thank you for this comment, we have changed the y-axis to: Fold-change relative to day 1.

* P18 - It is unclear which "two isoforms with posttranslational modifications which affected the migration rate of the protein (Fig. 1C)" were shown. Only two isoforms are mentioned throughout the paper. *

We have added text to clarify we are referring to TAL1-short and long (Lines 409-10).

*

P18 - "Our ChIP-seq results suggest that the isoforms bind at the same location (Fig. 4B)." - in 4B it seems most of TAL1-short bound positions are not bound by TAL1 long. Please clarify. *

* Worth mentioning that the Total TAL1 is taken from Jurkat cells but from a different experiment. * We have changed the statement and added the text referring to the experiments done independently (Lines 422-3).

*

Reviewer #2 (Significance (Required)): This paper used for the first time TAL1 isoform specific ChIP-seq, which enable accurate definition of isoform-specific targets in Jurkat cells. They demonstrated an interaction between choice of TSS and alternative splicing, and isoform specific functions. Given the clinical importance of TAL1 and the meticulous work performed to characterize its isoform specific regulation and function, I find this manuscript of interest, and only have minor suggestions to improve readability. *

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

Sharma et al. thoroughly characterized the regulation of TAL1 by mapping the use of its five promoters and enhancers, which together transcribe five transcripts, coding for two protein isoforms. For that purpose the authors used few cell lines: Jurkat as a T-ALL cell line, chronic myeloid leukemia (CML) cell line K562 and HEK293T with low TAL1 expression, as well as CutLL and MOLT4.

They profiled the chromatin marks H3K27ac and H3K4me3 at the TAL1 locus, and show that when a the -8 enhancer is compromised tha chromatin marks change, and not only the expression level of TAL1 is reduced, the level of exon 3 skipping is increased. When the -60 enhancr was activated, TAL1 expression increased, and exon 3 skipping was reduced. Those findings indicate that in tal1, transcription and alternative splicing are co-regulated, independent of RNAPII. The authors also show that as an autoregulator, TAL1-short has a preference to TSS1-3 of TAL1, which is not shared by TAL1-long, and that each of the 5' UTR affect Tal1 expression differently. TAL1-short binds E-proteins more strongly than TAL1-long, binds many more sites than TAL1-long and stronger, and each isoform has unique set of targets.

Finally, the authors set to identify the different functions of the TAL1 isoforms, and showed that Tal1-short slows cell growth and leads to TAL1-short but not TAL1-long leads to exhaustion of hematopoietic stem cells and promotes differentiation into erythroids.

This paper used for the first time TAL1 isoform specific ChIP-seq, which enable accurate definition of isoform-specific targets in Jurkat cells. They demonstrated an interaction between choice of TSS and alternative splicing, and isoform specific functions. Given the clinical importance of TAL1 and the meticulous work performed to characterize its isoform specific regulation and function, I find this manuscript of interest, and only have minor suggestions to improve readability.

Minor comments:

Add explicitly the motivation for choosing the cell line in each part.

Figure 1 - Consider marking the promoter numbers and the enhancers names in the same names as in text (-8,-60 etc.), to make it easier for the readers to understand which enhancers is being discussed.

P5, P18 - ProtParam is only a prediction tool, and does not supply an experimental measurement, as may be assumed from text. Please rephrase accordingly.

Figure 2B/D - y axis label unclear, not explained in text. In accordance, unclear if the change is in the amount of RNA, or the ratio between the long and short variants. Consider removing the bars and increasing the dots, to make the graphs cleaner.

P8 - The term '5C' may require more explanation, depending on target audience.

Figure 3 - the trend is that TAL1-short promotes transcription from all five TSSs. However, only in TSS1-3 is the difference significant, but the difference between the long and short forms is not significant. It is unclear if "The mean of three independent experiments done with three replicates" means overall there are three replicates per condition or nine. Please rephrase to clarify.

Fig 4 A - it seems that many of the sites bound by Tal1 total are not bind by either Tal1-short or Tal1-long. Indeed very little overlap between Tal1-short and Tal-1-total is seen in Fig 4I as well. It seems Tal1-long has very few peaks. Consider adding a discussion of possible reasons.

Fig 4c - it is hard to distinguish the different lines. Consider a more clear visualization. Also, some text is in a font size too small to read.

Fig 4 D-H - will be useful to see the numbers, not just the % divided by %.

Fig 4 legend - 'I&L' possibly means 'I-L'.

P14 - refer to where the results of the 'validation using real-time PCR' are shown.

P16 - symbol replaced by an empty rectangle 20 􀀀M

Figure 6D - Y axis value seem strange (fold change relative to day 0 should be 1 at day 0). Consider different Y axis label for C and D to clarify.

P18 - It is unclear which "two isoforms with posttranslational modifications which affected the migration rate of the protein (Fig. 1C)" were shown. Only two isoforms are mentioned throughout the paper.

P18 - "Our ChIP-seq results suggest that the isoforms bind at the same location (Fig. 4B)." - in 4B it seems most of TAL1-short bound positions are not bound by TAL1 long. Please clarify. Worth mentioning that the Total TAL1 is taken from Jurkat cells but from a different experiment.

Significance

This paper used for the first time TAL1 isoform specific ChIP-seq, which enable accurate definition of isoform-specific targets in Jurkat cells. They demonstrated an interaction between choice of TSS and alternative splicing, and isoform specific functions. Given the clinical importance of TAL1 and the meticulous work performed to characterize its isoform specific regulation and function, I find this manuscript of interest, and only have minor suggestions to improve readability.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript by Sharma and colleagues, the authors investigate the transcriptional regulation of the TAL1 isoforms - that derive from differential promoter usage and/or alternative splicing - and the contribution of TAL1 long and TAL1 short protein isoforms in normal haematopoietic development and disease.

The study suggests that TAL1 transcript isoforms are fine-tuned regulated. By using CRISPR/Cas9 techniques, the authors show that the enhancer -8 (MuTE) and enhancer -60 differentially regulate the TAL1 isoforms. Whether the remaining enhancers at the TAL1 locus (see Zhou Y et al, Blood 2013) also differentially regulate TAL1 transcription remains to be elucidated.

The authors found that TAL1 short isoform interacts strongly with T-cell specific transcription factors such as TCF3 and TCF12, as compared to TAL1 long isoform. TAL1 short shows an apoptotic transcription signature and it fails in rescuing cell growth as compared to TAL1 long in T-ALL. In addition, TAL1 short promotes erythropoiesis.

Lastly, the authors suggest that altering TAL1 long and TAL1 short protein isoforms ratio could have a potential therapeutic application in disease, but further studies are needed.

I have a number of comments:

Figure 1

It was not mentioned that MOLT4 cells also have MuTE. Do Jurkat and MOLT4 share a similar profile in terms of TAL1 transcript isoforms? It would have been very interesting to see whether the TAL1 transcript isoforms are similar in SIL-TAL1+ cells (e.g RPMI-8402). In these cells, TAL1 activation results from a deletion that fuses the 5' non-coding region of SIL with TAL1.

Figure 2

It is not very clear how the expression of the short isoform delta exon 3 is quantified. Detailed information and a schematic of the primer location could be helpful.

The results on Figure 2 derive from complex Cas9/CRISPR experiments. A schematic representation showing the location of the following elements is missing: CTCF sites, CTCF gRNA target region, dCas9-p300 gRNA target region and -60 enhancer. Are the levels of dCas9-p300 WT and dCas9-p300 MUT comparable in transfected HEK 293 cells? Were those possibly measured by qPCR or Western Blot? Why the authors chose to use 293T cells for the CTCF del as the enhancer usage around the locus must be so different from haematopoietic cells.

Is CPT - camptothecin? A control gene that is sensitive to CPT treatment would ensure the inhibitor is working.

In supplementary Figure 2D, the reduction in expression in Jurkat Del-12 is restricted to TSS2. There is no reduction in TAL1 TSS1 and TAL1 TSS4 (this is not clear from the result description section). As seen, these isoforms are upregulated and that could suggest a compensatory mechanism mediated by alternative promoter activation. The fact that Jurkat Del-12 express TAL1 from MSCV-TAL1 could also suggest that TSS1 and TSS4 are upregulated by TAL1 or indirectly, by other members of the TAL/LMO complex (see Sanda T et al, Cancer Cell 2012)

Figure 3

A. Are the levels of TAL1 short cDNA and TAL1 long cDNA comparable in the co-transfection luciferase experiments? The overexpression of the isoforms does not reflect the endogenous expression levels in cell lines where one of the isoforms is more predominantly expressed (e.g Jurkat cells express low levels of TAL1 short).

Figure 4

Are the levels of flag-TAL1 long and flag-TAL1 short comparable? The levels of expression could explain the low intensity signal for TAL1 long. Is there any reason for not performing a depletion of endogenous TAL1 prior to the ChIP seq flag experiment? Could the authors speculate about MAF motif enrichment in both isoforms and not in TAL1-total? Do TAL1 long and TAL1 short recognise the same DNA motif?

Figure 6

In A and B, are the levels of flag-TAL1 long and flag-TAL1 short in transduced K562 comparable? In C and D, are the TAL1 levels reduced at the protein level?

Minor points:

Figure 1

A. Include a scale bar C. Loading control such as GAPDH is missing in the Western Blot. Are CUTLL cells the same as CUTTL-1? D. Adjust scale of the CHIP seq tracks in K562 cells in order to see the peak summit. Include genome build In supplementary Figure 1B, the symbol scheme is not clear

Figure 2

A & C. Remove 'amount' from the Y axis. Is the total mRNA amount calculated as % of the reference genes? It could be specified on the y axis or figure legend.

In supplementary Figure 2C, a loading control is missing.

Figures 4, 5 and 6

Size of the figures should be increased.

Significance

The study from Sharma and colleagues is novel and it extends the knowledge on TAL1 regulation and the role of TAL1 in development and disease. Although the study suggests that there is a correlation between enhancers, chromatin mark deposition at exons and regulation of alternative splicing, the mechanistic link is not fully elucidated. The findings on TAL1 short protein are interesting but the data on TAL1 long lacks some refinement so then robust conclusions can be drawn.

The experimental data lacks a few controls. The text is clear and prior studies could be better referenced.

As TAL1 is a very crucial transcription factor oncogene in T-ALL, the study is important as it addresses a very relevant question in the field that is the regulation of the transcription of TAL1 and the functional relevance of both TAL1 short and TAL1 long isoforms.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We would like to thank the reviewers for their extensive review of our manuscript and constructive criticism. We have attempted to address the points raised in the reviewer's comments and have performed additional experiments and have edited the text of the manuscript to explain these points. Please see below, our point-by-point response to the reviewer’s comments. In the submitted revised manuscript, some figure numbers have changed from the prior reviewed version.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

In this MS, Mrj - a member of the JDP family of Hsp70 co-chaperones was identified as a regulator of the conversion of Orb2A (the Dm ortholog of CPEB) to its prion-like form.

In drosophila, Mrj deletion does not cause any gross neurodevelopmental defect nor leads to detectable alterations in protein homeostasis. Loss of Mrj, however, does lead to altered Orb2 oligomerization. Consistent with a role of prion-like characteristics of Orb2 in memory consolidation, loss of Mrj results in a deficit in long-term memory.

Aside from the fact that there are some unclarities related to the physicochemical properties of Orb2 and how Mrj affects this precisely, the finding that a chaperone could be important for memory is an interesting observation, albeit not entirely novel.

In addition, there are several minor technical concerns and questions I have that I feel the authors should address, including a major one related to the actual approach used to demonstrate memory deficits upon loss of Mrj.

Reviewer #1 (Significance (Required)):

Figure 1 (plus related Supplemental figures): • There seem to be two isoforms of Mrj (like what has been found for human DNAJB6). I find it striking to see that only (preferentially?) the shorter isoform interacts with Orb2. For DNAJB6, the long isoform is mainly related to an NLS and the presumed substrate binding is identical for both isoforms. If this is true for Dm-Mrj too, the authors could actually use this to demonstrate the specificity of their IPs where Orb2 is exclusively non-nuclear?

According to Flybase, Mrj has 8 predicted isoforms of which four are of 259 amino acids (PA, PB, PC, and PD), 3 are of 346 amino acids (PE, PG, and PH) and one is of 208 amino acids (PF) length (Supplementary data 1). We isolated RNA from flyheads and used this in RT-PCR experiments to check which Mrj isoforms express in the brain. Since both the 346 amino acid (1038 nucleotide long) and 259 amino acids (777 nucleotides long) form, which we refer to as the long and middle isoform, has the same N and C terminal sequences we used the same primer pair for this, but on RT-PCR the only amplicon we got corresponds to the 259 amino acid form. For the 208 amino acids (624 nucleotides long) form we designed a separate forward primer and attempted to amplify this using RT-PCR but were unable to detect this isoform also. This data is now presented in Supplemental Figure 4B. Since the only isoform detected from fly head cDNA corresponded to the 259 amino acid form, we think this is the predominant isoform of Mrj expressing in Drosophila and this is what is in our DnaJ library and what we have used in all our experiments here. This is also the same isoform described in previous papers on Drosophila Mrj (Fayazi et al, 2006; Li et al, 2016b). For this 259 amino acid Mrj isoform, we see its expression in both the nucleus and cytoplasm (Supplemental Figure 4C). As the long 346 AA fragment was undetectable in the brain, it was not feasible to address the reviewer’s point of using the long and short forms of Mrj for IP with Orb2. However, we have performed IP of human CPEB2 (hCPEB2) with the long and short isoforms of human DnaJB6 and have detected interaction of hCPEB2 with both the long and short isoforms of DnaJB6 (Supplemental Figure 6E).

• I would be interested to know a bit more about the other 5 JDPs that are interactors with Orb2: are the human orthologs of those known? It is striking that these other 5 JDPs interact with Orb2 in Dm (in IPs) but have no impact on Sup35 prion behavior. Importantly, this does not imply they may not have impact on the prion-like behavior of other Dm substrates, including Dm-Orb2.

We have performed BlastP analysis of CG4164, CG9828, CG7130, DroJ2, and Tpr2 protein sequences against Human proteins. Based on this we have listed the highest-ranking candidate identified here for each of these genes.

Drosophila Gene

Human gene

Query cover

Percent identity

E value

CG4164

dnaJ homolog subfamily B member 11 isoform 1

98 %

62.96%

2e-150

CG9828

dnaJ homolog subfamily A member 2

92%

39.41%

3e-84

CG7130

dnaJ homolog subfamily B member 4 isoform d

56%

69.44%

2e-30

Tpr2

dnaJ homolog subfamily C member 7 isoform 1

93%

46.22%

6e-139

DroJ2

dnaJ homolog subfamily A member 4 isoform 2

98%

60.60%

2e-169

In the context of the chimeric Sup35-based assay where Orb2A’s Prion-like domain (PrD) is coupled with the C-terminal domain of Sup35, the only protein which could convert Orb2A PrD-Sup35 C from its non-prion state to prion state was Mrj. Within the limitations of this heterologous-system based assay, the other 5 DnaJ domain proteins as well as the Hsp70’s were unable to convert the Orb2A PrD from its non-prion to prion-like state. What these other 5 interacting JDP proteins are doing through their interaction with Orb2A and if they are even expressing in the Orb2 relevant neurons will need to be tested separately and will be the subject of our future studies.

• The data in panels H, I indeed suggest that Mrj1 alters the (size of) the oligomers. It would be important to know what is the actual physicochemical change that is occurring here. The observed species are insoluble in 0.1 % TX100 but soluble in 0.1% SDS, which suggest they could be gels, but not real amyloids such as formed by the polyQ proteins that require much higher SDS concentrations (~2%) to be solubilized. This is relevant as Mrj1 reduces polyQ amyloidogenesis whereas is here is shown to enhance Orb2A oligomerization/gelidification. In the same context, it is striking to see that without Mrj the amount of Orb2A seems drastically reduced and I wonder whether this might be due to the fact that in the absence of Mrj a part of Orb2A is not recovered/solubilized due to its conversion for a gel to a solid/amyloid state? In other words: Mrj1 may not promote the prion state, but prevents that state to become an irreversible, non-functional amyloid?

On the reviewer’s point to address what is the actual physicochemical change occurring here, we will need to develop methods to purify the Orb2 oligomers in significant quantities to examine and distinguish if they are of gel or real amyloid-like nature. Currently, within the limitations of our ongoing work, this has not been possible for us to do and we can attempt to address this in our future work. Cryo-EM derived structure of endogenous Orb2 oligomers purified from a fly head extract from 3 million fly heads, made in the TritonX-100 and NP-40 containing buffer, the same buffer as what we have used here for the first soluble fraction, showed these oligomers as amyloids (Hervas et al, 2020). If the oligomers extracted using 0.1% and 2% SDS are structurally and physicochemically different, within the limitations of our current work, had not been possible to address.

The other point raised by the reviewer is, if in the absence of Mrj (in the context of Figure 4 of our previously submitted manuscript), a part of Orb2 is not solubilized due to us using a lower 0.1% SDS for extraction. To address this, we attempted to see how much of leftover Orb2 is remaining in the pellet after extraction with 0.1 % SDS. Towards this, according to the reviewers’ suggestion, we used a higher 2% SDS containing buffer to resuspend the leftover pellet after 0.1% SDS extraction, and post solubilisation ran all the fractions in SDD-AGE. We did this experiment with both wild-type and Mrj knockout fly heads. Under these different extractions, we first observed while there is more Orb2 in the soluble fraction (Triton X-100 extracted) of Mrj knockout, this amount is reduced in both the 0.1% SDS solubilized and 2% SDS solubilized fractions. So, even though there is leftover Orb2 after 0.1% SDS extraction, which can be extracted using 2% SDS, this amount is reduced in Mrj knockout. The other observation here is the Orb2 extracted using 2% SDS shows a longer smear in comparison to the 0.1% SDS extracted form suggesting a possibility of more and higher-sized oligomers present in this fraction. Since we do not have the exact physicochemical characterization of these oligomers detected with 0.1% and 2% SDS-containing buffer, we are not differentiating them by using the terms gels and real amyloids, but refer to them as 0.1% SDS soluble Orb2 oligomers and 2% SDS soluble Orb2 oligomers. Overall, our observations here suggest in absence of Mrj, both of these kinds of Orb2 oligomers are decreased and so Mrj is most likely promoting the formation of Orb2 oligomers. It is possible that the 0.1% SDS soluble Orb2 oligomers gradually accumulate and undergo a further transition to the 2% SDS soluble Orb2 oligomers, so if in absence of Mrj, the formation of the 0.1% SDS soluble Orb2 oligomers is decreased, the next step of formation of 2% SDS soluble Orb2 oligomers also be decreased. This data is now presented in Figure 5H, I and J).

On the other possibility raised by the reviewer that Mrj can prevent the oligomeric state of Orb2 to become an irreversible non-functional amyloid, we think it is still possible for Mrj to do this but this could not be tested under the present conditions.

• It may be good for clarity to refer to the human Mrj as DNAJB6 according to the HUGO nomenclature. Also, the first evidence for its oligomerization was by Hageman et al 2010.

We have now changed mentions of human Mrj to DNAJB6. We apologize for missing the Hageman et al 2010 reference and have now cited this reference in the context of Mrj oligomerization.

• It is striking to see that Mrj co-Ips with Hsp70AA, Hsp70-4 but not Hsp70Cb. The fact that interactions were detected without using crosslinking is also striking given the reported transient nature of J-domain-Hsp70 interactions Together, this may even suggest that Mrj-1 is recognized as a Hsp70 substrate (for Hsp70AA, Hsp70-4 but not Hsp70Cb) rather than as a co-chaperone. In fact, a variant of Mrj-1 with a mutation in the HPD motif should be used to exclude this option.

In IP experiments we notice Mrj interacts with Hsp70Aa and Hsc70-4 but not with Hsc70-1 and Hsc70Cb. In our previously submitted manuscript, we realized we made a typo on the figure, where we referred to Hsp70Aa as Hsc70Aa. We have corrected this in the current revised manuscript. On the crosslinking point raised by the reviewer, we reviewed the published literature for studies of immunoprecipitation experiments which showed an interaction between DnaJB6 and Hsp70. We noted while one of the papers (Kakkar et al, 2016) report the use of a crosslinker in the experiment which showed an interaction between GFP-Hsp70 and V5-DnaJB6, in another two papers the interaction between endogenous Mrj and endogenous Hsp/c70 (Izawa et al, 2000) and Flag-Hsp70 and GFP-DnaJB6 (Bengoechea et al, 2020) could be detected without using any crosslinker. Our observations of detecting the interaction of Mrj with Hsp70Aa and Hsc70-4 in the absence of a crosslinker are thus similar to the observations reported by (Izawa et al, 2000; Bengoechea et al, 2020).

On the point of if Mrj is a substrate for Hsp70aa and Hsc70-4 and not a co-chaperone, we feel in the context of this manuscript, since we are focussing on the role of Mrj in the regulation of oligomerization of Orb2 and in memory, the experiment with HPD motif mutant is probably not necessary here. However, if the reviewers suggest this experiment to be essential, we can attempt this experiment by making this HPD motif mutant.

• The rest of these data reconfirm nicely that Mrj/DNAJB6 can suppress polyQ-Htt aggregation. Yet note that in this case the oligomers that enter the agarose gel are smaller, not bigger. This argues that Mrj is not an enhancer of oligomerization, but rather an inhibitor of the conversion of oligomers to a more amyloid like state.

Figure 2 and Supplemental Figure 4 discuss the effect of Mrj on Htt aggregation. We have used 2 different Htt constructs here. For Figure 2I, we used only Exon1 of Htt with the poly Q repeats. Here in SDD-AGE, for the control lane, we see the Htt oligomers as a smear for the control. For Mrj, we see only a small band at the bottom which can be interpreted most likely as either a monomer or as small oligomers since we do not observe any smear here. However, for the 588 amino acid fragment of HttQ138 in the SDD-AGE we do not see a difference in the length of the smear but in the intensity of the smear of the Htt oligomers (Supplemental Figure 4E). Based on this we are suggesting in presence of Mrj, there are lesser Htt oligomers. On the point of Mrj is not an enhancer of oligomerization, but rather an inhibitor of the conversion of oligomers to a more amyloid-like state, our experiments with the Mrj knockout show reduced Orb2 oligomers (both for 0.1% and 2% SDS soluble forms), suggesting Mrj plays a role in the conversion of Orb2 to the oligomeric state. If Mrj inhibits the conversion of oligomers to a more amyloid-like state, this is possible but we couldn’t test this hypothesis here. However, for Htt amyloid aggregates, previous works done by other labs with DnaJB6 as well as our experiments with Mrj suggest this as a likely possibility.

Figure 3: • The finding that knockout of DNAJB6 in mice is embryonic lethal is related to a problem with placental development and not embryonic development (Hunter et al, 1999; Watson et al, 2007, 2009, 2011) as well recognized by the authors. Therefore, the finding that deletion of Dm-Mrj has no developmental phenotype in Drosophila may not be that surprising.

We agree with the reviewer’s point that DNAJB6 mutant mice have a problem with placental development. However, one of the papers cited here (Watson et al, 2009) suggests DNAJB6 also plays a crucial role in the development of the embryo independent of the placenta development defect. The mammalian DNAJB6 mutant embryos generated using the tetraploid complementation method show severe neural defects including exencephaly, defect in neural tube closure, reduced neural tube size, and thinner neuroepithelium. Due to these features seen in the mice knockout, and the lack of such developmental defects in the Drosophila knockout, we interpreted our findings in Drosophila as significantly different from the mammals.

• It is a bit more surprising that Mrj knockout flies showed no aggregation phenotype or muscle phenotype, especially knowing that DNAJB6 mutations are linked to human diseases associated with aggregation (again well recognized by the authors). However, most of these diseases are late-onset and the phenotype may require stress to be revealed. So, while important to this MS in terms of not being a confounder for the memory test, I would like to ask the authors to add a note of caution that their data do not exclude that loss of Mrj activity still may cause a protein aggregation-related disease phenotype. Yet, I also do think that for the main message of this MS, it is not required to further test this experimentally.

We agree with the reviewer and have added this suggestion in the discussion that loss of Mrj may still result in a protein aggregation-related disease phenotype, probably under a sensitized condition of certain stresses which is not tested in this manuscript.

Figure 4:

• IPs were done with Orb2A as bite and clearly pulled down substantial amounts of GFP-tagged Mrj. For interactions with Orb2B, a V5-tagged Mrj was use and only a minor fraction was pulled down. Why were two different Mrj constructs used for Arb2A and Orb2b?

In the previously submitted manuscript, we have used HA-tagged Mrj (not V5) for checking the interaction with full-length Orb2B tagged with GFP. This was done with the goal of using the same Mrj-HA construct as that used in the initial Orb2A immunoprecipitation experiment. Since this has raised some concern as in the IPs to check for interaction between truncated Orb2A constructs (Orb2A325-GFP and Orb2AD162-GFP) and Mrj (Mrj-RFP), we used a different GFP and RFP tag combination. To address this, we have now added the same tag combinations for the IPs (Mrj-RFP with Orb2A-GFP and Orb2B-GFP). In these immunoprecipitation experiments where Mrj-RFP was pulled down using RFP Trap beads, we were able to detect positive interaction with GFP-tagged Orb2A and Orb2B. This data is now added in Figure 4F and 4I. We also have added the IP data for interaction between Mrj-HA and untagged Orb2B in Figure 4K, similar to the combination of initial experiment between Mrj-HA and untagged Orb2A.

• In addition, I think it would be important what one would see when pulling on Mrj1, especially under non-denaturing conditions and what is the status of the Orb2 that is than found to be associated with Mrj (monomeric, oligomeric and what size).

We have now performed IP from wild-type fly heads using anti Mrj antibody and ran the immunoprecipitate in SDS-PAGE and SDD-AGE followed by immunoblotting them with anti-Orb2 antibody. Our experiments suggest that immunoprecipitating endogenous Mrj brings down both the monomeric and oligomeric forms of Orb2. This data is now added in Figure 4L, M and N.

• This also relates to my remark at figure 1 and the subsequent fractionation experiments they show here in which there is a slight (not very convincing) increase in the ratio of TX100-soluble and insoluble (0.1% SDS soluble) material. My question would be if there is a remaining fraction of 0.1% insoluble (2% soluble) Orb2 and how Mrj affects that? As stated before, this is (only) mechanistically relevant to understanding why there is less oligomers of Orb2 in terms of Mrj either promoting it or by preventing it to transfer from a gel to a solid state. The link to the memory data remains intriguing, irrespective of what is going on (but also on those data I do have several comments: see below).

We have addressed this in response to the reviewer’s comments on Figure 1. We find in both wild type and Mrj knockout fly heads, there are Orb2 oligomers that can be detected using 0.1% SDS extraction and with further extraction with 2% SDS. The 2% SDS soluble Orb2 oligomers were previously insoluble during 0.1% SDS-based extraction. However, the amounts of both of these oligomers are reduced in Mrj knockout fly heads. Since we do not have the physicochemical characterization of both of these kinds of oligomers, we are not using the terms gel or solid state here but referring to these oligomers as 0.1% SDS soluble Orb2 oligomers and 2% SDS soluble Orb2 oligomers. We speculate that the 0.1% SDS soluble Orb2 oligomers over time transition to the 2% SDS soluble Orb2 oligomers. As in the absence of Mrj in the knockout flies, both the 0.1% SDS soluble and 2% SDS soluble Orb2 oligomers are decreased, this suggests Mrj is promoting the formation of Orb2 oligomers. On the reviewer’s point, if Mrj can prevent the transition from 0.1% SDS soluble to 2% SDS soluble Orb2 oligomers, we think it is possible for Mrj to both promote oligomerization of Orb2 as well as prevent it from forming bigger non-functional oligomers, but the second point is not tested here. The relevant data is now presented in Figure 5H, I and J.

• I also find the sentence that "Mrj is probably regulating the oligomerization of endogenous Orb2 in the brain" somewhat an overstatement. I would rather prefer to say that the data show that Mrj1 affects the oligomeric behavior/status of Orb2.

Based on the reviewer’s suggestion we have now changed the sentence to Mrj is probably regulating the oligomeric status of Orb2

Figure 5:

• To my knowledge, the Elav driver regulates expression in all neurons, but not in glial cells that comprise a significant part of the fly heads/brain. The complete absence of Mrj in the fly-heads when using this driver is therefore somewhat surprising. Or do we need to conclude from this that glial cells normally already lack Mrj expression?

On driving Mrj RNAi with Elav Gal4, we did not detect any Mrj in the western. We attempted to address the glial contribution towards Mrj’s expression we used a Glia-specific driver Repo Gal4 line to drive the control and Mrj RNAi line and performed a western blot using fly head lysate with anti-Mrj antibody. In this experiment, we did not observe any difference in Mrj levels between the two sets. As the Mrj antibody raised by us works in western blots but not in immunostainings, we could not do a colocalization analysis with a glial marker. However, we used the Mrj knockout Gal4 line to drive NLS-GFP and performed immunostainings of these flies with a glial marker anti-Repo antibody. Here we see two kinds of cells in the brain, one which have GFP but no Repo and the other where both are present together. This suggest that while Glial cells have Mrj but probably majority of Mrj in the brain comes from the neurons. We also found a reference where it was shown that Elav protein as well as Elav Gal4 at earlier stages of development expresses in neuroblasts and in all Glia (Berger et al, 2007). So, another possibility is when we are driving Mrj RNAi using Elav Gal4, this knocks down Mrj in both the neurons as well as in the glia. This coupled with the catalytic nature of RNAi probably creates an effective knockdown of Mrj as seen in the western blot. This data is now added in Supplementary Figure 5G and H.

• Why not use these lines also for the memory test for confirmation? I understand the concerns of putative confounding effects of a full knockdown (which were however not reported), but now data rely only on the mushroom body-specific knockdown for the 201Y Gal4 line, for which the knockdown efficiency is not provided. But even more so, here a temperature shift (22oC-30oC) was required to activate the expression of the siRNA. For the effects of this shift alone no controls were provided. The functional memory data are nice and consistent with the hypothesis, but can it be excluded that the temperature shift (rather than the Mrj) knockdown has caused the memory defects? I think it is crucial to include the proper controls or use a different knockdown approach that does not require temperature shifts or even use the knockout flies.

We have now performed the memory experiments with Mrj knockout flies. Our experiments show at 16 and 24-hour time points Mrj knockout flies have significantly reduced memory in comparison to the control wildtype. This data is now added in Figure 6B.

Figure 6:

The finding of a co-IP between Rpl18 and Mrj (one-directional only) by no means suffices to conclude that Mrj may interact with nascent Orb2 chains here (which would be the relevant finding here). The fact that Mrj is a self-oligomerising protein (also in vitro, so irrespective of ribosomal associations!), and hence is found in all fractions in a sucrose gradient, also is not a very strong case for its specific interaction with polysomes. The finding that there is just more self-oligomerizing Orb2A co-sedimenting with polysomes in sucrose gradients neither is evidence for a direct effect of Mrj enhancing association of Orb2A with the translating ribosomes even though it would fit the hypothesis. So all in all, I think the data in this figure and non-conclusive and the related conclusions should be deleted.

We have now performed the reverse co-IP between Rpl18-Flag and Mrj-HA using anti-HA antibody and could detect an interaction between the two. This data is now added in Supplementary Figure 6A.

To address if Mrj is a self-oligomerizing protein that can migrate to heavier polysome fractions due to its size, we have loaded recombinant Mrj on an identical sucrose gradient as we use for polysome analysis. Post ultra-centrifugation we fractionated the gradients and checked if Mrj can be detected in the fraction numbers where polysomes are present. In this experiment, we could not detect recombinant Mrj in the heavier polysome fractions (data presented in Supplementary Figure 6B). Overall, our observations of Mrj-Rpl18 IPs, the presence of cellularly expressed Mrj in polysome fractions, and the absence of recombinant Mrj from these fractions, suggest a likelihood of Mrj’s association with the translating ribosomes.

On the reviewer’s point of us concluding Mrj may interact with nascent Orb2 chains, we have not mentioned this possibility in the manuscript as we don’t have any evidence to suggest this. We have also added a sentence: This indicates that in presence of Mrj, the association of Orb2A with the translating ribosomes is enhanced, however, if this is a consequence of increased Orb2A oligomers due to Mrj or caused by interaction between polysome-associated Orb2A and Mrj will need to be tested in future.

Based on these above-mentioned points, we hope we can keep the data and conclusions of this section.

Overall, provided that proper controls/additional data can be provided for the key experiments of memory consolidation, I find this an intriguing study that would point towards a role of a molecular chaperone in controlling memory functions via regulating the oligomeric status of a prion-like protein and that is worthwhile publishing in a good journal.

However, in terms of mechanistical interpretations, several points have to be reconsidered (see remarks on figure 1,4); this pertains especially to what is discussed on page 13. In addition, I'd like the authors to put their data into the perspective of the findings that in differentiated neurons DNAJB6 levels actually decline, not incline (Thiruvalluvan et al, 2020), which would be counterintuitive if these proteins are playing a role as suggested here in memory consolidation.

We have addressed the comments on Figures 1 and 4 earlier. We have also added new memory experiment’s data with the Mrj knockout in Figure 6.

We have attempted to put the observations with Drosophila Mrj in perspective to observations in Thiruvalluvan et al, on human DnaJB6 in the discussions as follows:

Can our observation in Drosophila also be relevant for higher mammals? We tested this with human DnaJB6 and CPEB2. In mice CPEB2 knockout exhibited impaired hippocampus-dependent memory (Lu et al, 2017), so like Drosophila Orb2, its mammalian homolog CPEB2 is also a regulator of long-term memory. In immunoprecipitation assay we could detect an interaction between human CPEB2 and human DnaJB6, suggesting the feasibility for DnaJB6 to play a similar role to Drosophila Mrj in mammals. However, as the human DnaJB6 level was observed to undergo a reduction in transitioning from ES cells to neurons, (Thiruvalluvan et al, 2020) how this can be reconciled with its possible role in the regulation of memory. We speculate, such a reduction if is happening in the brain will occur in a highly regulatable manner to allow precise control over CPEB2 oligomerization only in specific neurons where it is needed and the reduced levels of DnaJB6 is probably sufficient to aid CPEB oligomerization Alternatively, there may be additional chaperones that may function in a stage-specific manner and be able to compensate for the decline in levels of DNAJB6.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

Summary: The manuscript describes the role of the Hsp40 family protein Mrj in the prion-like oligomerization of Orb2. The authors demonstrate that Mrj promotes the oligomerization of Orb2, while a loss in Mrj diminishes the extent of Orb2 oligomerization. They observe that while Mrj is not an essential gene, a loss in Mrj causes deficiencies in the consolidation of long-term memory. Further, they demonstrate that Mrj associates with polysomes and increases the association of Orb2 with polysomes.

Major comments: None

Minor comments:

1. In the section describing the chaperone properties of Mrj in clearing Htt aggregates (Fig 2), the legend describes that "Mrj-HA constructs are more efficient in decreasing Htt aggregation compared to Mrj-RFP". It would be helpful to add Mrj-RFP to the quantification in Fig 2G to know exactly the difference in efficiency. Is there an explanation for why the 2 constructs behave differently?

We have added the quantitation of Htt aggregates in presence of Mrj-RFP in the revised version (Data presented in Figure 2G). While the efficiency of Mrj-RFP to decrease Htt aggregates is significantly less in comparison to Mrj-HA, it is still significantly better in comparison to the control CG7133-HA construct. It is possible, due to the tagging of Mrj with a larger tag (RFP), this reduces its ability to decrease the Htt aggregates in comparison to the construct where Mrj is tagged with a much smaller HA tag.

Figs A, B, C, G need to have quantification of the percentage of colocalization with details about the number of cells quantified for each experiment.

We have now added the intensity profile images and colocalization quantitation (pearson’s coefficient) in the Supplemental Figure 5A and B. This quantitation is done from multiple ROI’s taken from at 4-6 cells.

In Fig 6 B, C, F, G it would be helpful to label the 40S, 60S and 80S peaks in the A 254 trace.

We have now labeled the 80S, and polysome peaks in the Figure 7B, C, F and G. We could not separate the 40S and 60S peaks in the A254 trace.

It's interesting that Mrj has opposing functions with regard to aggregation when comparing huntingtin with Orb2. From the literature presented in the discussion, it appears as though chaperones including Mrj have an anti-aggregation role for prions. It would be helpful to have more discussion around why, in the case of Orb2, this is different. The discussion states that "The only Hsp40 chaperone which was found similar to Mrj in increasing Orb2's oligomerization is the yeast Jjj2 protein" - this point needs elaboration, as well as a reference.

In the discussions section we have now added the following speculations on this:

One question here is why Mrj behaves differently with Orb2 in comparison to other amyloids. Orb2 differs from other pathogenic amyloids in its extremely transient existence in the toxic intermediate form (Hervás et al, 2016). For the pathogenic amyloids, since they exist in the toxic intermediate form for longer, Mrj probably gets more time to act and prevent or delay them from forming larger aggregates. For Orb2, Mrj may help to quickly transition it from the toxic intermediate state, thereby helping this state to be transient instead of longer. An alternate possibility is post-transition from the toxic intermediate state, Mrj stabilizes these Orb2 oligomers and prevents them from forming larger aggregates. This can be through Mrj interacting with Orb2 oligomers and blocking its surface thereby preventing more Orb2 from assembling over it. Another difference between the Orb2 oligomeric amyloids and the pathogenic amyloids is in the nature of their amyloid core. For the pathogenic amyloids, this core is hydrophobic devoid of any water molecules, however for Orb2, the core is hydrophilic. This raises another possibility that if the Orb2 oligomers go beyond a certain critical size, Mrj can destabilize these larger Orb2 aggregates by targeting its hydrophilic core.

On the Jjj2 point raised by the reviewer, we have added the (Li et al, 2016a) reference now and elaborated as:

The only Hsp40 chaperone which was found similar to Mrj in increasing Orb2’s oligomerization is the yeast Jjj2 protein. In Jjj2 knockout yeast strain, Orb2A mainly exists in the non-prion state, whereas on Jjj2 overexpression the non-prion state could be converted to a prion-like state. In S2 cells coexpression of Jjj2 with Orb2A lead to an increase in Orb2 oligomerization (Li et al, 2016a). However, Jjj2 differs from Mrj, as when it is expressed in S2 cells, we do not detect it to be present in the polysome fractions.

The Jjj2 polysome data is now presented in Supplementary Figure 6C.

Reviewer #2 (Significance (Required)):

General assessment:

Overall, the work is clearly described and the manuscript is very well-written. The motivation behind the study and its importance are well-explained. I only have minor comments and suggestions to improve the clarity of the work. The study newly describes the interaction between the chaperone Mrj and the translation regulator Orb2. The experiments that the screen for proteins that interact with Orb2 and promote its oligomerization are very thorough. The experiments that delve into the role of Mrj in protein synthesis are a good start, and need to be explored further, but that is beyond the scope of this study.

Advance: The study describes a new interaction between the chaperone Mrj and the translation regulator Orb2. The study is helpful in expanding our knowledge of prion regulators as well factors that affect memory acquisition and consolidation.

Audience: This paper will be of most interest to basic researchers.

My expertise is in Drosophila genetics and neuronal injury.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

The manuscript submitted by Desai et al. identifies a chaperone of the Hsp40 family (Mrj) that binds Orb2 and modulates its oligomerization, which is critical for Orb2 function in learning and memory in Drosophila. Orb2 are proteins with prion-like properties whose oligomerization is critical for their function in the storage of memories. The main contribution of the article is the screen of Hsp40 and Hsp70-family proteins that bind Orb2. The authors show IP results for all the candidates tested, including those that bind Fig. 1) and those that don't (Supp Fig 3). There is also a figure devoted to examining the interaction of Mrj with polyglutamine models (Htt). They also generate a KO mutant that is viable and shows no gross defects or protein aggregation. Lastly, they show that the silencing of Mrj in the mushroom body gamma neurons results in weaker memories in a courtship paradigm. Although the data is consistent and generally supportive of the hypothesis, key details are missing in several areas, including controls. Additionally, the interpretation of some results leaves room for debate. Overall, this is an ambitious article that needs additional work before publication.

Specific comments:

1. General concern over the interpretation of IP experiments and colocalization. These experiments don't necessarily reflect direct interactions. They are consistent with direct interaction but not the only explanation for a positive IP or colocalization.

This paper is centred on the interaction between Orb2A and Mrj, which we have detected using immunoprecipitation. The reviewer’s concern here is, this experiment is not able to distinguish if this can be a direct protein-protein interaction or if the two proteins are part of a complex.

To address this concern we have used purified recombinant protein-based pulldowns. Our experiments with purified protein pulldowns (GST tagged Mrj from E.coli with Orb2A from E.coli or Orb2A-GFP from Sf9 cells) suggest Orb2A and Mrj can directly interact amongst themselves. This data is now presented in Figure 1J and K.

The Huntingtin section has a few concerns. The IF doesn't show all controls and the quantification is not well done in terms of what is relevant. A major problem is the interpretation of Fig 2F. The idea is that Mrj prevents the aggregation of Htt, which is the opposite of what is observed with Orb2. The panel actually shows a large Htt aggregate instead of multiple small aggregates. This has been reported before in Drosophila and other systems with different polyQ models. Mrj and other Hsp40 and Hsp70 proteins modify Htt aggregation, but in an unexpected way. This affects the model shown in Fig. 6H. Lastly, Fig 2H and 2I show very different level of total Htt.

In Figure 2F of the previously submitted manuscript, we have shown representative images of HttQ103-GFP cells coexpressing with a control DnaJ protein CG7133-HA and Mrj-HA. In Figure 2G we quantitated the number of cells showing aggregates within the population of doubly transfected cells. On the reviewer’s point of figure 2F showing large Htt aggregates instead of multiple small aggregates, we do not see a large Htt aggregate in presence of Mrj in this figure, the pattern looks diffused here and very different from the control CG7133 where the aggregates are seen. We have performed the same experiment with a different Htt construct (588 amino acids long fragment) tagged with RFP, and here also we notice in presence of Mrj, the aggregates are decreased and the expression pattern looks diffused (Supplementary Figure 4E, 4F).

If the comment on large Htt aggregates in presence of Mrj is concerning figure 2E, here we show Mrj-RFP to colocalize with the Htt aggregates. Here, even though Mrj-RFP colocalizes with Htt aggregates, it rescues the Htt aggregation phenotype as in comparison to the control CG7133, the number of cells with Htt aggregates is still significantly less here. We have added this quantitation of rescue by Mrj-RFP in the revised manuscript now. The observation of colocalization of Mrj-RFP with Htt aggregates is similar to previous reports of chaperones rescuing Htt aggregation and yet showing colocalization with the aggregates. Both Hdj-2 and Hsc70 suppress Htt aggregation and yet were observed to colocalize with Htt aggregates in the cell line model as well as in nuclear inclusions in the brain (Jana et al, 2000). In a nematode model of Htt aggregation, DNJ-13 (DnaJB-1), HSP-1 (Hsc70), and HSP-11 (Apg-2) were shown to colocalize with Htt aggregates and yet decrease the Htt aggregation (Scior et al, 2018). Hsp70 was also found to colocalize with Htt aggregates in Hela cells (Kim et al, 2002).

Regarding Figures 2H and 2I, while figure 2H is of an SDS-PAGE to show no difference in the levels of monomeric HttQ103 (marked with *) in presence of Mrj and the control CG7133, figure 2I is for the same samples ran in an SDD-AGE where reduced amount of Htt oligomers as seen with the absence of a smear in presence of Mrj. The apparent difference in Htt levels between 2H and 2I is due to the detection of Htt aggregates/oligomers in the SDD-AGE which are unable to enter the SDS-PAGE and hence undetected. In Supplementary Figure 4E, similar experiments were done with the longer Htt588 fragment and here we notice in the SDD-AGE reduced intensity of the smear made up of Htt oligomers, again suggesting a reduction in Htt aggregates. Thus our results are not in contradiction to previous studies where Mrj was found to rescue Htt aggregate-associated toxicity.

Endogenous expression of Mrj using Gal4 line: where else is it expressed in the brain / head and in muscle. Fig 3G shows no muscle abnormalities but no evidence is shown for muscle expression. It is nice that Fig 3E and F show no abnormal aggregates in the Mrj mutant, but this would be maybe more interesting if flies were subjected to some form of stress.

We have now added images of the brain and muscles to show the expression pattern of Mrj. Using Mrj Gal4 line and UAS- CD8GFP, we noticed enriched expression in the optic lobes, mushroom body, and olfactory lobes. We also noticed GFP expression in the larval muscles and neuromuscular junction synaptic boutons. This data is now presented in Supplementary Figure 5C, D, E and F.

On the reviewer’s point of subjecting the Mrj KO flies to some form of stress, we have not performed this. We have added in the discussions a note of caution, that loss of Mrj may still result in a protein aggregation-related disease phenotype, probably under a sensitized condition of certain stresses which is not tested in this manuscript.

Fig. 5B shows no Mrj detectable from head homogenates in flies silencing Mrj in neurons with Elav-Gal4. It would be nice if they could show that ONLY neurons express Mrj in the head. Also noted, Elav-Gal4 is a weak driver, so it is surprising that it can generate such robust loss of Mrj protein

We have used an X chromosome Elav Gal4 driver to drive the UAS-Mrj RNAi line and here we could not detect Mrj in the western. To address the reviewer’s point on the glial contribution towards expression of Mrj, we used a Glial driver Repo Gal4 to drive Mrj RNAi. In this experiment, we did not detect any difference in Mrj levels between the control and the Mrj RNAi line (presented now in Supplementary Figure 5G). We also used the Mrj knockout Gal4 line to drive NLS-GFP and immunostained these using a glial marker anti-Repo antibody. Here, we were able to detect cells colabelled by GFP as well as Repo, suggesting Mrj is likely to be present in the glial cells (presented now in Supplementary Figure 5H). We also looked in the literature and found a reference where it was shown that Elav protein as well as Elav Gal4 at earlier stages of development expresses in neuroblasts and in all Glia (Berger et al, 2007). So, another possibility is when we are driving Mrj RNAi using Elav Gal4, this knocks down Mrj in both the neurons as well as in the glia.

Fig 4-Colocalization of Orb2 with Mrj lacks controls. The quantification could describe other phenomena because the colocalization is robust but the numbers shown describe something else.

We have now added the intensity profile and colocalization quantitation (pearson’s coefficient) in Supplemental Figure 5A and B. This quantitation is done from multiple ROI’s taken from 4-6 cells. Also, to suggest the interaction of Orb2 isoforms with Mrj, we are not depending on colocalization alone and have used immunoprecipitation experiments to support our observations.

Fly behavior. The results shown for Mrj RNAi alleles is fine but it would be more robust if this was validated with the KO line AND rescued with Mrj overexpression.

We have now performed memory assays with the Mrj knockout. Our experiments showed Mrj knockouts to show significantly decreased memory in comparison to wild-type flies at 16 and 24-hour time points (presented in Figure 6B). We have not been able to make an Mrj Knockout-UAS Mrj recombinant fly, most likely due to the closeness of the two with respect to their genomic location in second chromosome.

Minor comments:

Please, revise minor errors, there are several examples of words together without a space.

We have identified the words without space and have corrected them now.

Intro: describe the use of functional prions. Starting the paragraph with this sentence and then explaining what prion diseases are is a little confusing. Also "prion proteins" can be confusing because the term refers to PrP, the protein found in prions.

We have now altered the introduction and have described functional prions.

Results, second subtitle in page 5. This sentence is quite confusing using prion-like twice

We have now changed the heading to “Drosophila Mrj converts Orb2A from non-prion to a prion-like state.”

Page 6: "conversion from non-prion to prion-like form...". This can be presented differently. Prion-like properties are intrinsic, proteins don't change from non-prion to prion-like. They may be oligomeric or monomeric or highly aggregated but the prion-like property doesn't change

We agree with the reviewer's point of Prion-like properties are intrinsic, but the protein might or might not exist in the prion-like state or confirmation. When we are using the term conversion from non-prion to prion-like form we mean to suggest a conformational conversion leading to the eventual formation of the oligomeric species. We also noted the terminology of non-prion to prion-like state change is used in several papers (Satpute-Krishnan & Serio, 2005; Sw & Yo, 2012; Uptain et al, 2001).

Scale bars and text are too small in several figures

We have now mentioned in the figure legends the size of the scale bars. For several images we have made the scale bars also larger.

Not sure why Fig 4C is supplemental, seems like an important piece of data.

We have kept this data in the supplemental data as we performed this experiment with recombinant protein which is tagged with 6X His and we are not sure if this high degree of oligomerization/aggregation of recombinant Mrj and further precipitation over time, happens inside the cells/ brain.

Intro to Mrj KO in page 7 is too long. Most of it belongs in the discussion

We have now moved the portions on mammalian DNAJB6 which were earlier in the results section to the discussions section.

Change red panels in IF to other color to make it easier for colorblind readers.

We have now changed the red panels to magenta. We apologize for our figures not being colorblind friendly earlier.

The discussion is a little diffuse by trying to compare Orb2 with mammalian prions and amyloids and yeast prions.

We looked into the functional prion data and couldn’t find much on chaperone mediated regulation of these. Also, we felt comparing with the amyloids and yeast prions brings out the contrast with respect to the Mrj mediated regulatory differences between the two.

Reviewer #3 (Significance (Required)):

This is a paper with a broad scope and approaches. The paper describes the role of Mrj in the oligomerization of Orb2 by protein biochemistry techniques and determine the role of loss of Mrj in the mushroom bodies in fly behavior.

The audience for this content is basic research and specialized. The role of Mrj in Orb2 aggregation and function sheds new light on the mechanisms regulating the function of this protein involved in a novel mechanism of learning and memory.

References:

Bengoechea R, Findlay AR, Bhadra AK, Shao H, Stein KC, Pittman SK, Daw JA, Gestwicki JE, True HL & Weihl CC (2020) Inhibition of DNAJ-HSP70 interaction improves strength in muscular dystrophy. J Clin Invest 130: 4470–4485

Berger C, Renner S, Lüer K & Technau GM (2007) The commonly used marker ELAV is transiently expressed in neuroblasts and glial cells in the Drosophila embryonic CNS. Dev Dyn 236: 3562–3568

Fayazi Z, Ghosh S, Marion S, Bao X, Shero M & Kazemi-Esfarjani P (2006) A Drosophila ortholog of the human MRJ modulates polyglutamine toxicity and aggregation. Neurobiol Dis 24: 226–244

Heinrich SU & Lindquist S (2011) Protein-only mechanism induces self-perpetuating changes in the activity of neuronal Aplysia cytoplasmic polyadenylation element binding protein (CPEB). Proc Natl Acad Sci U S A 108: 2999–3004

Hervás R, Li L, Majumdar A, Fernández-Ramírez MDC, Unruh JR, Slaughter BD, Galera-Prat A, Santana E, Suzuki M, Nagai Y, et al (2016) Molecular Basis of Orb2 Amyloidogenesis and Blockade of Memory Consolidation. PLoS Biol 14: e1002361

Hervas R, Rau MJ, Park Y, Zhang W, Murzin AG, Fitzpatrick JAJ, Scheres SHW & Si K (2020) Cryo-EM structure of a neuronal functional amyloid implicated in memory persistence in Drosophila. Science 367: 1230–1234

Izawa I, Nishizawa M, Ohtakara K, Ohtsuka K, Inada H & Inagaki M (2000) Identification of Mrj, a DnaJ/Hsp40 family protein, as a keratin 8/18 filament regulatory protein. J Biol Chem 275: 34521–34527

Jana NR, Tanaka M, Wang G h & Nukina N (2000) Polyglutamine length-dependent interaction of Hsp40 and Hsp70 family chaperones with truncated N-terminal huntingtin: their role in suppression of aggregation and cellular toxicity. Hum Mol Genet 9: 2009–2018

Kakkar V, Månsson C, de Mattos EP, Bergink S, van der Zwaag M, van Waarde MAWH, Kloosterhuis NJ, Melki R, van Cruchten RTP, Al-Karadaghi S, et al (2016) The S/T-Rich Motif in the DNAJB6 Chaperone Delays Polyglutamine Aggregation and the Onset of Disease in a Mouse Model. Mol Cell 62: 272–283

Kim S, Nollen EAA, Kitagawa K, Bindokas VP & Morimoto RI (2002) Polyglutamine protein aggregates are dynamic. Nat Cell Biol 4: 826–831

Li L, Sanchez CP, Slaughter BD, Zhao Y, Khan MR, Unruh JR, Rubinstein B & Si K (2016a) A Putative Biochemical Engram of Long-Term Memory. Curr Biol 26: 3143–3156

Li S, Zhang P, Freibaum BD, Kim NC, Kolaitis R-M, Molliex A, Kanagaraj AP, Yabe I, Tanino M, Tanaka S, et al (2016b) Genetic interaction of hnRNPA2B1 and DNAJB6 in a Drosophila model of multisystem proteinopathy. Hum Mol Genet 25: 936–950

Liebman SW & Chernoff YO (2012) Prions in yeast. Genetics 191: 1041–1072

Lu W-H, Yeh N-H & Huang Y-S (2017) CPEB2 Activates GRASP1 mRNA Translation and Promotes AMPA Receptor Surface Expression, Long-Term Potentiation, and Memory. Cell Rep 21: 1783–1794

Prusiner SB (2001) Neurodegenerative Diseases and Prions. New England Journal of Medicine 344: 1516–1526

Satpute-Krishnan P & Serio TR (2005) Prion protein remodelling confers an immediate phenotypic switch. Nature 437: 262–265

Scior A, Buntru A, Arnsburg K, Ast A, Iburg M, Juenemann K, Pigazzini ML, Mlody B, Puchkov D, Priller J, et al (2018) Complete suppression of Htt fibrilization and disaggregation of Htt fibrils by a trimeric chaperone complex. EMBO J 37: 282–299

Si K (2015) Prions: what are they good for? Annu Rev Cell Dev Biol 31: 149–169

Si K, Choi Y-B, White-Grindley E, Majumdar A & Kandel ER (2010) Aplysia CPEB can form prion-like multimers in sensory neurons that contribute to long-term facilitation. Cell 140: 421–435

Si K, Lindquist S & Kandel ER (2003) A neuronal isoform of the aplysia CPEB has prion-like properties. Cell 115: 879–891

Sw L & Yo C (2012) Prions in yeast. Genetics 191

Thiruvalluvan A, de Mattos EP, Brunsting JF, Bakels R, Serlidaki D, Barazzuol L, Conforti P, Fatima A, Koyuncu S, Cattaneo E, et al (2020) DNAJB6, a Key Factor in Neuronal Sensitivity to Amyloidogenesis. Mol Cell 78: 346-358.e9

Uptain SM & Lindquist S (2002) Prions as protein-based genetic elements. Annu Rev Microbiol 56: 703–741

Uptain SM, Sawicki GJ, Caughey B & Lindquist S (2001) Strains of [PSI(+)] are distinguished by their efficiencies of prion-mediated conformational conversion. EMBO J 20: 6236–6245

Watson ED, Mattar P, Schuurmans C & Cross JC (2009) Neural stem cell self-renewal requires the Mrj co-chaperone. Dev Dyn 238: 2564–2574

Wickner RB (2016) Yeast and Fungal Prions. Cold Spring Harb Perspect Biol 8: a023531

Wickner RB, Edskes HK, Maddelein ML, Taylor KL & Moriyama H (1999) Prions of yeast and fungi. Proteins as genetic material. J Biol Chem 274: 555–558

Wickner RB, Masison DC, Edskes HK & Maddelein ML (1996) Prions of yeast, [PSI] and [URE3], as models for neurodegenerative diseases. Cold Spring Harb Symp Quant Biol 61: 541–550

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript submitted by Desai et al. identifies a chaperone of the Hsp40 family (Mrj) that binds Orb2 and modulates its oligomerization, which is critical for Orb2 function in learning and memory in Drosophila. Orb2 are proteins with prion-like properties whose oligomerization is critical for their function in the storage of memories. The main contribution of the article is the screen of Hsp40 and Hsp70-family proteins that bind Orb2. The authors show IP results for all the candidates tested, including those that bind Fig. 1) and those that don't (Supp Fig 3). There is also a figure devoted to examining the interaction of Mrj with polyglutamine models (Htt). They also generate a KO mutant that is viable and shows no gross defects or protein aggregation. Lastly, they show that the silencing of Mrj in the mushroom body gamma neurons results in weaker memories in a courtship paradigm. Although the data is consistent and generally supportive of the hypothesis, key details are missing in several areas, including controls. Additionally, the interpretation of some results leaves room for debate. Overall, this is an ambitious article that needs additional work before publication.

Specific comments:

1. General concern over the interpretation of IP experiments and colocalization. These experiments don't necessarily reflect direct interactions. They are consistent with direct interaction but not the only explanation for a positive IP or colocalization.
2. The Huntingtin section has a few concerns. The IF doesn't show all controls and the quantification is not well done in terms of what is relevant. A major problem is the interpretation of Fig 2F. The idea is that Mrj prevents the aggregation of Htt, which is the opposite of what is observed with Orb2. The panel actually shows a large Htt aggregate instead of multiple small aggregates. This has been reported before in Drosophila and other systems with different polyQ models. Mrj and other Hsp40 and Hsp70 proteins modify Htt aggregation, but in an unexpected way. This affects the model shown in Fig. 6H. Lastly, Fig 2H and 2I show very different level of total Htt.
3. Endogenous expression of Mrj using Gal4 line: where else is it expressed in the brain / head and in muscle. Fig 3G shows no muscle abnormalities but no evidence is shown for muscle expression. It is nice that Fig 3E and F show no abnormal aggregates in the Mrj mutant, but this would be maybe more interesting if flies were subjected to some form of stress.
4. Fig. 5B shows no Mrj detectable from head homogenates in flies silencing Mrj in neurons with Elav-Gal4. It would be nice if they could show that ONLY neurons express Mrj in the head. Also noted, Elav-Gal4 is a weak driver, so it is surprising that it can generate such robust loss of Mrj protein
5. Fig 4-Colocalization of Orb2 with Mrj lacks controls. The quantification could describe other phenomena because the colocalization is robust but the numbers shown describe something else.
6. Fly behavior. The results shown for Mrj RNAi alleles is fine but it would be more robust if this was validated with the KO line AND rescued with Mrj overexpression.

Minor comments:

Please, revise minor errors, there are several examples of words together without a space.

Intro: describe the use of functional prions. Starting the paragraph with this sentence and then explaining what prion diseases are is a little confusing. Also "prion proteins" can be confusing because the term refers to PrP, the protein found in prions.

Results, second subtitle in page 5. This sentence is quite confusing using prion-like twice

Page 6: "conversion from non-prion to prion-like form...". This can be presented differently. Prion-like properties are intrinsic, proteins don't change from non-prion to prion-like. They may be oligomeric or monomeric or highly aggregated but the prion-like property doesn't change

Scale bars and text are too small in several figures

Not sure why Fig 4C is supplemental, seems like an important piece of data.

Intro to Mrj KO in page 7 is too long. Most of it belongs in the discussion

Change red panels in IF to other color to make it easier for colorblind readers.

The discussion is a little diffuse by trying to compare Orb2 with mammalian prions and amyloids and yeast prions.

Significance

This is a paper with a broad scope and approaches. The paper describes the role of Mrj in the oligomerization of Orb2 by protein biochemistry techniques and determine the role of loss of Mrj in the mushroom bodies in fly behavior.

The audience for this content is basic research and specialized. The role of Mrj in Orb2 aggregation and function sheds new light on the mechanisms regulating the function of this protein involved in a novel mechanism of learning and memory.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

The manuscript describes the role of the Hsp40 family protein Mrj in the prion-like oligomerization of Orb2. The authors demonstrate that Mrj promotes the oligomerization of Orb2, while a loss in Mrj diminishes the extent of Orb2 oligomerization. They observe that while Mrj is not an essential gene, a loss in Mrj causes deficiencies in the consolidation of long-term memory. Further, they demonstrate that Mrj associates with polysomes and increases the association of Orb2 with polysomes.

Major comments: None

Minor comments:

1. In the section describing the chaperone properties of Mrj in clearing Htt aggregates (Fig 2), the legend describes that "Mrj-HA constructs are more efficient in decreasing Htt aggregation compared to Mrj-RFP". It would be helpful to add Mrj-RFP to the quantification in Fig 2G to know exactly the difference in efficiency. Is there an explanation for why the 2 constructs behave differently?
2. Figs A, B, C, G need to have quantification of the percentage of colocalization with details about the number of cells quantified for each experiment.
3. In Fig 6 B, C, F, G it would be helpful to label the 40S, 60S and 80S peaks in the A 254 trace.
4. It's interesting that Mrj has opposing functions with regard to aggregation when comparing huntingtin with Orb2. From the literature presented in the discussion, it appears as though chaperones including Mrj have an anti-aggregation role for prions. It would be helpful to have more discussion around why, in the case of Orb2, this is different. The discussion states that "The only Hsp40 chaperone which was found similar to Mrj in increasing Orb2's oligomerization is the yeast Jjj2 protein" - this point needs elaboration, as well as a reference.

Significance

General assessment:

Overall, the work is clearly described and the manuscript is very well-written. The motivation behind the study and its importance are well-explained. I only have minor comments and suggestions to improve the clarity of the work. The study newly describes the interaction between the chaperone Mrj and the translation regulator Orb2. The experiments that the screen for proteins that interact with Orb2 and promote its oligomerization are very thorough. The experiments that delve into the role of Mrj in protein synthesis are a good start, and need to be explored further, but that is beyond the scope of this study.

Advance:

The study describes a new interaction between the chaperone Mrj and the translation regulator Orb2. The study is helpful in expanding our knowledge of prion regulators as well factors that affect memory acquisition and consolidation.

Audience:

This paper will be of most interest to basic researchers. My expertise is in Drosophila genetics and neuronal injury.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this MS, Mrj - a member of the JDP family of Hsp70 co-chaperones was identified as a regulator of the conversion of Orb2A (the Dm ortholog of CPEB) to its prion-like form.

In drosophila, Mrj deletion does not cause any gross neurodevelopmental defect nor leads to detectable alterations in protein homeostasis. Loss of Mrj, however, does lead to altered Orb2 oligomerization. Consistent with a role of prion-like characteristics of Orb2 in memory consolidation, loss of Mrj results in a deficit in long-term memory.

Aside from the fact that there are some unclarities related to the physicochemical properties of Orb2 and how Mrj affects this precisely, the finding that a chaperone could be important for memory is an interesting observation, albeit not entirely novel.

In addition, there are several minor technical concerns and questions I have that I feel the authors should address, including a major one related to the actual approach used to demonstrate memory deficits upon loss of Mrj.

Significance

Figure 1 (plus related Supplemental figures):

• There seem to be two isoforms of Mrj (like what has been found for human DNAJB6). I find it striking to see that only (preferentially?) the shorter isoform interacts with Orb2. For DNAJB6, the long isoform is mainly related to an NLS and the presumed substrate binding is identical for both isoforms. If this is true for Dm-Mrj too, the authors could actually use this to demonstrate the specificity of their IPs where Orb2 is exclusively non-nuclear?
• I would be interested to know a bit more about the other 5 JDPs that are interactors with Orb2: are the human orthologs of those known? It is striking that these other 5 JDPs interact with Orb2 in Dm (in IPs) but have no impact on Sup35 prion behavior. Importantly, this does not imply they may not have impact on the prion-like behavior of other Dm substrates, including Dm-Orb2.
• The data in panels H,I indeed suggest that Mrj1 alters the (size of) the oligomers. It would be important to know what is the actual physicochemical change that is occurring here. The observed species are insoluble in 0.1 % TX100 but soluble in 0.1% SDS, which suggest they could be gels, but not real amyloids such as formed by the polyQ proteins that require much higher SDS concentrations (~2%) to be solubilized. This is relevant as Mrj1 reduces polyQ amyloidogenesis whereas is here is shown to enhance Orb2A oligomerization/gelidification. In the same context, it is striking to see that without Mrj the amount of Orb2A seems drastically reduced and I wonder whether this might be due to the fact that in the absence of Mrj a part of Orb2A is not recovered/solubilized due to its conversion for a gel to a solid/amyloid state? In other words: Mrj1 may not promote the prion state, but prevents that state to become an irreversible, non-functional amyloid?

Figure 2 (plus related Supplemental figures):

• It may be good for clarity to refer to the human Mrj as DNAJB6 according to the HUGO nomenclature. Also, the first evidence for its oligomerization was by Hageman et al 2010.
• It is striking to see that Mrj co-IPs with Hsp70AA, Hsp70-4 but not Hsp70Cb. The fact that interactions were detected without using crosslinking is also striking given the reported transient nature of J-domain-Hsp70 interactions Together, this may even suggest that Mrj-1 is recognized as a Hsp70 substrate (for Hsp70AA, Hsp70-4 but not Hsp70Cb) rather than as a co-chaperone. In fact, a variant of Mrj-1 with a mutation in the HPD motif should be used to exclude this option.
• The rest of these data reconfirm nicely that Mrj/DNAJB6 can suppress polyQ-Htt aggregation. Yet note that in this case the oligomers that enter the agarose gel are smaller, not bigger. This argues that Mrj is not an enhancer of oligomerization, but rather an inhibitor of the conversion of oligomers to a more amyloid like state.

Figure 3:

• The finding that knockout of DNAJB6 in mice is embryonic lethal is related to a problem with placental development and not embryonic development (Hunter et al, 1999; Watson et al, 2007, 2009, 2011) as well recognized by the authors. Therefore, the finding that deletion of Dm-Mrj has no developmental phenotype in Drosophila may not be that surprising.
• It is a bit more surprising that Mrj knockout flies showed no aggregation phenotype or muscle phenotype, especially knowing that DNAJB6 mutations are linked to human diseases associated with aggregation (again well recognized by the authors). However, most of these diseases are late-onset and the phenotype may require stress to be revealed. So, while important to this MS in terms of not being a confounder for the memory test, I would like to ask the authors to add a note of caution that their data do not exclude that loss of Mrj activity still may cause a protein aggregation-related disease phenotype. Yet, I also do think that for the main message of this MS, it is not required to further test this experimentally.

Figure 4:

• IPs were done with Orb2A as bite and clearly pulled down substantial amounts of GFP-tagged Mrj. For interactions with Orb2B, a V5-tagged Mrj was use and only a minor fraction was pulled down. Why were two different Mrj constructs used for Arb2A and Orb2b?
• In addition, I think it would be important what one would see when pulling on Mrj1, especially under non-denaturing conditions and what is the status of the Orb2 that is than found to be associated with Mrj (monomeric, oligomeric and what size).
• This also relates to my remark at figure 1 and the subsequent fractionation experiments they show here in which there is a slight (not very convincing) increase in the ratio of TX100-soluble and insoluble (0.1% SDS soluble) material. My question would be if there is a remaining fraction of 0.1% insoluble (2% soluble) Orb2 and how Mrj affects that? As stated before, this is (only) mechanistically relevant to understanding why there is less oligomers of Orb2 in terms of Mrj either promoting it or by preventing it to transfer from a gel to a solid state. The link to the memory data remains intriguing, irrespective of what is going on (but also on those data I do have several comments: see below).
• I also find the sentence that "Mrj is probably regulating the oligomerization of endogenous Orb2 in the brain" somewhat an overstatement. I would rather prefer to say that the data show that Mrj1 affects the oligomeric behavior/status of Orb2.

Figure 5:

• To my knowledge, the Elav driver regulates expression in all neurons, but not in glial cells that comprise a significant part of the fly heads/brain. The complete absence of Mrj in the fly-heads when using this driver is therefore somewhat surprising. Or do we need to conclude from this that glial cells normally already lack Mrj expression?
• Why not use these lines also for the memory test for confirmation? I understand the concerns of putative confounding effects of a full knockdown (which were however not reported), but now data rely only on the mushroom body-specific knockdown for the 201Y Gal4 line, for which the knockdown efficiency is not provided. But even more so, here a temperature shift (22oC-30oC) was required to activate the expression of the siRNA. For the effects of this shift alone no controls were provided. The functional memory data are nice and consistent with the hypothesis, but can it be excluded that the temperature shift (rather than the Mrj) knockdown has caused the memory defects? I think it is crucial to include the proper controls or use a different knockdown approach that does not require temperature shifts or even use the knockout flies.

Figure 6:

The finding of a co-IP between Rpl18 and Mrj (one-directional only) by no means suffices to conclude that Mrj may interact with nascent Orb2 chains here (which would be the relevant finding here). The fact that Mrj is a self-oligomerising protein (also in vitro, so irrespective of ribosomal associations!), and hence is found in all fractions in a sucrose gradient, also is not a very strong case for its specific interaction with polysomes. The finding that there is just more self-oligomerizing Orb2A co-sedimenting with polysomes in sucrose gradients neither is evidence for a direct effect of Mrj enhancing association of Orb2A with the translating ribosomes even though it would fit the hypothesis. So all in all, I think the data in this figure and non-conclusive and the related conclusions should be deleted.

Overall, provided that proper controls/additional data can be provided for the key experiments of memory consolidation, I find this an intriguing study that would point towards a role of a molecular chaperone in controlling memory functions via regulating the oligomeric status of a prion-like protein and that is worthwhile publishing in a good journal.

However, in terms of mechanistical interpretations, several points have to be reconsidered (see remarks on figure 1,4); this pertains especially to what is discussed on page 13. In addition, I'd like the authors to put their data into the perspective of the findings that in differentiated neurons DNAJB6 levels actually decline, not incline (Thiruvalluvan et al, 2020), which would be counterintuitive if these proteins are playing a role as suggested here in memory consolidation.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the reviewers for their time and for the helpful comments. We felt the reviews overall were fair and quite positive. All three reviewers felt the manuscript could be of broad, general interest, especially given the relevance of the protein (Pbp1/ataxin-2) to neurodegenerative conditions and stress granule biology. Reviewer 3 seems to have some doubt there could be specificity for the types of transcripts regulated by Pbp1, given prior studies of mammalian ataxin-2 which implicated 16,000+ mRNAs that could bind via PAR-CLIP experiments! However, our study shows the power of utilizing a simpler model organism and thinking about the metabolic state of cells for elucidating the function of this interesting protein. Although our demonstration of the specificity of Pbp1 for regulating Puf3-target mRNAs involved in mitochondrial biogenesis and mitochondrial function may be surprising to this reviewer, we have the utmost confidence in our data and feel the study represents a highly significant finding that will be of interest to many researchers.

1. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

*Reviewer #1 (Evidence, reproducibility and clarity (Required)): *

*In this manuscript, van de Poll et al. aim to further establish the function of poly(A) binding protein-binding protein 1 (Pbp1) and its relationship to the RNA-binding protein Puf3 in regulating the expression of mitochondrial proteins. This work builds upon a solid body of previous studies from this group regarding the function of Puf3 and the role of Pbp1 in regulating TORC1 signaling. Here, the authors show that Pbp1 has a physical and functional interaction with Puf3 and that ablation or disruption of Pbp1 eliminates this interaction and reduces the expression of some Puf3 target proteins. This work provides reliable data supporting a role for Pbp1 in the regulation of mitochondrial protein expression and that there is a clear functional interaction between Puf3 and Pbp1. Nonetheless, there are several issues that should be addressed before the manuscript is suitable for publication. *

*1.) The model put forward suggests that Pbp1 works to recruit Puf3 to the vicinity of mitochondria, where Puf3 can then promote the expression of its target mRNAs. In Figure 3A, however, deletion of Pbp1 has a stronger affect on the expression of COX2 than deletion of Puf3 alone, which would not be expected if the role of Pbp1 is to modulate Puf3 function. Similarly, the expression of COX2 is higher in the double deletion versus the Pbp1 single deletion. The authors should attempt to clarify this experimentally, or at least make mention of alternative mechanisms for Pbp1 which may be causing this. *

We have mentioned alternative mechanisms in the text. We suggest that Puf3-target mRNAs also can be translated through Tom20, in the absence of Puf3 (p. 8, ref 18). In single deletion strains lacking Pbp1 (but with Puf3 present), Puf3 may direct some of its target mRNAs to decay pathways, leading to lower Cox2 expression compared to double deletion strains that also lack Puf3. We performed qPCR analysis of several Puf3-target and other mRNAs in pbp1∆puf3∆ double deletion strains and some transcripts (e.g., COX17) would support this possibility (Fig S4).

* 2.) The authors mention that the increased pull-down of Puf3 with Pbp1 in respiratory conditions suggests that the Pbp1-Puf3 interaction is responsive to the cellular metabolic state (Figure 3B). However, the increase in Puf3 expression makes it difficult to compare the interactions between the two conditions. *

Yes this is correct, we stated this in the legend of Fig 4B as: “Increased amounts of Puf3 are associated with Pbp1 in respiratory conditions.” We clarified in the text that Puf3 expression increases in respiratory conditions and this likely explains the increased pull-down of Puf3 with Pbp1 (p. 10).

* 3.) The authors only look at a very small subset of Puf3 target mRNAs using qPCR when it would be much more informative and overall convincing to examine a larger amount using RNA-seq experiments. *

We conducted an RNA-seq experiment to compare transcriptomes of WT vs pbp1∆ cells in fermentative (YPD) vs respiratory (YPL) conditions and observed mRNAs with functions associated with mito-translation and mito-respiration (i.e., Puf3-targets) to be most differentially expressed – and majority of these are lower in abundance in pbp1∆ cells (new Fig 2).

* 4.) The authors consistently mention that Pbp1 function is helping to stabilize Puf3 target mRNAs. However, if the authors wish to prove this particular mode-of-action, more direct evidence should be provided, such as a pulse-chase experiment. Otherwise, other models allowing for increased mRNA abundance should be noted. *

Using thiolutin which is the standard for such expts, we measured mRNA half-lives of several Puf3-target and other mRNAs (COX17, COX10, POR1, ACT1) by qPCR in WT vs pbp1∆ cells. However, these data turned out to be difficult to interpret, as several “control” mRNAs exhibited different decay profiles in pbp1∆ vs WT cells, and their behavior was different in respiratory conditions compared to what was reported in common glucose media. Nonetheless, the data are included as Fig S2, and the important observation is that each of the Puf3-target mRNAs tested behaves similarly following thiolutin treatment, compared to non Puf3-target mRNAs. Given that Puf3-target mRNAs were more stable in pbp1∆ cells (compared to PGK1) following thiolutin treatment, we have deleted the term “stabilize” throughout the text. The exact fate of these mRNAs in normal vs pbp1∆ mutant cells will require more sophisticated investigation in future studies.

* 5.) Given the proposed model, one would expect Puf3 to have reduced mRNA binding upon deletion of Pbp1. It would be interesting to examine Puf3 mRNA binding, perhaps through cross-linking immunoprecipitation (CLIP), to see if this indeed is the case. This would provide further direct evidence that Pbp1 is functioning through Puf3 and facilitating its function. Similarly, the authors mention that Pbp1 contains putative RNA binding domains, however, they make no mention if these domains may contribute to its function in mitochondrial protein expression. *

We performed an RNA-IP experiment to test whether Puf3 has reduced binding to its target mRNAs in the absence of Pbp1. In new Fig 7, Puf3 is still able to bind its mRNA targets in the absence of Pbp1. However, the association of Pbp1 with these mRNAs is reduced in puf3∆ knockouts. Such results are perhaps expected as Puf3 has been shown to bind in a sequence-specific manner to a ~8 nt motif in the 3’UTR of its target mRNAs. However, it is unclear whether the Lsm / LsmAD domains of Pbp1 actually bind RNAs directly (hence our use of the term “putative”) - they may be involved in protein-protein interactions. Moreover, Fig 5 shows that deletion of both domains has no apparent effect on mitochondrial protein expression. We prefer to address the role of the Lsm and LsmAD domains of Pbp1 in a future study.

* Reviewer #1 (Significance (Required)):

Overall, this manuscript provides a modest advance, but one that could prove to have important implications for the field-especially if the Pbp1 findings prove relevant to its human ortholog, ataxin-2. The advance is limited by the robustness of the specific molecular model proposed and the extent to which the Pbp1-Puf3 relationship is examined on the gene-expression level.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

In this manuscript, the authors found that the pbp1∆ mutant grew poorly on nonfermentable carbon source medium (YP Lactate medium) and that the pbp1∆ mutant had decreased amount of Cox2 protein, a cytochrome c oxidase. The pbp1∆ mutant also had decreased amounts of COX17, COX10, and MRP51 mRNAs. Since these mRNAs are target mRNAs of the RNA-binding protein Puf3, the authors next investigated the relationship between Pbp1 and Puf3. The analysis of the GFP reporter gene containing Puf3-binding sites in the 3' UTR showed that the levels of GFP reporter mRNA and protein were decreased in the pbp1∆ mutant strain. This reduction was dependent on the Puf3-binding site in the 3' UTR. Next, the authors examined the genetic interaction between the pbp1∆ and puf3∆ mutants, and found that the levels of Cox2 protein were reduced in the two mutants. Finally, the authors showed the interaction between Pbp1 and Puf3 by co-immunoprecipitation and determined the regions of Pbp1 and Puf3 required for the interaction. They also showed that these regions in Pbp1 and Puf3 proteins are also important for the regulation of Cox2 protein levels. The story is very clear and the data is reliable. However, the data from Western blotting should be shown quantitatively to make the results more reliable. Also, although the story is based on the reduction of Cox2 protein level, it would be better to discuss whether other proteins or mRNAs should be considered as well.

Major comments:

Figure 2C. Protein levels of GFP should be quantified and the data should be shown. *

These data have been quantified (now Fig 3C).

* Figure 3. Not only the Cox2 protein level but also mRNA levels of COX2, COX17, COX10, MRP51, etc in pbp1∆ mutant, puf3∆ mutant and pbp1∆ puf3∆ double mutant should be shown. Then the point of action of Pbp1 and Puf3 would become clearer. *

The mRNA levels have been determined by qPCR and are now in Fig S4.

Figure 4. * For the domain analysis of Pbp1 protein, showing differences in cell proliferation as in Figure 1B would indicate the physiological importance of the domain. *

Growth curves of the various Pbp1 domain deletions have been performed and shown in Fig 5D. They do support the physiological importance of the domain(s).

* Figure 4B. Quantification of the amount of co-immunoprecipitated proteins would indicate the strength of binding. *

These data have been quantified (now Fig 5B).

* Figure 4C. Protein levels should be quantified and the data presented. *

These data have been quantified (now Fig 5C).

* Line 153-8 The description of Line 153-8 is not appropriate for this position because it breaks up the flow of the story before and after. *

These text have been moved as requested.

* Minor comments:

Line 153 Isn't the following the first reference cited for Pbp1 is a negative regulator of TORC1? Transient sequestration of TORC1 into stress granules during heat stress Terunao Takahara 1, Tatsuya Maeda Mol Cell. 2012 Jul 27;47(2):242-52. doi: 10.1016/j.molcel.2012.05.019. Epub 2012 Jun 21.

Ref19. Ref 19 also shows that the pbp1∆ mutant strains grow poorly on the medium containing glycerol and lactate as carbon sources.

Overall, the gene is not italicized. *

These requested edits to the references and text have been made in the revised version of the manuscript.

* Reviewer #2 (Significance (Required)):

This manuscript analyzes the relationship between Pbp1 and Puf3 in yeast. Since these proteins are evolutionarily conserved from yeast to humans and are also associated with disease in humans, this reviewer believes this manuscript will be of interest to a wide audience. *

Reviewer #3 (Evidence, reproducibility and clarity (Required)): *

In this study, van de Poll et al. describe the involvement of a cytosolic RNA-binding protein (Pbp1) in the transcriptional regulation of mitochondrial biogenesis during the shift from fermentative to respiratory growth in budding yeast. Using the advantage of yeast genetics and molecular biology, they show that a number of mitochondrial transcripts and proteins are downregulated in pbp1Δ cells both under normal growth conditions and during respiratory shift. Since Puf3, another RNA-binding protein, is known to regulate the fate of these transcripts, they further investigate the interaction between Pbp1 and Puf3. Through a series of biochemical assays, they characterize the interaction between Pbp1 and Puf3 taking place through their low-complexity domains, which is suggested by authors to stabilize and promote the translation of Puf3-target transcripts.

As stated in the manuscript, Pbp1 is an evolutionarily conserved protein, encoded by ATXN2 gene in humans. Due to its involvement in multiple neurological disorders, it has been widely studied in mouse models and patient samples. The transcriptome profiles of Atxn2-KO mouse liver and cerebellum have been published (albeit having used a relatively older microarray technology for today's standards), revealing a prominent dysregulation of global translational machinery, and the ER-protein secretion pathway (Fittschen et al., 2015). These findings were followed by numerous studies showing dysregulations of distinct transcript pools under examination. Moreover, a PAR-CLIP study showed ATXN2 to associate with ~16.000 transcripts, 8000 of which depended on its interaction with PABPC1 (Yokoshi et al., 2014). In that study, ATXN2 was shown to preferentially bind to target transcripts on 3'-UTR AU-rich sequences. In addition to PABPC1, some other RNA-binding proteins, like TDP-43, were also shown to modulate the indirect interaction of ATXN2 with transcripts, while it also maintained direct interactions through its Lsm and LsmAD domains. Altogether, the mammalian data on ATXN2 thus far depicts it as an avid interactor of numerous RNA-binding proteins and countless transcripts, including microRNAs. The authors however have not cited or compared their findings with this vast array of mammalian literature (with the exception of two properly discussed papers in Discussion).

Considering its stress-responsive nature and association with RNP granules, it is plausible to assume that ATXN2/Pbp1 could regulate certain groups of transcripts in terms of their stability (in stress granules and p-bodies) or active translation under given environmental conditions and cellular state. The work of van de Poll et al. in this regard is an important step in expanding our knowledge about the many downstream effects of ATXN2/Pbp1. Yet, the following issues should be solved:

1) The pre-eminence of the proposed group of affected transcripts (i.e. those associated with mitochondrial biogenesis) has to be empirically established. Among all its interactions with other RNA-binding proteins, how important or dominant is the interaction of Pbp1 with Puf3 for mitochondrial biogenesis during the shift towards respiratory growth? Line 74 in the Results says "Analysis of a panel of nuclear- and mitochondrial-encoded mRNAs..."; how broad was this panel? how were the genes selected? Considering the fact that ATXN2/Pbp1 is associated with an immense number of transcripts, hand-picking a number of Puf3 targets and selectively analyzing their expression will surely give some significant dysregulations. Therefore, an unbiased transcriptomic survey is necessary to see all Pbp1-dependent dysregulations during respiratory shift. Since it's a process that requires heavy mitogenesis, one can assume that many dysregulations will concern mitochondrial factors. Indeed, proteome surveys in Atxn2-KO mouse tissues and pbp1Δ yeast (under fermentative growth and stress) point out to a strong mitochondrial dysregulation, so it raises hopes to see an even stronger Pbp1 impact during the respiratory shift in yeast. Then, bioinformatic analyses can reveal what proportion of those are Puf3 targets. If the authors' premise is valid, then the Puf3-targets will stand out in the transcriptome data, and give them an unbiased, solid and much stronger base for the following interaction analyses. They can then compare this dataset to the readily available mouse transcriptome from Atxn2-KO or polyQ-disease models, and strengthen their hypothesis about ATXN2/Pbp1 regulating mitochondrial biogenesis in association with Puf3.

The Results text about Figure 1 in its current state is an overstatement of the available data. Line 81-84 suggests mitochondrially encoded Cox2 levels are reduced because some mitoribosome subunits are Puf3 targets, but pbp1Δ itself is known to have altered mitochondrial membrane potential which negatively impacts mitochondrial import. So, the reduction of Cox2 levels (and potentially other mitochondrial-encoded proteins, it was never checked) may have nothing to do with Puf3, but rather be a direct consequence of reduced mitoribosome import in pbp1Δ. In order to make this statement, the total mitochondrial translation rates of WT and pbp1Δ strains would have to be compared, and if they are found the same, a selective effect on Puf3-target proteins has to be shown among many tested candidates. Same applies to line 86 "the specific decrease in Puf3-target mRNAs in pbp1Δ cells" referring to Figure 3C. This statement cannot be made without analyzing a larger group of transcripts including the targets of other RNA-binding proteins. The current data does not support any specific dysregulation of Puf3-targets, it just shows some Puf3 targets to be dysregulated, however without the knowledge of how many significant among all Puf3-targets, or how significant are Puf3-targets compared to others. An unbiased, high-throughput transcriptome data, and detailed bioinformatic analyses should replace Figure 1C. The high variation among replicates at 1h-3h-5h time points is also alarming, and puts the reproducibility of these experiments under question. *

As mentioned in the response to Reviewer 1, we performed an unbiased RNA-seq experiment comparing transcriptomes of WT vs pbp1∆ cells. The data are quite striking and strongly support our hypothesis (new Fig 2). To address the reviewer’s concern that pbp1∆ phenotypes may be due to altered mito membrane potential and mito protein import, we have performed an experiment to examine import of Cox4, which is a protein substrate that is commonly used for this purpose (note COX4 is not a Puf3-target mRNA). Steady-state Cox4 protein amounts are similar in WT vs pbp1∆ cells. Moreover, following treatment of cells with the uncoupler CCCP, there is more of the “pre”-processed Cox4 form in WT cells compared to pbp1∆ cells. These data would argue against the reviewer’s hypothesis and are included in the revised manuscript as Fig S1. Since every mitochondrial ribosomal subunit gene transcript is a target of Puf3 (PMID: 16254148) and therefore subject to regulation by Pbp1, we would argue that a defect in mito biogenesis (due to compromised translation of these mRNAs) precedes and may explain any subsequent defect in mito membrane potential. *

2) Stabilization of mRNAs The basal reduction in the mRNA levels of the reporter construct in pbp1Δ is a strong but not necessarily direct evidence of stabilization by Pbp1. mRNA half-life analyses (i.e. degradation curves) should be performed with desired targets to measure stability in WT and pbp1Δ strains. *

As mentioned above in the response to Reviewer 1, we performed mRNA half-life analyses for several transcripts in WT vs pbp1∆ strains using thiolutin treatment. The results are not straightforward to interpret as loss of Pbp1 led to a stabilization of Puf3-target mRNAs (relative to PGK1), however all Puf3-target mRNAs that we examined exhibited similar decay profiles. Thus, we deleted the term “stabilize” and further determination of the fate of these mRNAs in the absence of new transcription will require more careful and sophisticated experiments.*

3) Cox2 levels Regarding Line 125: "Cox2 protein levels, which are dependent on the translation of Puf3-target mRNAs": This may be generally true, but the data here suggests otherwise. pbp1Δ cells have completely diminished Cox2 levels, whereas puf3Δ have approx. 50% reduction. This means that Pbp1 is more important to maintain normal Cox2 levels, and does so independent of Puf3. In contrast, puf3Δ "rescues" some of the defect in pbp1Δ cells and increases Cox2 abundance to ~50%. As stated above, Cox2 reduction in pbp1Δ could be a direct consequence of mitochondrial membrane depolarization unrelated to Puf3, and could be accompanied by many other non-Puf3-targets being downregulated. Therefore, the authors should refrain from "Cox2 levels are dependent on the translation of Puf3-target mRNAs" statements throughout the text without an experimental proof in the context of pbp1Δ strain. *

See response to Reviewer 1. We believe that in the absence of Pbp1, Puf3 may now preferentially promote decay of various mito biogenesis transcripts, leading to apparently lower Cox2 levels. Moreover, per the results of our Cox4 experiment (Fig S1), we would respectfully disagree with the reviewer’s hypothesis. Nonetheless, we included an additional statement that mitochondrial membrane depolarization, as a consequence of reduced expression of numerous Puf3-targets, could also contribute to lower Cox2 abundance (p. 6 and 15). *

4) Pbp1-Puf3 interaction The authors state that Pbp1-Puf3 interaction is required for Puf3-target mRNA stabilization and translation. This suggests that Pbp1 stabilizes this pool of mRNAs because of its interaction with Puf3 primarily, not the mRNAs themselves. One general question while studying the interaction between two RNA-binding proteins is whether they interact in an RNA-dependent manner in vivo. The co-IP analyses show the interaction between Pbp1 and Puf3 to increase under respiratory shift as expected. However, in co-IPs from cell lysates, many RNA-binding proteins may seemingly interact due to their association with translation machinery at that given time. But this does not mean "direct" protein-protein interaction, just a co-existence around actively translating ribosomes. In order to ensure the direct interaction of these two proteins, the same co-IPs should be performed with/out RNase treatment of the lysate (many protocols available online). Only if Pbp1-Puf3 interaction persists in RNase+ samples, they can conclude a direct interaction. In addition, RNA-immunoprecipitation analyses should be performed in Pbp1-Flag and Pbp1-Flag/ puf3Δ strains to check if the association of Pbp1 with Puf3-target mRNAs indeed depends on Puf3. *

This is a good suggestion. We performed an experiment to test whether RNase treatment alters interaction between Pbp1 and Puf3 and there was minimal effect, supporting the hypothesis that the interaction may be direct. We also performed RNA-IP of Pbp1 in the presence of absence of Puf3 (new Fig 7). As the reviewer predicted, the RNA-IP enrichment of Puf3-target mRNAs was reduced in puf3∆ strains, suggesting that the association of Pbp1 with such transcripts depends on Puf3. It is known from work by others that Puf3 contains a PUF domain that enables sequence-specific binding to a motif in the 3’UTR of its target mRNAs, so these results are quite sensible.

* Minor comments: • Second sentence of the Abstract ("How mutations in its mammalian ortholog ataxin-2 are linked to neurodegenerative conditions remains unclear") is semantically incorrect. The term "linked" suggests an observed but uncharacterized effect of a genetic variation on a certain syndrome. Diseases can be linked to a chromosome or a locus without knowing the exact causative mutation. How the CAG repeat expansion mutations in ATXN2 are causative of SCA2, ALS or Parkinson-plus syndromes are very well known. One should also be careful with using "mutations" as a general term in the context of ATXN2, because there are certain variations in and around ATXN2 locus leading to a decrease in its activity and metabolic problems, which is far from its neurodegeneration-causing mutations. If the authors meant to state that the pathological mechanism is unclear by this sentence, that would also be a negligence of the extensive literature around this topic. Multiple disease models in mouse, Drosophila and C. elegans collectively point out to an RNA metabolism deficit, caused by toxic ATXN2 aggregates that sequestrate other RNA-binding proteins and their target transcripts. The specific downstream effects involve synaptic strength, Calcium-related action potentials, ER stress and cholesterol/sphingolipid synthesis. Therefore, this sentence could be rephrased to "PolyQ expansion mutations in its mammalian ortholog ataxin-2 lead to spinocerebellar dysfunction due to toxic protein aggregation." and simply avoid going into mechanistic details as it is not necessary for this manuscript.

• Lines 241-243: "Human sequencing" is also an incorrect term. Can be rephrased to "PolyQ expansion mutations in ATXN2 are associated with SCA2 and ALS". The references 22-24 are the first association of polyQ mutations with SCA2, however a reference for ALS is missing. Elden et al. Nature 2010 should be cited here.

• The authors should discuss the relevance of these findings to the mammalian ortholog of Pum3, namely PUM1/2. Afterall, it is also a very important conserved protein and well-studied in mammalian literature. *

These changes to the text and references have been made.

* Following a better characterization of the transcript pools selectively affected by Pbp1 (meaning a transcriptome survey), a graphical abstract sort of scheme could be useful in putting the findings in perspective and conveying the message.*

We decided not to include a graphical abstract at this time, since it is difficult for us to “picture” what is going on inside an Pbp1-containing RNP granule at this time. * Reviewer #3 (Significance (Required)):

The intricate experiments characterizing the nature of interaction between Pbp1 and Puf3 (Figures 3B, 4, 5, 6) are quite convincing. However, some fundamental questions remain especially regarding the primary rationale of studying Pbp1-Puf3 relationship and the breadth of some conclusions. The data are of general interest to a broad audience. The statistical tests in Figure 1 are of concern. The reviewer(s) has experience both with yeast molecular biology and with mammalian Atxn2 function. *

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this study, van de Poll et al. describe the involvement of a cytosolic RNA-binding protein (Pbp1) in the transcriptional regulation of mitochondrial biogenesis during the shift from fermentative to respiratory growth in budding yeast. Using the advantage of yeast genetics and molecular biology, they show that a number of mitochondrial transcripts and proteins are downregulated in pbp1Δ cells both under normal growth conditions and during respiratory shift. Since Puf3, another RNA-binding protein, is known to regulate the fate of these transcripts, they further investigate the interaction between Pbp1 and Puf3. Through a series of biochemical assays, they characterize the interaction between Pbp1 and Puf3 taking place through their low-complexity domains, which is suggested by authors to stabilize and promote the translation of Puf3-target transcripts.

As stated in the manuscript, Pbp1 is an evolutionarily conserved protein, encoded by ATXN2 gene in humans. Due to its involvement in multiple neurological disorders, it has been widely studied in mouse models and patient samples. The transcriptome profiles of Atxn2-KO mouse liver and cerebellum have been published (albeit having used a relatively older microarray technology for today's standards), revealing a prominent dysregulation of global translational machinery, and the ER-protein secretion pathway (Fittschen et al., 2015). These findings were followed by numerous studies showing dysregulations of distinct transcript pools under examination. Moreover, a PAR-CLIP study showed ATXN2 to associate with ~16.000 transcripts, 8000 of which depended on its interaction with PABPC1 (Yokoshi et al., 2014). In that study, ATXN2 was shown to preferentially bind to target transcripts on 3'-UTR AU-rich sequences. In addition to PABPC1, some other RNA-binding proteins, like TDP-43, were also shown to modulate the indirect interaction of ATXN2 with transcripts, while it also maintained direct interactions through its Lsm and LsmAD domains. Altogether, the mammalian data on ATXN2 thus far depicts it as an avid interactor of numerous RNA-binding proteins and countless transcripts, including microRNAs. The authors however have not cited or compared their findings with this vast array of mammalian literature (with the exception of two properly discussed papers in Discussion).

Considering its stress-responsive nature and association with RNP granules, it is plausible to assume that ATXN2/Pbp1 could regulate certain groups of transcripts in terms of their stability (in stress granules and p-bodies) or active translation under given environmental conditions and cellular state. The work of van de Poll et al. in this regard is an important step in expanding our knowledge about the many downstream effects of ATXN2/Pbp1. Yet, the following issues should be solved:

1. The pre-eminence of the proposed group of affected transcripts (i.e. those associated with mitochondrial biogenesis) has to be empirically established. Among all its interactions with other RNA-binding proteins, how important or dominant is the interaction of Pbp1 with Puf3 for mitochondrial biogenesis during the shift towards respiratory growth? Line 74 in the Results says "Analysis of a panel of nuclear- and mitochondrial-encoded mRNAs..."; how broad was this panel? how were the genes selected? Considering the fact that ATXN2/Pbp1 is associated with an immense number of transcripts, hand-picking a number of Puf3 targets and selectively analyzing their expression will surely give some significant dysregulations. Therefore, an unbiased transcriptomic survey is necessary to see all Pbp1-dependent dysregulations during respiratory shift. Since it's a process that requires heavy mitogenesis, one can assume that many dysregulations will concern mitochondrial factors. Indeed, proteome surveys in Atxn2-KO mouse tissues and pbp1Δ yeast (under fermentative growth and stress) point out to a strong mitochondrial dysregulation, so it raises hopes to see an even stronger Pbp1 impact during the respiratory shift in yeast. Then, bioinformatic analyses can reveal what proportion of those are Puf3 targets. If the authors' premise is valid, then the Puf3-targets will stand out in the transcriptome data, and give them an unbiased, solid and much stronger base for the following interaction analyses. They can then compare this dataset to the readily available mouse transcriptome from Atxn2-KO or polyQ-disease models, and strengthen their hypothesis about ATXN2/Pbp1 regulating mitochondrial biogenesis in association with Puf3.

The Results text about Figure 1 in its current state is an overstatement of the available data. Line 81-84 suggests mitochondrially encoded Cox2 levels are reduced because some mitoribosome subunits are Puf3 targets, but pbp1Δ itself is known to have altered mitochondrial membrane potential which negatively impacts mitochondrial import. So, the reduction of Cox2 levels (and potentially other mitochondrial-encoded proteins, it was never checked) may have nothing to do with Puf3, but rather be a direct consequence of reduced mitoribosome import in pbp1Δ. In order to make this statement, the total mitochondrial translation rates of WT and pbp1Δ strains would have to be compared, and if they are found the same, a selective effect on Puf3-target proteins has to be shown among many tested candidates. Same applies to line 86 "the specific decrease in Puf3-target mRNAs in pbp1Δ cells" referring to Figure 3C. This statement cannot be made without analyzing a larger group of transcripts including the targets of other RNA-binding proteins. The current data does not support any specific dysregulation of Puf3-targets, it just shows some Puf3 targets to be dysregulated, however without the knowledge of how many significant among all Puf3-targets, or how significant are Puf3-targets compared to others. An unbiased, high-throughput transcriptome data, and detailed bioinformatic analyses should replace Figure 1C. The high variation among replicates at 1h-3h-5h time points is also alarming, and puts the reproducibility of these experiments under question. 2. Stabilization of mRNAs The basal reduction in the mRNA levels of the reporter construct in pbp1Δ is a strong but not necessarily direct evidence of stabilization by Pbp1. mRNA half-life analyses (i.e. degradation curves) should be performed with desired targets to measure stability in WT and pbp1Δ strains. 3. Cox2 levels Regarding Line 125: "Cox2 protein levels, which are dependent on the translation of Puf3-target mRNAs": This may be generally true, but the data here suggests otherwise. pbp1Δ cells have completely diminished Cox2 levels, whereas puf3Δ have approx. 50% reduction. This means that Pbp1 is more important to maintain normal Cox2 levels, and does so independent of Puf3. In contrast, puf3Δ "rescues" some of the defect in pbp1Δ cells and increases Cox2 abundance to ~50%. As stated above, Cox2 reduction in pbp1Δ could be a direct consequence of mitochondrial membrane depolarization unrelated to Puf3, and could be accompanied by many other non-Puf3-targets being downregulated. Therefore, the authors should refrain from "Cox2 levels are dependent on the translation of Puf3-target mRNAs" statements throughout the text without an experimental proof in the context of pbp1Δ strain. 4. Pbp1-Puf3 interaction The authors state that Pbp1-Puf3 interaction is required for Puf3-target mRNA stabilization and translation. This suggests that Pbp1 stabilizes this pool of mRNAs because of its interaction with Puf3 primarily, not the mRNAs themselves. One general question while studying the interaction between two RNA-binding proteins is whether they interact in an RNA-dependent manner in vivo. The co-IP analyses show the interaction between Pbp1 and Puf3 to increase under respiratory shift as expected. However, in co-IPs from cell lysates, many RNA-binding proteins may seemingly interact due to their association with translation machinery at that given time. But this does not mean "direct" protein-protein interaction, just a co-existence around actively translating ribosomes. In order to ensure the direct interaction of these two proteins, the same co-IPs should be performed with/out RNase treatment of the lysate (many protocols available online). Only if Pbp1-Puf3 interaction persists in RNase+ samples, they can conclude a direct interaction. In addition, RNA-immunoprecipitation analyses should be performed in Pbp1-Flag and Pbp1-Flag/ puf3Δ strains to check if the association of Pbp1 with Puf3-target mRNAs indeed depends on Puf3.

Minor comments:

• Second sentence of the Abstract ("How mutations in its mammalian ortholog ataxin-2 are linked to neurodegenerative conditions remains unclear") is semantically incorrect. The term "linked" suggests an observed but uncharacterized effect of a genetic variation on a certain syndrome. Diseases can be linked to a chromosome or a locus without knowing the exact causative mutation. How the CAG repeat expansion mutations in ATXN2 are causative of SCA2, ALS or Parkinson-plus syndromes are very well known. One should also be careful with using "mutations" as a general term in the context of ATXN2, because there are certain variations in and around ATXN2 locus leading to a decrease in its activity and metabolic problems, which is far from its neurodegeneration-causing mutations. If the authors meant to state that the pathological mechanism is unclear by this sentence, that would also be a negligence of the extensive literature around this topic. Multiple disease models in mouse, Drosophila and C. elegans collectively point out to an RNA metabolism deficit, caused by toxic ATXN2 aggregates that sequestrate other RNA-binding proteins and their target transcripts. The specific downstream effects involve synaptic strength, Calcium-related action potentials, ER stress and cholesterol/sphingolipid synthesis. Therefore, this sentence could be rephrased to "PolyQ expansion mutations in its mammalian ortholog ataxin-2 lead to spinocerebellar dysfunction due to toxic protein aggregation." and simply avoid going into mechanistic details as it is not necessary for this manuscript.
• Lines 241-243: "Human sequencing" is also an incorrect term. Can be rephrased to "PolyQ expansion mutations in ATXN2 are associated with SCA2 and ALS". The references 22-24 are the first association of polyQ mutations with SCA2, however a reference for ALS is missing. Elden et al. Nature 2010 should be cited here.
• The authors should discuss the relevance of these findings to the mammalian ortholog of Pum3, namely PUM1/2. Afterall, it is also a very important conserved protein and well-studied in mammalian literature.
• Following a better characterization of the transcript pools selectively affected by Pbp1 (meaning a transcriptome survey), a graphical abstract sort of scheme could be useful in putting the findings in perspective and conveying the message.

Significance

The intricate experiments characterizing the nature of interaction between Pbp1 and Puf3 (Figures 3B, 4, 5, 6) are quite convincing. However, some fundamental questions remain especially regarding the primary rationale of studying Pbp1-Puf3 relationship and the breadth of some conclusions. The data are of general interest to a broad audience.<br /> The statistical tests in Figure 1 are of concern. The reviewer(s) has experience both with yeast molecular biology and with mammalian Atxn2 function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, the authors found that the pbp1∆ mutant grew poorly on nonfermentable carbon source medium (YP Lactate medium) and that the pbp1∆ mutant had decreased amount of Cox2 protein, a cytochrome c oxidase. The pbp1∆ mutant also had decreased amounts of COX17, COX10, and MRP51 mRNAs. Since these mRNAs are target mRNAs of the RNA-binding protein Puf3, the authors next investigated the relationship between Pbp1 and Puf3. The analysis of the GFP reporter gene containing Puf3-binding sites in the 3' UTR showed that the levels of GFP reporter mRNA and protein were decreased in the pbp1∆ mutant strain. This reduction was dependent on the Puf3-binding site in the 3' UTR. Next, the authors examined the genetic interaction between the pbp1∆ and puf3∆ mutants, and found that the levels of Cox2 protein were reduced in the two mutants. Finally, the authors showed the interaction between Pbp1 and Puf3 by co-immunoprecipitation and determined the regions of Pbp1 and Puf3 required for the interaction. They also showed that these regions in Pbp1 and Puf3 proteins are also important for the regulation of Cox2 protein levels.

The story is very clear and the data is reliable. However, the data from Western blotting should be shown quantitatively to make the results more reliable. Also, although the story is based on the reduction of Cox2 protein level, it would be better to discuss whether other proteins or mRNAs should be considered as well.

Major comments:

Figure 2C.

Protein levels of GFP should be quantified and the data should be shown.

Figure 3.

Not only the Cox2 protein level but also mRNA levels of COX2, COX17, COX10, MRP51, etc in pbp1∆ mutant, puf3∆ mutant and pbp1∆ puf3∆ double mutant should be shown. Then the point of action of Pbp1 and Puf3 would become clearer.

Figure 4.

For the domain analysis of Pbp1 protein, showing differences in cell proliferation as in Figure 1B would indicate the physiological importance of the domain.

Figure 4B.

Quantification of the amount of co-immunoprecipitated proteins would indicate the strength of binding.

Figure 4C.

Protein levels should be quantified and the data presented.

Line 153-8

The description of Line 153-8 is not appropriate for this position because it breaks up the flow of the story before and after.

Minor comments:

Line 153

Isn't the following the first reference cited for Pbp1 is a negative regulator of TORC1? Transient sequestration of TORC1 into stress granules during heat stress Terunao Takahara 1, Tatsuya Maeda Mol Cell. 2012 Jul 27;47(2):242-52. doi: 10.1016/j.molcel.2012.05.019. Epub 2012 Jun 21.

Ref19.

Ref 19 also shows that the pbp1∆ mutant strains grow poorly on the medium containing glycerol and lactate as carbon sources.

Overall, the gene is not italicized.

Significance

This manuscript analyzes the relationship between Pbp1 and Puf3 in yeast. Since these proteins are evolutionarily conserved from yeast to humans and are also associated with disease in humans, this reviewer believes this manuscript will be of interest to a wide audience.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript, van de Poll et al. aim to further establish the function of poly(A) binding protein-binding protein 1 (Pbp1) and its relationship to the RNA-binding protein Puf3 in regulating the expression of mitochondrial proteins. This work builds upon a solid body of previous studies from this group regarding the function of Puf3 and the role of Pbp1 in regulating TORC1 signaling. Here, the authors show that Pbp1 has a physical and functional interaction with Puf3 and that ablation or disruption of Pbp1 eliminates this interaction and reduces the expression of some Puf3 target proteins. This work provides reliable data supporting a role for Pbp1 in the regulation of mitochondrial protein expression and that there is a clear functional interaction between Puf3 and Pbp1. Nonetheless, there are several issues that should be addressed before the manuscript is suitable for publication.

1. The model put forward suggests that Pbp1 works to recruit Puf3 to the vicinity of mitochondria, where Puf3 can then promote the expression of its target mRNAs. In Figure 3A, however, deletion of Pbp1 has a stronger affect on the expression of COX2 than deletion of Puf3 alone, which would not be expected if the role of Pbp1 is to modulate Puf3 function. Similarly, the expression of COX2 is higher in the double deletion versus the Pbp1 single deletion. The authors should attempt to clarify this experimentally, or at least make mention of alternative mechanisms for Pbp1 which may be causing this.
2. The authors mention that the increased pull-down of Puf3 with Pbp1 in respiratory conditions suggests that the Pbp1-Puf3 interaction is responsive to the cellular metabolic state (Figure 3B). However, the increase in Puf3 expression makes it difficult to compare the interactions between the two conditions.
3. The authors only look at a very small subset of Puf3 target mRNAs using qPCR when it would be much more informative and overall convincing to examine a larger amount using RNA-seq experiments.
4. The authors consistently mention that Pbp1 function is helping to stabilize Puf3 target mRNAs. However, if the authors wish to prove this particular mode-of-action, more direct evidence should be provided, such as a pulse-chase experiment. Otherwise, other models allowing for increased mRNA abundance should be noted.
5. Given the proposed model, one would expect Puf3 to have reduced mRNA binding upon deletion of Pbp1. It would be interesting to examine Puf3 mRNA binding, perhaps through cross-linking immunoprecipitation (CLIP), to see if this indeed is the case. This would provide further direct evidence that Pbp1 is functioning through Puf3 and facilitating its function. Similarly, the authors mention that Pbp1 contains putative RNA binding domains, however, they make no mention if these domains may contribute to its function in mitochondrial protein expression.

Significance

Overall, this manuscript provides a modest advance, but one that could prove to have important implications for the field-especially if the Pbp1 findings prove relevant to its human ortholog, ataxin-2. The advance is limited by the robustness of the specific molecular model proposed and the extent to which the Pbp1-Puf3 relationship is examined on the gene-expression level.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

'The authors do not wish to provide a response at this time.'

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

The manuscript by Naeli et al., presents study on effects of SARS-CoV-2 NSP2 protein on miRNA mediated translational repression. Authors show in multiple ways using reporter mRNAs with various miRNA sites that NSP2 protein stimulates miRNA mediated repression. The increase in miRNA repression is likely due to the interaction of NSP2 with either GIGYF2 or Argonaute protein directly thus making more stable repressive complex on mRNA. The manuscript is written clearly and methods provide enough details for reproducibility. Only major comment would be that authors could have tested multiple endogenous targets of miRNAs for extent of miRNA mediated expression in the presence of NSP2. This could be achieved using either western blots for known targets (looking at protein levels) or using targeted qRT-PCR or distribution of mRNAs in polysome fractions (mRNA translational repression. An alternative would be Ribo-Seq experiment.

Minor comment:

Line in introduction arguing: "The GIGYF2/4EHP complex is recruited by a variety of factors including miRNAs" should state miRISC instead of miRNAs.

Significance

General assessment: The study presents sold evidence that NSP2 protein interacts with miRISC complex and increases miRNA-mediated translational repression of reporter mRNAs. Multiple target sites for miR20, let7 and miR92 are tested as well as two different human cell lines (Hek293 and U87) which gives strength to study and reproducibility of the results. Focus on the endogenous miRNA targets in the presence or absence of the NSP2 protein would make study even stronger.

The advance of the study is more rigorous analyses of NSP2 protein effects on miRNA-mediated gene expression regulation with some novel mechanistic insights. The study will be of interest for specialized and basic research audience with potential impact on translational research.

My field of expertise covers mechanisms of gene expression regulation by miRNAs and RBPs as well as impact of mRNAs, nascent peptides and ribosomes on protein synthesis.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this study, Naeli and collaborators propose a role for the SARS-CoV2 protein NSP2 in the miRNA-mediated translational repression. Their data show that NSP2 co-immunoprecipitate with the Argonaute AGO2 and the previously reported translation modulator GIGYF2. Using different reporters sensitive to miRNA repression, they show that overexpressing NSP2 enhances miRNA-mediated translational repression in different cell lines as well as one endogenous miRNA target CDK1 and thus without affecting the stability of targeted mRNAs. Interestingly, their data show that the effect of NSP2 depends on the number of miRNA binding sites suggesting that highly repressed mRNAs, due to the presence of several miRNA binding sites, are not affected by NSP2. Finally, using an AGO2 tethering reporter assay that does not rely on miRNA binding but needs two NSP2 interactors GIGYF2 and 4EHP, they demonstrate that overexpressing NSP2 still stimulates mRNA repression, suggesting that NSP2 can have a broader impact on miRISC-dependent silencing.

Overall, this is an interesting follow-up study from their previous paper (Xu et al., 2022) that have the potential to add another dimension to the modulation of mRNA translation by the SARS-CoV2 protein NSP2. But unfortunately, in its current form, the work falls short of supporting their claims appropriately and providing sufficient and relevant insights about the role of NSP2 in regulating the miRNA-mediate gene regulation. To overcome this, the authors should perform and add the following experiments to strengthen their study.

1. Interaction of NSP2 with the miRISC: With the data presented here, we can only conclude that NPS2 forms a complex with AGO2. Is their interaction direct or indirect? Is miRNA- and TNRC6-bound AGO2 (and thus miRISC) associated with NSP2? To address those important questions, the authors should perform NSP2 immunoprecipitations in GIGYF2 (and 4EHP) KO cell lines and monitor the presence of AGO2 and miRNAs in the immunoprecipitated complex. Those experiments will define the type of interaction NSP2 has with the miRISC (GIGYF2 dependent or not).
2. Along this line, is NSP2 action on miRNA-mediated gene regulation dependent or not of GIGYF2? Again, testing this by monitoring the repression of the different reporters upon NSP2 overexpression in the presence or not of GIGYF2 in cells will precise the contribution of NSP2 in miRNA-mediated gene silencing.

Besides these two sets of critical experiments that will define the relationship between NSP2 mRNA modulation and the microRNA pathway, it would be interesting for this study to readily test the effect of NSP2 on miRNA targets related to immune-regulatory processes and antiviral response. As the authors pointed out in their discussion, several mRNAs are involved in the control of genes found in those pathways. Demonstrating the contribution of NSP2 in the regulation of a few of them will strengthen their study's significance and interest by providing a possible mechanism at play during a SARS-CoV2 infection. Also, it would be interesting to test if the modulation of translation inhibition by NSP2 occurs in cells infected by the SARS-CoV2 virus. For instance, is NSP2 and miRISC interaction enhances during the viral infection? As the authors propose, NSP2 could also have improved antiviral microRNA repression, so it would be interesting to characterize this interplay in a relevant biological context.

Minor comment:

From the data presented in Figure S1A, it is impossible to conclude "that miR-20a represses the expression of the target mRNA in a GIGFY2-dependent manner" as its KO also affect the level/stability of 4EHP. Therefore, we cannot distinguish the contribution of GIGFY2 and 4EHP in this context as both protein levels decrease. The authors should tone down this statement on page 3.

Significance

With the addition of the proposed experiments, the revised study will better define the direct contribution of NSP2 with the miRISC. This work has the potential to provide another aspect of the mRNA translation modulation by the SARS-CoV2 protein NSP2 with the interesting angle of miRNA-mediated gene silencing.

As mentioned by the authors, another recent paper reports the potential impact of NSP2 on post-transcriptional silencing (Zou et al., iScience 2022). However, in contrast to the current study, this previous work did not directly test the interaction of NSP2 with the miRISC. Furthermore, it only used a single miRNA reporter (let-7) to support NSP2 contribution in miRNA-mediated gene silencing, which does not demonstrate the broader impact of NSP2 on this gene regulatory mechanism as tested in this current study. Upon revision, this study will provide more definitive proof of the involvement of NSP2 in miRNA-mediated gene regulation and thus will be of interest to experts in the miRNA field and scientists interested in understanding viruses/hosts interaction.

Although not essential, if the authors want to add data that addresses the function of NSP2 on this regulatory pathway in the context of viral infection (which seems feasible for this group), that will definitely increase the broader significance of their work.

I am an expert in molecular biology and molecular genetics, miRNA-mediated gene regulation, and small non-coding RNA biology.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

In this manuscript the authors examine whether SARS-CoV-2 protein, NSP2 mediate translation repression of cellular transcripts by increasing miRNA mediated suppression. The same authors previously demonstrated NSP2 interacts with GIGYF2 and this interaction suppresses the translation of the Ifnb mRNA. Here they extend this finding and using a series of reporters they illustrate that at least part of NSP2 translation suppression is mediated by increasing GIGYF2 mediated miRNA translation suppression. Overall the manuscript is clearly written and the experiments well executed.

Major comments:

Optional- The authors show that the ability of NSP2 to enhance miRNA-mediated repression is dependent on the extent of the initial repression and suggest this dependency on the initial repression levels may explain the discrepancy with published work showing NSP2 actually impairs (and not increase) microRNA-mediated silencing . Therefore, an important addition could be to examine what happen to translation in cells that express NSP2 and whether generally translation repression is significant in transcripts (native) that are enriched in miRNA binding site. This experiment will help to show NSP2 effect on native transcripts and potentially help to understand how much of these changes are likely explained by miRNAs.

Minor comments:

1. Personally I found the term repression fold quite confusing and unintuitive. Why not to present the actual measurements so that the control (blue bars) have high value and the red bars (representing repression) have lower values?
2. It seems inadequate to present on the same graph two values that are normalized to 1. For example, in figure 1D it will be important to show the mir20-Mut real values (at least mir20-Mut in NSP2 cells should not be normalized to 1). This will allow to show that the differences are indeed mediated by stronger translation repression of miR-20 WT luciferase in the presence of NSP2 and not by unexplained differences in mir20-mut luciferase expression. This is true for almost all the figures in the manuscript. Correspondingly, the statistical test should be two-factor ANOVA test examining if NSP2 expression significantly increase the difference between Luciferase miR20-WT and luciferase miR20-mut.

Significance

There is an urgent need for better molecular understanding of how SARS-CoV-2 proteins influence the machineries of the host cell.

This study investigates how NSP2 interaction with GIGYF2 mediate translation repression of cellular transcript. The authors also address the discrepancy with previously published work that showed NSP2 actually impairs (and not increase) microRNA-mediated silencing.

The paper would be of interest to RNA biologists and for molecular virologists that study SARS-CoV-2

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

I summarise the major findings of the work below. In my opinion the range and application of approaches has provided a broad evidence base that, in general, supports the authors conclusions. However, there are, in my opinion, particular failures to utilise and communicate this evidence. The manuscript may be much improved with attention in the following areas. In each case I will give general criticism with a few examples, but the principals of my comments could be applied throughout the work.

1) Insufficient quantification. The investigation combines various sources of qualitative data (EM, fluorescence microscopy, western blotting) to generate a reasonably strong evidence base. However, the work is over-reliant on representative images and should include more quantification from repeat experiments. When there are multiple fluorescence micrographs with intensity changes (not necessarily just representative images) (e.g. Figure 1 or 2) the authors should consider making measurements of these. Also the VLP production assays, which are assessed by western blotting would particularly benefit from a quantitative assessment (either by densitometry or, if samples remain, ELISA/similar approach).

We have performed quantification of immunofluoresence, western blotting and VLP experiments from existing data. These quantification are presented in our revised manuscript. An overview of new quantification is shown below:

Data shown

Quantification now shown in

Method

Analysis

Figure 1A

Supp F1C

IF

HAE (-/+ SARS-CoV-2)

• Tetherin total fluorescence intensity

Figure 1D

Supp F1E

IF

HeLa+ACE2 (-/+ SARS-CoV-2 )

• Tetherin total fluorescence intensity

Figure 2C

Supp F2B

IF

A549+ACE2 (-/+ SARS-CoV-2)

• Tetherin total fluorescence intensity

Figure 2G

Supp F2D

IF

T84 (-/+ SARS-CoV-2)

• Tetherin total fluorescence intensity

Supp F4A

Supp F4B

IF

HeLa + ss-HA-Spike transients (-/+ HA stained cells) - Tetherin total fluorescence intensity

Figure 4D

Supp F4E

IF

HeLa + TetOne ss-HA-Spike stables (-/+ Dox)

• Tetherin total fluorescence intensity

Figure 4F

Supp F4G

W blot

HeLa + TetOne ss-HA-Spike stables (-/+ Dox)

– Tetherin abundance

Figure 4G

Supp F4I

W blot – lysates

Spike VLP experiments

– tetherin abundance

Figure 4G

Supp F4J

W blot - VLPs

Spike VLP experiments

• N-FLAG abundance

Figure 6A

Supp F7A

W blot – lysates

ORF3a VLP experiments

– tetherin abundance

Figure 6A

Supp F7B

W blot - VLPs

ORF3a VLP experiments

• N-FLAG

For immunofluoresence anaysis, the mean, standard deviation, number of cells analysed and number of independent experiments are shown in the updated figure legends. Statistical analysis is also detailed in figure legends. Methods for the quantificaiton of fluoresence intensity is included in the Methods section.

Densitometry was performed on western blots and VLP experiments as suggested. The mean, standard devisation and number of independent expreiments analysed are expressed in figure legends. Methods for densityometry quantification is now included.

2) Insufficient explanation. I found some of the images and legends contained insufficient annotation and/or description for a non-expert reader to appreciate the result(s). Particularly if the authors want to draw attention to features in micrographs they should consider using more enlarged/inset images and annotations (e.g. arrows) to point out structures (e.g DMVs etc.). This short coming exacerbates the lack of quantification.

Additional detail has been provided to the figure legends, and we have updated several figures to draw attention to features in micrographs. Black arrowheads have been added to Figures 1E, 2D, 2H to highlight plasma membrane-associated virions, and asterisks to highlight DMVs in Figures 1E, 2D and Supplemental Figures 2C, 2E. Similarly, typical Golgi cisternae are highlighted by white arrowheads micrographs in Figure 2E. These figure legends have also been modified to highlight these additions.

3) Insufficient exploration of the data. I had a sense that some aspects of the data seem unconsidered or ignored, and the discussion lacks depth and reflection. For example the tetherin down-regulation apparent in Figures 1 and 2 is not really explained by the spike/ORF3a antagonism described later on, but this is not explicitly addressed.

We have made changes throughout the manuscript, but the discussion especially has been modified. We now discuss the ORF3a data in more depth, discuss possible mechanisms by which ORF3a alone enhances VLP release, and discuss our ORF7a data in context to previous reports.

The discussion has been updated to now include a better description of our data, and additional writing putting our work in to context with previously published work. See discussion section of revised manuscript.

Also, Figure 6 suggests that ORF3a results in high levels of incorporation of tetherin in to VLPs, but I don't think this is even described(?). The discussion should also include more comparison with previous studies on the relationship between SARS-2 and tetherin.

We have added a section to discuss how ORF3a may enhance VLP release,

‘We found that the expression of ORF3a enhanced VLP independently of its ability to relocalise tetherin (Figure 6A). This may be due to either the ability of ORF3a to induce Golgi fragmentation [38] which facilitates viral trafficking [39], or due to enhanced lysosomal exocytosis [37]. Tetherin was also found in VLPs upon co-expression with ORF3a (Figure 6A) which may also indicate to enhanced release via lysosomal exocytosis [37].

The secretion of lysosomal hydrolases has been reported upon expression of ORF3a [31] and whilst this may in-part be due to enhanced lysosome-plasma membrane fusion, our data highlights that ORF3a impairs the retrograde trafficking of CIMPR (Supplemental Figures 6B, 6F, 6G), which may similarly increase hydrolase secretion.’ – (Line 625-654).

The discussion has been developed to compare the relationship between SARS-CoV-2 and tetherin in previous studies,

‘SARS-CoV-1 ORF7a is reported to inhibit tetherin glycosylation and localise to the plasma membrane in the presence of tetherin [18]. We did not observe any difference in total tetherin levels, tetherin glycosylation, ability to form dimers, or surface tetherin upon expression of either SARS-CoV-1 or SARS-CoV-2 ORF7a (Figures 4A, 4B, 4C).

Others groups have demonstrated a role for ORF7a in sarbecovirus infection and both SARS-CoV-1 and SARS-CoV-2 virus lacking ORF7a show impaired virus replication in the presence of tetherin [18,41]. A direct interaction between SARS-CoV-1 ORF7a and SARS-CoV-2 ORF7a and tetherin have been described [18,41], although the precise mechanism(s) by which ORF7a antagonises tetherin remains enigmatic. We cannot exclude that ORF7a requires other viral proteins to antagonise tetherin, or that ORF7a antagonises tetherin via another mechanism. For example, ORF7a can potently antagonise IFN signalling [42] which would impair tetherin induction in many cell types.’ – (Line 667-704).

I have no minor comments on this draft of the manuscript.

Reviewer #1 (Significance (Required)):

Tetherin, encoded by the BST2 gene, is an antiviral restriction factor that inhibits the release of enveloped viruses by creating tethers between viral and host membranes. It also has a capacity for sensing and signalling viral infection. It is most widely understood in the context of HIV-1, however, there is evidence of restriction in a wide variety of enveloped viruses, many of which have evolved strategies for antagonising tetherin. This knowledge informs on viral interactions with the innate immune system, with implications for basic virology and translational research.

This study investigates tetherin in the context of SARS-CoV-2. The authors use a powerful collection of tools (live virus, gene knock out cells, recombinant viral and host expression systems) and a variety of approaches (microscopy, western blotting, infection assays), which is, itself, a strength. The study provides evidence to support a series of conclusions: I) BST2/tetherin restricts SARS-CoV-2 II) SARS-CoV-2 ablates tetherin expression III) spike protein can modestly down-regulate tetherin IV) ORF3A dysregulates tetherin localisation by altering retrograde trafficking. These conclusions are broadly supported by the data and this study make significant contributions to our understanding of SARS-CoV-2/tetherin interactions.

My enthusiasm is reduced by, in my opinion, a failure of the authors to fully quantify, explain and explore their data. I expect the manuscript could be significantly improved without further experimentation by strengthening these aspects.

This manuscript will be of interest to investigators in virology and/or cellular intrinsic immunity. Given the focus on SARS-CoV-2 it is possible/likely that it will find a slightly broader readership.

I have highly appropriate skills for evaluating this work being experienced in virology, SARS-CoV-2, cell biology and microscopy.

We wish to thank Reviewer #1 for their comments which have helped us to improve the quality of our revised manuscript.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

BST2/tetherin can restrict the release and transmission of many enveloped viruses, including coronaviruses. In many cases, restricted viruses have developed mechanisms to abrogate tetherin-restriction by expressing proteins that antagonize tetherin; HIV-1 Vpu-mediated antagonism of tetherin restriction is a particularly well studied example. In this paper, Stewart et al. report their studies of the mechanism(s) underlying SARS-CoV-2 antagonism of tetherin restriction. They conclude that Orf3a is the primary virally encoded protein involved and that Orf3a manipulates endo-lysosomal trafficking to decrease tetherin cycling and divert the protein away from putative assembly sites.

Major comments:- In my view some of the claims made by the authors are not fully supported by the data. For example, the bystander effect discussed in line 162 may suggest that infected cells can produce IFN but does not 'indicate' that they do

This text has now been edited,

‘The levels of tetherin in uninfected HAE cells is lower than observed in uninfected neighbours in infected wells demonstrating that infected HAE cells are able to generate IFN to act upon uninfected neighbouring cells, enhancing tetherin expression.’ - (Lines 163-172).

Most of the EM images show part of a cell profile, so statements such as (line 192) 'virus containing tubulovesicular organelles were often polarised towards sites of significant surface-associated virus' should be backed up with appropriate images, or indicated as 'not shown', or removed (the observation is not so important for this story). Line 196, DMVs can't be seen in these micrographs.

The statement 'virus containing tubulovesicular organelles were often polarised towards sites of significant surface-associated virus' has been removed. The micrographs in Figure 1E have been re-cropped, and image iii replaced with an image showing DMVs and budding virions. Plasma membrane-associated virions are highlighted by black arrowheads, DMVs by black asterisks, and intracellular virion by a white arrow.

Line 391, I can't see much change in CD63 distribution.

CD63 reproducibly appears clustered towards the nuclei in ORF3a expressing cells, whilst CD63 positive puncta are abundant in the periphery of mock cells. CD63 puncta are also larger, and the staining of CIMPR and VPS35 also appears to be associated with larger organelles. We have amended the text to now read,

‘Expression of ORF3a also disrupted the distribution of numerous endosome-related markers including CIMPR, VPS35, CD63, which all localised to larger and less peripheral puncta (Supplemental Figure 6B), and the mixing of early and late endosomal markers’ - (Line 469).

Quantification of the diameter of CD63 puncta indicate that they are larger in ORF3a expressing cells than in mock cells. Mock cells - 0.71μm (SD; 0.19), ORF3a - 1.15μm (SD;0.35). At least 75 organelles per sample, from 10 different cells. We have not included this data as we do not wish to labor this point but are happy to include this quantification if required to do so.

Line 321, the authors show that ORF7a does not affect tetherin localization, abundance, glycosylation or dimer formation, but they don't show that it doesn't restrict SARS-CoV-2. Can they be sure that epitope tagging this molecule does not abrogate function (or the functions of any of the other tagged proteins for that matter), or that ORF7a works in conjunction with one of the other viral proteins?

We are careful in the manuscript not to claim that ORF7a has no effect on tetherin. Our data indicate that ‘ORF7a does not directly influence tetherin localisation, abundance, glycosylation or dimer formation’ - (Line 361-362).

We were unable to reproduce an effect of ORF7a on tetherin glycosylation. Our data conflicts with that presented by Taylor et al, 2015, where ORF7a impaired tetherin glycosylation and ORF7a localised to the plasma membrane in tetherin expressing cells. The experiments performed by Taylor et al used HEK293 cells and ectopically expressed tagged tetherin. The differences in results may be attributed to the differences between cell lines or due to differences between endogenous or ectopic / tagged tetherin.

The study by Taylor et al uses SARS-CoV-1 ORF7a-HA from Kopecky-Bromberg et al., 2007 (DOI: 1128/JVI.01782-06), where the -HA tag is positioned at the C-terminus. Our ORF7a-FLAG constructs have a C-terminal epitope tag. While we cannot exclude the possibility that tagged proteins may act differently from untagged ones, the differences between our findings and previous work appear unlikely to be due to epitope tags.

Our manuscript states that although we cannot find any effect of ORF7a on tetherin localisation, abundance, glycosylation, or dimer formation, we cannot exclude that ORF7a impacts tetherin by another mechanism. For example, ORF7a has been found to antagonise interferon responses. Tetherin is abundantly expressed in HeLa cells and expression does not require induction through interferon. None of our experiments above would be impacted by interferon antagonism yet this could impact other cell types besides infection in vivo. These possibilities may explain the reported differential impact of ORF7a by different labs. An addition comment has been added to the discussion to reflect this,

’We cannot exclude that ORF7a requires other viral proteins to antagonise tetherin, or that ORF7a antagonises tetherin via another mechanism. For example, ORF7a potently antagonises IFN signalling [38], which would impair tetherin induction in many cell types. - (Line 701-704).

Note - Reference 38 has been added to the manuscript – Xia et al., Cell Reports DOI: 10.1016/j.celrep.2020.108234

In the ORF screen, a number of the constructs are expressed at low level, is it possible they [the authors] are missing something?

Some of the ORFs expressed in the miniscreen appear poorly expressed. We accept that in the use of epitope tagged constructs expression levels of individual viral proteins may impact upon a successful screen. However, this screen was performed to identify any potential changes in tetherin abundance or localisation, and the screen did successfully identify ORF3a, which we were able to follow-up and verify.

Line 376, the authors refer to ORF3a being a viroporin. A recent eLife paper (doi: 10.7554/eLife.84477; initially published in BioRxiv) refutes this claim and builds on other evidence that ORF3a interacts with the HOPS complex. The authors should at least mention this work, especially in the discussion, as it would seem to provide a molecular mechanism to support their conclusions.

This paper had not been peer reviewed at the time of our initial submission. We have now included the following text,

‘SARS-CoV-2 ORF3a is an accessory protein that localises to and perturbs endosomes and lysosomes [29]. It may do so by acting either as a viroporin [30] or by interacting with, and possibly interfering with the function of VPS 39, a component of the HOPS complex which facilitates tethering of late endosomes or autophagosomes with lysosomes [29,31]. Given ORF3a likely impairs lysosome function, the observed increased….’ - (Lines 444-449).

Fig 3, the growth curves illustrated in Fig3 C and D do not have errors bars; how many times were these experiments repeated?

These experiments require more repeats to include error bars. Infection and plaque assay (Figure 3C, 3D) are currently ongoing and we plan to complete them in the next 6-8 weeks and include them in the finalised manuscript.

In the new experiments, infections will additionally be performed at MOI 0.01, in addition to the previous MOIs (1 and 5).

Line 396, the authors show increased co-localization with LAMP1. As LAMP1 is found in late endosomes as well as lysosomes, they cannot claim the redistributed tetherin is specifically in lysosomes.

We have altered the text to now say:

‘The ORF3a-mediated increase in tetherin abundance within endolysosomes could be due to defective lysosomal degradation.’ - (Line 475).

There seems to be a marked difference in the anti-rb555 signal in the 'mock' cells in panels 5H and Suppl 6E. Is there a good reason for this, or does this indicate variability between experiments?

Antibody uptake experiments in Figure 5H and Supp Figure 6E were performed and acquired on different days. Relatively low levels of signal are available in these antibody uptake experiments, and the disperse labelling seen in the mocks does not aid this.

Fig 6a, why is there negligible VLP release from cells lacking BST2 and ORF3a-strep? How many times were these experiments performed? Is this a representative image? I think it confusing to refer to the same protein by two different names in the same figure (i.e. BST2 and tetherin). Do the authors know how the levels of ORF3a expressed in cells in these experiments compares to those seen in infected cells?

We have changed the blot in Figure 6A for one with clearer FLAG bands. Three independent experiments were performed for Figure 6A. Quantification of VLPs is now included in Supplemental Figure 7B.

We have changed ‘Bst2’ to ‘tetherin’ in all previous figures relating to protein; Figure 4G, Figure 6A, B, C.

We have no current information to compare ORF3a levels in these experiments versus in infected cells. We can investigate quantifying this if necessary.

My final point is, perhaps, the trickiest to answer, but nevertheless needs to be considered. As far as we know, SARS-CoV-2 and at least some other coronaviruses, bud into organelles of the early secretory pathway, often considered to be ERGIC. In the experiments shown here the authors provide evidence that ORF3a can influence tetherin recycling, but the main way of showing this is through its increased association with endocytic organelles. Do the authors have any evidence that Orf3a reduces tetherin levels in the ERGIC or whether the tetherin cycling pathway(s) involve the ERGIC?

This is an interesting point, and as the reviewer concedes, this is tricky to answer. Expression of ORF3a causes the redistribution or remodeling of various organelles (Figures 1E, 2D, 2F, Supp Figures 2C, 2E, 3E, 6B, 6C, 6D). We have been unable to test the direct involvement of ERGIC, despite attempts with a number of commercial antibodies. Given the huge rearrangements of organelles during SARS-CoV-2 infection, it is unclear exactly what will happen to the distribution of ERGIC.

Minor comments: Line 53, delete 'shell' its redundant and confusing when the authors have said coronaviruses have a membrane.

Deleted.

Line 61, delete 'the'

Deleted.

Line 72, delete 'enveloped'; coronaviruses already described as enveloped viruses (line 53)

Deleted.

Lines 93 - 100, lop-sided discussion of the viral life cycle; this paragraph is mostly about entry, which is not relevant to this paper, and does not really deal with the synthesis and assembly side of the cycle.

We have now added the following text,

‘….liberating the viral nucleocapsid to the cytosol of the cell. Upon uncoating, the RNA genome is released into the host cytosol and replication-transcription complexes assemble to drive the replication of the viral genome and the expression of viral proteins. Coronaviruses modify host organelles to generate viral replication factories - so-called DMVs (double-membrane vesicles) that act as hubs for viral RNA synthesis [10]. SARS-CoV-2 viral budding occurs at ER-to-Golgi intermediate compartments (ERGIC) and newly formed viral particles traffic through secretory vesicles to the plasma membrane where they are released to the extracellular space.’ - (Lines 95-104).

Line 103, why are the neighbouring cells 'naive'?

‘naïve’ removed.

Line 112 - 113, delete last phrase; tetherin is described as an IFN stimulated gene in line 111; to be accurate, the beginning of the sentence should be 'Tetherin is expressed from a type 1 Interferon stimulated gene ...'

Amended.

Line 118 - 119, should say 'For tetherin-restricted enveloped viruses' as not all enveloped viruses are restricted by tetherin.

Amended.

Line 131, coronaviruses are not the only family of tetherin-restricted viruses that assemble on intracellular membranes, e.g. bunyaviruses.

This has been modified and now reads,

‘In order for tetherin to tether coronaviruses, tetherin must be incorporated in the virus envelope during budding which occurs in intracellular organelles.’ - (Lines 133-135).

Line 192, there is no EM data in Supplemental Fig 1C.

This has now been removed.

Line 251, 'a synchronous infection event' should be 'synchronous infection' as there will be multiple infection events.

This has been changed.

Page 13 (and elsewhere), unlike Southern, 'Western' should not have a capital letter, except at the start of a sentence.

These have been updated throughout the manuscript (Lines 183, 341, 3549, 356, 392, 509, 763, 1330, 1399).

Lines 330 and 352, can the authors quantitate S protein-induced reduction in cell surface tetherin rather than using the somewhat subjective 'mild'?

These are now changed to,

‘Transient transfection of cells with ss-HA-Spike caused a 32% decrease in tetherin as observed by immunofluorescence (Supplemental Figure 4A, 4B), with…’ – (Line 370).

‘To explore whether the Spike-induced tetherin downregulation altered virus release, we performed experiments with virus like particles (VLPs) in HEK293T …’ – (Line 399).

Line 379, OFR, should be ORF.

Yes, changed.

Line 448, 'Tetherin retains the ability' - did it ever loose it?

This has been rephrased to,

‘Tetherin has the ability to restrict a number of different enveloped viruses that bud at distinct organelles.’ - (Line 547).

Line 451, 'luminal' is confusing in this context.

This has been modified to,

‘Tetherin forms homodimers between opposing membranes (e.g., plasma membrane and viral envelope) that are linked via disulphide bonds.’ - (Line 549).

Line 453, the process of virus envelopment is likely to be more than a 'single step'

This now reads,

‘…virus during viral budding, which occurs in modified ERGIC organelles.’ - (Line 552).

Line 457, in my view the notion that Vpu abrogation of tetherin restriction is just due to redistribution of tetherin to the TGN is somewhat simplistic and disregards a lot of other work.

We have removed mention of mechanisms of tetherin antagonism by other viruses. The key point we wish to make here is that tetherin is lost from the budding compartment. This now reads,

‘Many enveloped viruses antagonise tetherin by altering its localisation and removing it from the respective site of virus budding.’ – (Line 552-553).

Line 472, what is meant by 'resting states'?

This should have been ‘in the absence of stimulation’ and have now been re-written,

‘Tetherin is an IFN-stimulated gene (ISG) [13], and many cell types express low levels of tetherin in the absence of stimulation.’ - (Line 577).

Line 1204, how were 'mock infected cells .......... infected'?

This has now been re-written,

‘Differentiated nasal primary human airway epithelial (HAE) cells were embedded to OCT….’ - (Line 1385).

Reviewer #2 (Significance (Required)):

This study builds on published work supporting the notion that SARS-Cov-2 ORF3a is an antagonist for the restriction factor tetherin. Importantly, it provides insights to the the mechanism of ORF3a mediated tetherin antagonism, specifically to ORF3a inhibits tetherin cycling, diverting the protein to lysosomes and away from compartment(s) where virions assemble. Overall, the authors provide good supporting evidence for these conclusions, however there are issues that the authors need to address.

We wish to thank Reviewer #2 for their insightful comments and suggestions for improving this work.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Restriction factors are major barriers against viral infections. A prime example is Tetherin (aka BST2), which is able to physically tether budding virions to the plasma membrane preventing release of the infectious particles. Of note, tetherin has broad anti-viral activity and has been established as a crucial innate immune defense factor against HIV, IAV, SARS-CoV-2 and other important human pathogens. However, successful viruses like SARS-CoV-2 evolved strategies to counteract restriction factors and promote their replication. Important restriction factors, such as tetherin, may often be targeted by multiple viral strategies to ensure complete suppression of their anti-viral activities by the pathogen. Of note, it was previously published that the accessory protein ORF7a of SARS-CoV-2 binds to (Petrosino et al, Chemistry Europe, 2021) and antagonizes it (Martin-Sancho et al, Molecular Cell, 2021). Previous data on SARS-CoV also revealed that ORF7a promotes cleavage of tetherin (Taylor et al, 2015, J Virol). In this manuscript, the authors show that tetherin restricts SARS-CoV-2 by tethering virions to the plasma membrane and propose that tetherin is targeted by two proteins of SARS-CoV-2. Whereas the Spike protein promotes degradation of tetherin, the accessory protein ORF3a redirects tetherin away from newly forming SARS-CoV-2 virions. While the overall findings that both S and ORF3a are additionally targeting tetherin is both novel and intriguing, additional evidence is needed to support this. In addition, the authors show that in their experimental setups ORF7a does not induce cleavage of tetherin. This is in direct contrast to previously published data both on SARS-CoV(-1) and -2 (Taylor et al, 2015, J Virol; Petrosino et al, Chemistry Europe, 2021; Martin-Sancho et al, Molecular Cell, 2021). From my point of view that needs further experimental confirmation. While the authors state that the impact of Spike on tethrin is mild, the experiments should still allow the conclusion whether there is a (mild) effect or not. The mechanism of ORF3a is fortunately more robustly assessed and provides some novel insights. Unfortunately, the whole manuscript suffers from a striking lack of quantifications. In addition, it is not clear whether and how many times experiments were repeated to the same results. Overall, the data in this manuscript seem very speculative and preliminary and thus do not support the authors conclusions.

Major:

Much of the data seems like it was only done once. As I am sure that this is a writing issue, please clearly state how many times the individual assays were repeated, provide the quantification graphs and appropriate statistics. Some experiments may need additional quantification and confirmation by other methods to be convincing.

Quantification is provided throughout the revised manuscript. Figure legends have also been updated to provide information on quantification and statistical analysis.

For example, Figure 1A, C and D: Please quantify the levels of tetherin and use an alternative readout, e.g. Western blotting of infected cells.

Quantification has been performed and included in our revised manuscript in Supplemental Figures 1C, 1E. Tetherin is not shown in Figure 1C.

A table is provided (above) to highlight the additional quantification.

Figure 2A: Please quantify.

We are not sure we understand this point. The western blot shown in Figure 2A demonstrates the ectopic expression of ACE2 in our A549 cell line. A549 cells have been used by many labs to study SARS-CoV-2 infection, but express negligible ACE2.

Fig 3A: Please show and confirm successful tetherin KO in the cell lines that are used not only in microscopy.

A new blot is now shown in Figure 3A, including a blot demonstrating tetherin loss in both KO lines.

Figure 4C: Please quantify

Currently flow cytometry experiments have been performed twice each and this is now detailed in the figure legends. The data shown in each panel is representative and the data has been explored using analogous approaches. For example, Figure 4C is complemented by Figures 4A and 4B, Figures 4E is complemented by 4D and 4F. We do not feel that repeating these flow cytometry analysis will significantly improve the manuscript.

Figure 4D: Please quantify the effects are not obvious from the images provided.

Quantification is now provided in Supplemental Figure 4E.

Figure 4E, F Please provide a quantification of multiple independent repeats, the claimed differences are neither striking nor obvious.

Quantification of 4F is now provided in Supplemental Figure 4G. Tetherin levels were quantified to be reduced by 25% (SD: 8%) by addition of Doxycycline and induction of ss-HA-Spike. Information for quantification is provided in figure legends.

Figure 5A: Please quantify

These experiments have currently been performed twice and this is now described in the figure legends. Data shown is representative. We can perform one more repeat of these experiments to quantify if neccessary, but do not feel it will significantly alter the manuscript.

Figure 3C and D: At timepoint 0 the infection input levels are different. The initial infection levels have to be the same to draw the conclusion that tetherin KO affects virion release and not the initial infection efficiency. Can the authors either normalize or ensure that the initial infection is the same in all conditions and that variations in the initial infection efficiency do not correlated with the impact of tetherin on replication/release ? How often were those experiments repeated? Are the marginal differences in infectious titre significant? Overall the impact of tetherin on SARS-CoV-2 is very underwhelming but that may be due to efficient viral tetherin-counteraction strategies. Why is the phenotype inverted at 72 h?

Equal amounts of virus, as measured by plaque-forming units (PFU), were used for both HeLa cell lines and thus at 0 hpi the variation seen is within the parameters of the assay used. It remains possible that tetherin affects virus entry but this is unlikely and this assay was not designed to investigate that effect.

Growth curve assays are currently being repeated using an MOI of 0.01, 1 and 5. We are removing the 72 hpi sample from future experiments. At this time point, we find that the extensive cell death caused by viral replication (especially at higher MOIs) makes it difficult to accurately separate the released from intracellular fractions and conclusions cannot be accurately drawn from the data.

Additional repeats of these experiments are in progress and will be included in the finalised manuscript.

Figure 4B and C: Can the authors provide an explanation why SARS-CoV ORF7a is not inducing cleavage/removes glycosylation of tetherin. To show that the assays work, an independent positive control needs to be included. The FACS data in C is unfortunately not quantified.

See above comments (Reviewer #2) regarding discussion on ORF7a. Additional text has been included to discuss ORF7a data,

‘SARS-CoV-1 ORF7a is reported to inhibit tetherin glycosylation and localise to the plasma membrane in the presence of tetherin [18]. We did not observe any difference in total tetherin levels, tetherin glycosylation, ability to form dimers, or surface tetherin upon expression of either SARS-CoV-1 or SARS-CoV-2 ORF7a (Figures 4A, 4B, 4C).

Others groups have demonstrated a role for ORF7a in sarbecovirus infection and both SARS-CoV-1 and SARS-CoV-2 virus lacking ORF7a show impaired virus replication in the presence of tetherin [18,41]. A direct interaction between SARS-CoV-1 ORF7a and SARS-CoV-2 ORF7a and tetherin have been described [18,41], although the precise mechanism(s) by which ORF7a antagonises tetherin remains enigmatic. We cannot exclude that ORF7a requires other viral proteins to antagonise tetherin, or that ORF7a antagonises tetherin via another mechanism. For example, ORF7a can potently antagonise IFN signalling [42] which would impair tetherin induction in many cell types.’ – (Line 667-704).

Fig 4G: The rationale and result of this experiment are not clear.

The rationale for Spike VLP experiments is explained at Line 403. Given that Spike caused a reproducible decrease in cellular tetherin, we examined whether this downregulation was sufficient to antagonise tetherin and increase VLP yield.

Fig 6: What is the benefit of doing the VLP assays as opposed to genuine virus experiments? To me it rather seems to be making the data unnecessarily complex. Again, no quantifications or repeats are provided.

VLPs are used to separate the budding and release process from the replication process of RNA viruses. VLPs have been used in a number of SARS-CoV (DOI: 1002/jmv.25518) and HIV-1 (DOI: https://doi.org/10.1186/1742-4690-7-51) studies to analyse the impact of tetherin (and tetherin mutants) on release.

VLP experiment quantification are now included throughout.

Minor: Fig 1D: How do the authors explain the mainly intracellular Spike staining?

We do not understand this point. Spike staining is intracellular, whether expressed alone or in the context of infected cells.

Please add statistical analyses on the data e.g. Fig. 3 C and D

Additional repeats of these experiments are in progress and will be included in the finalised manuscript.

Fig. 4B and F: Why do the annotated sizes of tetherin differ between the blots?

Figures 4B and 4F are run in non-reduced and reduced conditions respectively. In order to best show the dimer deficient C3A-Tetherin, blots are typically run in non-reduced conditions to exemplify dimer formation and to highlight any defects in dimer formation. The rest of the blots in the manuscript are run in denaturing conditions to aid blotting of other proteins. (Lines 957-958) and now (Lines 1356-1357).

Fig. 5A: What is ORF6a? Do the authors mean ORF6?

Yes, this has been changed.

An MOI of 1 is NOT considered a low or relevant MOI. Can the authors either rephrase or repeat experiments with an actual low or relevant MOI i.e. 0.01 ?

We are currently repeating these experiments and are including MOIs of 0.01, 1 and 5.

Why were the cell models switched between Figure 1 and 2 and essentially the same experiments repeated?

HeLa cells express high levels of tetherin at steady state, whilst A549 cells require IFN stimulation. HeLa cells demonstrate that tetherin downregulation occurs via an IFN-independent manner. A549 and T84 cells are more physiologically relevant cell types for SARS-CoV-2 infection. These points are stated in Lines 230 and 261.

The manuscript may benefit a lot from streamlining and removing unessential deviations from the main message (e.g. discussions why multistep/single step growth curves are used/not relevant; why are they shown if the authors conclude that a single step is not relevant?). The discussion is extremely lengthy and does not provide sufficient discussion of the presented data.

The multistep/single step growth curve text will be adapted, but it will be re-written after additional infection experiments.

We have removed from the Discussion a small section discussing ORF7a mutants, given that the emphasis of our manuscript is not on ORF7a.

We have also removed a small section describing the rearrangements of intracellular organelles by SARS-CoV-2 as it does not directly relate to the central message of our manuscript.

According to my opinion, the current manuscript does not provide significant advancement for the field. While the intention was to update and expand our existing knowledge about tetherin restriction by SARS-CoV-2, the experiments do not support this yet. However, the general premise and approach/concept of the manuscript would be appealing to a broader audience. I especially like the notion that multiple proteins of SARS-CoV-2 could synergistically counteract an important innate immune defense factor, tetherin. My expertise is on SARS-CoV-2 and the interplay between the virus and host cell restriction factors.

Reviewer #3 (Significance (Required)):

According to my opinion, the current manuscript does not provide significant advancement for the field. While the intention was to update and expand our existing knowledge about tetherin restriction by SARS-CoV-2, the experiments do not support this yet. However, the general premise and approach/concept of the manuscript would be appealing to a broader audience. I especially like the notion that multiple proteins of SARS-CoV-2 could synergistically counteract an important innate immune defense factor, tetherin. My expertise is on SARS-CoV-2 and the interplay between the virus and host cell restriction factors.

We thank Reviewer#3 for their comments and suggestions for improving this work.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Restriction factors are major barriers against viral infections. A prime example is Tetherin (aka BST2), which is able to physically tether budding virions to the plasma membrane preventing release of the infectious particles. Of note, tetherin has broad anti-viral activity and has been established as a crucial innate immune defense factor against HIV, IAV, SARS-CoV-2 and other important human pathogens. However, successful viruses like SARS-CoV-2 evolved strategies to counteract restriction factors and promote their replication. Important restriction factors, such as tetherin, may often be targeted by multiple viral strategies to ensure complete suppression of their anti-viral activities by the pathogen. Of note, it was previously published that the accessory protein ORF7a of SARS-CoV-2 binds to (Petrosino et al, Chemistry Europe, 2021) and antagonizes it (Martin-Sancho et al, Molecular Cell, 2021). Previous data on SARS-CoV also revealed that ORF7a promotes cleavage of tetherin (Taylor et al, 2015, J Virol).

In this manuscript, the authors show that tetherin restricts SARS-CoV-2 by tethering virions to the plasma membrane and propose that tetherin is targeted by two proteins of SARS-CoV-2. Whereas the Spike protein promotes degradation of tetherin, the accessory protein ORF3a redirects tetherin away from newly forming SARS-CoV-2 virions.

While the overall findings that both S and ORF3a are additionally targeting tetherin is both novel and intriguing, additional evidence is needed to support this. In addition, the authors show that in their experimental setups ORF7a does not induce cleavage of tetherin. This is in direct contrast to previously published data both on SARS-CoV(-1) and -2 (Taylor et al, 2015, J Virol; Petrosino et al, Chemistry Europe, 2021; Martin-Sancho et al, Molecular Cell, 2021). From my point of view that needs further experimental confirmation. While the authors state that the impact of Spike on tethrin is mild, the experiments should still allow the conclusion whether there is a (mild) effect or not. The mechanism of ORF3a is fortunately more robustly assessed and provides some novel insights. Unfortunately, the whole manuscript suffers from a striking lack of quantifications. In addition, it is not clear whether and how many times experiments were repeated to the same results. Overall, the data in this manuscript seem very speculative and preliminary and thus do not support the authors conclusions.

Major

Much of the data seems like it was only done once. As I am sure that this is a writing issue, please clearly state how many times the individual assays were repeated, provide the quantification graphs and appropriate statistics. Some experiments may need additional quantification and confirmation by other methods to be convincing. For example, Figure 1A, C and D: Please quantify the levels of tetherin and use an alternative readout, e.g. Western blotting of infected cells. Figure 2A: Please quantify. Fig 3A: Please show and confirm successful tetherin KO in the cell lines that are used not only in microscopy. Figure 4C: Please quantify. Figure 4D: Please quantify the effects are not obvious from the images provided. Figure 4E,F Please provide a quantification of multiple independent repeats, the claimed differences are neither striking nor obvious. Figure 5A: Please quantify.

Figure 3C and D: At timepoint 0 the infection input levels are different. The initial infection levels have to be the same to draw the conclusion that tetherin KO affects virion release and not the initial infection efficiency. Can the authors either normalize or ensure that the initial infection is the same in all conditions and that variations in the initial infection efficiency do not correlated with the impact of tetherin on replication/release ? How often were those experiments repeated? Are the marginal differences in infectious titre significant? Overall the impact of tetherin on SARS-CoV-2 is very underwhelming but that may be due to efficient viral tetherin-counteraction strategies. Why is the phenotype inverted at 72 h? Figure 4B and C: Can the authors provide an explanation why SARS-CoV ORF7a is not inducing cleavage/removes glycosylation of tetherin. To show that the assays work, an independent positive control needs to be included. The FACS data in C is unfortunately not quantified.

Fig 4G: The rationale and result of this experiment are not clear.

Fig 6: What is the benefit of doing the VLP assays as opposed to genuine virus experiments? To me it rather seems to be making the data unnecessarily complex. Again, no quantifications or repeats are provided.

Minor

Fig 1D: How do the authors explain the mainly intracellular Spike staining?

Please add statistical analyses on the data e.g. Fig. 3 C and D

Fig. 4B and F: Why do the annotated sizes of tetherin differ between the blots?

Fig. 5A: What is ORF6a? Do the authors mean ORF6?

An MOI of 1 is NOT considered a low or relevant MOI. Can the authors either rephrase or repeat experiments with an actual low or relevant MOI i.e. 0.01 ?

Why were the cell models switched between Figure 1 and 2 and essentially the same experiments repeated? The manuscript may benefit a lot from streamlining and removing unessential deviations from the main message (e.g. discussions why multistep/single step growth curves are used/not relevant; why are they shown if the authors conclude that a single step is not relevant?). The discussion is extremely lengthy and does not provide sufficient discussion of the presented data.

According to my opinion, the current manuscript does not provide significant advancement for the field. While the intention was to update and expand our existing knowledge about tetherin restriction by SARS-CoV-2, the experiments do not support this yet. However, the general premise and approach/concept of the manuscript would be appealing to a broader audience. I especially like the notion that multiple proteins of SARS-CoV-2 could synergistically counteract an important innate immune defense factor, tetherin. My expertise is on SARS-CoV-2 and the interplay between the virus and host cell restriction factors.

Significance

According to my opinion, the current manuscript does not provide significant advancement for the field. While the intention was to update and expand our existing knowledge about tetherin restriction by SARS-CoV-2, the experiments do not support this yet. However, the general premise and approach/concept of the manuscript would be appealing to a broader audience. I especially like the notion that multiple proteins of SARS-CoV-2 could synergistically counteract an important innate immune defense factor, tetherin.

My expertise is on SARS-CoV-2 and the interplay between the virus and host cell restriction factors.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

BST2/tetherin can restrict the release and transmission of many enveloped viruses, including coronaviruses. In many cases, restricted viruses have developed mechanisms to abrogate tetherin-restriction by expressing proteins that antagonize tetherin; HIV-1 Vpu-mediated antagonism of tetherin restriction is a particularly well studied example. In this paper, Stewart et al. report their studies of the mechanism(s) underlying SARS-CoV-2 antagonism of tetherin restriction. They conclude that Orf3a is the primary virally encoded protein involved and that Orf3a manipulates endo-lysosomal trafficking to decrease tetherin cycling and divert the protein away from putative assembly sites.

Major comments:

In my view some of the claims made by the authors are not fully supported by the data. For example, the bystander effect discussed in line 162 may suggest that infected cells can produce IFN but does not 'indicate' that they do. Most of the EM images show part of a cell profile, so statements such as (line 192) 'virus containing tubulovesicular organelles were often polarised towards sites of significant surface-associated virus' should be backed up with appropriate images, or indicated as 'not shown', or removed (the observation is not so important for this story). Line 196, DMVs can't be seen in these micrographs. Line 391, I can't see much change in CD63 distribution.

Line 321, the authors show that ORF7a does not affect tetherin localization, abundance, glycosylation or dimer formation, but they don't show that it doesn't restrict SARS-CoV-2. Can they be sure that epitope tagging this molecule does not abrogate function (or the functions of any of the other tagged proteins for that matter), or that ORF7a works in conjunction with one of the other viral proteins? In the ORF screen, a number of the constructs are expressed at low level, is it possible they are missing something?

Line 376, the authors refer to ORF3a being a viroporin. A recent eLife paper (doi: 10.7554/eLife.84477; initially published in BioRxiv) refutes this claim and builds on other evidence that ORF3a interacts with the HOPS complex. The authors should at least mention this work, especially in the discussion, as it would seem to provide a molecular mechanism to support their conclusions.

Fig 3, the growth curves illustrated in Fig3 C and D do not have errors bars; how many times were these experiments repeated?

Line 396, the authors show increased co-localization with LAMP1. As LAMP1 is found in late endosomes as well as lysosomes, they cannot claim the redistributed tetherin is specifically in lysosomes.

There seems to be a marked difference in the anti-rb555 signal in the 'mock' cells in panels 5H and Suppl 6E. Is there a good reason for this, or does this indicate variability between experiments?

Fig 6a, why is there negligible VLP release from cells lacking BST2 and ORF3a-strep? How many times were these experiments performed? Is this a representative image? I think it confusing to refer to the same protein by two different names in the same figure (i.e. BST2 and tetherin). Do the authors know how the levels of ORF3a expressed in cells in these experiments compares to those seen in infected cells?

My final point is, perhaps, the trickiest to answer, but nevertheless needs to be considered. As far as we know, SARS-CoV-2 and at least some other coronaviruses, bud into organelles of the early secretory pathway, often considered to be ERGIC. In the experiments shown here the authors provide evidence that ORF3a can influence tetherin recycling, but the main way of showing this is through its increased association with endocytic organelles. Do the authors have any evidence that Orf3a reduces tetherin levels in the ERGIC or whether the tetherin cycling pathway(s) involve the ERGIC?

Minor comments:

Overall, the manuscript should be carefully edited to ensure the text reads clearly. A few examples of thing that need to be fixed are:-

Line 53, delete 'shell' its redundant and confusing when the authors have said coronaviruses have a membrane.

Line 61, delete 'the'

Line 72, delete 'enveloped'; coronaviruses already described as enveloped viruses (line 53)

Lines 93 - 100, lop-sided discussion of the viral life cycle; this paragraph is mostly about entry, which is not relevant to this paper, and does not really deal with the synthesis and assembly side of the cycle.

Line 103, why are the neighbouring cells 'naive'?

Line 112 - 113, delete last phrase; tetherin is described as an IFN stimulated gene in line 111; to be accurate, the beginning of the sentence should be 'Tetherin is expressed from a type 1 Interferon stimulated gene ...'

Line 118 - 119, should say 'For tetherin-restricted enveloped viruses' as not all enveloped viruses are restricted by tetherin.

Line 131, coronaviruses are not the only family of tetherin-restricted viruses that assemble on intracellular membranes, e.g. bunyaviruses.

Line 192, there is no EM data in Supplemental Fig 1C.

Line 251, 'a synchronous infection event' should be 'synchronous infection' as there will be multiple infection events

Page 13 (and elsewhere), unlike Southern, 'Western' should not have a capital letter, except at the start of a sentence.

Lines 330 and 352, can the authors quantitate S protein-induced reduction in cell surface tetherin rather than using the somewhat subjective 'mild'?

Line 379, OFR, should be ORF.

Line 448, 'Tetherin retains the ability' - did it ever loose it?

Line 451, 'luminal' is confusing in this context.

Line 453, the process of virus envelopment is likely to be more than a 'single step'

Line 457, in my view the notion that Vpu abrogation of tetherin restriction is just due to redistribution of tetherin to the TGN is somewhat simplistic and disregards a lot of other work.

Line 472, what is meant by 'resting states'?

Line 1204, how were 'mock infected cells .......... infected'?

Significance

This study builds on published work supporting the notion that SARS-Cov-2 ORF3a is an antagonist for the restriction factor tetherin. Importantly, it provides insights to the the mechanism of ORF3a mediated tetherin antagonism, specifically to ORF3a inhibits tetherin cycling, diverting the protein to lysosomes and away from compartment(s) where virions assemble. Overall, the authors provide good supporting evidence for these conclusions, however there are issues that the authors need to address.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

I summarise the major findings of the work below. In my opinion the range and application of approaches has provided a broad evidence base that, in general, supports the authors conclusions. However, there are, in my opinion, particular failures to utilise and communicate this evidence. The manuscript may be much improved with attention in the following areas. In each case I will give general criticism with a few examples, but the principals of my comments could be applied throughout the work.

1. Insufficient quantification. The investigation combines various sources of qualitative data (EM, fluorescence microscopy, western blotting) to generate a reasonably strong evidence base. However, the work is over-reliant on representative images and should include more quantification from repeat experiments. When there are multiple fluorescence micrographs with intensity changes (not necessarily just representative images) (e.g. Figure 1 or 2) the authors should consider making measurements of these. Also the VLP production assays, which are assessed by western blotting would particularly benefit from a quantitative assessment (either by densitometry or, if samples remain, ELISA/similar approach).
2. Insufficient explanation. I found some of the images and legends contained insufficient annotation and/or description for a non-expert reader to appreciate the result(s). Particularly if the authors want to draw attention to features in micrographs they should consider using more enlarged/inset images and annotations (e.g. arrows) to point out structures (e.g DMVs etc.). This short coming exacerbates the lack of quantification.
3. Insufficient exploration of the data. I had a sense that some aspects of the data seem unconsidered or ignored, and the discussion lacks depth and reflection. For example the tetherin down-regulation apparent in Figures 1 and 2 is not really explained by the spike/ORF3a antagonism described later on, but this is not explicitly addressed. Also, Figure 6 suggests that ORF3a results in high levels of incorporation of tetherin in to VLPs, but I don't think this is even described(?). The discussion should also include more comparison with previous studies on the relationship between SARS-2 and tetherin.

I have no minor comments on this draft of the manuscript.

Significance

Tetherin, encoded by the BST2 gene, is an antiviral restriction factor that inhibits the release of enveloped viruses by creating tethers between viral and host membranes. It also has a capacity for sensing and signalling viral infection. It is most widely understood in the context of HIV-1, however, there is evidence of restriction in a wide variety of enveloped viruses, many of which have evolved strategies for antagonising tetherin. This knowledge informs on viral interactions with the innate immune system, with implications for basic virology and translational research.

This study investigates tetherin in the context of SARS-CoV-2. The authors use a powerful collection of tools (live virus, gene knock out cells, recombinant viral and host expression systems) and a variety of approaches (microscopy, western blotting, infection assays), which is, itself, a strength. The study provides evidence to support a series of conclusions: I) BST2/tetherin restricts SARS-CoV-2 II) SARS-CoV-2 ablates tetherin expression III) spike protein can modestly down-regulate tetherin IV) ORF3A dysregulates tetherin localisation by altering retrograde trafficking. These conclusions are broadly supported by the data and this study make significant contributions to our understanding of SARS-CoV-2/tetherin interactions.

My enthusiasm is reduced by, in my opinion, a failure of the authors to fully quantify, explain and explore their data. I expect the manuscript could be significantly improved without further experimentation by strengthening these aspects.

This manuscript will be of interest to investigators in virology and/or cellular intrinsic immunity. Given the focus on SARS-CoV-2 it is possible/likely that it will find a slightly broader readership.

I have highly appropriate skills for evaluating this work being experienced in virology, SARS-CoV-2, cell biology and microscopy.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We thank the reviewers for their enthusiasm for our work and constructive feedback.

Below please find our point-by-point response to the comments:

Reviewer #1 -Key conclusions that were less convincing: -RhoA and NMII are in the title as mechanistic downstream regulators of CaM, but the results in Fig 8 call into question the role of RhoA. Why does RhoA activation not influence cell size and circularity? Can you overexpress MRLC-GFP and inhibit Rho and restore the wt phenotype? The role of NMII is also not clear - why does overexpressing MLCK-CA not have a phenotype but overexpressing MLCK downstream target MRLC show the phenotype? Are there any alternative pathways to regulate MRLC? It's not being discussed or described in 6D's schematic.

Response: Rho activation usually leads to the formation of more stress fibers and therefore does not lead to decreased cell size and increased circularity observed in MFN2 KO. The phenotype is restored by either ROCK or MLCK knockdown. We have discussed in the main text that the formation of PAB requires both RhoA and NMII activation under restricted spatiotemporal control.

MRLC has three major regulators (Ikebe & Hartshorne, 1985; Isotani et al., 2004). As we discussed, MLCK and ROCK phosphorylate MRLC at either Ser19 or Ser19 and Thr18. MRLC is dephosphorylated and inactivated by its phosphatase MLCP. We tried to knock down MLCP in wt MEF cells but failed to see any cell morphology changes (data not shown).

We were also surprised to see MRLC-GFP overexpression with Rho Activator can phenocopy PAB, but “MLCK-CA + Rho Activator” failed to. We believe it is because MLCK-CA constitutively over-activates a broad range of downstream effectors while overexpressing MRLC mimics endogenous activation or NMII alone. Also, only a proportion of cells acquired PAB structure under Rho Activator and MRLC overexpression, which indicates PAB formation also requires specific spatiotemporal controls.

*Rewrite for clarity -The role of ER/mito contacts in the system was unclear (since ER/mito contacts were not observed nor evaluated directly). *

__Response: __We have included additional data to measure ER/mito contacts in MEFs. Our result is consistent with numerous previous reports that MFN2 regulates ER/mito contacts. The data is now included in Fig. S3.

* -What role does focal adhesions have on PAB formation or any part of the model - There were results showing larger focal adhesions in the MFN2-/- cells, but not sure how this fits in with the bigger picture of contractility and PABs, and focal adhesions were not in the model in Figure 5.*

__Response: __Focal adhesion and actomyosin are tightly coupled, and our work focuses on the actin network. Our model did not include FAs since FA is not a significant focus in this study.

* -Whether regulating calcium impacts PAB formation*

__Response: __Calcium likely regulates PAB formation. We have shown PAB cell percentage decreases in mfn2-/- with ER-mito tethering contrast in Fig. S3.

-The role of PABs in migration is also unclear - can you affect PAB formation or get rid of PABs and quantify cell migration?

__Response: __Our data suggest that PAB formation and cell migration are inversely correlated. Since PAB results from a contractile actin band on the cell periphery, its role in defective cell spreading and migration is expected. We demonstrated that MLCK and ROCK knockdown reduced PABs and rescued cell spreading.

-It was hard to understand the correlation between the membrane tension of MFN2-/- cells and their ability to spread on softer substrates. How does this result fit in with the overarching model?

__Response: __Reduced membrane tension is presumably associated with decreased cell spreading. Softer substrates attenuate the mechanical force on focal adhesion proteins and the actin cytoskeleton (Burridge & Chrzanowska-Wodnicka, 2003; Pelham & Wang, 1997; Wong et al., 2015), which is required for focal adhesion maturation. As a result, softer substrates can reduce the over-contraction in the MFN2 KO cells. The results support the model that MFN2 KO cells have enhanced cell contraction on the substrates dependent on substrate interaction and force transduction on focal adhesions. Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

__Response: __We have removed the MFN2-related disease from the introduction to focus the paper more on cell biology in vitro.

*-There were a number of findings that did not seem to fit in with the paper, or they were included, but were not robustly described nor quantified. As an example, Paxillin-positive focal adhesions were evaluated in MFN2-/- and with various pharmacological approaches, but there is no quantification with respect to size, number, or distribution of focal adhesions, despite language in the main text that there are differences between conditions. *

__Response: __We have quantified the focal adhesions in the KO cells, and the data is now included in Fig. 5. We used the actin distribution to quantify the “PAB”; therefore, FAs are not a significant focus of this study.

*Also, the model was presented in figure 5, and then there were several pieces of data presented afterward, some that are included in the model (myosin regulation), and some that are not in the model (membrane tension, TFM, substrate stiffness, etc). * __Response: __Membrane tension and substrate stiffness dependence are physical properties of the cell. The model focuses on the molecular mechanism that leads to PAB formation.

The stiffness/tension figure was not clear to me, and it was difficult to make sense of the data since one would predict that an increase in actomyosin contractility at the cortex would lead to higher membrane tension, not lower, and then how membrane tension relates to spreading on soft matrices is also unclear.

__Response: __The result was surprising to us initially. However, the MFN2 KO cells have increased actomyosin contractility only at the cell-substrate interface but not throughout the entire cell cortex. A less spread cell would have a more relaxed membrane and display a lower membrane tension, consistent with our observation. Softer matrices reduce cell contractility at the cell-substrate interface, which allows MFN2 KO cells to relax and spread better. We have emphasized in the discussion of our manuscript that MFN2 KO cells have an increase in actomyosin contractility only at the cell-substrate interface.

The manuscript seems like an amalgamation of different pieces of data that do not necessarily fit together into a cohesive story, so the authors are encouraged to either remove these data, or shore them up and weave them into the narrative.

__Response: __We respectfully disagree with the reviewer since the cell morphology, actin structure, substrate interaction, and cell mechanics are tightly correlative and provide a complete picture of the role of MFN2 in regulating cell behavior.

* Request additional experiments 1. -The imaging used for a lot of the quantification (migration, circularity) is difficult to resolve. The cells often look like they are not imaged in the correct imaging plane. It would be helpful to have better representative images such that it is clear how the cells were tracked and how cell periphery regions of interest were manually drawn. Focal adhesions should also be shown without thresholding.*

__Response: __We used a TIRF microscope and imaged a Z stack. Imaris was used to combine the Z-stack images. The images in the manuscripts are now of the lowest stack with background subtraction.

2. -For most of the quantification, it appears that the experiment was only performed once and that a handful of cells were quantified. The figure legend indicates the number of cells (often reported as >12 cells or >25 cells), but the methods indicate that high content imaging was performed, and so the interpretation is that these experiments were only performed once. Biological replicates are required. If the data do represent at least 3 biological replicates, then more cell quantification is required (12 or 25 cells in total would mean quantifying a small number of cells per replicate).

Response: __We quantified more cells and indicated the number of cells quantified in the figure legends. The experiments are with three biological replicates.__

* 3. -Mitochondrial morphology and quantification should be performed in the MFN2 knockdown and rescue lines.*

__Response: __Mitochondrial morphology is well characterized in the Mfn2 KO and rescue MEF cells (Chen et al., 2003; Naon et al., 2016; Samanas et al., 2020). We observed a similar phenotype using mito-RFP to label mitochondrial structure (Fig. S1).

4.-Many of the comparisons throughout the figures is between MFN2 knockdown and MFN2 knockdown plus rescue or genetic/pharmacological approaches, but a comparison that is rarely made is between wildtype and experimental. These comparisons could be useful in comparing partial rescues and potential redundancies with the other mitofusin.

Response: We have included the WT in our assays (Fig. 2-6). We also confirmed that MFN1 could not rescue the MFN2 defects (Fig. 2). We observed partial and complete rescue in different assays. It would be difficult to conclude whether the phenotype is due to the redundancies with the other mitofusin because not all cells are rescued at the endogenous level.

* 5. -For the mito/ER tethering experiments, it is important to show that ER/mito contacts are formed and not formed in the various conditions with imaging approaches.*

__Response: __We adopted a previously established method to quantify ER-mitochondria contacts with the probe SPLICSL (Cieri et al., 2017; Vallese et al., 2020). Our results align with previous reports that Mfn2-null MEFs displayed significantly decreased ER-mitochondria contacts. MFN2 re-expression or ER-mitochondria tethering structure restored the contacts. (Fig. S3).

* 6. -For some of the pharmacological perturbations, it would be helpful to show that the perturbation actually led to the expected phenotype - as an example, in cells treated with different concentrations of A23187, what are the intracellular calcium levels and how do these treatments influence PAB formation? This aspect should be generally applied across the study - when a modification is made, that particular phenotype should first be evaluated, before dissecting how the perturbation affects downstream phenotypes.*

__Response: __We selected a collection of well-characterized inhibitors broadly used in the literature for pharmacological perturbations. For example, numerous studies used A23187 treatment to raise intracellular calcium to examine related actin cytoskeleton changes (Carson et al., 1994; Goldfine et al., 1981; Shao et al., 2015). We titrated the drugs in WT in preliminary experiments and observed similar phenotypes. (data not shown). We then use the same concentration to treat the MFN2 KO cells. Overall, we use pharmacological perturbations as supporting evidence. We use genetics (knockdown or overexpression) to validate our results.

7. -In Figures 4 and 5, the thresholding approaches in the images make the focal adhesions difficult to resolve, and therefore it is difficult to determine the size. As described above, these metrics should be defined and quantified.

__Response: __We used a TIRF microscope and imaged a Z stack. Imaris was used to combine the Z-stack images. The images in the manuscripts are now of the lowest stack with background subtraction.

8. -What is a PAB? How is it defined? What metrics make a structure a PAB versus regular cortical actin - are there quantifiable metrics? In figure 8, there are some structures that are labelled as a PAB, but some aren't (as an example, the left panel in 8b is a PAB, but the right panel in 8A is not, but they look the same), so a PAB should be defined with quantifiable measures, and then applied to the entire study.

__Response: __We developed an algorism to quantify PAB cells. We first used the ImageJ plugin FiloQuant (Jacquemet et al., 2019) to identify the cell border and cytoskeleton, then used our custom algorism to quantify the percentage of actin in the cell border area. The cellular circularity is also calculated at the same time. If the cell contains more than 50% actin in the cell border area, and the circularity is higher than 0.6, we then count it as a “PAB” cell (Fig. S2).

-As described above, why does RhoA activation not influence cell size and circularity? Can you overexpress MRLC-GFP and inhibit Rho and restore the wildtype phenotype?

__Response: __Rho activation usually leads to the formation of more stress fibers and therefore does not lead to decreased cell size and increased circularity observed in MFN2 KO.

We are sorry that we don’t understand the rationale of this experiment proposed by the reviewer. ROCK inhibition restored the wildtype phenotype in MFN2 KO cells (Fig.7). Figure 8 is to create the MFN2 KO phenotype in WT cells, which requires both Rho and MRLC overactivation.

* 10 Are the data and the methods presented in such a way that they can be reproduced? -We appreciate that the authors quantified many parameters, although some quantifications were missing. There are some missing methods - how was directionality quantified, was migration quantified by selecting the approximate center of cells using MTrackJ or were centroids quantified by outlining cells, for instance. Also, given that some of the phenotypes were somewhat arbitrarily assigned (ie. what constitutes a PAB?), it may be difficult to reproduce these approaches and interpret data appropriately.*

__Response: __We have clarified directionality quantification methods and other details. We used MTrackJ to track cell migration. And as we mentioned above, we came up with a customized algorithm to quantify PAB cells, which shows the critical effectors in a more quantifiable way.

* 11. Are the experiments adequately replicated and statistical analysis adequate? -Unfortunately, while the approximate number of cells was reported, the number of biological replicates were not reported, and therefore, the experimental information and statistical analyses are not adequate.*

__Response: __We have quantified more cells and indicated the number of repeats and cells quantified in the figure legend. Minor comments: Specific experimental issues that are easily addressable.

* 12. - for some of the graphs - mostly about calcium levels - fold change is shown, but raw values should also be included to determine whether the basal levels of calcium are different across the conditions.*

__Response: __Delta F/F0 is the standard method to normalize dye loading in cells for calcium concentration measurements (Kijlstra et al., 2015; Zhou et al., 2021).

13. - scale bars in every panel should also help make the points clearer.

__Response: __We have added scale bars in all panels.

* Reviewer #2 (Evidence, reproducibility and clarity (Required)): 1. Fig. 1A: The Mfn1 Western Blot is not of publication quality. Moreover, quantitation is necessary.*

__Response: __We performed additional western blots, changed the representative images, and quantified the level of knockdown and overexpression (Fig.2 and 7). We did not quantify the WB in Fig.1A since it was to confirm that the Mfn1-/- or Mfn2-/- were knock-out cell lines.

2. Fig. 1B (as well as Fig. 2G and others): the date do not reflect cellular size but instead spread cellular area.

__Response: __We thank the reviewer for this suggestion. We have changed all similar descriptions to “Spread Area” in the main text and figures.

3. Fig. 1C, D: Mfn1-null MEFs appear to be more spindle-shaped than wt cells, yet their circularity tends to be elevated. Do the authors have an explanation?

__Response: __The circularity of Mfn1-/- MEFs has a slight increase but is not significant compared to the wt cells. As we observed, Mfn1-/- MEFs have fewer protrusions than wt, which may contribute to the slight increase in its circularity (Fig. 5C). However, this is not the focus of this study.

* 4. Fig. 2A: The Mfn1 levels in Mfn2-/- + Mfn2 are lower than Mfn2-/-? Does this imply a crosstalk between Mfn1 and Mfn2 expression.*

__Response: __We agree with the reviewer that a compensatory change in MFN1 expression might happen in Mfn2-/- + MFN2 MEFs. Previous research also indicated crosstalk between MFN1 and MFN2 expression (Sidarala et al., 2022).

* 5. Fig. 2H: The authors should provide co-staining of mitochondria and Mfn2 as well.*

__Response: __Co-staining of mitochondria and MFN2 in Mfn2-/- MEFs or rescue lines has been done in numerous previous studies (Chen et al., 2003; Naon et al., 2016; Samanas et al., 2020). In this work, we transfected our cells with mito-RFP and showed mitochondria changes in Mfn2-/- and rescue MEF cells (Fig. S1G).

6. Fig. 4D-E: Western blots are not of publication quality. Looking at the blots provided in Fig. 4D, the reviewer is not convinced with the quantitative data shown in Fig. 4E. For instance, the intensity of pCaMKII band for "vec" does not look 3x higher than that of "+MFN2", whereas that of "+MFN2" looks much higher than that of WT.

__Response: __We have performed additional western blots and changed the representative images.

* 7. Fig. 5C: The authors should stain for vinculin, which are present in mature FAs only, rather than paxillin which are present in all FAs. This would strengthen the authors' conclusions. Also, FA size should be quantified.*

__Response: __We have quantified FA size in Figure 5. The maturity of FAs is not a major focus of this study. It is likely that most FAs here are mature since they are connected with stress fibers.

* 8. Fig. 6C - Why does the background have a grid and appear grey in color? Also, the cell interior appears in different colors in the different images. The authors should take a z-stack of images and provide the raw image files.*

__Response: __We used a TIRF microscope and imaged a Z stack. Imaris was used to combine the Z-stack images. The images in the manuscripts are now of the lowest stack with background subtraction.

9. Fig 7C: The MLCII Western blot is not of publication quality, and may affect the quantification provided in Fig. 7D.

__Response: __We have performed additional western blots and changed the representative images.

* 10. Fig 8: Do cell treatments with Rho Activator and MLCK-CA also impair migration velocity similar to Mfn2-null cells?*

__Response: __Our data indicated that Rho Activator and MRLC induced the “PAB” structure seen in MFN2 KO cells. It is likely that cell migration is impaired here. Spatiotemporal regulation of Rho Activation is important to cell migration, it is known that Rho overactivation can significantly inhibit cell migration (Nobes & Hall, 1999). Showing Rho Activator and MLCK-CA will reduce cell migration will not add new knowledge to the cell migration field. However not all cell migration defects are associated with the PAB. We, therefore, focused on PAB quantification in this figure.

11. Fig. 9A: The authors claim that wt cells have actin bundles that protrude against the membrane while Mfn2-null cells do not. This does not look convincing as the Mfn2-null actin bundle seem to be pushing against the membrane at the bottom of the image. No quantification is provided. It is unclear what conclusion can be drawn from the super-resolution images.

__Response: __We used super-resolution imaging to demonstrate the details of the peripheral actin band (PAB) structure. We have used two boxes to enlarge the regions where membrane parallel actin structures are predominant. The quantification of PAB is provided in other figures.

12. Suppl. Fig. 5C: The authors should take images using a confocal microscope for cells with Flipper-TR construct, eliminate the background and the cell center to only consider the cell periphery. Nikon TE2000 does not seem to be a confocal microscope.

Response: __The amount of Flipper-TR that MEF cells can take in was limited. With the current signal-to-noise ratio, complete background elimination is not feasible. A confocal microscope is not necessary for Flipper-TR FLIM imaging (García-Calvo et al., 2022). __* Reviewer #3 (Evidence, reproducibility and clarity (Required)):

General Comments (Major) 1. Data Presentation and analysis: The data analysis would benefit from using a method such as Super-Plots to show the data from separate biological repeats and to use N numbers that represent the number of biological repeats rather than the number of cells analysed. Please see the following reference for a suggestion on how to analyse the data: Lord SJ, Velle KB, Mullins RD, Fritz-Laylin LK. SuperPlots: Communicating reproducibility and variability in cell biology. J Cell Biol. 2020 Jun 1;219(6):e202001064. doi: 10.1083/jcb.202001064. PMID: 32346721; PMCID: PMC7265319.*

__Response: __We have changed the dot colors to show data from separate biological repeats.

• Another general comment is that many of the experiments show analysis of very few numbers of cells or maybe only one field of view in a microscopy quantification experiment. This seems unusually low - for example, in Figure 1E only 6 cells have been analysed. It seems like more could have been done and if statistical analysis like we suggest in 1 above is used, this might reveal that some of the differences are less significant than the authors think/report. This is important, as cells are noisy and it is unusual to have such high significance for experiments like cell migration and other parameters unless a lot of measurements are made. In Figure 1G, it appears that only one field of view has been used to quantify the data. We routinely use 3-5 fields of view to get a representative sample of what the cells are doing.*

__Response: __We have quantified more cells and indicated the number of repeats and cells quantified in the figure legend.

• Some of the micrographs appear to be missing scale bars- e.g. Fig. 2H, Fig. 8*

__Response: __We have included scale bars in the lower right corner of all panels.

* 4. In the cell tracking experiments, only some of the cells in the images appear to have been tracked. How were the tracked cells chosen? Normally, we would track every cell to avoid bias in selection.*

__Response: __We tracked all the cells in the view at the beginning of the experiments.

5. The western blot images do not show the molecular size of the bands. Show ladder position

__Response: __We have added bands to show molecular weights.

6. Mostly the graphs show individual data points, which is good, but in some cases only a bar is shown- it would be nice to have individual points overlaid on the bars- e.g. Figure 1I, 4E, 5B, 5E, 7B, 7D

__Response: __We have updated the graph to show individual points.

7. Many of the confocal images look very processed- they have no background and have a hazy black halo around the cell. I am not familiar with the type of processing that was done and I worry that the images are only showing a masked and processed version of the actual data. The authors need to explain what processing they have done and probably also to provide the unprocessed images in a supplementary figure or dataset for readers to see. The methods description is inadequate as it only says Image J was used to process the data.

__Response: __We used a TIRF microscope and imaged a Z stack. Imaris was used to combine the Z-stack images. The images in the manuscripts are now of the lowest stack with background subtraction.

* Individual comments on Figures: Figure 1: See general comments above- consider to use Superplots, more cells and more fields of view in quantifications. Show experimental points in bar graph.*

__Response: __We have quantified more cells and used super plots to display the data. The number of repeats and cells quantified are indicated in the figure legend.

Figure 2: In 2E, the colours are very similar for two of the experiments so it is difficult to distinguish them- e.g. the MFN1 vs MFN2 rescue data both appear dark blue. Response: We have changed the color for MFN1 rescue to distinguish the two samples better.

In 2H are the magnifications really the same for the WT as the +DOX and -DOX? The cell in the WT looks huge. Is this representative? Also, the phalloidin stain looks very spotty on the WT. This seems unusual.

__Response: __The images are of the same scale. The Mfn2-/- MEFs are smaller, and DOX-induced MFN2 expression can only partially rescue the cell size.

* Figure 3: Not many cells were analysed in 3B, especially the zero time point.*

__Response: __We have quantified more cells.

* Please define +T, we assume it is the tether construct, but it is not defined*

__Response: __We defined T as a tether in the main text and the figure legend. In 3F, how were the tracked cells chosen?

__Response: __We tracked all the cells in the view at the beginning of the experiments.

* Figure 4:* 4B: Why have they not tested FK506 and STO609 on the WT cells?

__Response: __We focused on understanding the MFN2 KO phenotype. Since neither FK506 nor STO609 altered the MFN2 KO phenotype, we did not include them in the WT group.

4C: How were the tracked cells chosen?

__Response: __We tracked all the cells in the view at the start of the experiments.

4D-E: The blot doesn't look representative of the quantification- were the numbers normalised to vinculin? The difference between WT and vector looks too large to be real if the amounts were normalised to the vinculin, as vinculin is increased in vector. Likewise, the pCAMKII looks to be substantially decreased from the +MFN2, but this is not what the quantification shows.

__Response: __We have performed additional western blots and changed the representative images.

Figure 5: 5B- please clarify which ratio is shown here. I assume it is the ratio of RhoA-GTP vs RhoA between the zero and 4 minute time points.

__Response: __Yes, we have clarified this point in the figure legend.

5C- These images appear to have a mask around the cell. It is hard to tell where the edge of the cell really is- what sort of processing was used? Especially for the paxillin staining, why is there no cytoplasm shown? Is this because the image is in TIRF?

__Response: __We used a TIRF microscope and imaged a Z stack. Imaris was used to combine the Z-stack images. The images in the manuscripts are now of the lowest stack with background subtraction.

Figure 6: Figure 6C- the blebbistatin treated cell looks very large- is this representative?

__Response: __Yes, Blebbistatin-treated cells are larger (Fig. 6A).

Figure 7: Fig 7C- The lanes for MLCII are all run together- is this from a different gel? Is this quantification accurate?

__Response: __We have performed additional western blots and changed the representative images.

Fig. 7F- What is the % level of knockdown achieved?

__Response: __The level of knockdown is labeled on the figure.

Figure 8: Fig 8A,B- does the scale bar represent all of the images in these two panels?

__Response: __Yes, the figure legend is updated to clarify this point.

Fig 8C,D- Superplots would be helpful here.

__Response: __We have used super plots to display the data.

Supplementary Data: The OCR data do not add much and are not discussed much in the manuscript. Perhaps they could be omitted.

__Response: __Our OCR data ruled out the possibility of metabolic regulation. Since MFN2 is a mitochondria protein with its typical functions in metabolic pathways, we cannot omit its influence on metabolism here. As we observed, shMLCK enhanced OCR, shROCK reduced OCR, and both knock-down rescued cell morphology and motility. We believe that PAB formation is independent of MFN2’s function in metabolic regulation.

Figure S3- The figure label doesn't match the manuscript test- was fibrinogen or collagen used?

__Response: __We tried cover glass alone, collagen, and fibronectin-coated glass. The PAB formation is independent of these extracellular substrates. We did not try fibrinogen because MEF cell is reported to prefer fibronectin (Lehtimäki et al., 2021).

Reference

Burridge, K., & Chrzanowska-Wodnicka, M. (2003). FOCAL ADHESIONS, CONTRACTILITY, AND SIGNALING. Http://Dx.Doi.Org/10.1146/Annurev.Cellbio.12.1.463, 12, 463–519. https://doi.org/10.1146/ANNUREV.CELLBIO.12.1.463

Carson, S. D., Perry, G. A., & Pirruccello, S. J. (1994). Fibroblast Tissue Factor: Calcium and Ionophore Induce Shape Changes, Release of Membrane Vesicles, and Redistribution of Tissue Factor Antigen in Addition to Increased Procoagulant Activity. Blood, 84(2), 526–534. https://doi.org/10.1182/BLOOD.V84.2.526.526

Chen, H., Detmer, S. A., Ewald, A. J., Griffin, E. E., Fraser, S. E., & Chan, D. C. (2003). Mitofusins Mfn1 and Mfn2 coordinately regulate mitochondrial fusion and are essential for embryonic development. The Journal of Cell Biology, 160(2), 189. https://doi.org/10.1083/JCB.200211046

Cieri, D., Vicario, M., Giacomello, M., Vallese, F., Filadi, R., Wagner, T., Pozzan, T., Pizzo, P., Scorrano, L., Brini, M., & Calì, T. (2017). SPLICS: a split green fluorescent protein-based contact site sensor for narrow and wide heterotypic organelle juxtaposition. Cell Death & Differentiation 2018 25:6, 25(6), 1131–1145. https://doi.org/10.1038/s41418-017-0033-z

García-Calvo, J., López-Andarias, J., Maillard, J., Mercier, V., Roffay, C., Roux, A., Fürstenberg, A., Sakai, N., & Matile, S. (2022). HydroFlipper membrane tension probes: imaging membrane hydration and mechanical compression simultaneously in living cells. Chemical Science, 13(7), 2086–2093. https://doi.org/10.1039/D1SC05208J

Goldfine, S. M., Schroter, E. H., & Izzard, C. S. (1981). Calcium-dependent shortening of fibroblasts induced by the ionophore, A23187. Journal of Cell Science, 50(1), 391–405. https://doi.org/10.1242/JCS.50.1.391

Ikebe, M., & Hartshorne, D. J. (1985). Phosphorylation of Smooth Muscle Myosin at Two Distinct Sites by Myosin Light Chain Kinase*. Journal of Biological Chemistry, 260, 10027–10031. https://doi.org/10.1016/S0021-9258(17)39206-2

Isotani, E., Zhi, G., Lau, K. S., Huang, J., Mizuno, Y., Persechini, A., Geguchadze, R., Kamm, K. E., & Stull, J. T. (2004). Real-time evaluation of myosin light chain kinase activation in smooth muscle tissues from a transgenic calmodulin-biosensor mouse. Proceedings of the National Academy of Sciences of the United States of America, 101(16), 6279–6284. https://doi.org/10.1073/PNAS.0308742101

Kijlstra, J. D., Hu, D., Mittal, N., Kausel, E., van der Meer, P., Garakani, A., & Domian, I. J. (2015). Integrated Analysis of Contractile Kinetics, Force Generation, and Electrical Activity in Single Human Stem Cell-Derived Cardiomyocytes. Stem Cell Reports, 5(6), 1226. https://doi.org/10.1016/J.STEMCR.2015.10.017

Lehtimäki, J. I., Rajakylä, E. K., Tojkander, S., & Lappalainen, P. (2021). Generation of stress fibers through myosin-driven reorganization of the actin cortex. ELife, 10, 1–43. https://doi.org/10.7554/ELIFE.60710

Naon, D., Zaninello, M., Giacomello, M., Varanita, T., Grespi, F., Lakshminaranayan, S., Serafini, A., Semenzato, M., Herkenne, S., Hernández-Alvarez, M. I., Zorzano, A., De Stefani, D., Dorn, G. W., & Scorrano, L. (2016). Critical reappraisal confirms that Mitofusin 2 is an endoplasmic reticulum-mitochondria tether. Proceedings of the National Academy of Sciences of the United States of America, 113(40), 11249–11254. https://doi.org/10.1073/PNAS.1606786113/SUPPL_FILE/PNAS.201606786SI.PDF

Nobes, C. D., & Hall, A. (1999). Rho GTPases Control Polarity, Protrusion, and Adhesion during Cell Movement. The Journal of Cell Biology, 144(6), 1235. https://doi.org/10.1083/JCB.144.6.1235

Pelham, R. J., & Wang, Y. L. (1997). Cell locomotion and focal adhesions are regulated by substrate flexibility. Proceedings of the National Academy of Sciences of the United States of America, 94(25), 13661. https://doi.org/10.1073/PNAS.94.25.13661

Samanas, N. B., Engelhart, E. A., & Hoppins, S. (2020). Defective nucleotide-dependent assembly and membrane fusion in Mfn2 CMT2A variants improved by Bax. Life Science Alliance, 3(5). https://doi.org/10.26508/LSA.201900527

Shao, X., Li, Q., Mogilner, A., Bershadsky, A. D., & Shivashankar, G. V. (2015). Mechanical stimulation induces formin-dependent assembly of a perinuclear actin rim. Proceedings of the National Academy of Sciences of the United States of America, 112(20), E2595–E2601. https://doi.org/10.1073/PNAS.1504837112/SUPPL_FILE/PNAS.1504837112.SM03.AVI

Sidarala, V., Zhu, J., Levi-D’Ancona, E., Pearson, G. L., Reck, E. C., Walker, E. M., Kaufman, B. A., & Soleimanpour, S. A. (2022). Mitofusin 1 and 2 regulation of mitochondrial DNA content is a critical determinant of glucose homeostasis. Nature Communications 2022 13:1, 13(1), 1–16. https://doi.org/10.1038/s41467-022-29945-7

Vallese, F., Catoni, C., Cieri, D., Barazzuol, L., Ramirez, O., Calore, V., Bonora, M., Giamogante, F., Pinton, P., Brini, M., & Calì, T. (2020). An expanded palette of improved SPLICS reporters detects multiple organelle contacts in vitro and in vivo. Nature Communications, 11(1). https://doi.org/10.1038/S41467-020-19892-6

Wong, S. Y., Ulrich, T. A., Deleyrolle, L. P., MacKay, J. L., Lin, J. M. G., Martuscello, R. T., Jundi, M. A., Reynolds, B. A., & Kumar, S. (2015). Constitutive activation of myosin-dependent contractility sensitizes glioma tumor-initiating cells to mechanical inputs and reduces tissue invasion. Cancer Research, 75(6), 1113–1122. https://doi.org/10.1158/0008-5472.CAN-13-3426

Zhou, W., Hsu, A. Y., Wang, Y., Syahirah, R., Wang, T., Jeffries, J., Wang, X., Mohammad, H., Seleem, M. N., Umulis, D., & Deng, Q. (2021). Mitofusin 2 regulates neutrophil adhesive migration and the actin cytoskeleton. Journal of Cell Science, 133(17). https://doi.org/10.1242/JCS.248880/VIDEO-11

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

This study explores the importance of MFN2, a known endoplasmic reticulum-mitochondria linker protein, in cell motility and actin-myosin organization. It is known that the mitofusin proteins MFN1 and MFN2 can tether the mitochondria to the ER and are connected with calcium regulation in muscle and non-muscle cell types. Calcium flux mediated by mitofusins has previously been implicated in apoptosis and ER stress. In this study, the authors show that loss or depletion of MFN2 (but not MFN1) can lead to aberrant calcium increase in the cytoplasm and trigger actin-myosin reorganisation. They show that the actin and myosin changes are linked with activation of RhoA and also that they can be suppressed by inhibiting myosin light chain phosphorylation/ activation. They also show that cells with reduced MFN2 are softer (using atomic force microscopy), which agrees with activation of RhoA and contractility. If cells are plated on softer substrata, it partially compensates for the over-activation of RhoA.

In many places, the data support the claims made, but in several places the experiments could be made more convincing or be more clearly presented. Some of the experiments appear to have been repeated only 1-2 times and very few cells or fields of view have been analysed.

General Comments (Major)

1. Data Presentation and analysis: The data analysis would benefit from using a method such as Super-Plots to show the data from separate biological repeats and to use N numbers that represent the number of biological repeats rather than the number of cells analysed. Please see the following reference for a suggestion on how to analyse the data: Lord SJ, Velle KB, Mullins RD, Fritz-Laylin LK. SuperPlots: Communicating reproducibility and variability in cell biology. J Cell Biol. 2020 Jun 1;219(6):e202001064. doi: 10.1083/jcb.202001064. PMID: 32346721; PMCID: PMC7265319.
2. Another general comment is that many of the experiments show analysis of very few numbers of cells or maybe only one field of view in a microscopy quantification experiment. This seems unusually low - for example, in Figure 1E only 6 cells have been analysed. It seems like more could have been done and if statistical analysis like we suggest in 1 above is used, this might reveal that some of the differences are less significant than the authors think/report. This is important, as cells are noisy and it is unusual to have such high significance for experiments like cell migration and other parameters unless a lot of measurements are made. In Figure 1G, it appears that only one field of view has been used to quantify the data. We routinely use 3-5 fields of view to get a representative sample of what the cells are doing.
3. Some of the micrographs appear to be missing scale bars- e.g. Fig. 2H, Fig. 8
4. In the cell tracking experiments, only some of the cells in the images appear to have been tracked. How were the tracked cells chosen? Normally, we would track every cell to avoid bias in selection.
5. The western blot images do not show the molecular size of the bands.
6. Mostly the graphs show individual data points, which is good, but in some cases only a bar is shown- it would be nice to have individual points overlaid on the bars- e.g. Figure 1I, 4E, 5B, 5E, 7B, 7D
7. Many of the confocal images look very processed- they have no background and have a hazy black halo around the cell. I am not familiar with the type of processing that was done and I worry that the images are only showing a masked and processed version of the actual data. The authors need to explain what processing they have done and probably also to provide the unprocessed images in a supplementary figure or dataset for readers to see. The methods description is inadequate as it only says Image J was used to process the data.

Individual comments on Figures:

Figure 1: See general comments above- consider to use Superplots, more cells and more fields of view in quantifications. Show experimental points in bar graph.

Figure 2: In 2E, the colours are very similar for two of the experiments so it is difficult to distinguish them- e.g. the MFN1 vs MFN2 rescue data both appear dark blue. In 2H are the magnifications really the same for the WT as the +DOX and -DOX? The cell in the WT looks huge. Is this representative? Also, the phalloidin stain looks very spotty on the WT. This seems unusual.

Figure 3: Not many cells were analysed in 3B, especially the zero time point. Please define +T, we assume it is the tether construct, but it is not defined In 3F, how were the tracked cells chosen?

Figure 4: 4B: Why have they not tested FK506 and STO609 on the WT cells? 4C: How were the tracked cells chosen? 4D-E: The blot doesn't look representative of the quantification- were the numbers normalised to vinculin? The difference between WT and vector looks too large to be real if the amounts were normalised to the vinculin, as vinculin is increased in vector. Likewise, the pCAMKII looks to be substantially decreased from the +MFN2, but this is not what the quantification shows.

Figure 5: 5B- please clarify which ratio is shown here. I assume it is the ratio of RhoA-GTP vs RhoA between the zero and 4 minute time points. 5C- These images appear to have a mask around the cell. It is hard to tell where the edge of the cell really is- what sort of processing was used? Especially for the paxillin staining, why is there no cytoplasm shown? Is this because the image is in TIRF?

Figure 6: Figure 6C- the blebbistatin treated cell looks very large- is this representative?

Figure 7: Fig 7C- The lanes for MLCII are all run together- is this from a different gel? Is this quantification accurate? Fig. 7F- What is the % level of knockdown achieved?

Figure 8: Fig 8A,B- does the scale bar represent all of the images in these two panels? Fig 8C,D- Superplots would be helpful here.

Supplementary Data:

The OCR data do not add much and are not discussed much in the manuscript. Perhaps they could be omitted. Figure S3- The figure label doesn't match the manuscript test- was fibrinogen or collagen used?

Significance

The main novelty here appears to be the connection between excess cytoplasmic calcium, MFN2 loss and RhoA/myosin activation. This is interesting and a useful addition to the literature. It is unclear what the significance is perhaps, as increased cytoplasmic calcium is likely to cause multiple effects in addition to these. So, this effect may be a side-effect of uncoupling the ER and the mitochondria. Nonetheless, it is important to know about this and it will likely inform future studies on the mitofusins.

This will be of interest to basic researchers studying mitochondria function and the connections between signaling and mitochondria function.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

In this manuscript, Wang et al. investigate the role of Mitofusin 2 (Mfn2) in MEF morphology and migration. MEFs lacking Mfn2 exhibit a rather circular shape and reduced migratory potential due to increased cytosolic Ca2+, which activates Ca2+/calmodulin-dependent protein kinase II, RhoA and myosin light-chain kinase (MLCK), triggering non-muscle myosin II overactivation and formation of an F-actin ring at the cell periphery. Overall, this is an interesting study. However, additional work is needed for this manuscript to reach publication quality.

Major Concerns:

• Several of the Western blots are not of publication quality. This may affect their quantification, and the authors' conclusions.
• The authors should examine how treatment of wt MEFs with Rho Activator and MLCK-CA affect cell motility (point #10).
• The reviewer is not convinced that the membrane tension measurements were performed correctly (see point #12 below). The reviewer would expect that the presence of an F-actin ring at the cell periphery would increase the membrane tension, which is opposite to the authors' findings.

Specific Comments:

1. Fig. 1A: The Mfn1 Western Blot is not of publication quality. Moreover, quantitation is necessary.
2. Fig. 1B (as well as Fig. 2G and others): the date do not reflect cellular size but instead spread cellular area.
3. Fig. 1C, D: Mfn1-null MEFs appear to be more spindle-shaped than wt cells, yet their circularity tends to be elevated. Do the authors have an explanation?
4. Fig. 2A: The Mfn1 levels in Mfn2-/- + Mfn2 are lower than Mfn2-/-? Does this imply a crosstalk between Mfn1 and Mfn2 expression.
5. Fig. 2H: The authors should provide co-staining of mitochondria and Mfn2 as well.
6. Fig. 4D-E: Western blots are not of publication quality. Looking at the blots provided in Fig. 4D, the reviewer is not convinced with the quantitative data shown in Fig. 4E. For instance, the intensity of pCaMKII band for "vec" does not look 3x higher than that of "+MFN2", whereas that of "+MFN2" looks much higher than that of WT.
7. Fig. 5C: The authors should stain for vinculin, which are present in mature FAs only, rather than paxillin which are present in all FAs. This would strengthen the authors' conclusions. Also, FA size should be quantified.
8. Fig. 6C - Why does the background have a grid and appear grey in color? Also, the cell interior appears in different colors in the different images. The authors should take a z-stack of images and provide the raw image files.
9. Fig 7C: The MLCII Western blot is not of publication quality, and may affect the quantification provided in Fig. 7D.
10. Fig 8: Do cell treatments with Rho Activator and MLCK-CA also impair migration velocity similar to Mfn2-null cells?
11. Fig. 9A: The authors claim that wt cells have actin bundles that protrude against the membrane while Mfn2-null cells do not. This does not look convincing as the Mfn2-null actin bundle seem to be pushing against the membrane at the bottom of the image. No quantification is provided. It is unclear what conclusion can be drawn from the super-resolution images.
12. Suppl. Fig. 5C: The authors should take images using a confocal microscope for cells with Flipper-TR construct, eliminate the background and the cell center to only consider the cell periphery. Nikon TE2000 does not seem to be a confocal microscope.

Significance

Delineating the role of Mfn2 in cell migration will represent a good, fundamental contribution to the field of cell migration. The reviewer finds this manuscript to be conceptually interesting. Unfortunately, there are technical issues, which need to be fixed so that the readers feel confident about the conclusion of this study.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Summary:

Wang et al use a combination of cell biological and biochemical approaches to show that Mitofusin2 (MFN2) - a protein typically known to regulate mitochondrial morphology - regulates cell morphology by regulating calcium levels and downstream cell contractility players. They show that cells deficient in MFN2 exhibit increased intracellular calcium levels, and overactivation of myosin II, leading to increased cell contractility. Furthermore, MFN2-deficient cells exhibit an F-actin ring around the cell periphery (which they call PABs). The study takes advantage of both pharmacological and genetic perturbations, as well a variety of assays to support many of their findings; however, it is unclear how these findings are related to each other. Furthermore, MFN2-related disease was raised a few times in the abstract and throughout the manuscript, but it's unclear how the findings in the paper relate to disease states, both in terms of the cell biology, as well as the model that was used (MEFs). This reviewer applauds the authors for exploring MFN2 function outside of its conventional role in mitochondria; but it was difficult to parse through the findings to resolve a mechanistic explanation for how MFN2 affect cell behavior, and what role, if any, PABs have on biological function. While it is important to dissect mitochondrial-independent functions for MFN2, given the whole scale changes in cells in MFN2-deficient cells, and the fact that there is a metabolic phenotype, it is difficult to know how many of the observed phenotypes are downstream of perturbed mitochondrial function versus on cytoskeletal dynamics directly.

Major comments:

Are the key conclusions convincing?

• Key conclusions that were convincing:
  • MFN2-/- cells exhibit decreased cell velocity, decreased cell size, increased cell circularity, and increased intracellular calcium
  • modifying the levels of calcium has an effect on cell circulariy.
  • MFN2-/- cells exhibit increased activation of contractility players
• Key conclusions that were less convincing:
  • RhoA and NMII are in the title as mechanistic downstream regulators of CaM, but the results in Fig 8 call into question the role of RhoA. Why does RhoA activation not influence cell size and circularity? Can you overexpress MRLC-GFP and inhibit Rho and restore the wt phenotype? The role of NMII is also not clear - why does overexpressing MLCK-CA not have a phenotype but overexpressing MLCK downstream target MRLC show the phenotype? Are there any alternative pathways to regulate MRLC? It's not being discussed or described in 6D's schematic.
  • The role of ER/mito contacts in the system was unclear (since ER/mito contacts were not observed nor evaluated directly).
  • What role does focal adhesions have on PAB formation or any part of the model - There were results showing larger focal adhesions in the MFN2-/- cells, but not sure how this fits in with the bigger picture of contractility and PABs, and focal adhesions were not in the model in Figure 5.
  • Whether regulating calcium impacts PAB formation
  • The role of PABs in migration is also unclear - can you affect PAB formation or get rid of PABs and quantify cell migration?
  • It was hard to understand the correlation between the membrane tension of MFN2-/- cells and its ability to spread on softer substrates. How does this result fit in with the overarching model?

Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?

• There were a number of findings that did not seem to fit in with the paper, or they were included, but were not robustly described nor quantified. As an example, Paxillin-positive focal adhesions were evaluated in MFN2-/- and with various pharmacological approaches, but there is no quantification with respect to size, number, or distribution of focal adhesions, despite language in the main text that there are differences between conditions. Also, the model was presented in figure 5, and then there were several pieces of data presented afterward, some that are included in the model (myosin regulation), and some that are not in the model (membrane tension, TFM, substrate stiffness, etc). The stiffness/tension figure was not clear to me, and it was difficult to make sense of the data since one would predict that an increase in actomyosin contractility at the cortex would lead to higher membrane tension, not lower, and then how membrane tension relates to spreading on soft matrices is also unclear. The manuscript seems like an amalgamation of different pieces of data that do not necessarily fit together into a cohesive story, so the authors are encouraged to either remove these data, or shore them up and weave them into the narrative.

Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.

• The imaging used for a lot of the quantification (migration, circularity) is difficult to resolve. The cells often look like they are not imaged in the correct imaging plane. It would be helpful to have better representative images such that it is clear how the cells were tracked and how cell periphery regions of interest were manually drawn. Focal adhesions should also be shown without thresholding.
• For most of the quantification, it appears that the experiment was only performed once and that a handful of cells were quantified. The figure legend indicates the number of cells (often reported as >12 cells or >25 cells), but the methods indicate that high content imaging was performed, and so the interpretation is that these experiments were only performed once. Biological replicates are required. If the data do represent at least 3 biological replicates, then more cell quantification is required (12 or 25 cells in total would mean quantifying a small number of cells per replicate).
• Mitochondrial morphology and quantification should be performed in the MFN2 knockdown and rescue lines.
• Many of the comparisons throughout the figures is between MFN2 knockdown and MFN2 knockdown plus rescue or genetic/pharmacological approaches, but a comparison that is rarely made is between wildtype and experimental. These comparisons could be useful in comparing partial rescues and potential redundancies with the other mitofusin.
• For the mito/ER tethering experiments, it is important to show that ER/mito contacts are formed and not formed in the various conditions with imaging approaches
• For some of the pharmacological perturbations, it would be helpful to show that the perturbation actually led to the expected phenotype - as an example, in cells treated with different concentrations of A23187, what are the intracellular calcium levels and how do these treatments influence PAB formation? This aspect should be generally applied across the study - when a modification is made, that particular phenotype should first be evaluated, before dissecting how the perturbation affects downstream phenotypes.
• In Figures 4 and 5, the thresholding approaches in the images make the focal adhesions difficult to resolve, and therefore it is difficult to determine the size. As described above, these metrics should be defined and quantified.
• What is a PAB? How is it defined? What metrics make a structure a PAB versus regular cortical actin - are there quantifiable metrics? In figure 8, there are some structures that are labelled as a PAB, but some aren't (as an example, the left panel in 8b is a PAB, but the right panel in 8A is not, but they look the same), so a PAB should be defined with quantifiable measures, and then applied to the entire study.
• As described above, why does RhoA activation not influence cell size and circularity? Can you overexpress MRLC-GFP and inhibit Rho and restore the wildtype phenotype?

Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.

• These suggested experiments described above will take a substantial amount of time and money, as it appears that some experiments were only performed once, and therefore many of these experiments might need to be performed 2-3 more times. Also, experiments showing that addition of drugs lead to expected outcomes prior to analyzing downstream phenotypes will also require a significant amount of time. It is hard to predict how long it will take, but we would guess, 6-8 months?

Are the data and the methods presented in such a way that they can be reproduced?

• We appreciate that the authors quantified many parameters, although some quantifications were missing. There are some missing methods - how was directionality quantified, was migration quantified by selecting the approximate center of cells using MTrackJ or were centroids quantified by outlining cells, for instance. Also, given that some of the phenotypes were somewhat arbitrarily assigned (ie. what constitutes a PAB?), it may be difficult to reproduce these approaches and interpret data appropriately.

Are the experiments adequately replicated and statistical analysis adequate?

• Unfortunately, while the approximate number of cells was reported, the number of biological replicates were not reported, and therefore, the experimental information and statistical analyses are not adequate.

Minor comments:

Specific experimental issues that are easily addressable. - for some of the graphs - mostly about calcium levels - fold change is shown, but raw values should also be included to determine whether the basal levels of calcium are different across the conditions. - scale bars in every panel should also help make the points clearer.

Are prior studies referenced appropriately?

• Yes

Are the text and figures clear and accurate?

• Some of the data in the figures were unclear - see above for more info.

Do you have suggestions that would help the authors improve the presentation of their data and conclusions?

• See above for more info.

Significance

Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field.

• There is a growing field of mitochondrial biology and how it relates to cell migration. This paper examines the function of a key mitochondrial morphology regulator, MFN2, and dissects a role for MFN2 in migration at the level of cytoskeletal regulation. We think that this is interesting, and that it's clear that MFN2 has multiple functions in the cell, but the phenotypes are so pleiotropic that it's difficult to parse out mechanistic understanding. The authors also describe a new actin architecture - a structure that they refer to as PABs - but there is no indication that PABs form in other cell types or tissues, or in other contexts, so it is unclear whether PABs are an important structure or an artifact of the system. Furthermore, part of the motivation of the work seems to be to understand MFN-related pathologies, but using a MEF system does not necessarily allow for that. One way to strengthen this part of the manuscript is to potentially use disease-relevant MFN2 mutations and determine downstream effects on cell morphology and migration.

State what audience might be interested in and influenced by the reported findings.

• This work would appeal to card-carrying cell biologists.

Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate.

• cell migration; actin; mitochondria

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

We would like to thank all the reviewers for their positive evaluations of our work and constructive comments, in particular for highlighting that our work “provides new insight into cancer metabolism knowledge”, is “conceptually interesting and experimentally well performed” and “the findings presented here will be very interesting to a broad range of researchers, including the cancer, metabolism and wider cell biology communities”.

Reviewer #1 (Evidence, reproducibility and clarity (Required)): Nazemi et al. show that the extracellular matrix (ECM) has a crucial role in sustaining the growth of invasive breast and pancreatic cells during nutrient deprivation. In particular, under amino acid starvation, cancer cells internalize ECM by macropinocytosis and activate phenylalanine and tyrosine catabolism, which in turn support cell growth in nutrient stress conditions. The paper is well written and the results shown are very interesting. The experimental plan is well designed to assess the hypothesis and the description of the methods is sufficiently detailed to reproduce the analyses, which are also characterized by appropriate internal controls. Finally, the data provided sustain the conclusions proposed by the authors.

* Major comment:*

Since the authors performed their experiments on invasive breast and pancreatic cancers and it has been noted that stress conditions could promote the escape of cancer cells from the site of origin (e.g., Jimenez and Goding, Cell Metabolism 2018; Manzano et al, EMBO Reports 2020), it would be interesting to evaluate how ECM internalization could have a role in sustaining the invasive abilities of cancer cells under amino acid starvation. Which is the impact of the inhibition of macropinocytosis and tyrosine catabolism on cell invasion? The authors could evaluate this aspect by in vitro 2D and 3D analysis.

This is a very important point, and we are planning to investigate this by using:

• 2D single cell migration assays on cell-derived matrices (we have extensively used this system to characterise invasive cell migration; Rainero et al., 2015; Rainero et al., 2012)
• 3D spheroids assays, to assess collective/3D cell invasion through collagen I and matrigel mixtures. Both experiments will be performed under amino acid starvation, in the presence of pharmacological inhibitors and siRNAs targeting macropinocytosis (FRAX597, PAK1) and tyrosine catabolism (Nitisinone, HPDL). Preliminary data suggest that both FRAX597 and Nitisinone reduce cell invasiveness.

In addition, to strengthen the paper and give a stronger significance in terms of clinical translatability, it could be useful to implement the analysis of breast and pancreatic patients by publicly dataset evaluating for example free survival, disease free survival, overall survival and metastasis free survival.

We have now included in the manuscript new data in figure 6 O-R showing that high HPDL expression correlates with reduces overall survival, distant metastasis-free survival, relapse-free survival and palliative performance scale in breast cancer patients. In response to other reviewers’ comments, we have removed the pancreatic cancer data from our manuscript.

Minor Comment:

The text and the figures are clear and accurate. The references cited support the hypothesis, rightly introduce the results and are appropriate for the discussion. However, the paragraph relative to figure 4 is a little confusing. Changing the order of the description of the results could be useful.

We apologise for the lack of clarity in this section. We have now re-organised the data both in the figure and in the result section, to describe the findings in a more logical way.

Reviewer #1 (Significance (Required)): Based on my metabolic background in tumour aetiology and progression, I think that this study provides new insights into cancer metabolism knowledge, in particular on how the stroma may drive metabolic reprogramming of cancer cells sustaining cell growth in nutrient stress conditions. Together with other similar studies on the stromal non-cellular components, the data here shown can contribute to expand the knowledge on the factors that promote cancer metabolic plasticity, which is exploited from cancer cells to obtain advantages in terms of growth, survival and progression. In conclusion, I think that the results shown are new and the manuscript is well presented. Following the short revision process suggested, it will be eligible for a final publication in a medium-high impact factor journal.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

• Please find enclosed my reviewing comments on the manuscript entitled "The extracellular matrix supports cancer cell growth under amino acid starvation by promoting tyrosine catabolism" by Nazemi et al.*

In this manuscript the group of Elena Romero and colleagues provides evidence that breast cancer cells, and pancreatic cancer cell, use matrix proteins degradation to feed their proliferative metabolic needs under amino acid starvation. Under this drastic condition, cancer cells use micropinocytosis to uptake matrix proteins, a process that requires mTORC1 activation and PAK1. Furthermore, a metabolomic study demonstrates that ECM-dependent cancer cell growth relies on tyrosine catabolism. Altogether, I found this study conceptually interesting and experimentally well performed. Experiments are well controlled and state-of-the-art technologies used in this manuscript make it a good candidate for publication. However, some aspects of the work need to be strengthen to reinforce the overall good quality of the manuscript, therefore, please find below some experimental propositions. 1. Despite the reviewer proposition, I believe that the additional experiments using the PDAC cancer cell does not improve the quality of the manuscript. Instead, it brings confusion to me, since the relative addition is minor compare to what is demonstrated using breast cancer cells.

We have decided to remove the pancreatic cancer cell data from the manuscript.

To importantly improve the potential impact of this manuscript, I suggest to add in vivo data using either syngenic mice model of breast cancer or xenografted human breast cancer cells in nude mice. What would be the impact of micropinocytosis and tyrosine catabolism inhibition on cancer growth, in vivo, should be demonstrated? If possible, it may be interesting to demonstrate that this micropinocytosis may interfere with cancer evolution toward a metastatic phenotype using, for example, the PyMT-MMTV mice model of breast cancer development?

We will perform orthotopic mammary fat pad injections in immunocompetent mice, to monitor primary tumour growth and localised invasion in the presence of Nitisinone or vehicle control. PyMT-driven breast cancer cells have been generated in the Blyth lab, from FVB-pure MMTV-PyMT mice and we have preliminary data indicating that these cells are able to internalise ECM and grow under starvation in an ECM-dependent manner. Prior to performing any in vivo work, we will perform further in vitro experiment to confirm the role of tyrosine catabolism in these cells. Nitisinone is an FDA-approved drug that has already been used in mouse models. Blood tyrosine levels can be measured to assess tyrosine catabolism inhibition by Nitisinone. These experiments will be conducted in collaboration with the Blyth lab at the CRUK Beatson Institute in Glasgow.

Data obtained using cancer cells with different metastatic property suggest that the ability to use ECM to compensate for soluble nutrient starvation is acquired during cancer progression. To further demonstrate that it is the case, would it be possible that non metastatic breast cancer cells are not able to perform micropinocytosis? Is PAK1 overexpressed with increase cancer cells metastatic ability, without affecting invasive capacity in 3D spheroids?

To address these points, we have started to measure PAK1 expression across the MCF10 series of cell lines, where MCF10A are non-transformed mammary epithelial cells, MCF10A-DCIS are ductal carcinoma in situ cells and MCF10CA1 are metastatic breast cancer cells. Our preliminary data show that there is no upregulation of PAK1 expression in the metastatic cells compared to non-transformed or non-invasive cancer cells. This suggest that the over-expression of PAK1 might not be a valuable strategy to address this point.

In addition, we found that collagen I uptake was upregulated in MCF10CA1 compared to MCF10A and MCF10A-DCIS. We will corroborate our preliminary data by quantifying collagen I and cell-derived matrices internalisation across the 3 cell lines.

What would be the efficacy to promote the ECM-dependent growth under starvation following a mTORC1 in non-invasive cancer cells?

We will measure the growth of MCF10A and MCF10A-DCIS on ECM under starvation in the presence of the mTOR activator MHY1485. Western blot analysis of downstream targets of mTORC1 (p-S6 and p-4EBP1) will confirm the extent of mTOR activation.

The discrepancy of cancer cells proliferation under starvation condition between plastic and ECM-based supports could be explained by the massive difference of support rigidity. This is also probably the case between CDM made by normal fibroblast and CAF. It brings the question of studying the role of matrix stiffness in regard to micropinocytosis-dependent cancer cells growth. It would also explain why this process is link to aggressive cancer cell behaviour, as ECM goes stiffer with time in cancer development. It may not be the case, but the demonstration that mechanical cues from the ECM could regulate the micropinocytosis-dependent cancer cells growth under amino acid starvation could bring additional value to the manuscript.

We will use 2 experimental approaches to address the effect of different stiffness in ECM-dependent cell growth:

1. Polyacrylamide hydrogels coated with different ECM components.
2. Collagen I gels in which the stiffness is modified by Ribose treatment (this approach has been published by the Parson’s lab). Our preliminary data confirmed that ribose cross-linking increased YAP nuclear localisation and collagen I can still be internalised under these conditions. We will assess ECM endocytosis and cell growth under starvation conditions (using EdU incorporation in conjunction with A and high throughput imaging with B)

Along with this, it has been demonstrated that matrix rigidity regulates glutaminolysis in breast cancer, resulting in aspartate production and cancer cells proliferation. Is asparate production increase by micropinocytosis? Could you rescue cancer cells growth by aspartate supplementation?

Our metabolomics experiments were performed under amino acid starvation; therefore glutamine was not present in the media. Nor glutaminolysis intermediates nor aspartate were upregulated on ECM compared to plastic in our datasets, suggesting that aspartate might not be involved in this system. We added this point in the discussion. However, glutamine, glutamate and aspartate were found to be upregulated on collagen I compared to plastic in complete media, where the most enriched pathway was alanine, aspartate and glutamate metabolism. Future work will address the role of the ECM in supporting cancer cell metabolism in the absence of nutrient starvation.

Data presented in Fig 1 and SF1 show that breast cancer cell lines growth in a comparable manner either they are cultured on plastic or on 3D ECM substrates in complete media. Again, on thick 3D substrates, in which the stiffness is lower compared to plastic, I would have thought that cancer cells would have grown slower. Could you please discuss this finding in regard to the literature?

Our experiments in full media were performed in the presence of dialysed serum, to represent a better control for the starvation conditions, which were in the presence of dialysed serum. This is consistent with a vast body of literature assessing nutrient starvation conditions in the presence of dialysed serum. This could explain the discrepancy between ours and published results. We have addressed this point in the discussion.

If you have the capacity to do so in your lab or in collaboration, would it be possible to measure the exact stiffness of the different matrix you use in this manuscript? Or using hydrogel, would it be possible to study the role of matrix stiffness in the ECM-dependent cancer cells growth under AA starvation? I would understand that this may be out of the scope of the present manuscript, but I again believe that such finding would reinforce the manuscript.

We don’t have the capacity to measure the stiffness in our lab, however NF-CDM and CAF-CDM, generated by the same cells and using the same protocol, have been previously measured at ~0.4kPa and ~0.8 kPa, respectively (Hernandez-Fernaud et al., 2017). We have now included this in the paper. As mentioned in response to point 4, we will use hydrogels to directly test the effect of matrix stiffness on ECM-dependent cell growth under nutrient starvation.

In SF 3A-C, it is shown that ECM does not affect caspase-dependent cell death under AA starvation. Did you considered a non-caspase dependent cell death that may be triggered by AA starvation?

We will complement the caspase 3/7 data by performing PI staining, to detect all forms of cell death. Preliminary data indicate that, consistent with our cas3/7 data, amino acid starvation promotes cell death, but the presence of the ECM doesn’t affect the percentage of PI positive cells, corroborating our conclusions that the ECM modulates cell proliferation and not cell death. We will complete these experiments in both MDA-MB-231 and MCF10CA1 cells and will include them in figure S3.In fig 5, it is shown that inhibition of Focal Adhesion Kinase (FAK) does not impair the ECM-dependent rescue of cancer cell growth under starvation. To further decipher the concept of adhesion dependent signalling, maybe the authors could also inhibit the Src kinase or ITG-beta1 activation?

Integrin b1 is also required for ECM internalisation (our unpublished data), therefore interfering with integrin function would make the interpretation of the data quite complex. As suggested by the reviewer, we will use the Src inhibitor PP2, which has been extensively used in the literature in MDA-MB-231 cells. Preliminary data indicate that, despite significantly reducing cell proliferation in complete media, Src inhibition does not affect cell growth on collagen I under amino acid starvation, consistent with our FAK inhibitor data. We will complete these experiments on both collagen I and cell-derived matrices and will include them in figure 5.

Minor comment, in F1B, it is written "AA free starvation" while in every others legend, it is written "AA starvation". I believe the "free" should be removed.

We apologise for this mistake; we have now removed “free” from the legend.

Reviewer #2 (Significance (Required)): Altogether, I found this study conceptually interesting and experimentally well performed. Experiments are well controlled and state-of-the-art technologies used in this manuscript make it a good candidate for publication. However, some aspects of the work need to be strengthen to reinforce the overall good quality of the manuscript.

Reviewer #3 (Evidence, reproducibility and clarity (Required)): In this manuscript the authors explore the mechanisms that metastatic cancer cells use to adapt their metabolism. The authors show that the growth of cancer cell lines is supported by uptake of ECM components in nutrient-starved conditions. The authors propose a very interesting mechanism in which the cells adapt their metabolism to ECM uptake as nutrient source via a PAK1-dependent macropinocytosis pathway which in turn increases tyrosine catabolism. Several key aspects of the authors complex hypothesis require further controls to fully support the authors ideas. As a disclaimer we do not feel qualified enough to comment on the metabolite experiments. Please find our detailed comments below.

* Major -The ECM mediated increase of cell growth under amino acid (AA) starvation is nicely shown In Fig.1 but the authors should include the full medium data from figure S1 in the graphs of Fig. 1 to enable the reader to evaluate the magnitude of rescue effect of the ECM components. The values should also be included in the results text*.

We have now moved all the complete media data into the main figure and highlighted the extent of the rescue in the result section.

Also the authors only glutamine starve in Fig1&2 and then don't mention it again can the authors please include a sentence to explain why this experiment was dropped.

As now highlighted in the result section, we focused on the amino acid starvation as it resulted in the strongest difference between normal and cancer. On the one hand, also non-invasive breast cancer cells can use ECM (namely matrigel) to grow under glutamine starvation, while this is not the case under amino acid starvation. On the other hand, only CAF-CDM, but not normal-CDM, could rescue cell growth under amino acid starvation. We reasoned that this condition was more likely to identify cancer-specific phenotypes.

- The evaluation of uptake pathways is very interesting. The focus on macropinocytosis is not entirely justified in our opinion looking at FigS4A. Caveolin1/2 and DNM1/3 seem to have strongest effect on uptake of Matrigel and not PAK1? Statements like "Since our data indicate that macropinocytosis is the main pathway controlling ECM endocytosis..." are not justified nor are they really needed in our opinion. Several pathways can be implicated in passive uptake.

We have now removed the statement, as suggested by the reviewer. In addition, we will assess CDM uptake upon caveolin 1/2 and DNM 2/3 knock-down, to test whether the effects are matrigel specific.

- The authors use FAK inhibition to evaluate the effect of focal adhesion signalling on their phenotypes and conclude that there is no connection between the observed increase of cell proliferation in presence of ECM and adhesion signalling. To make this assessment the authors need at the very least to show that their FAK inhibitor treatment at the used concentration results in changes in focal adhesions and the associated force transduction.

In the result section, we are including a western blot analysis showing that the concentration of FAK inhibitor used in sufficient so significantly reduced FAK auto-phosphorylation. Based on published evidence (Horton et al., 2016), FAK inhibition does not affect focal adhesion formation, but only the phosphorylation events. Therefore, we don’t think that we will be able to detected changes in focal adhesions regardless of the concentration of the inhibitor we use. To strengthen the observation that ECM-dependent cell growth in independent from adhesion signalling, as suggested by reviewer #2, we will use the Src inhibitor PP2, which has been extensively used in the literature in MDA-MB-231 cells. Preliminary data indicate that, despite significantly reducing cell proliferation in complete media, Src inhibition does not affect cell growth on collagen I under amino acid starvation, consistent with our FAK inhibitor data. We will complete these experiments on both collagen I and cell-derived matrices and will include them in figure 5.

-The pancreatic cancer data currently feels a bit like an afterthought. We suggest to remove this data from the manuscript. If this data is included we suggest the authors should expand this section and repeat key experiments of earlier figures.

We have now removed these data from the manuscript, as this was also the suggestion of reviewer #2.

-Was the fetal bovine serum (FBS) and Horse Serum (HS) the authors use in their experiments tested for ECM components? The authors mention that the FBS for MDA231 cells was dialysed but not the HS.

HS was used at a much lower concentration that FBS in our cell proliferation experiments (2.5% compared to 10%). We will characterise both sera components by mass spectrometry analysis, in collaboration with Dr Collins, biOMICS Facility, University of Sheffield.

Minor comments:

-Please can the authors provide experimental data directly comparing NF-CAM versus CAF-CDM on the same graph (Figure 1D-E).

In the experiments included in the manuscript, the two matrices were generated independently, and we don’t feel it is appropriate to combine the results in the same graph. We are now repeating these experiments by generating both matrices in the same plates, so that we can present the data in the same graph. -Please can the authors give more insight to the use of 25% Plasmax to mimic starved tumor microenvironment. Is there previous research that suggests the nutrient values are representative of TME?

Apologies for not clarifying this in the initial submission, the rationale behind this choice is based on the observation that, in pancreatic cancers, nutrients were shown to be depleted between 50-75% (Kamphorst et al., 2015). We have now stated this in the result section.

-Fig3E Can the authors please include example images of the pS6 staining in the supplementary figures and explain "mTOR endosomal index" in figure legend.

We have now included the representative images (new figure 3E) and we have described how the mTOR endosomal index was calculated both in the figure legend and in the method section. -Can the authors include a negative control for the mTORC1 localisation in Fig.3 (such as use of rapamycin/Torin)?

Amino acid starvation is the gold-standard control for mTORC1 lysosomal targeting, as described in a variety of publications, including Manifava et al., 2016; Meng et al., 2021; Averous et al., 2014. In addition, Torin 1 treatment has been shown to result in a significant accumulation of mTOR on lysosomes compared with untreated cells (Settembre et al., 2012). Consistent with this, we looked at mTOR localisation in the presence of Rapamycin and we did not detect any reduction in lysosomal targeting.

- The PAK1 expression level blots in the knockdown experiments should be quantified from N=3.

We have not included the quantification of the western blots in the new supplementary figure 5.

-What is the FA index in Fig.5, explain how it is calculated. Why not use FA size alone?

We have now defined this is the method section. We haven’t used FA size alone, as this measure can be affected by cell size. If a cell is bigger, the overall FA size will be bigger, but this doesn’t necessarily reflect a change in adhesions.

-Can the authors please use paragraphs on page 9 to improve readability. We apologise for overlooking this, we have now used paragraph in this section.

Reviewer #3 (Significance (Required)): The findings presented here will be very interesting to a broad range of researchers including the cancer, metabolism and wider cell biology communities. The Rainero lab has progressed the idea that ECM uptake supports cancer progression and the data presented here has the potential to significantly advance our understanding of the underlying cellular mechanisms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In this manuscript the authors explore the mechanisms that metastatic cancer cells use to adapt their metabolism. The authors show that the growth of cancer cell lines is supported by uptake of ECM components in nutrient-starved conditions. The authors propose a very interesting mechanism in which the cells adapt their metabolism to ECM uptake as nutrient source via a PAK1-dependent macropinocytosis pathway which in turn increases tyrosine catabolism. Several key aspects of the authors complex hypothesis require further controls to fully support the authors ideas. As a disclaimer we do not feel qualified enough to comment on the metabolite experiments. Please find our detailed comments below.

Major

• The ECM mediated increase of cell growth under amino acid (AA) starvation is nicely shown In Fig.1 but the authors should include the full medium data from figure S1 in the graphs of Fig. 1 to enable the reader to evaluate the magnitude of rescue effect of the ECM components. The values should also be included in the results text. Also the authors only glutamine starve in Fig1&2 and then don't mention it again can the authors please include a sentence to explain why this experiment was dropped.
• The evaluation of uptake pathways is very interesting. The focus on macropinocytosis is not entirely justified in our opinion looking at FigS4A. Caveolin1/2 and DNM1/3 seem to have strongest effect on uptake of Matrigel and not PAK1? Statements like "Since our data indicate that macropinocytosis is the main pathway controlling ECM endocytosis..." are not justified nor are they really needed in our opinion. Several pathways can be implicated in passive uptake.
• The authors use FAK inhibition to evaluate the effect of focal adhesion signalling on their phenotypes and conclude that there is no connection between the observed increase of cell proliferation in presence of ECM and adhesion signalling. To make this assessment the authors need at the very least to show that their FAK inhibitor treatment at the used concentration results in changes in focal adhesions and the associated force transduction.
• The pancreatic cancer data currently feels a bit like an afterthought. We suggest to remove this data from the manuscript. If this data is included we suggest the authors should expand this section and repeat key experiments of earlier figures.
• Was the fetal bovine serum (FBS) and Horse Serum (HS) the authors use in their experiments tested for ECM components? The authors mention that the FBS for MDA231 cells was dialysed but not the HS.

Minor comments:

• Please can the authors provide experimental data directly comparing NF-CAM versus CAF-CDM on the same graph (Figure 1D-E)
• Please can the authors give more insight to the use of 25% Plasmax to mimic starved tumor microenvironment. Is there previous research that suggests the nutrient values are representative of TME?
• Fig3E Can the authors please include example images of the pS6 staining in the supplementary figures and explain "mTOR endosomal index" in figure legend.
• Can the authors include a negative control for the mTORC1 localisation in Fig.3 (such as use of rapamycin/Torin)?
• The PAK1 expression level blots in the knockdown experiments should be quantified from N=3
• What is the FA index in Fig.5, explain how it is calculated. Why not use FA size alone?
• Can the authors please use paragraphs on page 9 to improve readability.

Significance

The findings presented here will be very interesting to a broad range of researchers including the cancer, metabolism and wider cell biology communities. The Rainero lab has progressed the idea that ECM uptake supports cancer progression and the data presented here has the potential to significantly advance our understanding of the underlying cellular mechanisms.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Please find enclosed my reviewing comments on the manuscript entitled "The extracellular matrix supports cancer cell growth under amino acid starvation by promoting tyrosine catabolism" by Nazemi et al.

In this manuscript the group of Elena Romero and colleagues provides evidence that breast cancer cells, and pancreatic cancer cell, use matrix proteins degradation to feed their proliferative metabolic needs under amino acid starvation. Under this drastic condition, cancer cells use micropinocytosis to uptake matrix proteins, a process that requires mTORC1 activation and PAK1. Furthermore, a metabolomic study demonstrates that ECM-dependent cancer cell growth relies on tyrosine catabolism.

Altogether, I found this study conceptually interesting and experimentally well performed. Experiments are well controlled and state-of-the-art technologies used in this manuscript make it a good candidate for publication. However, some aspects of the work need to be strengthen to reinforce the overall good quality of the manuscript, therefore, please find below some experimental propositions.

1. Despite the reviewer proposition, I believe that the additional experiments using the PDAC cancer cell does not improve the quality of the manuscript. Instead, it brings confusion to me, since the relative addition is minor compare to what is demonstrated using breast cancer cells.
2. To importantly improve the potential impact of this manuscript, I suggest to add in vivo data using either syngenic mice model of breast cancer or xenografted human breast cancer cells in nude mice. What would be the impact of micropinocytosis and tyrosine catabolism inhibition on cancer growth, in vivo, should be demonstrated? If possible, it may be interesting to demonstrate that this micropinocytosis may interfere with cancer evolution toward a metastatic phenotype using, for example, the PyMT-MMTV mice model of breast cancer development?
3. Data obtained using cancer cells with different metastatic property suggest that the ability to use ECM to compensate for soluble nutrient starvation is acquired during cancer progression. To further demonstrate that it is the case, would it be possible that non metastatic breast cancer cells are not able to perform micropinocytosis? Is PAK1 overexpressed with increase cancer cells metastatic ability, without affecting invasive capacity in 3D spheroids? What would be the efficacy to promote the ECM-dependent growth under starvation following a mTORC1 activation or PAK1 activation in non-invasive cancer cells?
4. The discrepancy of cancer cells proliferation under starvation condition between plastic and ECM-based supports could be explained by the massive difference of support rigidity. This is also probably the case between CDM made by normal fibroblast and CAF. It brings the question of studying the role of matrix stiffness in regard to micropinocytosis-dependent cancer cells growth. It would also explain why this process is link to aggressive cancer cell behaviour, as ECM goes stiffer with time in cancer development. It may not be the case, but the demonstration that mechanical cues from the ECM could regulate the micropinocytosis-dependent cancer cells growth under amino acid starvation could bring additional value to the manuscript.
5. Along with this, it has been demonstrated that matrix rigidity regulates glutaminolysis in breast cancer, resulting in aspartate production and cancer cells proliferation. Is asparate production increase by micropinocytosis? Could you rescue cancer cells growth by aspartate supplementation?
6. Data presented in Fig 1 and SF1 show that breast cancer cell lines growth in a comparable manner either they are cultured on plastic or on 3D ECM substrates in complete media. Again, on thick 3D substrates, in which the stiffness is lower compared to plastic, I would have thought that cancer cells would have grown slower. Could you please discuss this finding in regard to the literature? If you have the capacity to do so in your lab or in collaboration, would it be possible to measure the exact stiffness of the different matrix you use in this manuscript? Or using hydrogel, would it be possible to study the role of matrix stiffness in the ECM-dependent cancer cells growth under AA starvation? I would understand that this may be out of the scope of the present manuscript, but I again believe that such finding would reinforce the manuscript.
7. In SF 3A-C, it is shown that ECM does not affect caspase-dependent cell death under AA starvation. Did you considered a non-caspase dependent cell death that may be triggered by AA starvation?
8. In fig 5, it is shown that inhibition of Focal Adhesion Kinase (FAK) does not impair the ECM-dependent rescue of cancer cell growth under starvation. To further decipher the concept of adhesion dependent signalling, maybe the authors could also inhibit the Src kinase or ITG-beta1 activation?
9. Minor comment, in F1B, it is written "AA free starvation" while in every others legend, it is written "AA starvation". I believe the "free" should be removed.

Significance

Altogether, I found this study conceptually interesting and experimentally well performed. Experiments are well controlled and state-of-the-art technologies used in this manuscript make it a good candidate for publication. However, some aspects of the work need to be strengthen to reinforce the overall good quality of the manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

Nazemi et al. show that the extracellular matrix (ECM) has a crucial role in sustaining the growth of invasive breast and pancreatic cells during nutrient deprivation. In particular, under amino acid starvation, cancer cells internalize ECM by macropinocytosis and activate phenylalanine and tyrosine catabolism, which in turn support cell growth in nutrient stress conditions.

The paper is well written and the results shown are very interesting. The experimental plan is well designed to assess the hypothesis and the description of the methods is sufficiently detailed to reproduce the analyses, which are also characterized by appropriate internal controls. Finally, the data provided sustain the conclusions proposed by the authors.

Major comment:

Since the authors performed their experiments on invasive breast and pancreatic cancers and it has been noted that stress conditions could promote the escape of cancer cells from the site of origin (e.g., Jimenez and Goding, Cell Metabolism 2018; Manzano et al, EMBO Reports 2020), it would be interesting to evaluate how ECM internalization could have a role in sustaining the invasive abilities of cancer cells under amino acid starvation. Which is the impact of the inhibition of macropinocytosis and tyrosine catabolism on cell invasion? The authors could evaluate this aspect by in vitro 2D and 3D analysis. In addition, to strengthen the paper and give a stronger significance in terms of clinical translatability, it could be useful to implement the analysis of breast and pancreatic patients by publicly dataset evaluating for example free survival, disease free survival, overall survival and metastasis free survival.

Minor Comment:

The text and the figures are clear and accurate. The references cited support the hypothesis, rightly introduce the results and are appropriate for the discussion. However, the paragraph relative to figure 4 is a little confusing. Changing the order of the description of the results could be useful.

Significance

Based on my metabolic background in tumour aetiology and progression, I think that this study provides new insights into cancer metabolism knowledge, in particular on how the stroma may drive metabolic reprogramming of cancer cells sustaining cell growth in nutrient stress conditions. Together with other similar studies on the stromal non-cellular components, the data here shown can contribute to expand the knowledge on the factors that promote cancer metabolic plasticity, which is exploited from cancer cells to obtain advantages in terms of growth, survival and progression.

In conclusion, I think that the results shown are new and the manuscript is well presented. Following the short revision process suggested, it will be eligible for a final publication in a medium-high impact factor journal.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This study by Viedor et al. examines the role of TIS7 (IFRD1) and its ortholog SKMc15 (IFRD2) in the regulation of adipogenesis via their ability to modulate the levels of DLK1 (Pref-1), a well-known inhibitor of adipogenesis. They generate SKMc15 KO mice and cross them to previously published TIS7 KO mice. All 3 mutant strains show decreased fat mass, with the effect being most pronounced in double KO mice (dKO). Using mouse embryonic fibroblasts (MEFs) from mutant mice, they authors ascribe a defect in adipogenic differentiation of mutant cells to an upregulation of DLK-1. In the case of TIS7, they propose that this is due to its known inhibition of Wnt signaling, which regulates DLK-1 expression. In the case of SKMc15, they suggest a new mechanism linked to its ability to suppress translation. Overall, the work is of interest, with the finding, that SKMc15 regulates adipocyte differentiation being its novelty, and generally well done, though multiple aspects need to be improved to bolster the conclusions put forth.

**Major concerns:**

1)The main mechanism put forth by the authors to explain the inability of dKO cells to differentiate into adipocytes is the upregulation of DLK-1 levels. However, this notion is never directly tested. Authors should test if knockdown of DLK-1 in dKO cells is sufficient to correct the defect in differentiation, or if additional factors are involved.

Response: In response to the reviewer’s concerns, we have generated two stable cell lines expressing short hairpin RNAs directed against DLK1 in the TIS7 SKMc15 dKO MEFs. With these two and the parental dKO MEF cell line, we have performed adipogenesis differentiation experiments as explained in the manuscript before. Figure EV2C (left and right panels) shows that knockdown of DLK1 with two different DLK1 shRNA constructs (targeting DLK1 with or without the extracellular cleavage site) significantly (P2)There are multiple instances were the authors refer to "data not shown", such as when discussing the body length of dKO mice. Please show the data in all cases (Supplementary Info is fine) or remove any discussion of data that is not shown and cannot be evaluated.

Response: Following three results were in the initial version of our manuscript mentioned as “data not shown”:

• line 137: “body length, including the tail did not significantly differ between WT and dKO mice”
• line 307: “higher concentrations of free fatty acids in the feces of dKO mice”

• line 331: “effects of ectopic expression of TIS7, SKMc15 and their co-expression on DLK-1 levels” In the current version of the manuscript, we provide these results as:

• Figure EV1A shows no significant difference in body length.

• The significantly elevated levels of free fatty acids and energy determined by bomb calorimetry in the feces of dKO animals fed HFD are shown in Figures 6A and B, respectively.
• The significant inhibitory effect of ectopic expression of TIS7 and SKMc15 on DLK1 levels was identified by both qPCR and WB analyses, which are shown in Figure 3B. 3)Indirect calorimetry data shown in Fig. S1 should include an entire 24 hr cycle and plots of VO2, activity and other measured parameters shown (only RER and food intake are shown), not just alluded to in the legend.

Response: Based on the reviewer’s suggestion, we present here a table containing all parameters measured in the indirect calorimetry experiment.

Metabolic phenotyping presented in Figure EV1B containing 21 hours measurement was performed exactly according to the standardized protocol previously published by Rozman J. et al. [1]. All phenotyping tests were performed following the International Mouse Phenotyping Resource of Standardized Screens (IMPReSS) pipeline routines.

4)It is surprising that the dKO mice weight so much less than WT even though their food consumption and activity levels are similar, and their RER does not indicate a switch in fuel preference. An explanation could be altered lipid absorption. The authors indicate that feces were collected. An analysis of fat content in feces (NEFAs, TG) needs to be performed to examine this possibility. The discussion alludes to it, but no data is shown.

__Response: __We thank the reviewer for bringing up this important point that prompted us to present data clarifying this aspect of the metabolic phenotype of dKO mice. As shown in Figures 6A,B, while fed with HFD, dKO mice had higher concentrations of free fatty acids in the feces (109 ± 10.4 µmol/g) when compared to the WT animals (78 ± 6.5 µmol/g) and a consequent increase in feces energy content (WT: 14.442 ± 0.433 kJ/g dry mass compared to dKO: 15.497 ± 0.482 kJ/g dry mass). Thus, lack of TIS7 and SKMc15 reduced efficient free fatty acid uptake in the intestines of mice.

5)It would be important to know if increased MEK/ERK signaling and SOX9 expression are seen in fat pads of mutant mice, not just on the MEF system. Similarly, what are the expression levels of PPARg and C/EBPa in WAT depots of mutant mice?

Response: To address this point, we have now performed the MEK/ERK activity measurement for the revised version of the manuscript in gonadal WAT tissue (GWAT). As noted in samples from several mice, there was an increase in p42 and p44 MAPK phosphorylation in G WAT isolated from dKO mice compared with the G WAT from WT control mice (Figure 4G).).

The mRNA expression levels of PPARg and C/EBPa were significantly downregulated in GWAT samples isolated from dKO mice compared with levels from WT control animals (Figure 4H). However, we did not find any significant difference in SOX9 expression in fat pads. Total amounts of Sox9 mRNA in terminally differentiated adipocytes were very low and not within the reliable detection range, and the variation between animals within the same group was too great. Therefore, we provide these data only for the reviewer’s information here and do not present them in the manuscript.

6)Analysis of Wnt signaling in Fig. 3c should also include a FOPflash control reporter vector, to demonstrate specificity. Also, data from transfection studies should be shown as mean plus/minus STD and not SEM. This also applies to all other cell-based studies (e.g., Fig. 6b,c).

Response: To address the reviewer’s concerns, we performed FOPflash control reporter measurements in MEFs of all four genotypes. As expected, in every tested cell line the luciferase activity of the FOPflash reporter was substantially lower than that of TOPflash, confirming the specificity of this reporter system.

We also thank the reviewer for this important reference to our statistical analyses. We have revised the original data and found that the abbreviation SEM was inadvertently used in the legends instead of STD. STD was always used in the original analyses and therefore we have corrected all legends accordingly in the new version of the manuscript.

7)It is unclear why the authors used the MEF model rather than adipocyte precursors derived from the stromal vascular fraction (SVF) of fat pads from mutant mice. If they did generate data from SVF progenitors, they should include it.

__Response: __We agree with this comment, although performing the experiments was challenging enough for us. Therefore, we isolated inguinal fat pads and obtained SVF cells from mice of all four genotypes (WT, TIS7, SKMc15 single and double KOs) and have repeated crucial experiments, i.e. adipocyte differentiation, DLK1, PPARg and C/EBPa mRNA and protein analyses in these cells. Novel data gained in this cell system fully confirmed our previous observations in MEFs. Therefore, in the current version of the manuscript we have replaced figures describing the effects of lacking TIS7 and SKMc15 in MEFs by adipose tissues samples (Figures 2D,E, 4G,H,I and 6C) or SVF cells from inguinal WAT (Figures 2A,B,F,G,H, 3C,D,E and F). In addition to the results obtained from SVF cells of inguinal WAT, we also obtained comparable data from SVF cells isolated from fat pads of gonadal WAT. We provide the results from gonadal WAT hereafter for the reviewers' information only.

amido black

• 60 kDa

• 50 kDa

__G WAT __tissues

dKO

WT

GAPDH

DLK-1

• 60 kDa

• 50 kDa

• 35 kDa

WT

dKO

DLK-1

• 60 kDa

• 50 kDa

__G WAT __undifferentiated cells

undifferentiated G WAT cells

The only experiments where we have still used data obtained in MEFs are those where the ectopic expression or effects of shRNA were necessary (e.g. Figures 2C, 3B,H,I, 5F,G EV2B,C and EV3 A-F).

8)Given that the authors' proposed mechanism involves both, transcriptional and post-transcriptional regulation of DLK-1 by TIS7 and SKMc15, Fig. 4d should be a Western blot capturing both of these events, and not just quantitation of mRNA levels.

Response: As requested by the reviewer, we have added in Figure 3B the Western blot analysis of DLK1 expression. Secondly, this experiment was entirely redone and we now show the effects of ectopic expression of SKMc15, TIS7 alone and their combination side by side with the control GFP. We present here the effects of stable expression of ectopic TIS7 and SKMc15 in dKO MEFs following the viral delivery of expression constructs, antibiotic selection and 8 days of adipocyte differentiation.

9)There is no mention of the impact on brown adipose tissue (BAT) differentiation of KO of TIS7, SKMc15, or the combination. Given the role of BAT in systemic metabolism beyond energy expenditure, the authors need to comment on this issue.

Response: We thank the reviewer for bringing up this important point that prompted us to better describe the phenotype of TIS7, SKMc15 and double knockout mice. We measured DLK1 protein levels in BAT isolated from WT, TIS7, and SKMc15 mice with single and double knockout and detected a significant increase in DLK1 protein levels in all three knockout genotypes. Five mice per genotype were analyzed, and the statistical analysis in Figure 4I represents the mean ± STD. The p-values are based on the results of the Student's t-test and one-way Anova analysis (p-value = 0.0241).

**Minor comments:**

10)The y axis in Fig. 2c is labeled as gain of body weight (g). Is it really the case that WT mice gained 30 g of body weight after just 3 weeks of HFD? This rate of increase seems extraordinary, and somewhat unlikely. Please re-check the accuracy of this panel.

Response: We thank the reviewer for drawing our attention to the apparent mislabeling of the y-axis. The correct labeling is: "Increase in body weight in %" and Figure 1F has been corrected accordingly.

11)The Methods indicates all statistical analysis was performed using t tests, but this is at odds with some figure legends that indicate additional tests (e.g., ANCOVA).

Response: This inaccurate information in the manuscript was corrected.

12)Please specify in all cases the WAT depot used for the analysis shown (e.g., Fig. 3d is just labeled as WAT, as are Fig. 4a,e, etc.).

Response: This information was added at all appropriate places of the manuscript.

13)Fig. 5d is missing error bars, giving the impression that this experiment was performed only once (Fig. 5c). The legend has no details. Please amend.

__Response: __We thank the reviewer for this important point regarding the statistical analyses. In the new version of the manuscript, we have included a graph (now Figure 4D) depicting results of three independent experiments including the results of the statistical analysis performed. Statistical analysis was performed using One-Way ANOVA (P=0.0016).

Reviewer #1 (Significance (Required)):

The role of TIS7 in adipocyte differentiation is well established. The only truly novel finding in this work is the observation that SKMc15 also plays a role in adipogenesis. The molecular mechanisms proposed (modulation of DLK-1 levels) are not novel, but make sense. However, they need to be bolstered by additional data.

**Referees cross-commenting**

I think we are all in agreement that the findings in this work are of interest, but that significant additional work is required to discern the mechanisms involved. In my view, a direct and specific link between SKMc15 and translation of DLK-1 needs to be established and its significance for adipogenesis in cells derived from the SVF of fat pads determined. Reviewer 2 has suggested some concrete ways to provide evidence of a direct link.

__Response: __We agree with the reviewer's comment and have also noted that this point will be crucial in assessing the novelty value of our manuscript, as was also expressed in the referees cross-commenting. Therefore, we have now additionally performed a polysomal RNA analysis, which has of course been included in the current version of the manuscript.

We analyzed the differences in DLK-1 translation between wild-type control cells and SKMc15 knockout cells in the gradient-purified ribosomal fractions by DLK-1 qPCR. Our analysis identified significantly (pSimilarly, as proposed by the reviewer, we have established stromal vascular fraction cell cultures from inguinal fat pads. In SVF cells of TIS7 and SKMc15 single and double knockout mice, we found increased DLK1 mRNA and protein levels (Figures 2F,G and H) as well as decreased PPARg and C/EBPa levels (Figures 3C,D,E and F). Specifically, we found that the ability of knockout SVF cells to differentiate into adipocytes was significantly downregulated (Figures 2A and B), fully confirming our original findings in TIS7 and SKMc15 knockout MEFs.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

**Summary:**

In the current study, Vietor et al. aimed to explore the regulation of Delta-like homolog 1 (DLK-1), an inhibitor of adipogenesis, and demonstrated a role for TIS7 and its orthologue SKMc15 in the regulation of adipogenesis by controlling the level of DLK-1. Using mouse models with whole body deficiency of TIS7 (TIS7 KO) or SKMc15 (SKMc15KO) and double KO (TIS7 and SKMc15 dKO) mice, the authors used a combination of in-vivo experiments and cell culture experiments with mouse embryonic fibroblasts derived from the KO animals, to show that the concurrent depletion of TIS7 and SKMc15 dramatically reduced the amount of adipose tissues and protected against diet-induced obesity in mice, which was associated with defective adipogenesis in vitro.

**Major Comments:**

Overall, this study presents convincing evidence that TIOS7 and SKMc15 are necessary for optimal adipogenesis, and proposes a novel mechanism for the control of DLK1 abundance via coordinated regulation of DLK-1 transcription and translation. However, a number of questions remain largely unanswered. In particular, the direct ability of SKMc15 to regulate the translation of DLK-1 is lacking, and this claim remains speculative. SKMc15 being a general inhibitor of translation, SKMc15 may have an effect on adipogenesis independently of its regulation of DLK-1. Thus, addressing the following comments would further improve the quality of the manuscript:

Response:

We have been very attentive to these comments to improve the novelty and quality of our manuscript and have tried to address them experimentally. Therefore, this thorough revision of our manuscript took a longer time. First, we identified polysomal enrichment of DLK-1 RNA in SKMc15 KO MEFs, demonstrating that SKMc15 translationally affects DLK-1 levels (Figure 3I). Second, treatment with a recombinant DLK-1 protein as well as its ectopic expression quite clearly blocked adipocyte differentiation of WT MEFs (Figures EV3B,C). In addition, two different shRNA constructs targeting DLK-1 significantly induced adipocyte differentiation of TIS7 SKMc15 dKO MEFs (Figure EV2C, left and right panels). We believe that these results, taken together, sufficiently support our proposed mechanism, namely that TIS7 and SKMc15 control adipocyte differentiation through DLK-1 regulation.

• The experimental evidence supporting that SKMc15 controls DLK-1 protein levels comes primarily from the observations that DLK-1 abundance is further increased in SKMc15 KO and dKO WAT than in TIS7KO WAT (Fig 3d), and that translation is generally increased in SKMc15 KO and dKO cells (Fig 5a). However, since the rescue experiment is performed in dKO cells, by restoring both TIS7 and SKMc15 together, it is impossible to disentangle the effects on DLK-1 transcription, DLK-1 translation and on adipogenesis. A more detailed description of the TIS7 and SKM15c single KO cells, with or without re-expression of TIS7 and SKMc15 individually, at the level of DLK-1 mRNA expression and DLK-1 protein abundance would be necessary. In addition, polyribosome fractioning followed by qPCR for DLK-1 in each fraction, and by comparison with DLK-1 global expression in control and SKMc15 KO cells, would reveal the efficiency of translation for DLK-1 specifically, and directly prove a translational control of DLK-1 by SKMc15. Alternatively, showing that DLK-1 is among the proteins newly translated in SKMc15 KO cells (Fig. 5a) would be helpful. Response: As suggested by the reviewer we used single TIS7 and SKMc15 knockout cells and demonstrated that both, TIS7 and SKMc15, affect Dlk-1 mRNA levels. We identified a highly significant effect on total DLK-1 mRNA levels in SKMc15 knockout MEFs as presented in Figure 3H. We also show that DLK-1 mRNA is specifically enriched in polysomal fractions obtained from proliferating SKMc15 knockout MEFs when compared to WT MEFs. However, the strong accumulation of DLK-1 mRNA in polysomes cannot be explained by transcriptional upregulation of DLK-1 alone, suggesting that regulation also occurs at the translational level. We took up this suggestion and ectopically expressed TIS7 and SKMc15 separately or together. For this purpose, we used not only MEF cell lines with double knockout but also with single knockout. Our recent data showed that stable ectopic expression of SKMc15 significantly increased adipocyte differentiation in both, single and double TIS7 and SKMc15 knockout MEF cell lines (Figures EV1C,D and EV2A). Ectopic expression of TIS7 significantly induced the adipocyte differentiation in TIS7 single knockout MEFs (Figure EV1C). In addition, both genes down regulated DLK-1 mRNA expression in dKO MEFs (Figure EV2A, bar chart on the right). We fully agree with the opinion of both reviewers and as already explained above we identified by qPCR in the polysomes that SKMc15 directly regulates DLK-1 translation (Figure 3I).

• While the scope of the study focuses on the molecular control of adipogenesis by TIS7 and SKMc15 via the regulation of DLK-1, basic elements of the metabolic characterization of the KO animals providing the basis for this study would be useful. Since the difference in body weight between WT and dKO animals is already apparent 1 week after birth (Fig 1a), it would be interesting to determine whether the fat mass is decreased at an earlier age than 6 months (Fig 1b). The dKO mice are leaner despite identical food intake, activity and RER (Sup Fig 1). It remains unclear whether defective fat mass expansion is a result or consequence of this phenotype. Is the excess energy stored ectopically? The authors mention defective lipid absorption, however, these data are not presented in the manuscript. It would be interesting to investigate the relative contribution of calorie intake and adipose lipid storage capacity in the resistance to diet-induced obesity. In addition, data reported in Fig 1c seem to indicate a preferential defect in visceral fat development, as compared to subcutaneous fat. It would be relevant if the authors could quantify it and comment on it. Are TIS7 and SKMc15 differentially expressed in various adipose depots? The authors used embryonic fibroblasts as a paradigm to study adipogenesis. It would be important to investigate, especially in light of the former comment, whether pre-adipocytes from subcutaneous and visceral stroma-vascular fractions present similar defects in adipogenesis. Response: We addressed the issue of lipid storage capacity raised by the reviewer using two experimental methods. First, we have analyzed feces of mice fed with high fat diet. The free fatty acids content in dKO mice feces was significantly (PConcerning the question of younger animals, we have repeated microCT fat measurements on a group of 1-2 months old WT and dKO male mice (n=4 per group). The total amount of abdominal fat was in WT mice significantly higher than in dKO mice (P=0.019; Student’s T-test). We provide these data only for the reviewer’s information here and do not present them in the manuscript.

We have also followed the reviewer’s advice and revisited our microCT measurements of abdominal fat and anylyzed the possible differences between subcutaneous and visceral fat. In all three types of abdominal fat mass measurement (total, subcutaneous and visceral) there was always significantly (ANOVA P=0.034 subcutaneous, P=0.002 total and P=0.002 visceral fat) less fat in the dKO group (n=8) of mice when compared to WT (n=12) mice. However, the difference was more prominent in visceral (P=0.001; Student’s T-test) than in subcutaneous fat (P=0.027; Student’s T-test). We provide these data only for the reviewer’s information here and do not present them in the manuscript.

In addition, we have analyzed the expression of TIS7 and SKMc15 mRNA expression in both, inguinal and gonadal WAT. Our qPCR result showed that both genes are expressed in different types of WAT. The qPCR analysis was performed on RNA isolated from undifferentiated SVF cells isolated from several animals. The expression of TIS7 and SKMc15 was normalized on GAPDH. Data represent mean and standard deviation of technical replicates from several mice as labeled in the graph. We provide these data only for the reviewer’s information here and do not present them in the manuscript.

Topics of a) stromal vascular fraction as a source of pre-adipocytes and b) comparison of TIS7 and SKMc15 roles in gonadal vs. inguinal fat pads we answered in response to the Reviewer #1, point 7. The results are presented in Figures 2, 3 and 4 and in this document.

Both data and methods are explained clearly. The experiments are, for the most part, adequately replicated. However, whenever multiple groups are compared, ANOVA should be employed instead of t-test for statistical analysis.

Response: Thank you for pointing this out. Wherever it was applicable, we used ANOVA for the statistical analysis of data.

**Minor comments:**

• Figure 4 d. The appropriate control would be WT with empty vector Response: this experiment was entirely replaced by the new Figure 3B where stably transfected MEF cells expressing TIS7 or SKMc15 were used.

• Figure 7c/d. The appropriate control would be WT with empty vector Response: We have now generated new, confirmatory data in MEF cells stably expressing TIS7 or SKMc15 following lentiviral expression.

• Figure 5C. An additional control would be WT with WT medium __Response: __We agree with your suggestion and therefore we have incorporated this control in all experimental repeats presented in the new Figure 4C.

• Figure 2: In the legends, the "x" is missing for the dKO regression formula __Response: __Thank you, we have corrected this mistake. In the current version of the manuscript it is Figure 1D.

• Since the role of SKMc15 in adipogenesis has never been described, the authors could consider describing the single SKMc15 KO in addition to the dKO, or explain the rationale for focusing the study on dKO. __Response: __The original reason for focusing on dKO mice and cells was the obvious and dominant phenotype in this animal model. However, we have sought to address the reviewer's concerns and have now also examined DLK-1 mRNA levels in proliferating SKMc15 knockout MEFs (Figure 3H). In addition to this experiment, we measured DLK-1 mRNA levels also during the process of adipocyte differentiation of single knockout cells. In WT MEFs we observed a transient increase of DLK-1 mRNA only on day 1. In contrast, significantly elevated DLK-1 mRNA levels were found in TIS7 single-knockout MEFs throughout the differentiation process, with the highest level reached at day 8. Interestingly, in SKMc15 single knockout MEFs we found an upregulation of DLK-1 mRNA level in proliferating cells but not a further increase during the differentiation. This supported our idea that SKMc15 acts mainly via translational regulation of DLK-1. We provide these data only for the reviewer’s information here and do not present them in the manuscript.

To emphasize this point, we revised the entire manuscript accordingly and added data on SKMc15 knockout mice. In particular, experiments presenting data characterizing SKMc15 single knockout mice are presented in: Figures 1C,D,E and F, Figures 2A,B,C and D, Figures 3E,F,H and I, Figures 4A and I and in Figure EV1D.

Reviewer #2 (Significance (Required)):

While the effects of DLK-1 on adipogenesis have been widely documented, the factors controlling DLK-1 expression and function remain poorly understood. Here the authors propose a novel mechanism for the regulation of DLK-1, and how it affects adipocyte differentiation. This study should therefore be of interest for researchers interested in the molecular control of adipogenesis and cell differentiation in general. Furthermore, the characterization of the function of SKMc15 in the control of translation may be of interest to a broader readership.

**Referees cross-commenting**

I agree with all the comments raised by the other reviewers. Addressing the often overlapping but also complementary questions would help to clarify the molecular mechanisms by which TIS7 and SKMc15 control adipogenesis, and support the conclusions raised by the authors.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

In the article, "The negative regulator DLK1 is transcriptionally regulated by TIS7 (IFRD1) and translationally by its orthologue SKMc15 (IFRD2)", the authors performed a double knockout (dKO) of TIS7 and its orthologue SKMc15 in mice and could show that those dKO mice had less adipose tissue compared to wild-type (WT) mice and were resistant to a high fat-diet induced obesity. The study takes advantage of number of different methods and approaches and combines both in vivo and in vitro work. However, some more detailed analysis and clarifications would be needed to fully justify some of the statements. Including the role of TIS7 as a transcriptional regulator of DLK1, SKMc15 as translational regulator of DLK1 and overall contribution of DLK1 in the observed differentiation defects. The observed results could still be explained by many indirect effects caused by the knock-outs and more direct functional connections between the studied molecules would be needed. Moreover, some assays appear to be missing biological replicates and statistical analysis. Please see below for more detailed comments:

**Major comments:**

-Are the key conclusions convincing? Yes.

-Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? No.

-Would additional experiments be essential to support the claims of the paper? Yes. Please see my comments.

-Are the suggested experiments realistic in terms of time and resources? Recombinant DLK1 10 μg - Tetu-bio - 112€ ; 8 days of adipocyte differentiation in 3 biological replicate ~ 1 month.

__Response: __We followed the advice of the individual reviewers as expressed in “Referees cross-commenting” and tested this idea experimentally. Since the manufacturer couldn’t suppy information on biological activities of recombinant DLK-1 proteins, we analyzed in vivo the effects of two different ones, namely RPL437Mu01 and RPL437Mu02. The 8-day adipocyte differentiation protocol showed that the RPL437Mu02 protein was cytotoxic to WT MEF cells and therefore could not be used for analysis. On the other hand, treatment with the Mu01 recombinant DLK-1 protein did not result in a substantial cell death. According to oil red O staining, incubation with 3.3 mg/ml (final concentration) RPL437Mu01 led to 75% inhibition of adipocyte differentiation when compared to not treated WT MEFs (Figure EV3B and C).

-Are the data and the methods presented in such a way that they can be reproduced? Yes.

-Are the experiments adequately replicated and statistical analysis adequate?

Adequately reproduced yes. Please see my comments concerning the statistical analysis.

1)Fig1a: In the method section it is written that an unpaired 2-tailed Student's t test was used for all statistical comparisons. However, here something like Multivariate analysis of variance (MANOVA) should rather be used to assess statistical significance between the mice. Moreover, the details of this should be clearly stated in the corresponding Figure legend.

__Response: __Based on this suggestion, we have revised all of our statistical analyses. In several cases, (Figures 1F, 2B and C) we have replaced the statistical analysis using Student’s T test with Anova. However, based on the definition “the difference between ANOVA and MANOVA is merely the number of dependent variables fit. If there is one dependent variable then the procedure ANOVA is used”, in case of Figure 1A we used ANOVA.

2)Fig2a: please use an appropriate title for Fig2a instead of "Abdominal fat vs. body mass".

Response: Title of the Figure 1D (formerly Figure 2a) we changed to “Effect of TIS7 and SKMc15 on the abdominal fat mass”.

3)Fig2c: in the method section it is written that an unpaired 2-tailed Student's t test was used for all statistical comparisons. However, in Fig2c 4 groups are compared (WT, TIS7 KO, SKMc15 KO and dKO) and thus something like Multivariate analysis of variance (MANOVA) should rather be used to assess statistical significance.

Response: For Figure 1F (formerly Figure 2c), in the revised version of the manuscript, we applied the ordinary one-way ANOVA with Holm-Šidák's multiple comparison test. This analysis gave us statistically even more significant results concerning the difference between WT and dKO mice than previously found by Student's T test. The results in detail were as follows:

Holm-Šidák's multiple comparisons test Summary Adjusted P Value

WT vs. TIS7 KO ** 0,0096

WT vs. SKMc15 KO * 0,0308

WT vs. dKO **** 4)Fig2 conclusion: Additive or just showing stronger effect?

Response: We have re-phrased the concluding summary for Figure 1F (formerly Figure 2c). We agree that the precise description of differences found between the weight of single and double knockout animals should be described as “stronger” and not additive effect of knockout of both genes.

5)Fig3a: the microscope picture for SKMc15 KO shows that cells might have died. Please state the percentage of cell death.

Response: We would like to comment on these concerns of the reviewer as follows: In the image in Figure 3 of the original manuscript, the density of SKMc15 KO MEF cells after the adipocyte differentiation protocol was lower than in the WT control. Regarding the possible cell death, the cells stained with Oil Red O were adherent and alive. The adipocyte differentiation protocol consists of 3 days proliferation and further 5 days of differentiation including three changes of media during which dead cells are washed away and their vitality cannot be checked. However, in the meantime, we have repeated this protocol and the density of SKMc15 knockout MEFs was now not substantially lower than those of controls. Despite the comparable cell density, we have seen a substantial negative effect of the SKMc15 knockout on the adipogenic differentiation ability of these cells. Several examples are shown here:

TIS7 +/+ SKMc15 +/+ MEFs

TIS7 +/+ SKMc15 -/- MEFs

oil red O staining; 8d differentiated cells

Importantly, in the current version of our manuscript we replaced MEFs (shown in the former Figure 3a) by SVF cells (Figure 2A in the current manuscript). In these cells we did not see any significant difference in their density after 8 days of the adipocyte differentiation protocol.

6)Fig3b: It would be informative to additionally observe some of marker genes for adipogenesis and whether all of them are affected.

Response: In our newly established SVF cell lines, derived from inguinal WAT we have confirmed data previously identified in MEFs. As shown in the new Figure 3, PPARg and C/EBPa mRNA levels were downregulated in all knockout SVF cell lines, both undifferentiated (Figures 3C and D) and adipocyte differentiated (Figures 3E and F). On the other hand, DLK-1 mRNA and protein levels, both in undifferentiated (Figures 2F and G) and adipocyte differentiated (Figure 2H) SVF cells were significantly upregulated in dKO cells when compared to WT cells.

7)Fig3b: instead of using an unpaired 2-tailed Student's t test with proportion, an one-way ANOVA would be more appropriate.

__Response: __On the recommendation of the reviewer, we applied a simple ANOVA to our new data from SVF cells using the Holm-Šidák test for multiple comparisons. The Anova summary using GraphPad Prism Ver. 9.2 identified statistically highly significant (P value 8)Fig3c: Same comment as for Fig3b.

__Response: __Also, in this experiment (now Figure 2C) we used ordinary one-way ANOVA with Holm-Šidák's multiple comparisons test. The ANOVA summary identified statistically highly significant (P value 9)Fig3d: A representative Western blot for 3 independent experiments is shown. Please add the other two as supplementary materials.

__Response: __Here we provide examples of the requested two additional, independent experiments. These refer now to the Figure 2D in the revised version of the manuscript:

31 07 2020

= manuscript

b____-catenin

22 07 2020

WT

TIS7 KO

dKO

SKMc15 KO

WT

TIS7 KO

dKO

SKMc15 KO

actin

b____-catenin

30 07 2020

actin

b____-catenin

WT

TIS7 KO

dKO

SKMc15 KO

WT

TIS7 KO

dKO

SKMc15 KO

actin

10)Fig3d:Is this distinguishing between the active and inactive catenin?

__Response: __No, the b-catenin antibody, that we used is not discriminating between active and inactive b-catenin forms.

11)Fig4a: Please perform qPCR for measuring DLK-1 mRNA levels in TIS7 KO and SKMc15 KO samples to check whether there is a correlation between mRNA and protein level as the statement of the authors is that "DLK1 is transcriptionally regulated by TIS7 (IFRD1) and translationally by its orthologue SKMc15".

Response: Similar questions were raised by Reviewer 2 on p. 11 “Since the role of SKMc15 in adipogenesis has never been described, the authors could consider describing the single SKMc15 KO in addition to the dKO, or explain the rationale for focusing the study on dKO.” Please see our reply to his comment.

12)Fig4c: please add the other two western blots as supplementary materials.

__Response: __Here we provide data from two additional, independent experiments.

13)Fig4d: The effects in MEFs appear quite modest. What about a rescue with TIS7 or SKMc15 alone?

__Response: __As mentioned already in response to the question 2 of Reviewer #1, in our newly performed experiments we found significant inhibitory effects of ectopic TIS7 and SKMc15 expression on DLK1 levels, identified both by qPCR and WB analyses (Figure 3B).

14)Page 12, row 207: I would not call histones transcription factors.

__Response: __We re-phrased this sentence accordingly.

15)Fig4e: Would be good to see a schematic overview of the locations of the ChIP primers in relation to the known binding sites and the gene (TSS, gene body). Moreover, the results include an enrichment for only one region while in the text two different regions are discussed. Importantly, to confirm the specificity of the observed enrichment, a primer pair targeting an unspecific control region not bound by the proteins should be included.

__Response: __The selection of oligonucleotide sequences used for ChIP analyses of the binding of b-catenin, TIS7 and SKMc15 to the Dlk-1 promoter was, based on the following reference, as mentioned in Methods section of our original manuscript on p.21, line 494: Paul C, Sardet C, Fabbrizio E. “The Wnt-target gene Dlk-1 is regulated by the Prmt5-associated factor Copr5 during adipogenic conversion”. Biol Open. 2015 Feb 13;4(3):312-6. doi: 10.1242/bio.201411247.

We used two regions of the Dlk-1 promoter: a proximal one, encompassing the TCF binding site 2 (TCFbs2) and a more distal one, annotated as “A”:

Oligonucleotide sequences used for ChIP PCR:

Dlk-1 TCFbs2 5'f CATTTGACGGTGAACATATTGG

5'r GCCCAGACCCCAAATCTGTC

Dlk-1 region A (-2263/-2143) 5'f TTGTCTAACCACCCTACCTCAAA

5’r CTCTGAGAAAAGATGTTGGGATTT

We observed specific binding at the proximal site.

16)Fig5a: Has this experiment been replicated? That is no mention about the reproducibility or quantification of this result. This is the main experiment regarding the role of SKMc15 as a translational regulator of DLK1, also mentioned in the title of the manuscript.

__Response: __This relates to the Figure 4A in the revised manuscript. Yes, we repeated this experiment several times. Here we provide images and quantifications of three independent experiments.

17)Fig5b: Showing another unaffected secreted protein would be an appropriate control here.

Response: As recommended by the reviewer, we have performed an additional WB with a recombinant anti-Collagen I antibody [Abcam, [EPR22209-75] ab255809]. Medium from 8 days adipocyte differentiated WT and dKO MEFs was concentrated using Centriprep 30K and resolved on 10% SDS-PAGE gel. Western blot presented in the new Fig 4 B shows even slightly higher amounts of Collagen-1 protein in medium from WT than in dKO MEFs. Mw of the detected band was approximately 35 kDa, which corresponded to the manufacturer’s information.

18)Fig5c: I would recommend to perform additional experiments to prove that DLK-1 secreted in the medium can contribute and is responsible for the inhibition of the differentiation. Indeed, a time course of adipocyte differentiation followed by the addition of soluble DLK-1 would confirm that DLK-1 can inhibit adipocyte differentiation in this experimental setup. Moreover, silencing (for example RNAi) of DLK1 in the dKO cells before harvesting the conditioned media would allow to estimate the contribution of DLK1 to the observed inhibition of differentiation by the media. This is important because many other molecules could also be mediating this inhibition.

__Response: __We agree with this reviewer’s concern, which are shared by other reviewers. Similarly, as in response to Reviewer #2 and as already mentioned above, in response to “major comments” of Reviewer #3, in our novel experiments we found that treatment with recombinant DLK-1 protein as well as ectopic expression of DLK-1 blocked adipocyte differentiation of WT MEFs (Figures EV3B,C,D and E) as well as medium from dKO shDLK-1 391 cells (Figure EV3F).

19)Fig5c: The details and the timeline of the experiment with conditioned media are not provided in the figure or in the methods. At what time point was conditioned media changed? How long were the cells kept in conditioned media? How does this compare to the regular media change intervals? Could the lower differentiation capacity relate to turnover of the differentiation inducing compounds in the media due to longer period between media change? Moreover, is the result statistically significant after replication?

__Response: __Based on the reviewer`s comment we have added technical information concerning the experimental protocol of the treatment with conditioned media. In general, the treatment for adipocyte differentiation was identical with the previous experiments. The only difference was that after three days in proliferation medium, we used either fresh differentiation medium or 2-day-old differentiation medium from dKO control or dKO-shDLK-1 391 cell cultures then for wild-type cells, as shown in the figure (Figure EV3F). Cells were incubated additional five days with the differentiation medium with two changes of media, every second day. The adipocyte differentiation of medium “donor” cells and the DLK-1 protein levels in these cells were monitored by oil red O staining and Western blot analysis, respectively.

Additionally, we show now in Figure 4C representative images from three independent biological repeats and in Figure 4D the statistical analysis confirming a significant decrease in adipocyte differentiation ability of WT MEFs following their incubation with a conditioned differentiation medium from dKO MEFs.

20)Fig5d: please add a statistical analysis of the oil-red-o quantification.

__Response: __As requested, we included statistical analysis of at least three independent experiments. In Figure 4D we present the statistical analysis confirming a significant decrease in adipocyte differentiation of WT MEFs following their incubation with the differentiation medium from dKO cells. Additionally, Figure 4C shows representative images of oil red O staining from several independent experiments.

21)Fig7c-d: Does overexpression also rescue the PPARg and CEBPa induction during differentiation. The importance of their induction in undifferentiated MEFs is a little difficult to judge.

__Response: __We have focused our attention primarily on the ability of TIS7 and SKMc15 to “rescue” the adipocyte differentiation phenotype of dKO MEFs. dKO MEFs stably expressing SKMc15, TIS7 or both genes were differentiated into adipocytes for 8 days and afterwards stained with oil red O. There was a statistically significant increase in oil red O staining following the individual ectopic expression of SKMc15 (p=5.7E-03), a negative effect of TIS7 ectopic expression and a significant (p=9.3E-03), positive effect of co-expression of both genes (Figure EV2A). We found a significant decrease in Dlk-1 mRNA expression following the ectopic expression of TIS7 and/or SKMc15 (Figure EV2A, very right panel). However, C/EBPa mRNA levels were only partially rescued in 8 days differentiated MEFs by TIS7 and/or SKMc15 ectopic expression, and PPARg mRNA levels were not significantly altered.

22)Fig8: it is not surprising that PPARg targets are not induced in the absence of PPARg. What is the upstream event explaining this defect? Is DLK1 alone enough to explain the results? Could there be additional mediators of the differences? How big are transcriptome-wide differences between WT MEFs and dKO MEFs?

__Response: __We agree with the reviewer that the lean phenotype of dKO mice most likely cannot be explained by simple transcriptional regulation of PPARg. Although we showed that in undifferentiated MEFs, the levels of PPARg and C/EBPa are controlled (or upregulated) by both TIS7 and SKMc15, we also expected differences in the expression of genes regulating fat uptake. To determine changes in expression of lipid processing and transporting molecules, we performed transcriptome analyses of total RNA samples isolated from the small intestines of HFD-fed WT type and dKO animals. Cluster analyses of lipid transport-related gene transcripts revealed differences between WT type and dKO animals in the expression of adipogenesis regulators. Those included among other genes the following, mentioned as examples:

• peroxisome proliferator-activated receptors γ (PPARγ) and d [2], fatty acid binding proteins 1 and 2 (FABP1, 2) [3],
• cytoplasmic fatty acid chaperones expressed in adipocytes,
• acyl-coenzyme A synthetases 1 and 4 (ACSL1,4) found to be associated with histone acetylation in adipocytes, lipid loading and insulin sensitivity [4],
• SLC27a1, a2 fatty acid transport proteins, critical mediators of fatty acid metabolism [5],
• angiotensin-converting enzyme (ACE) playing a regulatory role in adipogenesis and insulin resistance [6],
• CROT, a carnitine acyltransferase important for the oxidation of fatty acids, a critical step in their metabolism [7],
• phospholipase PLA2G5 robustly induced in adipocytes of obese mice [8]; [9]. Parts of the following text are embedded in the manuscript.

We decided to study in more detail the regulation of CD36 that encodes a very long chain fatty acids (VLCFA) transporter because CD36 is an important fatty acid transporter that facilitates fatty acids (FA) uptake by heart, skeletal muscle, and also adipose tissues [10]. PPARγ induces CD36 expression in adipose tissue, where it functions as a fatty acid transporter, and therefore, its regulation by PPARγ contributes to the control of blood lipids. Diacylglycerol acyltransferase 1 (DGAT1), a protein associated with the enterocytic triglyceride absorption and intracellular lipid processing is besides CD36 another target gene of adipogenesis master regulator PPARγ [11]. DGAT1 mRNA levels are strongly up regulated during adipocyte differentiation [12], its promoter region contains a PPARγ binding site and DGAT1 is also negatively regulated by the MEK/ERK pathway. DGAT1 expression was shown to be increased in TIS7 transgenic mice [13] and its expression was decreased in the gut of high fat diet-fed TIS7 KO mice [14]. Importantly, DGAT1 expression in adipocytes and WAT is up regulated by PPARγ activation [11].

Heatmap of hierarchical cluster analysis of intestinal gene expression involved in lipid transport altered in TIS7 SKMc15 dKO mice fed a high-fat diet for 3 weeks.

What is the upstream event explaining this defect?

Wnt pathway causes epigenetic repression of the master adipogenic gene PPARγ. There are three epigenetic signatures implicated in repression of PPARγ: increased recruitment of MeCP2 (methyl CpG binding protein 2) and HP-1α co-repressor to PPARγ promoter and enhanced H3K27 dimethylation at the exon 5 locus in a manner dependent on suppressed canonical Wnt. These epigenetic effects are reproduced by antagonism of canonical Wnt signaling with Dikkopf-1.

Zhu et al. showed that Dlk1 knockdown causes suppression of Wnt and thereby epigenetic de-repression of PPARγ [15]. Dlk1 levels positively correlate with Wnt signaling activity and negatively with epigenetic repression of PPARγ [16]. Activation of the Wnt pathway caused by DLK1 reprograms lipid metabolism via MeCP2-mediated epigenetic repression of PPARγ [17]. Blocking the Wnt signaling pathway abrogates epigenetic repressions and restores the PPARγ gene expression and differentiation [18].

**Minor comments:**

1)Please use the same font in the main text for the references.

Response: We thank the reviewer for the remark. This issue was corrected.

Reviewer #3 (Significance (Required)):

The study provides interesting insights into the role of these factors in adipocyte differentiation that would be relevant especially to researchers working on adipogenesis and cellular differentiation in general. The authors find the studied factors to have additive contribution to the differentiation efficiency. However, the exact nature of the roles and whether they are strictly speaking additive or synergistic is not clear. More detailed analysis of their contribution and molecular interplay would add to the broader interest of the study on molecular networks controlling cellular differentiation.

**Referees cross-commenting**

I very much agree on the different points raised by the other reviewers, some of which are also matching my own already raised concerns. And therefore it makes sense to request these modifications from the authors.

References

1. Rozman, J., M. Klingenspor, and M. Hrabe de Angelis, A review of standardized metabolic phenotyping of animal models. Mamm Genome, 2014. 25(9-10): p. 497-507.
2. Lefterova, M.I., et al., PPARgamma and the global map of adipogenesis and beyond. Trends Endocrinol Metab, 2014. 25(6): p. 293-302.
3. Garin-Shkolnik, T., et al., FABP4 attenuates PPARgamma and adipogenesis and is inversely correlated with PPARgamma in adipose tissues. Diabetes, 2014. 63(3): p. 900-11.
4. Joseph, R., et al., ACSL1 Is Associated With Fetal Programming of Insulin Sensitivity and Cellular Lipid Content. Mol Endocrinol, 2015. 29(6): p. 909-20.
5. Anderson, C.M. and A. Stahl, SLC27 fatty acid transport proteins. Mol Aspects Med, 2013. 34(2-3): p. 516-28.
6. Riedel, J., et al., Characterization of key genes of the renin-angiotensin system in mature feline adipocytes and during in vitro adipogenesis. J Anim Physiol Anim Nutr (Berl), 2016. 100(6): p. 1139-1148.
7. Zhou, S., et al., Increased missense mutation burden of Fatty Acid metabolism related genes in nunavik inuit population. PLoS One, 2015. 10(5): p. e0128255.
8. Wootton, P.T., et al., Tagging SNP haplotype analysis of the secretory PLA2-V gene, PLA2G5, shows strong association with LDL and oxLDL levels, suggesting functional distinction from sPLA2-IIA: results from the UDACS study. Hum Mol Genet, 2007. 16(12): p. 1437-44.
9. Sergouniotis, P.I., et al., Biallelic mutations in PLA2G5, encoding group V phospholipase A2, cause benign fleck retina. Am J Hum Genet, 2011. 89(6): p. 782-91.
10. Coburn, C.T., et al., Defective uptake and utilization of long chain fatty acids in muscle and adipose tissues of CD36 knockout mice. J Biol Chem, 2000. 275(42): p. 32523-9.
11. Koliwad, S.K., et al., DGAT1-dependent triacylglycerol storage by macrophages protects mice from diet-induced insulin resistance and inflammation. J Clin Invest, 2010. 120(3): p. 756-67.
12. Cases, S., et al., Identification of a gene encoding an acyl CoA:diacylglycerol acyltransferase, a key enzyme in triacylglycerol synthesis. Proc Natl Acad Sci U S A, 1998. 95(22): p. 13018-23.
13. Wang, Y., et al., Targeted intestinal overexpression of the immediate early gene tis7 in transgenic mice increases triglyceride absorption and adiposity. J Biol Chem, 2005. 280(41): p. 34764-75.
14. Yu, C., et al., Deletion of Tis7 protects mice from high-fat diet-induced weight gain and blunts the intestinal adaptive response postresection. J Nutr, 2010. 140(11): p. 1907-14.
15. Zhu, N.L., et al., Hepatic stellate cell-derived delta-like homolog 1 (DLK1) protein in liver regeneration. J Biol Chem, 2012. 287(13): p. 10355-10367.
16. Zhu, N.L., J. Wang, and H. Tsukamoto, The Necdin-Wnt pathway causes epigenetic peroxisome proliferator-activated receptor gamma repression in hepatic stellate cells. J Biol Chem, 2010. 285(40): p. 30463-71.
17. Tsukamoto, H., Metabolic reprogramming and cell fate regulation in alcoholic liver disease. Pancreatology, 2015. 15(4 Suppl): p. S61-5.
18. Miao, C.G., et al., Wnt signaling in liver fibrosis: progress, challenges and potential directions. Biochimie, 2013. 95(12): p. 2326-35.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

In the article, "The negative regulator DLK1 is transcriptionally regulated by TIS7 (IFRD1) and translationally by its orthologue SKMc15 (IFRD2)", the authors performed a double knockout (dKO) of TIS7 and its orthologue SKMc15 in mice and could show that those dKO mice had less adipose tissue compared to wild-type (WT) mice and were resistant to a high fat-diet induced obesity. The study takes advantage of number of different methods and approaches and combines both in vivo and in vitro work. However, some more detailed analysis and clarifications would be needed to fully justify some of the statements. Including the role of TIS7 as a transcriptional regulator of DLK1, SKMc15 as translational regulator of DLK1 and overall contribution of DLK1 in the observed differentiation defects. The observed results could still be explained by many indirect effects caused by the knock-outs and more direct functional connections between the studied molecules would be needed. Moreover, some assays appear to be missing biological replicates and statistical analysis. Please see below for more detailed comments:

Major comments:

-Are the key conclusions convincing? Yes.

-Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether? No.

-Would additional experiments be essential to support the claims of the paper? Yes. Please see my comments.

-Are the suggested experiments realistic in terms of time and resources? Recombinant DLK1 10 μg - Tetu-bio - 112€ ; 8 days of adipocyte differentiation in 3 biological replicate ~ 1 month.

-Are the data and the methods presented in such a way that they can be reproduced? Yes.

-Are the experiments adequately replicated and statistical analysis adequate? Adequately reproduced yes. Please see my comments concerning the statistical analysis.

1)Fig1a: In the method section it is written that an unpaired 2-tailed Student's t test was used for all statistical comparisons. However, here something like Multivariate analysis of variance (MANOVA) should rather be used to assess statistical significance between the mice. Moreover, the details of this should be clearly stated in the corresponding Figure legend.

2)Fig2a: please use an appropriate title for Fig2a instead of "Abdominal fat vs. body mass".

3)Fig2c: in the method section it is written that an unpaired 2-tailed Student's t test was used for all statistical comparisons. However, in Fig2c 4 groups are compared (WT, TIS7 KO, SKMc15 KO and dKO) and thus something like Multivariate analysis of variance (MANOVA) should rather be used to assess statistical significance.

4)Fig2 conclusion: Additive or just showing stronger effect?

5)Fig3a: the microscope picture for SKMc15 KO shows that cells might have died. Please state the percentage of cell death.

6)Fig3b: It would be informative to additionally observe some of marker genes for adipogenesis and whether all of them are affected.

7)Fig3b: instead of using an unpaired 2-tailed Student's t test with proportion, an one-way ANOVA would be more appropriate.

8)Fig3c: Same comment as for Fig3b.

9)Fig3d: A representative Western blot for 3 independent experiments is shown. Please add the other two as supplementary materials.

10)Fig3d:Is this distinguishing between the active and inactive catenin?

11)Fig4a: Please perform qPCR for measuring DLK-1 mRNA levels in TIS7 KO and SKMc15 KO samples to check whether there is a correlation between mRNA and protein level as the statement of the authors is that "DLK1 is transcriptionally regulated by TIS7 (IFRD1) and translationally by its orthologue SKMc15".

12)Fig4c: please add the other two western blots as supplementary materials.

13)Fig4d: The effects in MEFs appear quite modest. What about a rescue with TIS7 or SKMc15 alone?

14)Page 12, row 207: I would not call histones transcription factors.

15)Fig4e: Would be good to see a schematic overview of the locations of the ChIP primers in relation to the known binding sites and the gene (TSS, gene body). Moreover, the results include an enrichment for only one region while in the text two different regions are discussed. Importantly, to confirm the specificity of the observed enrichment, a primer pair targeting an unspecific control region not bound by the proteins should be included.

16)Fig5a: Has this experiment been replicated? That is no mention about the reproducibility or quantification of this result. This is the main experiment regarding the role of SKMc15 as a translational regulator of DLK1, also mentioned in the title of the manuscript.

17)Fig5b: Showing another unaffected secreted protein would be an appropriate control here.

18)Fig5c: I would recommend to perform additional experiments to prove that DLK-1 secreted in the medium can contribute and is responsible for the inhibition of the differentiation. Indeed, a time course of adipocyte differentiation followed by the addition of soluble DLK-1 would confirm that DLK-1 can inhibit adipocyte differentiation in this experimental setup. Moreover, silencing (for example RNAi) of DLK1 in the dKO cells before harvesting the conditioned media would allow to estimate the contribution of DLK1 to the observed inhibition of differentiation by the media. This is important because many other molecules could also be mediating this inhibition.

19)Fig5c: The details and the timeline of the experiment with conditioned media are not provided in the figure or in the methods. At what time point was conditioned media changed? How long were the cells kept in conditioned media? How does this compare to the regular media change intervals? Could the lower differentiation capacity relate to turnover of the differentiation inducing compounds in the media due to longer period between media change? Moreover, is the result statistically significant after replication?

20)Fig5d: please add a statistical analysis of the oil-re3d-o quantification.

21)Fig7c-d: Does overexpression also rescue the PPARg and CEBPa induction during differentiation. The importance of their induction in undifferentiated MEFs is a little difficult to judge.

22)Fig8: it is not surprising that PPARg targets are not induced in the absence of PPARg. What is the upstream event explaining this defect? Is DLK1 alone enough to explain the results? Could there be additional mediators of the differences? How big are transcriptome-wide differences between WT MEFs and dKO MEFs?

Minor comments:

1)Please use the same font in the main text for the references.

Significance

The study provides interesting insights into the role of these factors in adipocyte differentiation that would be relevant especially to researchers working on adipogenesis and cellular differentiation in general. The authors find the studied factors to have additive contribution to the differentiation efficiency. However, the exact nature of the roles and whether they are strictly speaking additive or synergistic is not clear. More detailed analysis of their contribution and molecular interplay would add to the broader interest of the study on molecular networks controlling cellular differentiation.

Referees cross-commenting

I very much agree on the different points raised by the other reviewers, some of which are also matching my own already raised concerns. And therefore it makes sense to request these modifications from the authors.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Summary:

In the current study, Vietor et al. aimed to explore the regulation of Delta-like homolog 1 (DLK-1), an inhibitor of adipogenesis, and demonstrated a role for TIS7 and its orthologue SKMc15 in the regulation of adipogenesis by controlling the level of DLK-1. Using mouse models with whole body deficiency of TIS7 (TIS7 KO) or SKMc15 (SKMc15KO) and double KO (TIS7 and SKMc15 dKO) mice, the authors used a combination of in-vivo experiments and cell culture experiments with mouse embryonic fibroblasts derived from the KO animals, to show that the concurrent depletion of TIS7 and SKMc15 dramatically reduced the amount of adipose tissues and protected against diet-induced obesity in mice, which was associated with defective adipogenesis in vitro.

Major Comments:

Overall, this study presents convincing evidence that TIOS7 and SKMc15 are necessary for optimal adipogenesis, and proposes a novel mechanism for the control of DLK1 abundance via coordinated regulation of DLK-1 transcription and translation. However, a number of questions remain largely unanswered. In particular, the direct ability of SKMc15 to regulate the translation of DLK-1 is lacking, and this claim remains speculative. SKMc15 being a general inhibitor of translation, SKMc15 may have an effect on adipogenesis independently of its regulation of DLK-1. Thus, addressing the following comments would further improve the quality of the manuscript:

•The experimental evidence supporting that SKMc15 controls DLK-1 protein levels comes primarily from the observations that DLK-1 abundance is further increased in SKMc15 KO and dKO WAT than in TIS7KO WAT (Fig 3d), and that translation is generally increased in SKMc15 KO and dKO cells (Fig 5a). However, since the rescue experiment is performed in dKO cells, by restoring both TIS7 and SKMc15 together, it is impossible to disentangle the effects on DLK-1 transcription, DLK-1 translation and on adipogenesis. A more detailed description of the TIS7 and SKM15c single KO cells, with or without re-expression of TIS7 and SKMc15 individually, at the level of DLK-1 mRNA expression and DLK-1 protein abundance would be necessary. In addition, polyribosome fractioning followed by qPCR for DLK-1 in each fraction, and by comparison with DLK-1 global expression in control and SKMc15 KO cells, would reveal the efficiency of translation for DLK-1 specifically, and directly prove a translational control of DLK-1 by SKMc15. Alternatively, showing that DLK-1 is among the proteins newly translated in SKMc15 KO cells (Fig. 5a) would be helpful.

•While the scope of the study focuses on the molecular control of adipogenesis by TIS7 and SKMc15 via the regulation of DLK-1, basic elements of the metabolic characterization of the KO animals providing the basis for this study would be useful. Since the difference in body weight between WT and dKO animals is already apparent 1 week after birth (Fig 1a), it would be interesting to determine whether the fat mass is decreased at an earlier age than 6 months (Fig 1b). The dKO mice are leaner despite identical food intake, activity and RER (Sup Fig 1). It remains unclear whether defective fat mass expansion is a result or consequence of this phenotype. Is the excess energy stored ectopically? The authors mention defective lipid absorption, however, these data are not presented in the manuscript. It would be interesting to investigate the relative contribution of calorie intake and adipose lipid storage capacity in the resistance to diet-induced obesity. In addition, data reported in Fig 1c seem to indicate a preferential defect in visceral fat development, as compared to subcutaneous fat. It would be relevant if the authors could quantify it and comment on it. Are TIS7 and SKMc15 differentially expressed in various adipose depots? The authors used embryonic fibroblasts as a paradigm to study adipogenesis. It would be important to investigate, especially in light of the former comment, whether pre-adipocytes from subcutaneous and visceral stroma-vascular fractions present similar defects in adipogenesis.

We estimate the suggested experiments above realistic in terms of time and resources and important to support the major conclusions of the study. Both data and methods are explained clearly. The experiments are, for the most part, adequately replicated. However, whenever multiple groups are compared, ANOVA should be employed instead of t-test for statistical analysis.

Minor comments:

•Figure 4 d. The appropriate control would be WT with empty vector

•Figure 7c/d. The appropriate control would be WT with empty vector

•Figure 5C. An additional control would be WT with WT medium

•Figure 2: In the legends, the "x" is missing for the dKO regression formula

•Since the role of SKMc15 in adipogenesis has never been described, the authors could consider describing the single SKMc15 KO in addition to the dKO, or explain the rationale for focusing the study on dKO.

Significance

While the effects of DLK-1 on adipogenesis have been widely documented, the factors controlling DLK-1 expression and function remain poorly understood. Here the authors propose a novel mechanism for the regulation of DLK-1, and how it affects adipocyte differentiation. This study should therefore be of interest for researchers interested in the molecular control of adipogenesis and cell differentiation in general. Furthermore, the characterization of the function of SKMc15 in the control of translation may be of interest to a broader readership.

Referees cross-commenting

I agree with all the comments raised by the other reviewers. Addressing the often overlapping but also complementary questions would help to clarify the molecular mechanisms by which TIS7 and SKMc15 control adipogenesis, and support the conclusions raised by the authors.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This study by Viedor et al. examines the role of TIS7 (IFRD1) and its ortholog SKMc15 (IFRD2) in the regulation of adipogenesis via their ability to modulate the levels of DLK1 (Pref-1), a well-known inhibitor of adipogenesis. They generate SKMc15 KO mice and cross them to previously published TIS7 KO mice. All 3 mutant strains show decreased fat mass, with the effect being most pronounced in double KO mice (dKO). Using mouse embryonic fibroblasts (MEFs) from mutant mice, they authors ascribe a defect in adipogenic differentiation of mutant cells to an upregulation of DLK-1. In the case of TIS7, they propose that this is due to its known inhibition of Wnt signaling, which regulates DLK-1 expression. In the case of SKMc15, they suggest a new mechanism linked to its ability to suppress translation. Overall, the work is of interest, with the finding that SKMc15 regulates adipocyte differentiation being its novelty, and generally well done, though multiple aspects need to be improved to bolster the conclusions put forth.

Major concerns:

1)The main mechanism put forth by the authors to explain the inability of dKO cells to differentiate into adipocytes is the upregulation of DLK-1 levels. However, this notion is never directly tested. Authors should test if knockdown of DLK-1 in dKO cells is sufficient to correct the defect in differentiation, or if additional factors are involved.

2)There are multiple instances were the authors refer to "data not shown", such as when discussing the body length of dKO mice. Please show the data in all cases (Supplementary Info is fine) or remove any discussion of data that is not shown and cannot be evaluated.

3)Indirect calorimetry data shown in Fig. S1 should include an entire 24 hr cycle and plots of VO2, activity and other measured parameters shown (only RER and food intake are shown), not just alluded to in the legend.

4)It is surprising that the dKO mice weight so much less than WT even though their food consumption and activity levels are similar, and their RER does not indicate a switch in fuel preference. An explanation could be altered lipid absorption. The authors indicate that feces were collected. An analysis of fat content in feces (NEFAs, TG) needs to be performed to examine this possibility. The discussion alludes to it, but no data is shown.

5)It would be important to know if increased MEK/ERK signaling and SOX9 expression are seen in fat pads of mutant mice, not just on the MEF system. Similarly, what are the expression levels of PPARg and C/EBPa in WAT depots of mutant mice?

6)Analysis of Wnt signaling in Fig. 3c should also include a FOPflash control reporter vector, to demonstrate specificity. Also, data from transfection studies should be shown as mean plus/minus STD and not SEM. This also applies to all other cell-based studies (e.g., Fig. 6b,c).

7)It is unclear why the authors used the MEF model rather than adipocyte precursors derived from the stromal vascular fraction (SVF) of fat pads from mutant mice. If they did generate data from SVF progenitors, they should include it.

8)Given that the authors' proposed mechanism involves both, transcriptional and post-transcriptional regulation of DLK-1 by TIS7 and SKMc15, Fig. 4d should be a Western blot capturing both of these events, and not just quantitation of mRNA levels.

9)There is no mention of the impact on brown adipose tissue (BAT) differentiation of KO of TIS7, SKMc15, or the combination. Given the role of BAT in systemic metabolism beyond energy expenditure, the authors need to comment on this issue.

Minor comments:

10)The y axis in Fig. 2c is labeled as gain of body weight (g). Is it really the case that WT mice gained 30 g of body weight after just 3 weeks of HFD? This rate of increase seems extraordinary, and somewhat unlikely. Please re-check the accuracy of this panel.

11)The Methods indicates all statistical analysis was performed using t tests, but this is at odds with some figure legends that indicate additional tests (e.g., ANCOVA).

12)Please specify in all cases the WAT depot used for the analysis shown (e.g., Fig. 3d is just labeled as WAT, as are Fig. 4a,e, etc.).

13)Fig. 5d is missing error bars, giving the impression that this experiment was performed only once (Fig. 5c). The legend has no details. Please amend.

Significance

The role of TIS7 in adipocyte differentiation is well established. The only truly novel finding in this work is the observation that SKMc15 also plays a role in adipogenesis. The molecular mechanisms proposed (modulation of DLK-1 levels) are not novel, but make sense. However, they need to be bolstered by additional data.

Referees cross-commenting

I think we are all in agreement that the findings in this work are of interest, but that significant additional work is required to discern the mechanisms involved. In my view, a direct and specific link between SKMc15 and translation of DLK-1 needs to be established and its significance for adipogenesis in cells derived from the SVF of fat pads determined. Reviewer 2 has suggested some concrete ways to provide evidence of a direct link.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

'The authors do not wish to provide a response at this time.'

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary: The manuscript analyzes how the constriction of a tissue by an enveloping basement membrane alters the migration of cells migrating through that tissue. The tissue analyzed is the Drosophila egg chamber, an important model for basement membrane studies in vivo, and the cells migrating through it are the border cells. The border cells migrate through the center of the egg chamber, moving as a cluster between the nurse cells, which are in turn surrounded by follicle cells, which secrete the basement membrane on the outside of the egg chamber. The authors decrease and increase the basement membrane stiffness with various genetic perturbations, and they find that the border cells move more rapidly when the stiffness is reduced. They then investigate how basement membrane stiffness is communicated to the border cells several cell layers inside, by measuring cortical tension with laser-recoil. They found that external basement membrane stiffness alters the cortical tension of the nurse cells and the follicle cells, such that reduced matrix stiffness causes reduced cortical tension; further, reducing cortical tension directly within the cells also results in increased border cell migration rates. They conclude that basement membrane stiffness can alter cell migration in a new way, by altering constriction and cortical tension, with an inverse relationship between stiffness and migration rate. This is a strong manuscript and I would request very few changes.

The authors are commended on the rigor and completeness of their study. Several independent methods are used to alter basement membrane stiffness (loss of laminin, knock-down of laminin, knock-down of collagen IV, over-production of collagen IV - all of which end up changing collagen IV levels) and all show the same result. Further, they are extremely rigorous about testing and excluding an attractive alternative hypothesis, that the basement membrane of the border cell cluster itself controls its migration rate. The use of mirror-Gal4 is very elegant and convincing, as it expressed only in the central part of the egg chamber, and they found border cells responded differently only in that region. Moreover, the authors were exceptionally thorough in reproducing the basement membrane mechanical data in their own hands using the bursting assay. Overall, the experimental data support the claims of the paper. There is only one more control I would like to see, for the knockdown of laminin in the border cell cluster with a triple-Gal4 combination. Presumably using all three Gal4 lines was necessary to get complete knockdown, and it would be nice to see anti-laminin for the border cell cluster under these knockdown conditions.

Despite the rigor, because all of the manipulations to the basement membrane alter the levels of collagen IV, the authors cannot formally exclude the possibility that collagen IV in the basement membrane has another function besides stiffness, perhaps sequestering a signaling ligand, and that this other function somehow alters the cortical tension of the egg chamber. In the paper by Crest et al, externally applied collagenase served as a control for this possibility, but collagenase will not work for the authors because this study is in vivo. I suggest the authors bring up this caveat in the discussion. If they wanted to extend the study (optional), they could knock down the crosslinking enzyme peroxidasin in the egg chamber, which ought to reduce basement membrane stiffness without changing the collagen content. The problem here is that it hasn't already been shown to work that way in the egg chamber, and so both stiffness and collagen levels would need to be measured. Testing the stiffness directly would be difficult, since the bursting assay is not actually a measurement of stiffness (more on that below). Rather than go this route, I suggest just acknowledging the formal possibility, which seems to me unlikely anyway.

In terms of clarity, the manuscript absolutely needs a schematic at the beginning to introduce the egg chamber and border cell migration, labeling the cell types, showing the route and direction of border cell migration, and labeling the A/P axis. Without this the non-expert reader cannot readily understand the study.

Finally, in terms of clarity, the authors repeatedly use statements such as "stiffness influences migration rate". Influences how? These results are not intuitive to me, and it would help enormously if the authors would make statements like, decreasing stiffness increases migration (as I tried to in my summary). Here are two examples of statements to refine: • Line 189 - "We found that reducing laminin levels affected the migration speed of both phases (Fig.1F, G)." Please say increased, not affected. • Line 245 -"Altogether, these results demonstrate that the stiffness of the follicle BM influences dynamics and mode of BC migration." Again, be specific about how. There are many such statements, from the abstract to the results to the discussion, where it would help the clarity to be more precise about what kind of influence.

Minor comments: • The movies are beautiful! • All the quantitative data are shown in bar charts with means and errors. It is much better to show the individual data points, superimposing the means and distributions on top of the individual points. • The bursting assay does not actually measure basement membrane stiffness; rather, it measures failure after elastic expansion. These are related, as was found by Crest et al and the authors say that at one point, but stiffness and failure are not the same thing. Please change the language discussing this assay to "mechanical properties" rather than stiffness. • The laser-recoil assays are done well and are convincing. Throughout the results section, the authors describe these as measuring "cortical tension", which is correct. However, in the figure legends the language changes to "membrane tension" which is only one component of cortical tension. Change them all to cortical tension. • In the Discussion, it would be nice to include something on the two different modes of migration (tumbling and not tumbling). • I suggest changing the title to remove the word "forces", because forces are never directly measured from basement membrane. • Although Dai et al (Science 2020) is discussed near the end, I suggest bringing this reference up to the introduction, so the reader can have the background on the mechanical aspects of border cell migration at the start of this study. • Two typos (there may be more): At the bottom of Fig. 2, text turns strangely white that should probably be black; and in line 260, you mean Fig. S5 not S4 (laser ablation).

Significance

Mechanobiology, and mechanobiology of the basement membrane, is a vibrant area of study now, arising from the intersection of biophysics/engineering and genetics. There is general interest in how the basement membrane alters forces within the tissue, and this study is the first to my knowledge to relate basement membrane mechanics to migration via constriction and cortical tension. The authors do a great job of discussing the broader significance of their work in the Discussion. To greatly broaden the scope of this work in the future, the authors could collaborate with a mouse team to look for similar responses in a mammalian tissue, as they discuss. It is worth noting that there is a lot of work on matrix stiffness and migration showing that stiffness promotes migration speed; in these cases, matrix is a substrate, not a compression mechanism. But the opposite nature of the result in interesting and makes this work non-intutive and perhaps hard for some readers to grasp.<br /> As the paper is written now, I think the audience for this work would mostly be oogenesis, border cell migration, and/or basement membrane researchers in the Drosophila community, of which there are many (I am in this camp). With some rewriting to make it more accessible to other audiences, I think it would be interesting to a larger developmental biology audience. The content is not like any other paper I know, but it may be similar in scope and subject matter to the papers detailing how follicle cells and basement membrane interact during follicle rotation.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

Ester and her colleagues described a force model that controls border cell migration by varying the stiffness of the basement membrane. It's based on the modification of laminin and Coll IV which are components of the basement membrane. To reduce BMs stiffness, they introduced LanB1RNAi or the LanB1 mutant to inhibit laminin production or vkgRNAi to reduce Coll IV. They also applied EHBP1mCh to enhance BM stiffness. Furthermore, they applied laser ablition to confirm that the BM stiffness affects the tension of nurse cells and follicle cells, thus regulating border cell behavour by changing environmental properties. It is a nice work revealing how the environment controls border cell migration; however, there are several points that concern me: 1. It's reported that actin polymerization at the front of the cell generates protrusions, as well as that myosin contractility helps to suppress lateral random protrusions, thus leading to a directed and efficient cell migration. So why do more lateral protrusions (tj>LanB1RNAi) produce a faster migration speed? 2. We know some labs also did experiments with those Kel/Dic mutant flies. And the Kel mutant is very sick, which sometimes leads to NC degeneration. As a result, we have serious doubts that this mutant's border cell migration will remain normal. 3. From figure4, we noticed that with mirrGal4, the vertex distance increase is much lower than tjGal4 (control of D, H and K), and even with expressing the EHPB1mCh, the distance is still lower than the tjGal4 control. These indicate the NC cortical tension is lower with mirrGal4 expression, which is patially against the paper's main point. (Similar issue in figure5 D and E). 4. Sfigure1 A and B seem not to have the right contrast (the blue and the red should have the same brightness), so the comparison of the intensities might be inaccurate and needs to be requantified after adjustment of the images. 5. Sfigure2 A-E showed that the vkgRNAi has the highest bursting frequency, whereas F and G do not. And the majority of the data from F does not fit with A-E, and it is unclear what timepoint sF.G. is at. 6. SFigure 6 only displayed a representative image of the control condition; the lack of representative images for the other conditions resulted in unconvincing results. 7. Some figures and movies have prominent variation of migrating stages, such as not-detached border cells compared with detached border cells. This might strongly cause the results inconsistent with each other. 8. There are numerous typos in both the manuscript and the figures. Based on all these concerns, I recommend authors to do some improvement before this manuscript is accepted by some reputed journals.

Significance

Strengths: the manuscript is well written and organised; limitation: figures and results are not supportive enough, and thus conclusion is not completely convincing, statistical quantification is not clear and somehow confusing.

Advice: If the conclusion is solid, this story will fill the unclear importance of surrounding environment on cell-rich tissue for collective cell migration. Concept is very novel while needing more supporting data. It is a fundamental study for development biology.

Audience: The story will fit well for developmental and cell biology, as well as people with biomechanical background.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

The manuscript by Lopez et al have employed the ex vivo model of Border cell migration in Drosophila ovaries to examine the constriction forces imposed by basement membrane on migrating border cells. The authors have extensively employed live cell imaging coupled with genetics to demonstrate that basement membrane encasing fly eggs modulates the dynamics of migrating border cells. Through laser ablation experiments they show that basement influences the tension of the underlying follicle cell and nurse cells which in turn affects the migration efficiency of collectively moving border cells. Over all the experiments are well quantified with good degree of statistics that drive the claim of the authors.

Major comments: Over in all the images, it is very hard to appreciate the overall contour of the egg chamber. This is important to get an insight regarding the stage of the egg chamber being evaluated.

1. By depleting the constituents of basement membrane, the authors show that the speed of the migrating border cells increases. However in Supplementary Figure 1A and B where that authors have depleted LanB1, the migrating border cell cluster seems to lag while the control has reached the oocyte boundary. Is this a single off phenomena?

2. This is regarding the osmotic swelling experiments. The frequency and speed of bursting of egg chambers in deionized water was used to evaluate the stiffness of basement membrane in different genetic background. As egg chambers of different stages have variable sizes, it would be fair to evaluate egg chambers of only a particular stage for this analysis as the tonocity of the egg chambers may depend on their size.

3. Line 211, "Live time lapse imaging, showed that the overexpression of EHBP1mCh in all FCs delayed BC migration (tslGFP; tj> EHBP1mCh, Figure S4A-B', Movie S4, n=6)." Though the border cell cluster hasn't moved significantly in Fig S4B', the egg chamber development seems to be stalled as the movement of main body follicle cells is affected. My concern if over expression of EHBP1mCh in the follicle cells is stalling the oogenesis itself could that indirectly affect the border cell movement. Secondly though EHBP1 has been shown to affect secretion of the basement membrane constituents, it could also modulate asymmetric secretion of other components. Can the authors evaluate if over expression of EHBP1mCh rescues the delay in migrating border cells in Lanb1 heterozygous background to render stronger support to their claim.

4. In Supplementary Fig 7 B and B' the nurse cell morphology seems to be affected. Could the distorted nurse cell morphology in the abi-depleted germline cell affecting the migration efficiency of border cells.

5. Line 313-314 The authors state that "The radius of curvature of a spherical interface is inversely proportional to the difference in pressure between the two sides of the interface." This may be applicable to a smooth surface but may be not directly applicable to the cell membrane as there are local regional variations and thus any inference on the cytoplasmic pressure of nurse cells may be misleading.

Minor comment:

1. In supplementary figure 6D, the square boxes are obscuring the border cell membrane and it will be better if the authors can modify the figure to render more clarity.
2. There are couple of places where sentence structure needs to be corrected.

Referees cross-commenting

I agree with all the comments of other reviewers. Overall I also feel that results do not strongly support the main conclusions. The authors draw major conclusions based on experiments that are merely suggestive rather than being conclusive. Some of the concerns are listed. Like Reviewer 3 raising the concern that Collagen IV may have other functions in the basement membrane other than providing stiffness. A similar concern I too have raised regarding over-expression of EHBP1. I agree with Reviewer 3 that there are several other factors that can affect the outcome of bursting assay besides the stiffness of the basement membrane itself. So the authors need to be careful in linking the bursting frequency of the egg chamber with the stiffness of the basement membrane itself.

I agree with other reviewers that the quality of the images need to be better. In addition, the image presented should be representative of the population and should fit with the over claim made by the authors (Point No 3 of Reviewer 2 and Point No 1 of Reviewer1). I also agree that authors need to explain Reviewer 2's concern (Point No-1) as to why the lateral protrusion in tj>LanB1RNAi doesn't inhibit the movement of border cell clusters but rather produce faster migration speeds?

Lastly it is important for the authors to verify that like Kel/Dic mutants are indeed effective or any genetic perturbation like overexpression of EHBP1mCh is not the stalling the oogenesis progression perse, thus giving a false impression of altered migratory speed of border cell clusters.

Significance

The role basement membrane is well documented in affecting the shape of neighbouring cells Here the authors claim that the stiffness of basement membrane is regulating the migration efficiency of the border cells. I believe that basement membrane encasing the follicle and underlying germline cells provides a very narrow passage for the border cells to migrate. Any mechanical perturbation that releases or increases the pressure or make the nurse cells membranes less or more taut will affect the dynamics of migrating border cells. Though the authors have demonstrated this with very elegant experiments, I am afraid that their findings are standard outcomes in any physically constrained system and somehow doesn't significantly advance the field.

Transparent Peer Review

---

Download the complete Review Process [PDF] including:

• reviews
• authors' reply
• editorial decisions

</br>

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

Manuscript number: RC-2022-01776

Corresponding author(s): David Bryant

1. General Statements [optional]

We describe an ARF6 GTPase module that controls integrin recycling to drive invasion in PTEN-null Ovarian Cancer (OC). We used high-throughput, time-lapse imaging and machine learning to characterise spheroid behaviours from a series of cell lines modelling common genetic lesions in OC patients. We identified that PTEN loss was associated with increased invasion, the formation of invasive protrusions enriched for the PTEN substrate PI(3,4,5)P3, and enhanced recycling of integrins in an ARF6-dependent matter. We utilised Mass Spectrometry proteomics and unbiased labelling to investigate the interactome of ARF6, identifying a single ARF GAP (AGAP1) and a single ARF GEF (CYTH2). Importantly, this ARF6-AGAP1-CYTH2 modality was associated with poor clinical outcome in patients.

We thank all Reviewers for their highly complementary assessment of our manuscript, describing our paper as a "very impressive study, very well done and controlled with rigorous statistical analyses that uses sophisticated methods", a study that is "stunning in its thoroughness and depth and breadth of its molecular analysis", with "experiments are properly designed, and the data are well presented. The conclusions are appropriate and supported by the data". Finally, we would like to thank the reviewers for appreciating that our results are "of significance for both scientific discovery and clinical application, which will interest the broad audience in both basic and clinical research".

2. Description of the planned revisions

Insert here a point-by-point reply that explains what revisions, additional experimentations and analyses are planned to address the points raised by the referees.

Reviewer comments in bold. Our response in non-bold.

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

This paper by Konstantinou et al aims at deciphering the mechanisms by which PTEN loss could be driving poorer prognosis in patients. The authors use their great high-throughput 3D screening method coupled to an unbiased proteomic method and a CRISPR screen to uncover a new pro-invasive axis driving collective invasion of high-grade serous ovarian carcinoma (HGSOC) cells. Overall, this is a very impressive study, very well done and controlled with rigorous statistical analyses that uses sophisticated methods to convincingly show that the CYTH2-ARF6-AGAP1-ITGA6/ITGB1 module is required for the pro-invasive effect of PTEN depletion and discriminates patients with poorest prognosis.

__

MAJOR COMMENTS __

Below are listed all the claims that, in my opinion, are not adequately supported by the data.

1) Choice of the cell line: More justification on the use of the ID8 cell line and on the p53 deletion is needed. The authors need to clearly state that most p53 mutations in ovarian cancer are missense mutations that lead to a strong accumulation of a p53 protein devoid of transcriptional activity. Nevertheless, it seems that p53 mutations are not associated to differences in patient survival. Hence the choice of studying PTEN loss in the complete absence of p53, a situation that does not mirror the clinical situation, needs to be explained. Moreover, the in vivo experiments already performed in the literature mentioned in the discussion should be mentioned in the introduction to provide more context and physiological relevance to this study (especially regarding the special focus on the p53 null/ dKO cells throughout the study).

We will update the manuscript with a detailed explanation of the cell line of choice. Briefly, while indeed Tp53 is found mutated in HGSOC, approximately 30-35 % of these are classified as null mutations (PMID: 21552211), making models with null Trp53 representative of the clinical situation. Further, there is no difference in patient outcome in HGSOC by Tp53 mutation type (PMID: 20229506), while gene expression data from TCGA suggest that HGSC is marked by loss of wild-type P53 signalling regardless of Tp53 mutation type (PMID 25109877). Thus, we believe our choice of model can faithfully mirror the clinical situation.

2) "Therefore, PTEN loss in ovarian cancer, particularly at the protein level, occurs in the tumour epithelium and is associated with upregulated AKT signalling and poor overall survival". This claim is an over-interpretation and over-generalisation of the data presented. I appreciate the honesty of the authors in showing all the ovarian datasets that are available and highlight the discrepancies in expression of the proteins they study in stroma and epithelium. I think the way to present these data in the text without over-interpreting and generalizing would be to show that there is a clear epithelial-specific downregulation of PTEN at the mRNA level. Most likely due to the contribution to other cell types in the stroma, only 3 out of 5 bulk tumour mRNA datasets show a tumour specific downregulation of PTEN and no association with survival based on a median split of PTEN mRNA expression. Nevertheless, although there is no direct correlation between PTEN mRNA and protein levels, patients with low PTEN protein levels have poorer survival that is associated to an upregulation of Akt signalling. This allows to have a clearer conclusion, based solely on the protein data presented and no over-generalisation using the mRNA data. This, to me, makes a stronger case for studying PTEN loss in ovarian cancer and is fully supported by the data presented.

We will incorporate this reviewer suggestion into the modified manuscript.

3) PTEN loss induces modest effects in 2D culture. The authors make claims regarding the fact that some of the phenotypes they look at happen after PTEN depletion alone or in combination with p53 loss and are more prominent in 3D vs 2D. Many of these are insufficiently backed up by data. A few key experiments are also only performed in 2D and should be done in 3D. Finally, some clarifications if the role of PTEN is most prominent on either collective, ECM-induced or 3D-dependent invasion.

some clarifications if the role of PTEN is most prominent on either collective, ECM-induced or 3D-dependent invasion

We believe that the reviewer may be confused. Both of our models, either spheroids or invading monolayers, are events occurring inside gels of ECM. Therefore, these are all are 3D, ECM-induced, collective invasion. We have not performed 2D migration assays. We apologise that the this was not clearer in the first submission. We will correct this in the updated manuscript.

First, the authors claim that PTEN loss alone (i.e. without p53 deletion) leads to changes in Akt signalling. Supp fig 1H clearly shows that there is no significant increase in Akt activation, although there seems to be one in the Western Blot (WB) presented in supp fig 1G. There is a clear, significant increase in the Akt activation in all the PTEN KO clones when in association with p53 loss though. This claim is hence not backed up by data and the conclusion seems to be that the effect on Akt signalling requires both deletion of p53 and PTEN.

The reviewer is correct: that the increase to pAKT levels upon PTEN KO is more robust with co-KO of TP53, thereby indicating synergy with p53. We will update the manuscript to note this, accordingly.

It will be interesting to see a quantification of the pS473-Akt staining (supp fig S1J), as it seems from these images that pAkt is preferentially found on rounded cells. It should also be performed in 3D conditions to see if there is an enrichment at invasive tips and back-up the invasion data.

This observation made us realise that the images we had included were giving the wrong impression (that pAkt levels would be highest in round cells). Based on the quantitation in Fig. S1M, PTEN KO cells (which have elevated pAkt levels), show a marked depletion of rounded cells. Therefore, pAkt elevated is not associated with being enriched in rounded cells. We will replace this image with cells mirroring the phenotypes quantified in Fig S1M.

We used 2D for quantitation of pAKT staining, as we perform a like for like comparison. We cannot compare pAkt in 3D protrusions accurately between genotypes because of the frequency of protrusions: in p53 KO protrusion are rare. In 3D, therefore, it is not a situation where protrusions are present in both genotypes and we compare enrichment or depletion in a stable structure. Rather, what we can provide is whether when protrusions form, there is clear pAkt labelling in a protrusion. We will include for the revision a representative image of each phenotype in 3D, including a 3D Trp53-/-;Pten-/- spheroid stained for pAKT S473.

Arf6 is recruited to the invasive tips of cells invading a 2D wound (fig4D). How do the authors reconcile the fact that all the machinery required for 3D invasion is present but that PTEN loss has a modest effect on cells in 2D? If the wound assay was done on glass, it should be done again on ECM coated glass to see if it recapitulates the effects seen in 3D. This experiment will help deconvolute if the effect of PTEN loss is more linked to collective behaviour than 3D organization or presence of ECM.

We again apologise for not being clearer in our description. Both the wound assays and the IF of invading monolayer were performed with cell monolayers invading into Matrigel. Monolayers are grown on top of Matrigel, wounded, and then overlayed with Matrigel. Therefore, this is orthogonal to our spheroid assay, and completely 3D. We will address this comment by changing the text in the results section to highlight the 3D nature of the method.

The recycling assays are all done in 2D, condition under which the authors claim that the PTEN phenotype is weakest. Although I understand that it is not possible to do this assay in 3D, its contribution to elucidating the mechanism by which integrins participate in the PTEN loss invasive phenotype is not clear. The requirement of integrins relies on the data showing that ITGB1 KO results in no collagen4-positive basement membrane of the cysts and greatly impaired invasion. Experiments looking at the integrin localisation would be helpful: can an enrichment at the invasive tips can be seen? Are ITGA6 and/or ITGB1 repartitions homogeneous between the cysts membranes and the invasive tips? In my opinion the Src/FAK data is not enough to draw the conclusions of fig7I schematic.

We will endeavour to include images of 3D spheroids of Trp53-/-;Pten-/- cells and stained for β1 integrin (total and active) and α5 integrin to interrogate localisation at the tips.

4) Expression of AGAP1 isoforms do not alter ARF6 levels. Data in fig 6C, D show a significant downregulation of Arf6 and Akt signalling after expression of AGAP1S. Can the authors clarify what they mean?

We thank the reviewer for picking up that discrepancy between the results and the text. We will change the relevant text to highlight that expression of AGAP1S is associated with a statistically significant reduction of roughly 30% in ARF6 levels and 10% in p:t AKT. We do not know why AGAP1s may enact such an effect.

5) Arf6 is not modulated in the different cell lines: data in fig4B (far right graph) and supp fig 4B, J seem to indicate otherwise. Can the authors clarify what they mean?

It is not clear exactly what the reviewer is referring to here. If the reviewer is referring to Supplementary Figure 4B, this is an experiment examining the levels of ARF5 or ARF6 upon knockdown, so levels would be expected to vary. Fig S4B does not correspond to the experiment performed in S4J. Our interpretation is that loss of p53 alone or in combination with Pten does not seem to be consistently be accompanied with an increase in either the levels of total or bulk GTP-bound ARF6 that could explain the dependency of Trp53-/-;Pten-/- on the GTPase for the invasive phenotype. We will make our interpretation clearer in the text

6) Immunofluorescence panels without quantifications: Quantifications for the different stainings shown in fig3A; 4D, E; 5H; 7B and supp fig S1L, J; S3 need to be included to fully back the conclusions of the authors. Indeed, these images are used to draw conclusions and not only as illustrations.

It is not possible to do a direct comparison between protrusion vs no protrusion (see our response above). We will include a line scan to show clear enrichment at the end of the tip for image shown. Quantitation for Figure S1L is already included (S1K and M), quantitation for Figure S1J is presented in Fig S1I and for Fig 5H quantitation of the phenotype is present in Fig 5I.

7) Quantifications of invasion show that WT cysts become hyper-protrusive at around the half experiment mark (around 30-40hrs). Nevertheless, all movies or galleries show spherical cysts, which does not seem representative. Can the authors change this or explain why these images/movies were chosen?

We present the fold change at each time point because that is intuitively easier to understand rather than the raw number. The quantitation does not show that the cysts necessarily become hyper-protrusive at the specific timepoint, but rather that the proportion of hyper-protrusive cysts observed in this genotype peaks at the specific timepoint. This phenotype may still be in the minority of behaviours. As an example, something that occurs 5% of the time in the control, with a two-fold increase in behaviours, might still only be 10% of the population. Therefore, adding in a picture that may be representative of a small proportion of the population may not be a realistic depiction of what is happening across the entire population. We will provide the reviewer with the exact percentage of spheroids that are classified as hyper-protrusive at the specific cell line across timepoints, to make this clearer.

8) Since it seems that the main effect of PTEN is to drive the localisation and intensity of recycling of Arf6 cargoes, it will be helpful to confirm that all the proteins involved in the Arf6 module be shown to be accumulated/present at the pro-invasive tips. Immunofluorescence stainings showing the presence of AGAP1 (could be done with the AGAP1S isoform that is mNeon-tagged), pS473-Akt, ITGB1 (active integrin if possible, otherwise total integrin), ITGA5, PI3K should be included if possible. A quantification comparing signal in the cysts and in the invasive tips should also be included to see if there is an accumulation to PIP3-enriched areas.

We will endeavour to include the requested images.

9) Data in fig5I convincingly show that PTEN loss induces a fragmented collagen4-positive basement membrane. The authors use this data to claim that this is one of the ways that PTEN could be driving invasion but no correlation between these structures and the hyper-protrusive phenotype is made. This experiment needs to be done to support this claim.

This comment made us realise that in an attempt to make images simpler (displayed nuclei and COL4 only), we omitted a staining for where protrusions were moving through gaps in the ECM. We will update these times to demonstrate such events.

__

MINOR COMMENTS __

1) Data visualization: I think that the heatmap representation is overkill when only 2 or 3 conditions are presented. A graph showing the evolution of area or spherical/Hyper-protrusive phenotype proportions across time would be easier to read and more impactful: each genotype could be presented with a colour and the spherical/hyper-protrusive phenotypes as either plain or dashed lanes across time. I understand that this representation allows for the stats to be done at each time points but they are generally pretty clear (especially for the PTEN KO or dKO phenotypes) and do not need to be done for each time point in my opinion. These heatmaps could be put in supplementary figures if the authors feel strongly about putting stats for each time points.

We thank their reviewer for their suggestion. We believe that our approach, while complex, is the best visualisation to reflect both the changes across time but also between conditions while allowing appreciation of the statistical significance. This visualisation has been optimised by our lab over years of working with this type of data and we would prefer that they remain consistent with the accepted standard of our other publications. We are, however, happy to expand the explanation in the text on how to interpret the bubble heatmaps.

Fig supp S1M, fig 5I should be presented as a stacked histogram to improve readability and merged with fig supp S1K.

We will merge Figures S1M and S1K. We believe that Figure 5I is easier to read as is.

Displaying fold change as antilog rather than log values would be easier for the reader to realise the magnitude of the differences.

We disagree with the reviewer.

A bar graph would be easier to read than the matrix representation for fig 6B.

We disagree with the reviewer as we feel it makes it easier to directly compare each lipid between the two cell lines.

The way Area data is presented throughout to me makes it very difficult to understand what is going on. Could the authors at least give some explanations in figure legends. A curve graph displaying the evolution of the area across time would be easier to read and see the differences between conditions.

Please see our response to Minor point 1

2) It is confusing that, in fig supp S1M, there is a significant decrease of the rounded phenotype after PTEN loss that is not associated to a significant change in another of the categories. Could the authors explain how?

This can be simply explained from our data: while the rounded phenotype was reduced in a consistent way across replicate experiments (therefore resulting in significance), the effect on the other two phenotypes was not consistent (not set in magnitude and directionality). This therefore does not lead to a significant (i.e. consistent) effect on the latter two phenotypes. PTEN loss therefore seems to allow cells to undergo – at the expense of being round - a range of shape changes, rather than a set phenotype.

3) One of the big differences of the PTEN KO cells seems their ability to invade through the matrigel bed and migration on the glass below (supp movie S2). From what I gather, these cysts would be considered out of focus and excluded from the analysis. Would it be possible that this would minimize some of the results? Would it be possible to include a quantification of this particular phenotype to confirm it is specific to PTEN KO cells?

In the same spirit, could the authors provide the percentage of non-classified cysts, to make sure that the same proportion of cysts is quantified across all different genotypes.

Indeed, we cannot exclude that we under-estimate the magnitude of the effect on the PTEN null. We will include this point in the discussion. We can include a reviewer-only figure showing the proportion of cysts and levels of the ‘OutOfFocus’ objects across cell lines.

__

4) Can the authors clarify how a 0 fold change (in log value) in fig 2D can be highly significant? __

We believe that the reviewer is equating statistical significance with something being biologically meaningful. Statistical analysis does not indicate a priori whether something is biologically meaningful. Rather, it assesses the likelihood that an observed result is occurring by chance (or not). For instance, if a small change (e.g 0.04 in a log2 fold change) occurred repeatedly across experimental replicates this is unlikely to be a result of chance, and therefore could be statistically significant. Yet, such a small magnitude of effect is probably biologically minor. This is why our heatmaps provide both statistical significance, fold change, and consistency in magnitude of effect.

5) Delta isoform of PI3K seems to have an effect on area in the middle of the experiment, but has no effect at all on invasion. Could the authors comment? Are these smaller cysts still as invasive? There might be an interesting uncoupling between proliferation and invasion there.

The cysts are actually slightly larger with PI3Kδ inhibition and there is no change in invasion. We will expand our comments in text as well to account for this observation.

6) ITGB1 depletion seems to induce a downregulation of Akt protein. Is that right? Does it change Akt localisation? Is there a dose effect whereby there is not enough Akt protein to mediate invasion?

The p:t AKT ratio does not change consistently across all gRNAs (Figure 5C) but we can look at Akt (total) protein levels and include this information if needed.

__

7) Stats should be added directly on the graphs for the recycling assays, doing a pairwise comparison of the different genotypes for each time points. Can the authors clarify what the t-32min quantification graphs adds (fig7E, supp fig S8G-I)? I would advise to remove them, as this data is already presented in the recycling assay graphs. __

We don't include these because although they are technical replicates, they are demonstrative of a single experiment. What we include instead is the quantitation across independent biological experiments (which each have their own internal multiple technical replicates), where it is appropriate to include statistical analysis.

8) There is a substantial amount of typos and erroneous references to figures. I listed below the ones that I spotted and I encourage the authors to carefully check.

1. there are some mistakes in referencing the number of cysts in supp table 1. There is for example no cysts experiments in Figure 1 but yet there are some references to figure 1 in supp table 1. Please correct it. I think it will be easier for the reader if the number of cysts quantified for each conditions was also indicated in the figure legends. Supp table 1 can still be included for readers that want additional details.
2. comma missing page 3
3. page 3 and 4: PI(3,4)2 means PI(3,4)P2? Can be shorten to PIP2 for ease of read and specify if it is another PIP2 specie otherwise
4. define CYTH abbreviation: I suppose this is for cytohesin?
5. fig1F-I: don't understand why TCGA.OV is specified on some but not all the graphs. It seems to me that all the data are from TCGA.OV? Makes it seems it is nit the case
6. legend of fig1H, I: y axis is -Log10 values in 1I, not Log10 values
7. page 6: dKO abbreviation is already specified above and should be used to avoid repetition and for ease of read
8. supp fig S1D: missing legend for the second bar (after Wild Type)
9. supp fig S1N: legend of the X-axis should be below the axis
10. supp fig S1O: the numerotation of the X-axis needs to be below the line of the axis for ease of read, not above it
11. legend of S2A: clones 1.12 and 1.15 are p53-/-;PTEN-/- and not PTEN-/-
12. supp figS2C can the authors specify the different stages of matrigel (liquid or gel) that are used for the invasion assay, to make it easier for the non-specialist to understand what is going on. Please confirm that the 50% GFR matrigel makes a gel on top of the cells and fill in the wound to produce the 3D invasion assay setup.
13. page 7: no parental cells are used in S3A, B only p53 null and p53 null and dKO. Please also specify what cells are being compared in the text
14. description of arrow heads and colours need to be moved to figure legends and not in main text (page 7)
15. fig 2D: the signification of the dot in the circles needs to be in the legends (since it is its first apparition in the manuscript). It only appears later on, in supp2A legend. Additional description of the matrices is necessary, as they contain a lot of information to digest to understand fully what is going on
16. legend of fig3: error in figure reference: area data is D and not E, protrusive phenotypes are E and not F
17. arrow missing in fig3B
18. fig 3D,E, G, H: please indicate the cell line studied
19. fig 3I: the different genotypes need to be stated on the galleries for clarity
20. page 8: define Arf6-mNG in the text
21. __ page 9: "We thank the reviewer for their careful examination of the manuscript. We will go through all above points and make the corresponding careful adjustments to the manuscript.

OPTIONAL SUGGESTIONS

1) Choice of cell line: There is a high number of patients (around 9% according to (Cole et al. 2016)) that present the R248Q gain-of-function mutation. A recent study has shown that this mutant p53 protein is associated to an activation of Akt signalling and an increase of the intercellular trafficking of EGFR (Lai et al. 2021). Given that EGFR was also a hit in this screen, that is seems to have a central role in Arf6 cargoes (fig 4G), I think it would be a great addition to this study. It could hence cooperate with PTEN loss to drive strong, robust invasion.

This is an excellent observation and one we will likely follow-up in an independent study.

2) Are MAPK involved in the PTEN KO pro-invasive phenotype? In particular Erk1/2, since EGFR is one of the PTEN loss induced Arf6 cargoes.

This is an excellent observation and one we will likely follow-up in an independent study.

__

REFERENCE Cole, Alexander J., Trisha Dwight, Anthony J. Gill, Kristie-Ann Dickson, Ying Zhu, Adele Clarkson, Gregory B. Gard, et al. 2016. « Assessing Mutant P53 in Primary High-Grade Serous Ovarian Cancer Using Immunohistochemistry and Massively Parallel Sequencing ». Scientific Reports 6 (1): 26191. _https://doi.org/10.1038/srep26191_.

Lai, Zih-Yin, Kai-Yun Tsai, Shing-Jyh Chang, et Yung-Jen Chuang. 2021. « Gain-of-Function Mutant TP53 R248Q Overexpressed in Epithelial Ovarian Carcinoma Alters AKT-Dependent Regulation of Intercellular Trafficking in Responses to EGFR/MDM2 Inhibitor ». International Journal of Molecular Sciences 22 (16): 8784. _https://doi.org/10.3390/ijms22168784_. __

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The authors have conducted a study of the molecular requirements for cancer invasion that is stunning in its thoroughness and depth and breadth of its molecular analysis. The writing is exceptionally precise though also very dense (see below). The molecular model proposed is that PTEN loss (in a p53 null background) leads to reliance upon ARF6 for invasion, with regulation through interactions with AGAP1 and beta1-integrin and it is convincingly demonstrated. They focus on interpreting the consequences of genetic and pharmacologic manipulations in a cell line, using a series of 2D and 3D assays. The phenotypes are more prominent in 3D assays.

Concerns and Suggestions:

• There is a disconnect between the essentially complete loss of protrusions and invasion in 3D (e.g. 4A) and the reduction in magnitude of protrusive invasion but the continued presence of elongated cells with protrusions in 2D (e.g. S4C). This discrepancy is present in a couple of comparisons and is glossed over in quick callouts to many figure panels.

  We thank the reviewer for mentioning this as this comment was very helpful in determining that we needed to clarify our description of the role of ARF6 to protrusion formation vs maturation. In the Trp53-/- genotype, protrusions can form, but they rapidly retract, failing to mature into structures that drive invasion through ECM (e.g. Figure S2E). This protrusion maturation occurs upon PTEN KO. When ARF6 depleted, PTEN-null cells can form protrusions, but now again lack the ability to mature into invasion-inducing structures.

This concept of needing ARF6 for protrusion maturation and maintenance is underpinned by our model of ARF6 regulating recycling of active integrin back to the protrusion front. Indeed, we have observed ARF6 being required not for protrusion initiation, but rather ensuring protrusions are not retracted in other contexts (i.e. upon loss of the ARF6 GEF protein IQSEC1 in invading 3D culture of PC3 cells; PMID: 33712589).

We also note that, as responded to Reviewer 1, the assay is a 3D invasion rather than 2D migration assay, with cells sandwiched between Matrigel.

We will update the relevant sections of the results and discussion with the point above.

Once a journal has been identified, it would be wise for the editor to allow some flexibility in word limit to enable some very dense sections to be expanded slightly to guide the reader through the experiments and results more clearly. For example, in the section "ARF6 regulates active integrin pools...", there are callouts like (Fig. 7C,E; S8A-C; G-I) and then (Fig. 7D,E; Fig. S8E-F, H-I). It takes a lot of time to unpack these different experimental claims based on a single sentence.

We greatly appreciate the refreshing comments of this reviewer to advocate for actions to improve clarity in our reporting. We would take glad advantage of such a possibility.

The patient data on CYTH2 and its relationship to survival is modestly convincing.

In Ovarian Cancer, effects on survival are often minor. This is not a disease where one often sees large shifts in survival, which is why we are so excited about the large shifts that we do see with the ARF GTPase module we identified. However, we concede that the effects on CYTH2, although significant, are not vast changes. We will point this out and tone down our language.

Very minor- search on %- there are a few inconsistencies in terms of spaces and commas vs. periods. The Methods also have some inconsistencies in terms of spaces between numbers and units or numbers and degrees Celsius. References are also in a different font. Overall it was extremely carefully written though (just dense).

We thank the reviewer for their careful inspection of our manuscript. We will carefully go over the sections flagged before resubmission

Reviewer #2 (Significance (Required)):

One limitation of the experimental design is that the depth of molecular analysis in vitro comes at the expense of any in vivo validation, which the authors acknowledge in the Discussion. They attempt to make similar points using analysis of patient survival data from public databases but these analyses generally yielded small magnitude differences. The main audience for this study is likely to be cell biologists interested in cell migration, cell-ECM adhesion, cancer invasion, and GTPases. I don't see any need for new experiments- what can be done has been done and then some. I do think that it would benefit readers if the text could be made less dense.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

Summary: Using a murine HGSOC 3D cell model, in combination with analysis of human ovarian cancer datasets, the authors uncover a CYTH2-ARF6-AGAP1 signaling module regulated by PTEN and identify a biomarker for tumor invasion and targeted therapy.

Major comments:

__The findings of this study are significant as they reveal a critical signaling module that controls tumor invasion by mediating tumor cell interaction with the extracellular matrix. The experiments are properly designed, and the data are well presented. The conclusions are appropriate and supported by the data. The limitation of the study has also been discussed properly.

One suggestion regarding the survival analysis in Fig. 6 and 7. __

The authors noted that the CYTH2-ARF6-AGAP1 module is not specifically or only induced in Pten-null contexts, but rather that Pten-null cells become more dependent on the module for enacting the invasive phenotype. Based on this, it would be interesting to evaluate how the PTEN status impacts the survival difference by integrating the PTEN genomic status (WT versus mutation) or its expression level (protein or mRNA) into the survival analysis of patient cohorts in Fig. 6 and Fig. 7.

We thank the reviewer for this excellent point. We will include such analysis, where possible. One consideration will be that extensive division of patients based on these molecular characteristics may results in patient numbers too low to draw conclusions of significance.

**Referees cross-commenting**

Gene deletion and mutation may elicit different functional outcomes. I therefore agree with Reviewer #1 that "the choice of studying PTEN loss in the complete absence of p53, a situation that does not mirror the clinical situation, needs to be explained".

We will make our reasons for this choice clear in the text before submission. Please refer to response to Reviewer 1, Major comment 1.

Reviewer #3 (Significance (Required)):

The model used and data presented in this study are of significance for both scientific discovery and clinical application, which will interest the broad audience in both basic and clinical research.

3. Description of the revisions that have already been incorporated in the transferred manuscript

4. Description of analyses that authors prefer not to carry out

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

Summary:

Using a murine HGSOC 3D cell model, in combination with analysis of human ovarian cancer datasets, the authors uncover a CYTH2-ARF6-AGAP1 signaling module regulated by PTEN and identify a biomarker for tumor invasion and targeted therapy.

Major comments:

The findings of this study are significant as they reveal a critical signaling module that controls tumor invasion by mediating tumor cell interaction with the extracellular matrix. The experiments are properly designed, and the data are well presented. The conclusions are appropriate and supported by the data. The limitation of the study has also been discussed properly.

One suggestion regarding the survival analysis in Fig. 6 and 7. The authors noted that the CYTH2-ARF6-AGAP1 module is not specifically or only induced in Pten-null contexts, but rather that Pten-null cells become more dependent on the module for enacting the invasive phenotype. Based on this, it would be interesting to evaluate how the PTEN status impacts the survival difference by integrating the PTEN genomic status (WT versus mutation) or its expression level (protein or mRNA) into the survival analysis of patient cohorts in Fig. 6 and Fig. 7.

Referees cross-commenting

Gene deletion and mutation may elicit different functional outcomes. I therefore agree with Reviewer #1 that "the choice of studying PTEN loss in the complete absence of p53, a situation that does not mirror the clinical situation, needs to be explained".

Significance

The model used and data presented in this study are of significance for both scientific discovery and clinical application, which will interest the broad audience in both basic and clinical research.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The authors have conducted a study of the molecular requirements for cancer invasion that is stunning in its thoroughness and depth and breadth of its molecular analysis. The writing is exceptionally precise though also very dense (see below). The molecular model proposed is that PTEN loss (in a p53 null background) leads to reliance upon ARF6 for invasion, with regulation through interactions with AGAP1 and beta1-integrin and it is convincingly demonstrated. They focus on interpreting the consequences of genetic and pharmacologic manipulations in a cell line, using a series of 2D and 3D assays. The phenotypes are more prominent in 3D assays.

Concerns and Suggestions:

1. There is a disconnect between the essentially complete loss of protrusions and invasion in 3D (e.g. 4A) and the reduction in magnitude of protrusive invasion but the continued presence of elongated cells with protrusions in 2D (e.g. S4C). This discrepancy is present in a couple of comparisons and is glossed over in quick callouts to many figure panels.
2. Once a journal has been identified, it would be wise for the editor to allow some flexibility in word limit to enable some very dense sections to be expanded slightly to guide the reader through the experiments and results more clearly. For example, in the section "ARF6 regulates active integrin pools...", there are callouts like (Fig. 7C,E; S8A-C; G-I) and then (Fig. 7D,E; Fig. S8E-F, H-I). It takes a lot of time to unpack these different experimental claims based on a single sentence.
3. The patient data on CYTH2 and its relationship to survival is modestly convincing.
4. Very minor- search on %- there are a few inconsistencies in terms of spaces and commas vs. periods. The Methods also have some inconsistencies in terms of spaces between numbers and units or numbers and degrees Celsius. References are also in a different font. Overall it was extremely carefully written though (just dense).

Significance

One limitation of the experimental design is that the depth of molecular analysis in vitro comes at the expense of any in vivo validation, which the authors acknowledge in the Discussion. They attempt to make similar points using analysis of patient survival data from public databases but these analyses generally yielded small magnitude differences. The main audience for this study is likely to be cell biologists interested in cell migration, cell-ECM adhesion, cancer invasion, and GTPases. I don't see any need for new experiments- what can be done has been done and then some. I do think that it would benefit readers if the text could be made less dense.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #1

Evidence, reproducibility and clarity

This paper by Konstantinou et al aims at deciphering the mechanisms by which PTEN loss could be driving poorer prognosis in patients. The authors use their great high-throughput 3D screening method coupled to an unbiased proteomic method and a CRISPR screen to uncover a new pro-invasive axis driving collective invasion of high-grade serous ovarian carcinoma (HGSOC) cells. Overall, this is a very impressive study, very well done and controlled with rigorous statistical analyses that uses sophisticated methods to convincingly show that the CYTH2-ARF6-AGAP1-ITGA6/ITGB1 module is required for the pro-invasive effect of PTEN depletion and discriminates patients with poorest prognosis.

Major comments

Below are listed all the claims that, in my opinion, are not adequately supported by the data.

1. Choice of the cell line: More justification on the use of the ID8 cell line and on the p53 deletion is needed. The authors need to clearly state that most p53 mutations in ovarian cancer are missense mutations that lead to a strong accumulation of a p53 protein devoid of transcriptional activity. Nevertheless, it seems that p53 mutations are not associated to differences in patient survival. Hence the choice of studying PTEN loss in the complete absence of p53, a situation that does not mirror the clinical situation, needs to be explained. Moreover, the in vivo experiments already performed in the literature mentioned in the discussion should be mentioned in the introduction to provide more context and physiological relevance to this study (especially regarding the special focus on the p53 null/ dKO cells throughout the study).
2. "Therefore, PTEN loss in ovarian cancer, particularly at the protein level, occurs in the tumour epithelium and is associated with upregulated AKT signalling and poor overall survival". This claim is an over-interpretation and over-generalisation of the data presented. I appreciate the honesty of the authors in showing all the ovarian datasets that are available and highlight the discrepancies in expression of the proteins they study in stroma and epithelium. I think the way to present these data in the text without over-interpreting and generalizing would be to show that there is a clear epithelial-specific downregulation of PTEN at the mRNA level. Most likely due to the contribution to other cell types in the stroma, only 3 out of 5 bulk tumour mRNA datasets show a tumour specific downregulation of PTEN and no association with survival based on a median split of PTEN mRNA expression. Nevertheless, although there is no direct correlation between PTEN mRNA and protein levels, patients with low PTEN protein levels have poorer survival that is associated to an upregulation of Akt signalling. This allows to have a clearer conclusion, based solely on the protein data presented and no over-generalisation using the mRNA data. This, to me, makes a stronger case for studying PTEN loss in ovarian cancer and is fully supported by the data presented.
3. PTEN loss induces modest effects in 2D culture. The authors make claims regarding the fact that some of the phenotypes they look at happen after PTEN depletion alone or in combination with p53 loss and are more prominent in 3D vs 2D. Many of these are insufficiently backed up by data. A few key experiments are also only performed in 2D and should be done in 3D. Finally, some clarifications if the role of PTEN is most prominent on either collective, ECM-induced or 3D-dependent invasion.

First, the authors claim that PTEN loss alone (i.e. without p53 deletion) leads to changes in Akt signalling. Supp fig 1H clearly shows that there is no significant increase in Akt activation, although there seems to be one in the Western Blot (WB) presented in supp fig 1G. There is a clear, significant increase in the Akt activation in all the PTEN KO clones when in association with p53 loss though. This claim is hence not backed up by data and the conclusion seems to be that the effect on Akt signalling requires both deletion of p53 and PTEN.

It will be interesting to see a quantification of the pS473-Akt staining (supp fig S1J), as it seems from these images that pAkt is preferentially found on rounded cells. It should also be performed in 3D conditions to see if there is an enrichment at invasive tips and back-up the invasion data.

Arf6 is recruited to the invasive tips of cells invading a 2D wound (fig4D). How do the authors reconcile the fact that all the machinery required for 3D invasion is present but that PTEN loss has a modest effect on cells in 2D? If the wound assay was done on glass, it should be done again on ECM coated glass to see if it recapitulates the effects seen in 3D. This experiment will help deconvolute if the effect of PTEN loss is more linked to collective behaviour than 3D organization or presence of ECM.

The recycling assays are all done in 2D, condition under which the authors claim that the PTEN phenotype is weakest. Although I understand that it is not possible to do this assay in 3D, its contribution to elucidating the mechanism by which integrins participate in the PTEN loss invasive phenotype is not clear. The requirement of integrins relies on the data showing that ITGB1 KO results in no collagen4-positive basement membrane of the cysts and greatly impaired invasion. Experiments looking at the integrin localisation would be helpful: can an enrichment at the invasive tips can be seen? Are ITGA6 and/or ITGB1 repartitions homogeneous between the cysts membranes and the invasive tips? In my opinion the Src/FAK data is not enough to draw the conclusions of fig7I schematic. 4. Expression of AGAP1 isoforms do not alter ARF6 levels. Data in fig 6C, D show a significant downregulation of Arf6 and Akt signalling after expression of AGAP1S. Can the authors clarify what they mean? 5. Arf6 is not modulated in the different cell lines: data in fig4B (far right graph) and supp fig 4B, J seem to indicate otherwise. Can the authors clarify what they mean? 6. Immunofluorescence panels without quantifications: Quantifications for the different stainings shown in fig3A; 4D, E; 5H; 7B and supp fig S1L, J; S3 need to be included to fully back the conclusions of the authors. Indeed, these images are used to draw conclusions and not only as illustrations. 7. Quantifications of invasion show that WT cysts become hyper-protrusive at around the half experiment mark (around 30-40hrs). Nevertheless, all movies or galleries show spherical cysts, which does not seem representative. Can the authors change this or explain why these images/movies were chosen? 8. Since it seems that the main effect of PTEN is to drive the localisation and intensity of recycling of Arf6 cargoes, it will be helpful to confirm that all the proteins involved in the Arf6 module be shown to be accumulated/present at the pro-invasive tips. Immunofluorescence stainings showing the presence of AGAP1 (could be done with the AGAP1S isoform that is mNeon-tagged), pS473-Akt, ITGB1 (active integrin if possible, otherwise total integrin), ITGA5, PI3K should be included if possible. A quantification comparing signal in the cysts and in the invasive tips should also be included to see if there is an accumulation to PIP3-enriched areas. 9. Data in fig5I convincingly show that PTEN loss induces a fragmented collagen4-positive basement membrane. The authors use this data to claim that this is one of the ways that PTEN could be driving invasion but no correlation between these structures and the hyper-protrusive phenotype is made. This experiment needs to be done to support this claim.

Minor comments

1. Data visualization: I think that the heatmap representation is overkill when only 2 or 3 conditions are presented. A graph showing the evolution of area or spherical/Hyper-protrusive phenotype proportions across time would be easier to read and more impactful: each genotype could be presented with a colour and the spherical/hyper-protrusive phenotypes as either plain or dashed lanes across time. I understand that this representation allows for the stats to be done at each time points but they are generally pretty clear (especially for the PTEN KO or dKO phenotypes) and do not need to be done for each time point in my opinion. These heatmaps could be put in supplementary figures if the authors feel strongly about putting stats for each time points.

Fig supp S1M, fig 5I should be presented as a stacked histogram to improve readability and merged with fig supp S1K.

Displaying fold change as antilog rather than log values would be easier for the reader to realise the magnitude of the differences.

A bar graph would be easier to read than the matrix representation for fig 6B.

The way Area data is presented throughout to me makes it very difficult to understand what is going on. Could the authors at least give some explanations in figure legends. A curve graph displaying the evolution of the area across time would be easier to read and see the differences between conditions. 2. It is confusing that, in fig supp S1M, there is a significant decrease of the rounded phenotype after PTEN loss that is not associated to a significant change in another of the categories. Could the authors explain how? 3. One of the big differences of the PTEN KO cells seems their ability to invade through the matrigel bed and migration on the glass below (supp movie S2). From what I gather, these cysts would be considered out of focus and excluded from the analysis. Would it be possible that this would minimize some of the results? Would it be possible to include a quantification of this particular phenotype to confirm it is specific to PTEN KO cells?

In the same spirit, could the authors provide the percentage of non-classified cysts, to make sure that the same proportion of cysts is quantified across all different genotypes. 4. Can the authors clarify how a 0 fold change (in log value) in fig 2D can be highly significant? 5. Delta isoform of PI3K seems to have an effect on area in the middle of the experiment, but has no effect at all on invasion. Could the authors comment? Are these smaller cysts still as invasive? There might be an interesting uncoupling between proliferation and invasion there. 6. ITGB1 depletion seems to induce a downregulation of Akt protein. Is that right? Does it change Akt localisation? Is there a dose effect whereby there is not enough Akt protein to mediate invasion? 7. Stats should be added directly on the graphs for the recycling assays, doing a pairwise comparison of the different genotypes for each time points. Can the authors clarify what the t-32min quantification graphs adds (fig7E, supp fig S8G-I)? I would advise to remove them, as this data is already presented in the recycling assay graphs. 8. There is a substantial amount of typos and erroneous references to figures. I listed below the ones that I spotted and I encourage the authors to carefully check.

• a. there are some mistakes in referencing the number of cysts in supp table 1. There is for example no cysts experiments in Figure 1 but yet there are some references to figure 1 in supp table 1. Please correct it. I think it will be easier for the reader if the number of cysts quantified for each conditions was also indicated in the figure legends. Supp table 1 can still be included for readers that want additional details.
• b. comma missing page 3
• c. page 3 and 4: PI(3,4)2 means PI(3,4)P2? Can be shorten to PIP2 for ease of read and specify if it is another PIP2 specie otherwise
• d. define CYTH abbreviation: I suppose this is for cytohesin?
• e. fig1F-I: don't understand why TCGA.OV is specified on some but not all the graphs. It seems to me that all the data are from TCGA.OV? Makes it seems it is nit the case
• f. legend of fig1H, I: y axis is -Log10 values in 1I, not Log10 values
• g. page 6: dKO abbreviation is already specified above and should be used to avoid repetition and for ease of read
• h. supp fig S1D: missing legend for the second bar (after Wild Type)
• i. supp fig S1N: legend of the X-axis should be below the axis
• j. supp fig S1O: the numerotation of the X-axis needs to be below the line of the axis for ease of read, not above it
• k. legend of S2A: clones 1.12 and 1.15 are p53-/-;PTEN-/- and not PTEN-/-
• l. supp figS2C can the authors specify the different stages of matrigel (liquid or gel) that are used for the invasion assay, to make it easier for the non-specialist to understand what is going on. Please confirm that the 50% GFR matrigel makes a gel on top of the cells and fill in the wound to produce the 3D invasion assay setup.
• m. page 7: no parental cells are used in S3A, B only p53 null and p53 null and dKO. Please also specify what cells are being compared in the text
• n. description of arrow heads and colours need to be moved to figure legends and not in main text (page 7)
• o. fig 2D: the signification of the dot in the circles needs to be in the legends (since it is its first apparition in the manuscript). It only appears later on, in supp2A legend. Additional description of the matrices is necessary, as they contain a lot of information to digest to understand fully what is going on
• p. legend of fig3: error in figure reference: area data is D and not E, protrusive phenotypes are E and not F
• q. arrow missing in fig3B
• r. fig 3D,E, G, H: please indicate the cell line studied
• s. fig 3I: the different genotypes need to be stated on the galleries for clarity
• t. page 8: define Arf6-mNG in the text
• u. page 9: "<" symbol should be an alpha symbol
• v. fig 4A: indicate the cell line used on the figure
• w. supp fig S4E: why is it specified mouse-specific for the shArf6?
• x. 4H, I, J: indicate on the figure if these interactors are mostly unchanged, strong interactors or weak interactors for clarity
• y. legend of fig4H: "coloured spots underneath denote the protein complex that each interactor belongs (in J)" should indicate panel G and not J
• z. fig4I, J: are you sure of the legend for the fold change coloring? Log2 of 1 is a 0 fold change, i don't see how these could show any significant difference (i.e. some of the pale red circles are significant)
• aa. page 11: description of the assay (starting with "Machine learning classification of...") is very confusing, please clarify
• bb. page16: figure 4H should be 4I (PTEN-null specific association of Arf6 with ITGA5)
• cc. supp fig S5H-P: choose Tumour or cancer to homogeneise naming across the graphs
• dd. fig 5H: box are difficultly visible in green, change color to yellow or something more visible
• ee. page 13: Fig6E, F should also refer to 6G
• ff. LCM abbreviation on page 10 and 12 refers to LCMD? Otherwise please define it.

Optional suggestions

1. Choice of cell line: There is a high number of patients (around 9% according to (Cole et al. 2016)) that present the R248Q gain-of-function mutation. A recent study has shown that this mutant p53 protein is associated to an activation of Akt signalling and an increase of the intercellular trafficking of EGFR (Lai et al. 2021). Given that EGFR was also a hit in this screen, that is seems to have a central role in Arf6 cargoes (fig 4G), I think it would be a great addition to this study. It could hence cooperate with PTEN loss to drive strong, robust invasion.
2. Are MAPK involved in the PTEN KO pro-invasive phenotype? In particular Erk1/2, since EGFR is one of the PTEN loss induced Arf6 cargoes.

Reference

Cole, Alexander J., Trisha Dwight, Anthony J. Gill, Kristie-Ann Dickson, Ying Zhu, Adele Clarkson, Gregory B. Gard, et al. 2016. « Assessing Mutant P53 in Primary High-Grade Serous Ovarian Cancer Using Immunohistochemistry and Massively Parallel Sequencing ». Scientific Reports 6 (1): 26191. https://doi.org/10.1038/srep26191.

Lai, Zih-Yin, Kai-Yun Tsai, Shing-Jyh Chang, et Yung-Jen Chuang. 2021. « Gain-of-Function Mutant TP53 R248Q Overexpressed in Epithelial Ovarian Carcinoma Alters AKT-Dependent Regulation of Intercellular Trafficking in Responses to EGFR/MDM2 Inhibitor ». International Journal of Molecular Sciences 22 (16): 8784. https://doi.org/10.3390/ijms22168784.

Significance

It has only been recently appreciated that PTEN loss is a driver in ovarian cancer (Martins et al. 2020) but no studies to data have aimed at understanding the mechanisms. This study is hence the first to propose one and as such provides a very valuable advance for researchers interested in ovarian cancer. The authors also propose that the CYTH2-ARF6-AGAP1 high mRNA be used as a signature of worsen prognosis. This hence paves the way to better understanding and stratify patients with ovarian cancers.

One of the main difference after PTEN loss is the accumulation of PIP3 in pro-invasive tips that correlates with the recruitment of Arf6 to these tips. The authors have developed a very powerful automated quantification pipeline to follow the behaviour of cysts grown in 3D that they have coupled to an unbiased proteomic method to identify interactors and a CRISPR screen to test their functional relevance. This is clearly the strongest aspect of the paper that allows them to gather very robust data and identify the machinery driving invasion in PTEN KO cells. The authors' model claims that this in turns recruit the Arf6 machinery, composed of CYTH2 2G (the only CYTH2 isoform correlated to a poorer prognosis, preferentially binding PIP3) and AGAP1 that leads to a local increase in active integrin recycling that mediates the more invasive phenotype of PTEN depleted cells. It is rightfully mentioned in the discussion that PTEN depletion only leads to a modest change in Arf6 interactors, and that most likely PTEN loss acts by locally directing the Arf6 machinery to the invasive tips. Indeed, the authors convincingly show that Arf6, AGAP1 and ITGB1 are required for the formation of these invasive protrusions.

The limitation of this study is the combination of 2D and 3D experiments to drive general conclusions on the mechanism. These are listed in the previous section. Another big limitation, in my opinion, is the choice of the cell model: indeed, nearly all patients present a vast increase in the amount of the p53 protein present due to a very large number of mutations that in most cases prevent its binding to DNA. Throughout this paper the authors have used a p53 null cell line that expresses no p53 protein. This is not compatible with the clinical situation. Moreover, since p53 also present frequent gain-of-function mutations that have been shown to be associated to an increase of Akt signalling and intercellular trafficking of EGFR. Studying the implication of the Arf6 module identified here in a context of p53 WT or mutant protein overexpression would be of great interest.

Reference

Martins, Filipe Correia, Dominique-Laurent Couturier, Anna Paterson, Anthony N. Karnezis, Christine Chow, Tayyebeh M. Nazeran, Adekunle Odunsi, et al. 2020. « Clinical and Pathological Associations of PTEN Expression in Ovarian Cancer: A Multicentre Study from the Ovarian Tumour Tissue Analysis Consortium ». British Journal of Cancer 123 (5): 793‑802. https://doi.org/10.1038/s41416-020-0900-0.

Note: This rebuttal was posted by the corresponding author to Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Reply to the reviewers

1. General Statements [optional]

We are grateful to the reviewers for highlighting the value and power of our 3D chimeric dataset to explore cancer/stellate interactions in pancreatic cancer invasion. We also appreciate their support of our findings identifying divergent roles for the two related enzymes ADAMTS2 and ADAMTS14. We thank the reviewers for their detailed comments, which have allowed us to prepare a significantly stronger and clearer manuscript.

Following the reviewers comments we have made three major changes to the manuscript, which we will outline here in addition to the point-by-point rebuttal.

1. i) Revised manuscript structure. We have modified the structure of the manuscript, which we hope improves the clarity and accessibility of the work.

Figure 1 remains the description of our 3D invasion model and our approach to identify stellate cell and cancer cell transcriptomic information from this context.

Figure 2 describes our focus on proteases and now includes concordance of our data with clinical data sets. This is also now where we describe the strikingly opposing roles for ADAMTS2 and ADAMTS14 in regulating invasion.

Figure 3 is now the figure demonstrating that ADAMTS2 and ADAMTS14 have an equal contribution to collagen processing from stellate cells. This is an important experiment given that the main physiological roles for these enzymes are in the processing of collagen, and the importance of collagen for cancer progression. It was therefore reasonable to hypothesise that the effect of these enzymes on invasion could be due to differences in their collagen processing in this context. The finding that both have an equal effect on collagen processing points towards a wider, and more diverse, role for these enzymes in regulating biology.

Figure 4 describes the divergent roles of these two enzymes on myofibroblast differentiation, and by extension TGFβ bioavailability. In this figure we now include experiments with TGFβ reporter constructs, which demonstrate an increase in active TGFβ following loss of ADAMTS14 and a reduction in TGFβ activity following loss of ADAMTS2.

Figure 5 is our matrisomic experiment to identify enriched enzyme-specific substrates following knockdown of either ADAMTS2 or ADAMTS14.

Figure 6 details our investigation into the substrate responsible for the reduction in invasion following loss of ADAMTS2. As the previous matrisomic experiment identified only two enriched ADAMTS2 substrates, we investigated both in our 3D assays, identifying SERPINE2 as the responsible substrate. Further analysis identified a reduction in plasmin activity in ADAMTS2 deficient cells. This was rescued with co-knockdown of SERPINE2, implicating this pathway as being crucial for mediating the effect of ADAMTS2. Additionally, we now include experiments demonstrating that concomitant knockdown of SERPINE2 alongside ADAMTS2 rescues the reduction in TGFβ activity observed with ADAMTS2 loss alone.

Figure 7 describes our analysis of ADAMTS14 substrates. As the matrisomics identified a large change in proteins following ADAMTS14 knockdown, we performed an siRNA screen of candidates to identify those responsible for ADAMTS14 phenotype. This, followed by further validation in our 3D invasive assay, revealed Fibulin2 as the responsible substrate. Fibulin2 has a well-established role in regulating TGFβ release from the matrix. In accordance with this we present new data using TGFβ reporter constructs, which demonstrate that the increase in active TGFβ following ADAMTS14 knockdown can be reversed with co-knockdown of Fibulin2.

1. ii) Improvement of the clinical significance of our chimeric data set and ADAMTS proteins. Ideally, we would like to present IHC images of ADAMTS2 and ADAMTS14 expression in PDAC tissue samples to corroborate our in vitro findings. However as these enzymes are secreted, this precludes antibody based imaging, as it would not provide cell type specific information. RNA scope presents an alternative, however we have experienced technical issues with this technique due to RNA degradation in PDAC tissue and unavailability of ADAMTS2/14 specific probes. In place of this we have used a range of publically available resources.

We have compared our chimeric data set with human clinical data using the resource published by Maurer and colleagues (PMID: 30658994). This paper presents transcriptomic data from PDAC tumour and stromal compartments using laser microdissection of clinical tissue. In accordance with our data set, the majority of metzincins, including ADAMTS2 and ADAMTS14, are expressed in the stromal compartment. These data are presented in updated figure 2.

We have also examined ADAMTS2 and ADAMTS14 expression in PDAC and CAF subtypes using publically available data sets. Using the TCGA dataset, we identified that ADAMTS2 and ADAMTS14 are highly expressed in PDAC tumours compared to normal counterparts. As the majority of PDAC is comprised of stroma, the bulk transcriptomic data from TCGA, combined with the results from the Maurer publication, lead us to conclude that this expression reflects the stromal origin of these proteases. In addition, using publically available single cell RNA sequencing data published by Luo and colleagues (PMID: 36333338), we identified ADAMTS2 and ADAMTS14 expression in the prominent PDAC CAF subtypes, inflammatory and myofibroblastic CAFs. Together these data demonstrate that these enzymes are enriched in clinical disease, which when combined with our mechanistic 3D studies implies a greater role for these enzymes in disease progression than previously appreciated.

iii) Improved mechanistic link between ADAMTS2 and ADAMTS14 with TGFβ bioavailability

To strengthen the association between ADAMTS2 and ADAMTS14 function, their substrates SERPINE2 and Fibulin2, and TGFβ bioavailability, we have performed the following experiments using TGFβ reporter constructs:

We have taken conditioned media from stellate cells lacking either ADAMTS2 or ADAMTS14, along with co-knockdown of their substrate, and stimulated a recipient cell line expressing a SMAD Luciferase reporter. These cells express luciferase in response to TGFβ stimulation. In accordance with a role for ADAMTS14 and Fibulin2 in regulating TGFβ, we demonstrate that following ADAMTS14 knockdown there is a strong increase in active TGFβ in the media (Figure 4I), which is abrogated with co-knockdown of Fibulin2 (Figure 7F).

We have also obtained a fluorescent reporter, CAGA-eGFP, which expresses GFP in response to TGFβ stimulation in order to examine TGFβ activity in 3D cultures. Stellate cells expressing this construct were embedded in collagen: Matrigel hydrogels following knockdown of either ADAMTS2 or ADAMTS14 and CAGA fluorescence recorded after 72 hours of culture. In accordance with our data, stellate cells deficient in ADAMTS14 showed increased fluorescence in 3D, indicative of increased TGFβ activity, which was abrogated with co-knockdown of Fibulin2 (Figure 4J, K and 7G, H). Equally, loss of ADAMTS2 reduced TGFβ activity in 3D culture, which was rescued with co-knockdown of SERPINE2 (Figure 4J, K and 6 D, E).

These experiments confirm a link between the ADAMTS enzyme, its relevant substrate, and TGFβ bioavailability. Together with extensive published work linking SERPINE2 and Fibulin2 with TGFβ release we are confident in our proposed mechanism for the dichotomic relationship of ADAMTS2 and ADAMTS14 in regulating TGFβ and thus myofibroblast action.

2. Point-by-point description of the revisions

This section is mandatory. *Please insert a point-by-point reply describing the revisions that were already carried out and included in the transferred manuscript. *

• *

Reviewer #1 (Evidence, reproducibility and clarity (Required)):

• *

This study aims to explain the opposing contributions of stromal stellate cells/CAFs to PDAC. By first identifying stroma-specific proteases, followed by a process of candidate selection and elimination, the authors find that two specific metalloproteases that share enzymatic activity against collagen in fact have differential activity on TGFb availability. This could be interpreted as a way of shaping the CAF population and tumor-promotin or -restricting properties of the stroma.

There are several flaws that the authors could address to improve the manuscript:

1. In the flow of experiments and analyses, there is a strange mix of fully unbiased discovery phases followed by functional experiments that do not consider all possible candidates to test, and vice versa. For instance, from the mixed-species transcript analysis, ADAMTS2 and -14 are chosen based on their shared collagenase activity based on literature. However, the authors then perform again a proteomics analysis to identify things from the entire matrisome that are cleaved by these enzymes? Then, for ADAMTS2 a co-silencing approach is done on one selected candidate (Serpine2), but for ADAMTS14 an siRNA screen is performed? The problem of this approach is that the rationale for some studied enzymes is very strong, where as for others it is not.

We thank the reviewer for their comment and trust the revised manuscript provides more clarity for the rationale of our approach. We performed the chimera sequencing as a discovery experiment to reveal the communication between cancer and stellate cells in a 3D, invasive context. We present the chimera experiment and data here as a resource for the community, with our analysis of ADAMTS2 and ADAMTS14 function serving as a first example of the biological insight this data set can reveal. Other insights revealed from this dataset are active avenues of research in our group.

Our finding that ADAMTS2 and ADAMTS14 have dramatically opposing roles in regulating invasion was especially striking given their equal contribution to collagen processing in this context. This led us to conclude that the divergent nature of these enzymes must be due to enzyme-specific substrates. A substrate repertoire for these enzymes has been previously published (PMID: 26740262) and we reasoned that the responsible substrate would be enriched following knockdown of the relevant enzyme. Thus we preformed matrisomics on cells lacking either of these enzymes, which did indeed reveal enrichment of known, enzyme-specific substrates that we could use for further analysis.

The matrisome following ADAMTS2 knockdown was minimally changed and only presented enrichment of two ADAMTS2 substrates. As there was only a minimal cellular phenotype in 2D following loss of ADAMTS2, we decided to concentrate our studies on the two identified substrates in our 3D assay. Conversely as the matrisome following ADAMTS14 knockdown was dramatically different from control cells, and ADAMTS14 knockdown presented a clear phenotype in αSMA expression, we decided to perform a screen of all matrisome hits. This highlighted the role of IL-1β in mediating myofibroblast differentiation, which has been reported elsewhere and validated our approach. Further, this refined the number of enriched ADAMTS14 substrates to two, MMP1 and Fibulin2, with Fibulin2 being identified as the responsible candidate in our 3D assays.

The ECM is more than just collagen. Choosing these two metalloproteases based on their shared collagen substrate is an approach that perhaps oversimplifies the ECM a bit, and again, does not provide the strongest rationale that these metalloproteases are most likely to explain counteracting stromal activities on tumor growth and progression.

We fully agree with the reviewers comment and feel our work acutely demonstrates this point. Loss of either ADAMTS2 or ADAMTS14 had similar effects on collagen processing; implicating their divergent roles on invasion was independent of their effects on collagen regulation. This work therefore showcases the incredible complexity of ECM regulation in tumour progression. As discussed in the manuscript, collagen along with other elements of the ECM can regulate tumour progression and we believe our work adds an additional facet to this.

Related to the above: How were the stellate cells used for the matrisome analysis grown? In the suspension setup or adherent? This will have a large impact on the outcome. Is there for instance hyaluronic acid in this matrix?

The matrisome analysis was conducted on cells cultured in 2D. Vitamin C was added to the media to promote matrix production. We agree that this is not truly reflective of the in vivo situation but as a discovery tool this led us to identify the ADAMTS2 and ADAMTS14 substrates responsible for the function observed in 3D.

1. Performing the species-specific transcript analysis both ways is a neat approach, but why did the authors ignore the opportunity to formally overlay/compare the two stromal gene sets to define likely candidates based on statistics?

We primarily used this approach as a discovery tool to identify key differences between cancer and stellate cell compartments. Comparing the two species data sets is problematic as the murine cancer cells express many elements found in the stellate cells, while the human data set presents a cleaner comparison. This is evident from comparing metzincin expression in the two data sets. The human data set (Figure 2A) shows clear separation between cancer and stellate compartments, which is less evident in the murine data set (Supp figure 2A). As noted in supplementary figure 1A, unlike the human cancer cells used in this study, the murine cancer cells are capable of invading without stellate support (although when cultured with stellate cells invasive projections are always stellate led). Nevertheless the murine data set matches the human, although with less clarity.

Minor comments: The bioinformatics Methods need more details on how reads were mapped to the different genomes. How many mismatches were allowed and was the mapping done separately or using for instance Xenofilter?

We have improved the methodology section to include more detail for this separation. Using STAR aligner, reads were mapped to host species using a combined human and mouse genome. Ambiguous reads were subsequently discarded from the analysis. While there are bioinformatic packages that seek to match ambiguous reads to parent species we did not use these for our analysis.

The authors use the knowledge on the activities of both ADAMTS2 and -14 on collagen as a rationale to choose these two. Is there really a need for the paragraph (and associated figures) from line 102 on?

Given the prominent role collagen has been shown to have in regulating PDAC progression and the primary role for ADAMTS2 and ADAMTS14 being collagen processing, we initially hypothesised that the divergent role for these enzymes on invasion could be due to differences in collagen processing in this context. The fact that both equally contribute to collagen processing is surprising and adds to the novelty of our findings that these enzymes have a more complex role in regulating stromal biology.

We have altered the structure of the manuscript to emphasise this point. The divergent roles of ADAMTS2 and ADAMTS14 on invasion are now presented in Figure 2, with their equal role in collagen processing now presented after in Figure 3. Figure 4 onwards now details the opposing roles of these enzymes in myofibroblast differentiation and our investigation into the enzyme-specific substrates responsible for this.

Abstract, line 21; some words are missing?

We thank the reviewer for bringing this to our attention and have now amended the abstract.

Were the siRNA screen hits validated?

---

Yes, hits relevant for our further investigations, MMP1 and Fibulin2, are presented in the manuscript.

What is the genotype of the mouse cancer cells? KPC-derived?

DT6066 are KPC derived while R254 are derived from KPF mice. This has been added to the methods with relevant reference.

Reviewer #1 (Significance (Required)):

The trick of dissecting tumor from stromal signals in spheroid cocultures by RNA-Seq is a cool trick, but not new and the authors should probably cite some prior work.

We have included reference to other work where researchers have used species deconvolution to explore heterocellular interactions (Lines 68-72). However, we believe our work is one of the first to use this approach to explore cellular interactions in an in vitro, 3D, invasive context.

What this all means for patients (or in vivo tumors even) remains unclear. There is some debate on whether highly activated CAFs (ACTA2/aSMA+ cells, some call them myCAFs) are indeed tumor-restrictive or whether they promote invasion. The authors appear to argue the latter (which I can agree with) but without any translational work to show what the net outcome of this mechanism is, the study remains descriptive and perhaps of limited interest.

We contend that our 3D invasion model is a powerful tool to understand the role of stellate cells in leading invasion. We have shown the utility of this model in several studies to dissect the biology of this cell type, revealing the importance of the nuclear translocation of FGFR1 in stellate invasion (PMID: 36357571), the role of the kinase PKN2 in regulating stellate heterogeneity (PMID: 35081338) and the influence of cancer cell-derived exosomes on stellate invasion (PMID: 33592190).

CAFs within PDAC stroma are highly plastic and can adopt multiple functions depending on distinct environmental cues. Thus, identifying how they are regulated is of paramount importance if they are to be therapeutically targeted. We contend that our mechanistic studies using heterocellular 3D models can aid in the dissection of the biology of these cells with more granularity than offered by clinical or in vivo studies, particularly in the context of secreted proteases. To add clinical relevance for our findings we have compared our chimera data set with previously published laser microdissected tumour and stroma PDAC tissue (Figure 2B), and identified ADAMTS2 and ADAMTS14 expression in prominent CAF subtypes (inflammatory and myofibroblastic) from published single cell RNA seq data taken from tumours (Supp figure 2C). As these enzymes are produced in multiple CAF subtypes, genetically targeting them in vivo appears prohibitive. The generation of ADAMTS2 and ADAMTS14 specific inhibitors would be required to assess their roles in vivo.

Reviewer #2 (Evidence, reproducibility and clarity (Required)):

The manuscript by Carter and colleagues examines that role of cancer-associated fibroblasts (CAFs) in regulation of invasion in a 3D co-culture assay with epithelial cells. The authors propose that invasive chains of cancer cells are led by fibroblasts. The authors utilise a system of co-culture to create chimeric human-mouse fibroblast-cancer cell spheroids (both directions utilised, to eliminate species bias) to allow for in situ sequencing of the co-operating transcriptional programmes of each cell type during 3D invasion. From this powerful approach, this allowed the authors to identify two key two collagen-processing enzymes, ADAMTS2 and ADAMTS14, as contributing to CAF function in their system. The authors identify that these two enzymes have opposing roles in invasion, and map some of their key substrates in invasion, which extend beyond collagen-processing. The authors propose that one key function is to control the processing of TGFbeta, the latter of which is a regulator of the myofibroblast subpopulation of CAF. Overall, the findings of the manuscript are interesting, but need some further proof-of-principal demonstrations to extend their findings to support the claims within.

• The authors demonstrate a clear role of ADAMTS2 and ADAMTS14 in stellate function during differentiation and invasion. Is there any evidence of such changes in patient materials? Could the authors query publicly available databases of micro-dissected stroma vs epithelium to validate the translational relevance of their findings?

We thank the reviewer for their suggestion; we have now explored clinical relevance of ADAMTS2 and ADAMTS14 expression in two ways. We have used previously published work by Maurer and colleagues (PMID: 30658994), which descibes transcriptomic analysis of laser microdissected tumour and stroma from pancreatic cancer tissue. In accordance with our chimeric data set the majority of metzincins, including ADAMTS2 and ADAMTS14, are expressed in the sromal compartment (Figure 2B). We have also used publically available scRNA seq data to examine ADAMTS2 and ADAMTS14 expression in distinct CAF subtypes (Supp Figure 2C). Both ADAMTS2 and ADAMTS14 are expressed in inflammatory and myofibroblastic CAFs, with ADAMTS14 expression lower than that of ADAMTS2. Given the complexity of CAF heterogeneity it is possible that ADAMTS2/14 secretion by one population regulates the resulting phenotype of surrounding CAFs, however this hypothesis if beyond the scope of our current work.

Major comments: - Page no. 4, Line 71, The authors conclude that the invasion in the chimeric spheroids is "led by" stellate cells. This is a key concept in the manuscript. How do the authors define the "led by" phenomena? What is the frequency that this occurs?

In our experience all invasive projections are stellate led, defined as a stellate-labelled nucleus present at the tip of invasive projections. Indeed the human cancer cells used in this study are incapable of invading in the absence of stellate cells (Supp figure 1 A). We have previously reported this model where we demonstrated FGFR1 activity in the stellate cells is crucial for invasion (PMID: 36357571). Others have demonstrated the general importance for fibroblasts in leading invasion (PMID: 18037882, 28218910). Interestingly in our study, mouse cancer cells were capable of invading in the absence of stellate cells. However, when cultured with stellate cells, projections were predominantly stellate led.

• For Figure 2A and S2A, the text suggests that the heatmap represents the stellate vs cancer cell expression (as shown in Figure 1B and S1B) in the respective species but the labelling below the heatmap suggests they are all cancer cells (Mia, Pan, R2 and DT). Is this a typo? Could the authors clarify this?

We use Mia, Pan, R2 and DT to define the sphere combination from which the data originated. We have improved the clarity of the heatmaps by colour coding the different cell types within each sphere, and matching it with the cell type data presented in the heat map. We hope this improved labelling makes the heatmaps more accessible.

• The text and the figures are lacking information about the cell line names used in the experiment, e.g, Figure 2C, 2D, 2SB, 2SC and 2SD does not indicate what cell line was used in the study. This is the same with other figures as well. Please indicate in all instances exactly which samples are queried.

We have now included reference to the cell type and stellate cell species used in each experiment in relevant figure legends. Key 3D invasive experiments were conducted with both human and mouse stellate cells.

• It's mentioned in the text that the authors have used the cancer and stellate cells in a 1:2 ratio but the numbers of stellate cells look different between different spheroids confocal images. e.g. The numbers look very different between the Miapaca2:PS1 vs Miapaca2:mPSC spheroids. Is this simply the representative images, or are their bona fide differences. This, in turn, would impact on claims of cells being 'led' by stellate cells. Can the authors clarify?

This is a consequence of the method by which the stellate cells were immortalised. Human PS1 stellate cells were immortalised with hTERT, while mouse stellate cells were immortalised with SV40. A consequence of this is that the mouse stellate cells proliferate faster in 3D than the human stellate cells, with both proliferating slower than the cancer cell compartment. So while spheroids start at 1000 cells (666 stellate, 333 cancer) with stellate cells as the prominent component they are quickly overtaken by the cancer cells. Despite this difference in proliferation we find no difference in the invasive capacity of the stellate cells, with invasive projections always stellate led irrespective of whether they are human or mouse.

• While for most of the experiments the authors generated the chimeric spheroids first and then performed the respective experiments, it appears that for the invasion assay simply co-culture of Cancer cells and stellate cells was done. Is this correct? Have the authors tried performing the assay with the chimeric spheroids to see if the stellate cells still invade?

The Boyden chamber migration assay was conducted by seeding a co-culture of stellate and cancer cells in the apical compartment then imaging their migration to the basolateral side. This provided a second method to predominantly showcase the enhanced migration of cells lacking ADAMTS14 in a manner that could be quantified over time. We have not tried placing spheroids in the apical compartment and imaging invasion through the pores.

• The authors claim that ADAMTS2 and ADAMTS14 regulate the bioavailability of TGFB, and this is a key reason that these regulate CAF differentiation. However, there is no direct demonstration of this concept, which is conspicuous by absence. Could the authors either directly demonstrate this, or remove such notions from the results, and explicitly state that this is an untested speculation in discussion? Examples of this are:

o Line 173, authors state "ADAMTS2 facilitates TGFβ release through degradation of the plasmin inhibitor, SERPINE2 (Figure 5D)"

o Line 196 authors conclude "Together these data implicate 197 ADAMTS14 as a key regulator of TGFβ bioavailability (Figure 6F)."

o Line 240 states "This reduces the activation of Plasmin, preventing the release of TGFβ (Figure 5C)." Since this is just a model without detailed experiments, It will be better to propose rather than conclude.

We appreciate the reviewer’s concern and have now added additional experiments to strengthen the association of ADAMTS enzymes and TGFβ bioavailability.

Using a TGFβ-responsive luciferase reporter we demonstrate that the media from stellate cells lacking ADAMTS14 has greatly increased amounts of active TGFβ (Figure 4), which is abrogated when Fibulin2 is knocked down alongside (Figure 7). This links ADAMTS14 and Fibulin2 to TGFβ activity. Given the extensive literature detailing a role for Fibulin2 in regulating matrix TGFβ release through interactions with fibrillin (e.g, PMID: 19349279, 12598898, 12429738) we believe this is how ADAMTS14 is regulating myofibroblast differentiation. As we do not directly examine the association of Fibulin2 with fibrillin in this manuscript we have amended the associated statements to reflect this.

We have also used a TGFβ-responsive fluorescent reporter to examine TGFβ activity of stellate cells in 3D. Consistent with our results, loss of ADAMTS2 reduces, while loss of ADAMTS14 enhances, TGFβ activity (Figure 4), which can be reversed with concomitant knockdown of their respective substrates SERPINE2 (Figure 6) and Fibulin2 (Figure 7).

• Figure S5C shows a less invasive phenotype in the NTCsi + ADAMTS14si spheroids compared to the NTCsi + NTCsi control. However, there appears no appreciable difference between NTCsi + ADAMTS14si and NTCsi + NTCsi spheroids' brightfield images in Figure 5SD.

Could the authors comment on this?

We thank the reviewer for bringing this to our attention and apologise for our mistake. The images were positioned erroneously. This has now been corrected and the images reflect the quantification that demonstrates a clear increase in invasion following loss of ADAMTS14, which is abrogated with co-knockdown of Fibulin2.

Minor Comments: - Page no. 2, Line 20 has an incomplete sentence "Crosstalk between cancer and stellate cells is pivotal in pancreatic cancer, resulting in differentiation 21 of stellate cells into myofibroblasts that drive."

Apologies for the error. This has been rectified.

• Figure 2C; Figure S2C and Figure S5E lack quantification for the western blots.

We have now included densitometry for all western blots, presenting values relative to the respective loading control and normalised to the experimental control. Values are averages taken from all biological repeats with significance indicated where relevant.

• Why did the authors choose to investigate the Metzincin family? Could the authors provide their reasoning to investigate these proteins, to the exclusion of other candidates?

We focused on the metzincin family, as they are best known for their involvement in cancer invasion. A goal for this manuscript is to present our chimera data set as a discovery tool for the community. While this initial manuscript focuses on protease activity, we have further projects on-going that have used this data set to identify important elements of cancer/stellate communication.

• Info about the number of fields imaged per sample for the microscopy data is missing in the figure legends (e.g. Figure 2F and 2I, Figure 5SF).

We have now included a statement in each relevant figure legend to indicate that quantification was performed on at least five fields of view per biological repeat.

• Any particular reason why the ADAMTS2 expression was not checked through Western blotting like ADAMTS14 in Figure S2B.

We attempted to examine ADAMTS2 by western blotting but were unable to find an antibody that produced consistent results with our samples, and corroborated consistent knockdown by PCR.

• The legends for Figure 3SC and 5SF mention that "Images are representative of at least two biological replicates". How many technical replicates were used? It would be useful if the relative intensity of the images is measured and plotted in a graph.

We have now moved these images to the main figure alongside quantification of αSMA intensity. Images are collected from two biological repeats with quantification obtained from at least five fields of view per image. Together these data strongly demonstrate that loss of ADAMTS14 increases αSMA fibre intensity, which is blocked by either an inhibitor of TGFβ signalling (Figure 4), or co-knockdown of Fibulin2 (Figure 7).

Reviewer #2 (Significance (Required)):

This work provides an examination of the cross talk between fibroblasts and cancer cells in a 3-Dimensional culture model of pancreatic tumour cell invasion. By using chimeric human-mouse spheroids, the authors are able to identify cell-type specific transcripts by bulk RNA sequencing in situ. This advance is not to be underestimated as a number of existing approaches for cell type-specific profiling (eg. single-cell sequencing) relies upon dissociation of cell communities prior to sequencing. It is very likely that transcriptional programmes change during this isolation process. This approach allows the authors to identify transcriptional co-operating programmes in situ. This data provides a resource to understand this key co-operation of these two cell types during tumourigenesis, and will be of interest to the pancreatic cancer field. In addition, the mapping of the key substrate of these enzymes provides further insights that may be useful in understanding the expanded target repertoire of these enzymes beyond collagen processing.

We thank the reviewer for their strong support of our chimeric spheroid approach and resulting investigation into the dichotomic roles of ADAMTS2 and ADAMTS14.

Reviewer #3 (Evidence, reproducibility and clarity (Required)):

AMADTS2 and ADAMTS14 belong to the disintegrin and metalloproteinase with thrombospondin motif protein family, mainly produced by pancreatic stellate cells (PSC) and are related to cancer cell invasion. This study reveals that ADAMTS2 and ADAMTS14 have opposite roles in myofibroblast differentiation based on experiments testing HSC driven cancer cell invasion, variant expression of HSC activation makers and the related downstream targets analysed upon RNA sequencing analyses. The authors (TA) established PSC/cancer cell chimeric spheroids for investigating the crosstalk between these cell types in 3D in vitro. Based on their findings, they claim that ADAMTS2 and ADAMTS14 have different functions regarding PSC and TGF-β activation. However, their conclusions mainly rely on quantitatve data of invasion and mechanistic details are completely lacking.

Comments: Typos, even in the abstract, e.g. first sentence incomplete

We apologise for the error in the abstract and have rectified this in the revised manuscript.

Introduction is rather sparce with one third of the text repeating the results of the study

Our manuscript details a discovery experiment using chimeric spheroids to identify cancer cell and stellate cell transcriptomes in a 3D invasive context. We then showcase the power of this data set by using it to identify and then describe divergent roles for ADAMTS2 and ADAMTS14 in shaping stellate cell biology. Given this two-tiered approach we incorporated text that would normally be placed in the introduction into the results section (e.g. our description of the importance of collagen processing in PDAC, presented as a prelude to the results from figure 3). We feel this improves the flow of the manuscript, rather than having information that isn’t necessarily relevant to the reader at the outset.

Some citations do not at all fit with the position where they are placed; needs approval

We have examined this in detail and are confident in our use of appropriate references throughout.

In this study, it is said that TS2 (AMADTS2) and TS14 (ADAMTS14) have opposite functions on myofibroblast differentiation, with individual depletion leading to distinct matrisomal phenotypes in PSC. However, both similarly contribute to collagen processing. As we know, collagen is increased in response to TGFβ signaling, since TS2 depletion (knock down, kd) inhibits and TS14 kd is suggested to promote TGFβ activation, it is expected that this has impact on the available collagen levels. How do the authors explain that nevertheless the kd effect on collagen is very similar?

The primary effects of these enzymes are on the processing of pro-collagen to its mature form, rather than on the production of collagen. This is evidenced in figure 3B where collagen expression in the whole cell lysate is the same following ADAMTS2 knockdown, and slightly reduced with loss of ADAMTS14, but the mature form is lost in the cell culture supernatant.

While myofibroblast differentiation is associated with increased collagen production, it is possible that this is perturbed in a situation where the cell is surrounded by collagen that is incompletely processed (e.g. through biomechanical feedback). Given that our results clearly indicated that the effect of ADAMTS2 and ADAMTS14 on invasion is independent of their roles in collagen processing, this avenue is beyond the scope of the current manuscript.

The authors claim that TS2 facilitates TGFβ release and TS14 is a key regulator of TGFβ bioavailability. However, throughout the whole data, there is no experimental evidence for this conclusion. TGFβ activation, LAP concentration and downstream effects should be provided.

Most of the conclusions in the manuscript are based on effects to invasion and the estimated quantification histograms. "Black boxes in between the treatment, e.g. knockdown and readout, that relate to the signals and mechanisms remain black boxes throughout. For example, the impact of the treatments on stellate cell activation markers, the cancer cells invasion signaling, the SERPINE2- and Fibulin2-dependent myofibroblast differentiation pathways should be mechanistically investigated.

We disagree with this comment. Our invasive model shows a clear role for ADAMTS2 and ADAMTS14 in regulating invasion, which is mitigated by disrupting their substrates SERPINE2 and Fibulin2.

ADAMTS2 loss is associated with a reduction in plasmin activity, which again is mitigated with concurrent loss of SERPINE2. Equally, inhibition of plasmin activity with Aprotinin matches the loss of invasion observed with loss of ADAMTS2. Plasmin has a well-established role in mediating TGFβ release from the matrix. We have now included additional experiments using a TGFβ fluorescent reporter in 3D culture. This demonstrates that loss of ADAMTS2 reduces TGFβ activity, which can be rescued with co-knockdown of SERPINE2 (Figure 6). Our data therefore support a mechanism where ADAMTS2 blocks TGFβ release from the matrix, and therefore myofibroblast differentiation, through its regulation of SERPINE2 activity.

We have strengthened our proposed mechanism for ADAMTS14 regulation of TGFβ through Fibulin2 with the use of both luciferase and fluorescent TGFβ reporter constructs. Using these reporters, we demonstrate that stellate cells lacking ADAMTS14 exhibit increased TGFβ activity (Figure 4), which is mitigated with co-knockdown of Fibulin2 (Figure 7). Combined with the effects on αSMA expression and 3D invasion, our data fit with a model where ADAMTS14 regulates TGFβ bioavailability through Fibulin2.

The authors investigate one cell line each for their conclusions; we know that different cell lines behave differently; can they confirm that the finding they present is of general validity or a finding that is specific for the tested cancer/PSC cell lines. Can the principle findings also be proven in primary cells. More importantly, the authors should proof their findings in PaCa tissue of patients as follows: Expression of the proteases in the tissue, related variation of matrisome signatures, e.g. by snRNASeq, to confirm relevance of the finding.

All our key 3D invasive experiments are repeated with both human and mouse stellate cells, adding strength to our proposed association with ADAMTS2 and SERPINE2, and ADAMTS14 and Fibulin2, on the invasive capacity of stellate cells. As detailed above we have explored the clinical relevance of our findings by examining laser dissected tumour and stromal data from PDAC tissue, and scRNA fibroblast data. These data confirm that ADAMTS2 and ADAMTS14 are predominantly expressed in the stromal compartment of the tumour and are associated with key CAF subtypes present in the PDAC environment, inflammatory and myofibroblastic CAFs.

Details related to the figures: Figure 1: Are the numbers of PSC and PaCa cells integrated in the spheres related to the numbers found in patients?

The 2:1 ratio of stellate to cancer cells used to produce spheres is a technical requirement and reflects the numbers in patients (PMID: 23359139). Cancer cells will proliferate substantially faster than the stellate cells so at the end of the experiment (day 3) the spheres are predominantly cancer cells. Nevertheless the stellate cells are able to drive invasion of the cancer cells, which can be quantitatively assessed in this model.

B, it seems that the PSC in the spheroid are not equally distributed but instead are all located in close vicinity to eachother in a cloud; is that the representative situation for the spheres and is this similar in the PaCa cancer tissue? Does this have influence on the results?

We have replaced this image with a more representative image that shows mouse stellate cells dispersed throughout the sphere.

Figure 2: It is interesting to hear that BMP1, which is actually a ligand for BMP signaling is a protease for Collagen. How does this work?

While the BMP family generally belong to the TGFβ superfamily, BMP1 is the exception in that it is a C-terminal collagenase. Please refer to reference 21 in the manuscript (PMID: 33879793), which details the role of BMP1 on collagen processing and the resulting effect on PDAC progression.

C, Quantification of all blots should be presented.

We have now included densitometry for all western blots, presenting values relative to the loading control and normalised to the experimental control. Values are averages taken from all biological repeats with significance indicated by stars.

Figure S2: TS2 kd and TS14 kd should be confirmed and provided by both qrt PCR and WB data.

We were unable to assess ADAMTS2 knockdown by western blot due to the quality of available antibodies. We are confident that either western or PCR confirmation of knockdown is sufficient, especially given the strong phenotype observed with the resulting knockdown.

Figure 3: F; this result is arguing against the conclusion that TGFb bioavailability is a function of the ADAMs, since the kd impacts on the treatment result with exogenous TGFb. This suggests an effect downstream of ligand activation by proteasomal cleavage, e.g. receptor activation or signal transduction; this needs clarification. H, I: TGFβR inhibitor reduces TS14kd enhanced αSMA expression. How is unclear and needs clarification, since from F we know that already activated TGFβ needs TS2 to fully induce αSMA expression.

SupplFig.3: B, C, as above!

αSMA expression in stellate cells requires continuous exposure to TGFβ over 48 hours. Active TGFβ has an incredibly short half-life (minutes) and so requires positive feedback to maintain signalling. We propose that following ADAMTS2 knockdown the cells are incapable of releasing further TGFβ to maintain the phenotype. Equally following ADAMTS14 knockdown the cells are able to release more TGFβ, which is incapable of initiating signalling when the receptor is blocked.

Figure 4: TIMP1 is a canonical TGFb signaling target gene in fibrosis. How the authors explain that TIMP1 is upregulated in both knockdowns, when they claim that TS2 and 14 have opposing functions on TGFb activation. This result as well puts their conclusions as regards TGFb and also the myofibroblast phenotype into question. Especially, since TIMP1 signifies stellate cell activation not only in the pancreas, but also in the liver and kidney. C, D, E should be explained in more detail and all details of the results should be presented.

TIMP1 is a substrate for both ADAMTS2 and ADAMTS14, so its enrichment following knockdown of either is unsurprising, reflective of reduced cleavage of TIMP1. Both our 3D invasive assessment in Figure 6 and αSMA imaging in supplementary figure 5 demonstrate that TIMP1 is not responsible for the effect observed as a consequence from loss of either ADAMTS2 or ADAMTS14.

This holds also for the different myofibroblast phenotypes. All data should be included. From recent scRNASeq investigations, several myofibroblast populations were described and compared, e.g my-stellate cells vs i-stellate cells. To which of these phenotypes the identified populations belong?

As mentioned above, we have interrogated publically available data sets and identified ADAMTS2 and ADAMTS14 expression in multiple CAF subtypes. As these proteases are secreted it is probable that one CAF subtype can control the phenotype of surrounding CAFs through ADAMTS2 and ADAMTS14 production. While intriguing, this hypotheses is beyond the scope of the current work.

Figure 5: C, Only brightfield images are provided, confocal images are suggested for comparison of +/- Aprotinin treatment.

We do not think the addition of confocal images will add to the comparison. Aprotinin clearly reduces invasion, which coupled with the action of stellate-derived SERPINE2 on invasion, and reduced plasmin activity following ADAMTS2 knockdown, suggests that plasmin is important for regulating the effects of ADAMTS2 on invasion.

The efficiency of TS2 and Serpine2 kd should be provided by qrt PCR and WB.

TS2 kd promoted SERPINE2 expression should also be presented by qrt PCR and WB.

We are confident that either western or PCR confirmation of knockdown is sufficient. Of note is that following ADAMTS2 knockdown, SERPINE2 expression is unchanged (sup figure 4C). This would indicate that the enrichment of SERPINE2 observed in the matrisome following loss of ADAMTS2 is reflective of reduced cleavage, rather than a change in expression.

Figure 6: A, why ta use aSMA and not invasive activity as a readout here?

Increased αSMA expression following ADAMTS14 knockdown provides a strong, clear, 2D phenotype to act as a readout for an siRNA screen with high-content imaging. Performing such a screen with our 3D invasive model is currently impractical.

There are many parameters leading to decreased aSMA expression upon kd; (1) why only MMP1 and Fibulin were selected as candidates?

From our αSMA screen, MMP1 and Fibulin2 knockdown were the only candidates that were able to both prevent an increase in αSMA seen with ADAMTS14 loss alone, and are known ADAMTS14 substrates. Further validation in our 3D invasive model demonstrated that Fibulin2 and not MMP1 was responsible for the effect of ADAMTS14 loss on invasion.

(2) the single kd control of the screen candidates is missing!

We feel this control is not needed, as the goal of the experiment was to establish which candidate was responsible for mediating the effects brought about by ADAMTS14 knockdown. Increased αSMA expression with IL-1β loss validates our approach, as this is a known negative regulator of TGFβ signalling.

(3) Can it be expected that all these matrisomal proteins are involved in aSMA expression regulation? I have doubts.

We agree with the reviewers comment, from the siRNA screen (sup figure 5B) it is clear that the majority of the identified matrisome proteins have a minimal effect on αSMA expression following loss of ADAMTS14.

C, D, E, why MMP1 was not also tested in these assays?

Our spheroid assay clearly demonstrated that invasion was enhanced following ADAMTS14 knockdown even with co-knockdown of MMP1. Given the strong rescue observed with co-knockdown of Fibulin2 we proceeded to further analyse this candidate over MMP1.

F, Fibrillin is shown in the figure but not described in the text. It would be quite interesting to see whether Fibrillin kd has the same effect as TS14 kd on LTGF-β activation (which of course need to be shown experimentally).

The association of fibrillin with TGFβ release is well established as it underpins the biology behind Marfan syndrome. Loss of fibrillin, or mutations to its TGFβ binding sites results in a phenotype consistent with super active TGFβ signalling.

E, what is the meaning of αSMA intensity quantification? By IF staining of αSMA? PSC αSMA expression should be quantified by qrt PCR and WB.

We have now incorporated the confocal images analysing αSMA expression into the main figure and labelled the quantification accordingly. We feel this improves the clarity of the figures. Every western blot is now presented with quantification.

Also here, kd efficiency of TS14 and Fibulin2 should be provided by qrt PCR and WB.

Figure S5E should be part of figure 6, qrt PCR of Fibulin2 should be added.

We have moved this western blot to the main figure (Fig 7C). We feel additional PCR validation of Fibulin 2 knockdown is not necessary.

Figure 5/6 and throughout: It is claimed that ADAMTS2 and ADAMTS14 regulate TGFβ bioavailability through SREPINE2-Plasmin and Fibulin2. As mentioned above, TGFβ activation is only mentioned in the schemes, but no experimental evidence is given. In addition, according to previous studies, ADAMTSs can activate latent TGFβ directly by interaction with the LAP of latent TGFβ. .

We have now included extra experimental evidence to support an association of ADAMTS proteins with TGFβ bioavailability. Using a TGFβ luciferase reporter construct, we demonstrate that active TGFβ is increased following loss of ADAMTS14, which is abrogated with concomitant loss of Fibulin2. This provides further evidence that ADAMTS14 is mediating its effects on myofibroblast differentiation / invasion through TGFβ release.

Figure 3B, C, and 6D: We are confused from the migration/invasion assays. Invasion should be based on migration of tumor cells, whereas in the migration assays only stellate cells seem to be active? Can you explain this to us? According to Figure 3B, stellate and cancer cells are cocultured in the chamber. Is this the same condition as for the experiment presented as figure 6D?

In our migration assay, stellate and cancer cells are co-cultured in the apical chamber and cell migration imaged over time. We pooled data of both cancer and stellate cell migration following stellate specific knockdown of either ADAMTS2 or ADAMTS14, which showed an increase in cell migration following loss of ADAMTS14. In figure 7, we again use this assay to demonstrate that Fibulin2 expression accounts for the phenotype observed from loss of ADAMTS14.

In summary, this study for the first time found that ADAMTS2 and ADAMTS14 have opposite roles on myofibroblast differentiation, which is shown by using chimeric spheroids of stellate and pancreatic cancer cells. The authors claim a therapeutic potential for pancreatic cancer by regulating ADAMTS2/14-mediated stellate cell activation, which should avoid cancer cell invasion. The approach is interesting and there is preliminary evidence, however the study has many gaps and requires substantive workload.

We thank the reviewer for their support of our findings. We hope the additional data, combined with the known role for these substrates in the regulation of TGFβ, strengthens the clarity of our manuscript.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #3

Evidence, reproducibility and clarity

AMADTS2 and ADAMTS14 belong to the disintegrin and metalloproteinase with thrombospondin motif protein family, mainly produced by pancreatic stellate cells (PSC) and are related to cancer cell invasion. This study reveals that ADAMTS2 and ADAMTS14 have opposite roles in myofibroblast differentiation based on experiments testing HSC driven cancer cell invasion, variant expression of HSC activation makers and the related downstream targets analysed upon RNA sequencing analyses. The authors (TA) established PSC/cancer cell chimeric spheroids for investigating the crosstalk between these cell types in 3D in vitro. Based on their findings, they claim that ADAMTS2 and ADAMTS14 have different functions regarding PSC and TGF-β activation. However, their conclusions mainly rely on quantitatve data of invasion and mechanistic details are completely lacking.

Comments:

Typos, even in the abstract, e.g. first sentence incomplete Introduction is rather sparce with one third of the text repeating the results of the study Some citations do not at all fit with the position where they are placed; needs approval

In this study, it is said that TS2 (AMADTS2) and TS14 (ADAMTS14) have opposite functions on myofibroblast differentiation, with individual depletion leading to distinct matrisomal phenotypes in PSC. However, both similarly contribute to collagen processing. As we know, collagen is increased in response to TGFβ signaling, since TS2 depletion (knock down, kd) inhibits and TS14 kd is suggested to promote TGFβ activation, it is expected that this has impact on the available collagen levels. How do the authors explain that nevertheless the kd effect on collagen is very similar?

The authors claim that TS2 facilitates TGFβ release and TS14 is a key regulator of TGFβ bioavailability. However, throughout the whole data, there is no experimental evidence for this conclusion. TGFβ activation, LAP concentration and downstream effects should be provided.

Most of the conclusions in the manuscript are based on effects to invasion and the estimated quantification histograms. "Black boxes in between the treatment, e.g. knockdown and readout, that relate to the signals and mechanisms remain black boxes throughout. For example, the impact of the treatments on stellate cell activation markers, the cancer cells invasion signaling, the SERPINE2- and Fibulin2-dependent myofibroblast differentiation pathways should be mechanistically investigated.

The authors investigate one cell line each for their conclusions; we know that different cell lines behave differently; can they confirm that the finding they present is of general validity or a finding that is specific for the tested cancer/PSC cell lines. Can the principle findings also be proven in primary cells. More importantly, the authors should proof their findings in PaCa tissue of patients as follows: Expression of the proteases in the tissue, related variation of matrisome signatures, e.g. by snRNASeq, to confirm relevance of the finding.

Details related to the figures:

Figure 1: Are the numbers of PSC and PaCa cells integrated in the spheres related to the numbers found in patients? B, it seems that the PSC in the spheroid are not equally distributed but instead are all located in close vicinity to eachother in a cloud; is that the representative situation for the spheres and is this similar in the PaCa cancer tissue? Does this have influence on the results?

Figure 2: It is interesting to hear that BMP1, which is actually a ligand for BMP signaling is a protease for Collagen. How does this work? C, Quantification of all blots should be presented.

Figure S2: TS2 kd and TS14 kd should be confirmed and provided by both qrt PCR and WB data.

Figure 3: F; this result is arguing against the conclusion that TGFb bioavailability is a function of the ADAMs, since the kd impacts on the treatment result with exogenous TGFb. This suggests an effect downstream of ligand activation by proteasomal cleavage, e.g. receptor activation or signal transduction; this needs clarification. H, I: TGFβR inhibitor reduces TS14kd enhanced αSMA expression. How is unclear and needs clarification, since from F we know that already activated TGFβ needs TS2 to fully induce αSMA expression.

SupplFig.3: B, C, as above!

Figure 4: TIMP1 is a canonical TGFb signaling target gene in fibrosis. How the authors explain that TIMP1 is upregulated in both knockdowns, when they claim that TS2 and 14 have opposing functions on TGFb activation. This result as well puts their conclusions as regards TGFb and also the myofibroblast phenotype into question. Especially, since TIMP1 signifies stellate cell activation not only in the pancreas, but also in the liver and kidney. C, D, E should be explained in more detail and all details of the results should be presented. This holds also for the different myofibroblast phenotypes. All data should be included. From recent scRNASeq investigations, several myofibroblast populations were described and compared, e.g my-stellate cells vs i-stellate cells. To which of these phenotypes the identified populations belong?

Figure 5: C, Only brightfield images are provided, confocal images are suggested for comparison of +/- Aprotinin treatment. The efficiency of TS2 and Serpine2 kd should be provided by qrt PCR and WB. TS2 kd promoted SERPINE2 expression should also be presented by qrt PCR and WB.

Figure 6: A, why ta use aSMA and not invasive activity as a readout here? There are many parameters leading to decreased aSMA expression upon kd; (1) why only MMP1 and Fibulin were selected as candidates? (2) the single kd control of the screen candidates is missing! (3) Can it be expected that all these matrisomal proteins are involved in aSMA expression regulation? I have doubts. C, D, E, why MMP1 was not also tested in these assays? F, Fibrillin is shown in the figure but not described in the text. It would be quite interesting to see whether Fibrillin kd has the same effect as TS14 kd on LTGF-β activation (which of course need to be shown experimentally). E, what is the meaning of αSMA intensity quantification? By IF staining of αSMA? PSC αSMA expression should be quantified by qrt PCR and WB. Also here, kd efficiency of TS14 and Fibulin2 should be provided by qrt PCR and WB.

Figure S5E should be part of figure 6, qrt PCR of Fibulin2 should be added.

Figure 5/6 and throughout: It is claimed that ADAMTS2 and ADAMTS14 regulate TGFβ bioavailability through SREPINE2-Plasmin and Fibulin2. As mentioned above, TGFβ activation is only mentioned in the schemes, but no experimental evidence is given. In addition, according to previous studies, ADAMTSs can activate latent TGFβ directly by interaction with the LAP of latent TGFβ. . Figure 3B, C, and 6D: We are confused from the migration/invasion assays. Invasion should be based on migration of tumor cells, whereas in the migration assays only stellate cells seem to be active? Can you explain this to us? According to Figure 3B, stellate and cancer cells are cocultured in the chamber. Is this the same condition as for the experiment presented as figure 6D?

In summary, this study for the first time found that ADAMTS2 and ADAMTS14 have opposite roles on myofibroblast differentiation, which is shown by using chimeric spheroids of stellate and pancreatic cancer cells. The authors claim a therapeutic potential for pancreatic cancer by regulating ADAMTS2/14-mediated stellate cell activation, which should avoid cancer cell invasion. The approach is interesting and there is preliminary evidence, however the study has many gaps and requires substantive workload.

Significance

AMADTS2 and ADAMTS14 belong to the disintegrin and metalloproteinase with thrombospondin motif protein family, mainly produced by pancreatic stellate cells (PSC) and are related to cancer cell invasion. This study reveals that ADAMTS2 and ADAMTS14 have opposite roles in myofibroblast differentiation based on experiments testing HSC driven cancer cell invasion, variant expression of HSC activation makers and the related downstream targets analysed upon RNA sequencing analyses.

The authors (TA) established PSC/cancer cell chimeric spheroids for investigating the crosstalk between these cell types in 3D in vitro. Based on their findings, they claim that ADAMTS2 and ADAMTS14 have different functions regarding PSC and TGF-β activation. However, their conclusions mainly rely on quantitatve data of invasion and mechanistic details are completely lacking.

Comments:

Typos, even in the abstract, e.g. first sentence incomplete Introduction is rather sparce with one third of the text repeating the results of the study Some citations do not at all fit with the position where they are placed; needs approval

In this study, it is said that TS2 (AMADTS2) and TS14 (ADAMTS14) have opposite functions on myofibroblast differentiation, with individual depletion leading to distinct matrisomal phenotypes in PSC. However, both similarly contribute to collagen processing. As we know, collagen is increased in response to TGFβ signaling, since TS2 depletion (knock down, kd) inhibits and TS14 kd is suggested to promote TGFβ activation, it is expected that this has impact on the available collagen levels. How do the authors explain that nevertheless the kd effect on collagen is very similar?

The authors claim that TS2 facilitates TGFβ release and TS14 is a key regulator of TGFβ bioavailability. However, throughout the whole data, there is no experimental evidence for this conclusion. TGFβ activation, LAP concentration and downstream effects should be provided.

Most of the conclusions in the manuscript are based on effects to invasion and the estimated quantification histograms. "Black boxes in between the treatment, e.g. knockdown and readout, that relate to the signals and mechanisms remain black boxes throughout. For example, the impact of the treatments on stellate cell activation markers, the cancer cells invasion signaling, the SERPINE2- and Fibulin2-dependent myofibroblast differentiation pathways should be mechanistically investigated.

The authors investigate one cell line each for their conclusions; we know that different cell lines behave differently; can they confirm that the finding they present is of general validity or a finding that is specific for the tested cancer/PSC cell lines. Can the principle findings also be proven in primary cells. More importantly, the authors should proof their findings in PaCa tissue of patients as follows: Expression of the proteases in the tissue, related variation of matrisome signatures, e.g. by snRNASeq, to confirm relevance of the finding.

Details related to the figures:

Figure 1: Are the numbers of PSC and PaCa cells integrated in the spheres related to the numbers found in patients? B, it seems that the PSC in the spheroid are not equally distributed but instead are all located in close vicinity to eachother in a cloud; is that the representative situation for the spheres and is this similar in the PaCa cancer tissue? Does this have influence on the results?

Figure 2: It is interesting to hear that BMP1, which is actually a ligand for BMP signaling is a protease for Collagen. How does this work? C, Quantification of all blots should be presented.

Figure S2: TS2 kd and TS14 kd should be confirmed and provided by both qrt PCR and WB data.

Figure 3: F; this result is arguing against the conclusion that TGFb bioavailability is a function of the ADAMs, since the kd impacts on the treatment result with exogenous TGFb. This suggests an effect downstream of ligand activation by proteasomal cleavage, e.g. receptor activation or signal transduction; this needs clarification. H, I: TGFβR inhibitor reduces TS14kd enhanced αSMA expression. How is unclear and needs clarification, since from F we know that already activated TGFβ needs TS2 to fully induce αSMA expression.

SupplFig.3: B, C, as above!

Figure 4: TIMP1 is a canonical TGFb signaling target gene in fibrosis. How the authors explain that TIMP1 is upregulated in both knockdowns, when they claim that TS2 and 14 have opposing functions on TGFb activation. This result as well puts their conclusions as regards TGFb and also the myofibroblast phenotype into question. Especially, since TIMP1 signifies stellate cell activation not only in the pancreas, but also in the liver and kidney. C, D, E should be explained in more detail and all details of the results should be presented. This holds also for the different myofibroblast phenotypes. All data should be included. From recent scRNASeq investigations, several myofibroblast populations were described and compared, e.g my-stellate cells vs i-stellate cells. To which of these phenotypes the identified populations belong?

Figure 5: C, Only brightfield images are provided, confocal images are suggested for comparison of +/- Aprotinin treatment. The efficiency of TS2 and Serpine2 kd should be provided by qrt PCR and WB. TS2 kd promoted SERPINE2 expression should also be presented by qrt PCR and WB.

Figure 6: A, why ta use aSMA and not invasive activity as a readout here? There are many parameters leading to decreased aSMA expression upon kd; (1) why only MMP1 and Fibulin were selected as candidates? (2) the single kd control of the screen candidates is missing! (3) Can it be expected that all these matrisomal proteins are involved in aSMA expression regulation? I have doubts. C, D, E, why MMP1 was not also tested in these assays? F, Fibrillin is shown in the figure but not described in the text. It would be quite interesting to see whether Fibrillin kd has the same effect as TS14 kd on LTGF-β activation (which of course need to be shown experimentally). E, what is the meaning of αSMA intensity quantification? By IF staining of αSMA? PSC αSMA expression should be quantified by qrt PCR and WB. Also here, kd efficiency of TS14 and Fibulin2 should be provided by qrt PCR and WB.

Figure S5E should be part of figure 6, qrt PCR of Fibulin2 should be added.

Figure 5/6 and throughout: It is claimed that ADAMTS2 and ADAMTS14 regulate TGFβ bioavailability through SREPINE2-Plasmin and Fibulin2. As mentioned above, TGFβ activation is only mentioned in the schemes, but no experimental evidence is given. In addition, according to previous studies, ADAMTSs can activate latent TGFβ directly by interaction with the LAP of latent TGFβ. .

Figure 3B, C, and 6D: We are confused from the migration/invasion assays. Invasion should be based on migration of tumor cells, whereas in the migration assays only stellate cells seem to be active? Can you explain this to us? According to Figure 3B, stellate and cancer cells are cocultured in the chamber. Is this the same condition as for the experiment presented as figure 6D?

In summary, this study for the first time found that ADAMTS2 and ADAMTS14 have opposite roles on myofibroblast differentiation, which is shown by using chimeric spheroids of stellate and pancreatic cancer cells. The authors claim a therapeutic potential for pancreatic cancer by regulating ADAMTS2/14-mediated stellate cell activation, which should avoid cancer cell invasion. The approach is interesting and there is preliminary evidence, however the study has many gaps and requires substantive workload.

Note: This preprint has been reviewed by subject experts for Review Commons. Content has not been altered except for formatting.

Learn more at Review Commons

---

Referee #2

Evidence, reproducibility and clarity

The manuscript by Carter and colleagues examines that role of cancer-associated fibroblasts (CAFs) in regulation of invasion in a 3D co-culture assay with epithelial cells. The authors propose that invasive chains of cancer cells are led by fibroblasts. The authors utilise a system of co-culture to create chimeric human-mouse fibroblast-cancer cell spheroids (both directions utilised, to eliminate species bias) to allow for in situ sequencing of the co-operating transcriptional programmes of each cell type during 3D invasion. From this powerful approach, this allowed the authors to identify two key two collagen-processing enzymes, ADAMTS2 and ADAMTS14, as contributing to CAF function in their system. The authors identify that these two enzymes have opposing roles in invasion, and map some of their key substrates in invasion, which extend beyond collagen-processing. The authors propose that one key function is to control the processing of TGFbeta, the latter of which is a regulator of the myofibroblast subpopulation of CAF. Overall, the findings of the manuscript are interesting, but need some further proof-of-principal demonstrations to extend their findings to support the claims within.

• The authors demonstrate a clear role of ADAMTS2 and ADAMTS14 in stellate function during differentiation and invasion. Is there any evidence of such changes in patient materials? Could the authors query publicly available databases of micro-dissected stroma vs epithelium to validate the translational relevance of their findings?

Major comments:

• Page no. 4, Line 71, The authors conclude that the invasion in the chimeric spheroids is "led by" stellate cells. This is a key concept in the manuscript. How do the authors define the "led by" phenomena? What is the frequency that this occurs?
• For Figure 2A and S2A, the text suggests that the heatmap represents the stellate vs cancer cell expression (as shown in Figure 1B and S1B) in the respective species but the labelling below the heatmap suggests they are all cancer cells (Mia, Pan, R2 and DT). Is this a typo? Could the authors clarify this?
• The text and the figures are lacking information about the cell line names used in the experiment, e.g, Figure 2C, 2D, 2SB, 2SC and 2SD does not indicate what cell line was used in the study. This is the same with other figures as well. Please indicate in all instances exactly which samples are queried.
• It's mentioned in the text that the authors have used the cancer and stellate cells in a 1:2 ratio but the numbers of stellate cells look different between different spheroids confocal images. e.g. The numbers look very different between the Miapaca2:PS1 vs Miapaca2:mPSC spheroids. Is this simply the representative images, or are their bona fide differences. This, in turn, would impact on claims of cells being 'led' by stellate cells. Can the authors clarify?
• While for most of the experiments the authors generated the chimeric spheroids first and then performed the respective experiments, it appears that for the invasion assay simply co-culture of Cancer cells and stellate cells was done. Is this correct? Have the authors tried performing the assay with the chimeric spheroids to see if the stellate cells still invade?

• The authors claim that ADAMTS2 and ADAMTS14 regulate the bioavailability of TGFB, and this is a key reason that these regulate CAF differentiation. However, there is no direct demonstration of this concept, which is conspicuous by absence. Could the authors either directly demonstrate this, or remove such notions from the results, and explicitly state that this is an untested speculation in discussion? Examples of this are:

  • Line 173, authors state "ADAMTS2 facilitates TGFβ release through degradation of the plasmin inhibitor, SERPINE2 (Figure 5D)"
  • Line 196 authors conclude "Together these data implicate 197 ADAMTS14 as a key regulator of TGFβ bioavailability (Figure 6F)."
  • Line 240 states "This reduces the activation of Plasmin, preventing the release of TGFβ (Figure 5C)." Since this is just a model without detailed experiments, It will be better to propose rather than conclude.

• Figure S5C shows a less invasive phenotype in the NTCsi + ADAMTS14si spheroids compared to the NTCsi + NTCsi control. However, there appears no appreciable difference between NTCsi + ADAMTS14si and NTCsi + NTCsi spheroids' brightfield images in Figure 5SD. Could the authors comment on this?

Minor Comments:

• Page no. 2, Line 20 has an incomplete sentence "Crosstalk between cancer and stellate cells is pivotal in pancreatic cancer, resulting in differentiation 21 of stellate cells into myofibroblasts that drive."
• Figure 2C; Figure S2C and Figure S5E lack quantification for the western blots.
• Why did the authors choose to investigate the Metzincin family? Could the authors provide their reasoning to investigate these proteins, to the exclusion of other candidates?
• Info about the number of fields imaged per sample for the microscopy data is missing in the figure legends (e.g. Figure 2F and 2I, Figure 5SF).
• Any particular reason why the ADAMTS2 expression was not checked through Western blotting like ADAMTS14 in Figure S2B.
• The legends for Figure 3SC and 5SF mention that "Images are representative of at least two biological replicates". How many technical replicates were used? It would be useful if the relative intensity of the images is measured and plotted in a graph.

Significance

This work provides an examination of the cross talk between fibroblasts and cancer cells in a 3-Dimensional culture model of pancreatic tumour cell invasion. By using chimeric human-mouse spheroids, the authors are able to identify cell-type specific transcripts by bulk RNA sequencing in situ. This advance is not to be underestimated as a number of existing approaches for cell type-specific profiling (eg. single-cell sequencing) relies upon dissociation of cell communities prior to sequencing. It is very likely that transcriptional programmes change during this isolation process. This approach allows the authors to identify transcriptional co-operating programmes in situ. This data provides a resource to understand this key co-operation of these two cell types during tumourigenesis, and will be of interest to the pancreatic cancer field. In addition, the mapping of the key substrate of these enzymes provides further insights that may be useful in understanding the expanded target repertoire of these enzymes beyond collagen processing.