The concept of "bridges" and "hubs" in curvature in this work is really interesting to me. It suggests that you can identify places along a developmental trajectory where fate transitions are narrowed vs. robust and redundant. Have you considered applying this metric to scRNA-seq datasets focused on tumor progression, e.g. something like Liu et al. 2024? I wonder if you could predict what cell populations are likely to undergo critical state transitions during the development of tumors and identify therapeutically relevant cell populations to target.

Given that Eupatilin seems to promote cell quiescence, would pursuing this molecule directly as a therapeutic have a poor side effect profile in juveniles, who are still growing? How are prolferating non-kidney cells affected by Eupatilin?

When thinking about identifying new promoters for transgenic gene expression, has anyone generated maps of the chromatin landscape in tardigrades? e.g. using ATAC-Seq? This might help you identify active/open promoters at different developmental stages.

We recently released a snakemake pipeline for exploring protein structural spaces, generating using TM-score similarity matrices, called ProteinCartography. It's cool to see your convergent solution to the problem of visualizing protein similarity applied to embeddings!

Might it be possible that the specific cells used for the half-cell experiment were in a rare or atypical cell state, which is underrepresented in the bulk data?

For ease of comparison to Fig. 1, perhaps you could use the same color scheme to demarcate A/P in the PCA plots, and use shape to indicate RNAi control vs. treatment.

Are there candidate gene names associated with these transcripts? For the reader, using those aliases may be easier to understand/remember as opposed to a variety of barcodes.

How many genes would be considered significantly varying if the q-value was instead 0.05 or 0.01?

To control for possible off-target effects, would it be reasonable to design a separate set of primers?

Are the samples in this example all labeled with the same 39-plex markers? If so, would it be possible to generate a joint embedding across all of the images? This would enable you to more easily relate the contents of the images to each other.

Will the underlying scripts and RNA-Seq data for this publication be made available at a later date?

I really appreciated how this manuscript takes advantage of one of the tasks LLMs are really good at—text summarization—and applies it to a scientific problem. Overall, the model's performance seems quite accurate for most questions (>90% in many cases). I imagine that additional prompt engineering and repeated prompting might continue to improve performance, as demonstrated by your updates to question 5. Thanks for sharing these results!

Have you considered using OpenAI's API to automate the submission of these queries? You can try out a GUI version of the API using ChatGPT's playground here. This interface has several advanced features that could improve your prompt responses. For example, you can set a System prompt to refine the behavior of ChatGPT, for example, telling it to respond only in exact quotations. You can also adjust the Temperature parameter of the model, which allows you to control how variable the responses are.

OpenAI has a really helpful overview of additional strategies you could take to improve the prompts, such as asking the model if it missed any information in previous responses. This might allow you to increase the accuracy of prompt responses where the model told you that no information was found, or to correct itself when it made a mistake.

without error?

How reproducible were the model's outputs? If you queried it repeatedly, did you observe different responses?

Do search-based methods, e.g. BLAST or Foldseek, suggest an origin for these genes outside of metazoa? Are there other non-metazoan organisms that have this pair of genes present in their genomes?

Was there a specific technical reason to validate the expression using a different species than was used for snRNA-Seq?

It's impressive how many different lepidopteran species were used in this study. Was there a technical reason to use Heliconius for the snRNA-Seq and Vanessa / Plodia / Junonia for the remaining experiments?

Is the smoothness of virtually stained fluorescence due to a denoising/ deconvolution effect of the model, a loss of information in the prediction process, or missing information in the original training data (e.g. stained nuclear fluorescence may not contain enough information to fully explain the brightfield intensities, or vice versa)? A combination of all of these factors?

When repeatedly performing a prediction on the same label-free image, do you observe different predicted virtual staining outputs each time? How much variability do you observe between predictions?

Do the authors plan to share the code that underpins these analyses at a future date?

It would be interesting in future studies to include transcriptome information from more rodents, to be able to determine what changes in gene expression have occurred along the gerboa lineage which may not be associated with changes in morphology. This would allow for winnowing the list down to fewer candidates.

Given that the morphology of the jerboa hindfoot is quite different from mouse, how expected would it be that atrophy is responsible for loss of these muscles? Are these muscles under less stress / use within the jerboa physiology as opposed to mouse?

What do the dark grey bars mean in this analysis? They do not seem to be reflected in the color scale.

Slightly decreasing the opacity of each point might make it clearer to see that there are multiple points for each species / stage of development / tissue.

In this manuscript, the authors use comparative transcriptomics across mouse and gerboa to investigate gene pathways that may be associated with loss of hindfoot muscles in gerboas. The authors generate RNA-seq data from forelimb and hindlimb muscles in mouse and gerboa and perform both inter-specific and intra-specific differential expression analyses to narrow down their list of genes to those that may be particularly associated with the gerboa trait. The authors then cross-reference these data with expression data from human disease contexts, identifying muscle atrophy as a potentially shared feature between developmental muscle loss in gerboas and human muscle phenotypes.

The authors acknowledge that the data presented may be incomplete. Nevertheless, I think this study provides an interesting framework for performing cross-species expression analyses. I think future studies including expression data from more species, including members of each lineage that either have or lack the phenotype of interest (hindfoot muscle loss) would help clarify which genes are truly important for this trait, as opposed to simply different across species due to evolutionary changes shared along each lineage that are unrelated to the phenotype of interest. Correcting for fore- and hind-foot differences in intraspecific analyses may not remove those confounding genes, particularly if gene expression is evolving independently in these tissues in each lineage.

In this manuscript, the authors characterize the expression of Hox genes in two morphs of the marine annelid Streblospio benedicti through in situ hybridization chain reaction (HCR) and time-course RNA-Seq analysis.

Despite some important morphological and life history differences between the two morphs, the authors observe broad similarity of Hox gene expression, with the exception of a few Hox genes that are expressed in swimming setae, which are found in one morph but not another.

Due to differences in the developmental rate of these two morphs, as well as physiological differences, such as embryo size, the authors find it difficult to definitively determine the source of differences between the two life histories. The authors conclude that Hox expression is not the primary driver of life history differences between these morphs. The authors note some heterochronies in Hox expression, but acknowledge that some of these may be a result of differences in staging between the two morphs.

It would be interesting to see if the authors could perform a genetic or chemical manipulation of each morph to cause it to alter its life history. Such a perturbation might make it easier to compare developmental trajectories within a shared physiological background, without the confounding properties of the two morphs examined in this study, which also came from geographically distinct populations.

How similar or different are the genomes of these two morphs? Are there any polymorphisms that distinguish the populations and could be relevant to differences in their development? Are there manipulations that can be performed to change the developmental trajectory of one population into the other, or vice versa? (I'm unfortunately not able to access the Zakas review, these may be obviously answered there).

Are there other kinds of genetic circuits, aside from the segmentation/regionalization hierarchy that you could examine which could be more easily analogized across the two morphs to create a "clock" which could be compared across the morphs? Perhaps eye development and looking at the cell types and morphology of the eyes?

upstream of?

For example, the swimming setae appear to emerge prior to the expression of Pb, suggesting that processes other than Hox, which precede (upstream of) Hox expression, drive the development of this structure, which then later expresses regional identity markers such as Pb.

The autofluorescent regions in the samples for the lecithotrophic and plantotrophic regions differ at the same developmental stage (e.g. 2-eye, Hox1/Lab). Is this a result specifically of variability in sample preparation, or a consequence of differences in embryo / cell size, and therefore different quantities of yolk autofluorescence?

Based on the methods detailed below, my understanding is that some of these hybridizations were imaged using different fluorescent channels. To aid understanding of autofluorescence, it might be helpful to false-color these images based on the fluorescent channel used for the hairpin (e.g. green for 488, red for 546, etc.) This could make it more obvious that the setae in the Post2 early plantotrophic stage are visible due to autofluorescence rather than genuine staining. You might also consider adding that detail directly to the figure legend, as the other reference to this artifact is only found in the methods.

Were DAPI stains also performed on these samples? If they were, providing those might make it somewhat easier to interpret precise segment boundaries and relative expression patterns across the samples.

Were there any plausible splice isoforms of the Hox genes identified in your RNA-Seq data?

Was the BLAST also performed reciprocally? Could you perhaps use a phylogenomic tool for ortholog detection, such as OrthoFinder, to systematically group genes across species into orthogroups?

We've also developed a phylogenomic workflow called NovelTree that builds on top of OrthoFinder to perform a variety of phylogenomic analyses, which might help systematically identify conserved genes across the spiralian species you work on.

Is there a specific value for this parameter that might be missing?

It might be helpful, if available, to share the cluster number ranges you tried to provide more clarity into the range of possible clustering outcomes.

It's cool to see that stepping along the hierarchy of predicted binding can actually lead to accurate predictions of gene expression!

I think this should say "H"? The temporal relationship might be a bit more legible if the RNA-Seq and peak accessibility heatmaps were stacked vertically rather than placed side-by-side.

labels in the figure image say "tropopyosin" rather than "tropomyosin"

Is it possible to visualize the footprints generated by TOBIAS for each timepoint which are likely to match the RUNT binding site at these loci? That might make it easier to evaluate whether RUNT binding is correlated with accessibility.

How well-conserved are the DNA binding domains of the Hofstenia hits for JASPAR TFs? I'm curious whether you'd expect the motifs to be comparable over large evolutionary distances – from my understanding, JASPAR motifs tend to be enriched for arthropod and vertebrate TFs.

It appears that many of the GO functions reported are higher-level biological functions; were there any more granular terms that showed up as significant in your analysis?

I'd be curious to know about the distribution of distances between promoter peaks and nearby intergenic peaks. Are putative regulatory regions mostly proximal to promoters, or are there lots of distal regulatory elements as well?

"principal"

In this manuscript, the authors generate a developmental timecourse of ATAC-Seq data for the acoel worm Hofstenia miamia. The authors use a variety of computational approaches to characterize the position and function of regulatory elements, identify clusters of regulatory elements with similar accessibility trajectories over time, and predict relationships between transcription factor binding and gene expression.

The authors test their predictions by identifying a hierarchy of predicted binding sites in the muscle lineage based on scRNA-seq data and confirming that genes predicted to be bound by transcription factors expressed in the muscle lineage are expressed in tropomyosin+ cells, suggesting that TF binding predictions can be used to identify lineage-specific genes.

The data and resources generated by this manuscript will be useful to both the community of researchers working on this specific organism, as well as scientists interested in performing cross-species analyses on gene regulatory evolution.

It would be interesting to know whether previously identified regulatory elements/ transgenic drivers in Hofstenia are also identified by ATAC-Seq, and whether the authors were able to use ATAC-Seq peaks to develop new reporters in Hofstenia.

Are Hand2 or its downstream targets known to result in changes in chromatin state? It would be interesting to determine what aspects of Hand2 regulation block posterior -> anterior switches.

This is a very nice demonstration of the consistency of the model and experimental results – effectively, due to the mosaicism of the insertion, the authors generated an ALM transplantation experiment in situ.

I wonder if decreasing the amount of the construct used for injection or using electroporation later in development (or as a juvenile), rather than injection, would result in a higher rate of accessory limbs due to a greater chance of discontinuity in identity? It seems surprising to me that this phenotype was only observed in cases of high Hand2 misexpression.

In this manuscript, the authors use RNA-sequencing, lineage tracing, and pharmacological treatments to identify the molecular basis of anterior-posterior positional memory in salamanders.

The experiments performed in this manuscript and the presentation of the results are strikingly elegant, both in visual and narrative clarity. I found Fig. 5, in which the authors visually demonstrate reprogramming of cell identity by recruitment to the blastema, to be particularly beautiful in terms of the simplicity and clarity of the results.

The results of these experiments – indicating that axolotl limbs use Hand2 as a memory marker for positional identity – would be interesting to explore in other regenerative, partially-regenerative, and non-regenerative organisms. Given that the axolotl limb seems to have somewhat different contribution of cell types to the limb compared to other well-studied vertebrates (for example, the proportion of cells descended from the Shh population), it would also be interesting to know whether this particularity of axolotl limb development plays any role in their capacity for limb regeneration.

To improve clarity, the arrows in panel A could also have dotted lines extended across panels B and C; currently, a casual glance could lead to the misinterpretation that they might be marking some feature of the JH-III molecule.

It's cool to see that, concurrent with degradation of limb morphogenesis, you could simultaneously observe differentiation.

did not form any?

It's surprising how quickly the JHm + 7EP treated embryos recovered the extraembryonic fluid reabsorption, given that the amount of gene expression rescue by JHm in Fig. 3B seemed to be more limited.

morphogenesis?

missing verb in this clause?

In this manuscript, the authors examined the role of juvenile hormone (JH) in the development of an ametabolous insect, the firebrat Thermobia domestica. JH is produced during the post-embryonic development of hemimetabolous and holometabolous insects and is necessary for metamorphosis. The authors suppress JH production and precociously induce JH signaling through exogenous application of synthetic JH mimics to understand the role of JH in Thermobia development.

The authors demonstrate that JH is necessary for Thermobia embryos to undergo proper differentiation – embryos treated with 7EP, which inhibits the production of endogenous JH, demonstrate delayed development of various body features, such as the ommatidia.

Conversely, embryos treated precociously with exogenous JHm (juvenile hormone mimic) showed developmental defects consistent with premature differentiation, such as the development of myofibrils earlier than in vehicle-treated embyros.

Overall, the authors provide multiple lines of evidence for their findings, from measuring hormone concentrations through chemical assays, evaluating gene expression using qPCR, and analyzing embryonic morphology using stains and brightfield classification of phenotypes. The authors propose that JH was coopted from an ancestral state, in which it regulated the morphogenesis to differentiation switch during embryonic development, to its derived state in hemi- and holometabolous insects, where it performs an analogous role in the maturation of nymph- and larval-stage insects towards adulthood.

In this thorough and clever study, the authors identify the diverse targets of the translational repressor Nanos, an RNA binding protein that secondarily promotes transcript degradation. While Nanos is known chiefly for regulating the transcription factor hunchback, its other targets have not been well-identified. The authors took the unconventional approach of identifying Nos targets by enhancing its transcript-degrading abilities through fusion to the Upf protein. Through a combination of morphological analysis, RNA-seq and differential expression, re-analysis of existing datasets, and even yeast hybrid experiments, the authors thoroughly identify thousands other binding targets of Nos.

Throughout the manuscript, the authors conduct a variety of sophisticated and multi-layered experiments that leverage a variety of techniques and existing datasets, allowing them to quite comprehensively examine all of the features of their dataset to understand how Upf fusion affects their results or to identify the Bruno gene as an additional candidate binding partner of Nanos. This work is creative, resourceful, and deep, and it will be exciting to see how others might take similar approaches to more deeply investigate other proteins important for the regulation of maternal gene expression.

To more clearly demonstrate rescue caused by the L7 mutation in Upf1-NosL7, it might be helpful to have images from a WT (non-transgenic) embryo.

For ease of comprehension, it might be helpful to indicate with text on top of the corners of each image the precise channel (bcd mRNA, DAPI, etc.)

It might be helpful to include the raw numbers for these figures in the figure legend or the figure itself, rather than having a separate supplementary file.

Also, what are these percentage in WT embryos from the parent strain of these transgenic organisms?

It is very impressive that the authors followed up on the observations that emerged from this control experiment with such rigor.

Are the authors able to directly compare Upf1-NosL7 / Upf1-Nos / Upf / Nos to each other via differential expression analysis? That might more specifically reveal the differences caused by addition Nos to the Upf protein, etc.

Fig. 2A shows the expression of bcd in the data already, but it would be nice to include the other three as well, since they are referenced here.

This is a very clean result!

It would be interesting to see some images of the bicaudal phenotype generated by this experiment, perhaps as part of Fig. 1.

duplication of "Nos"?

missing reference

What kind of protein is produced by g30822? It would be interesting to determine if this is a transcription factor, secreted factor, etc. and what orthologs it has in other arthropods.

Given the large number of embryo images in several of these figures, it might be helpful to provide larger versions of relevant groups of images as supplementary figures. As is, the images are quite small and features of the embryos can be difficult to distinguish. There also appears to be some debris (perhaps fibers?) in some of these images; it would be helpful to indicate those where present.

It might be helpful to repeat the cluster number label beneath the bar chart, as well as below the bubble chart, for ease of reading.

It might just be a consequence of file compression or something by bioRXiv, but it's very hard to interpret some of the expression data presented due to their low resolution in this format.

Did the authors look for enrichment of other genes homologous to insect gap genes? It would be really interesting to see whether there is a differential AP distribution of those ohnologs and how their expression relates to Hox expression.

It appears that the data used for performing the analysis are included in the methods. It might be helpful to mark that this information is included by adding "(described in Methods, Cell Cycle Scoring)" or some other parenthetical to this sentence.

I'm curious how different one would expect the stage 9.1 cells to be from the other stages. Might it be the case that all cell types undergo some global changes in gene expression that result in the cells appearing more separated, and that Harmony better captures that separation?

In this study, the authors examine spider development using ACME dissociation and SPLiT-seq at three developmental stages associated with segmentation and regionalization. The authors cluster cells in their data to identify groupings of cell identity across timepoints. The authors examine the expression of AP genes such as Hox genes and DV genes as well as newly-identified markers from their scRNA-seq data using in situ hybridization and fluorescent in situ hybridization.

In generating and analyzing their data, the authors uncover expression of genes in the precheliceral region and in the posterior SAZ, which gives rise to the opisthosomal segments.

This study and the data it generates provide an exciting window into spider development and should greatly accelerate future investigations.

One thing that could be added to the manuscript to provide a greater understanding of its impact would be a more thorough engagement with and discussion of the current arthropod comparative developmental literature.

For example, it would be interesting to consider how the data presented for the SAZ corresponds to the sequential addition of segments during development of the flour beetle Tribolium castaneum, a system for which thorough investigations of this process have been conducted.

It would also be interesting to hear consider how the authors might decode the logic of Hox gene co-expression in the spider appendages based on their RNA-Seq expression data, or how the data from precheliceral patterning might provide some additional insights into the arthropod head problem.

Overall, this study provides a wealth of data for future developmental biology work and will be a valuable resource for other researchers.

This study uses prdm1a -/- mutant zebrafish to evaluate the function of human PRDM1 variants within patients presenting craniofacial defects. By transiently expressing hPRDM1 variants and comparing their rescue potential to wildtype hPRDM1, the authors demonstrate that the variants have reduced capacity for rescue of craniofacial defects.

It would be really interesting to see how this kind of rare disease allele functional assessment in model organisms could be scaled to improve diagnosis of rare diseases. It's great that the authors performed these experiments and reported this result, which seems to stem from a previous study on related alleles.

How different in sequence are the zebrafish and human versions of prdm1a?

Would it be possible to generate mutant versions of the zebrafish gene based on specific mutations found in the human variants and examine their capacity for rescue?

This is a really cool outcome suggesting that you could determine how human variants from individual patients contribute to disease etiology.

How common is it to explore clinical variants using mutant rescue experiments in model organisms? Would it be possible to use this kind of strategy as a diagnostic approach for rare diseases?

In this impressive study, the authors use analysis of single-cell RNA-seq data to identify marker genes of the multiple celltypes found in the tunicate papillae. The authors build reporter genes and use combinatorial reporter expression to disambiguate the identity of cells within the papillae. The authors go further to use tissue-specific CRISPR perturbations to construct a genetic hierarchy of factors that determines many of the different cell identities within the papillae, and also functionally characterize cell types by assessing their ability to produce cement, characterizing the gene expression of knockouts using bulk RNA-Seq, and validating that observed gene expression changes (such as for the cytoskeletal gene villin) are functionally relevant for cell identity.

It's very impressive that the authors were able to build this many different reporter constructs and to perform a range of different tissue-specific CRISPR knockouts to map the differentiation hierarchy of this tissue with such depth.

I'm curious to know what genes are expressed in these individual cell types and how one could understand their evolution. Since sequence-based homology seems not to have been effective in identifying homologs to other organisms, it would be interesting if the authors could explore structure-based approaches or leverage cross-species analysis approaches that can be performed using existing scRNA-Seq data.

If you took the markers you genetically described here and looked back at the scRNA-Seq data, could you identify the core genes specifically responsible for producing the cement? Based on figure 5, it seems that ICs and OCs both contribute to cement production.

You could also leverage existing single-cell RNA-seq datasets and see if you can integrate expression from multiple species. Software such as SAMap or Seurat's CCA algorithm might allow you to merge cells from different species to look for shared gene expression signatures.

It's really impressive that this kind of tissue-specific knockout in just a few cells works well enough to see a measurable effect. I imagine the knockout efficiency must be very high!

If you were to align the orthologous sequences, how different are they? Are there particular motifs that seem to be mutually enriched in those sequences?

How was homology assessment determined? I can't seem to find mention of this in the methods.

I'm curious if functional analogs of these proteins could be identified using structural similarity approaches, such as Foldseek. The Foldseek web server provides a user-friendly GUI that could be used to search for similar proteins across the Alphafold (or other) databases and could provide a means to identify functionally relevant proteins in other organisms that aren't easily detected using homology approaches.

You could also try using HMMs to determine annotations for these proteins, if they're not available. HMMER has a web interface that searches existing HMM databases against your query sequence.

The programmatic version (noGUI) version of this tool could be made a little easier to use by accepting inputs from the command line, such as with argparse.

It should be somewhat straightforward to modify the code to make it accept command line arguments. A tutorial can be found here: https://realpython.com/command-line-interfaces-python-argparse/

Might there be physical constraints on the tissue itself that prevent an increased number of tooth fields from forming? Are there examples of total tooth number increasing without a concurrent increase in the pharyngeal tooth plate surface?

Additionally, as this experiment only marks the number of teeth at two timepoints – the start and the end of the experiment – is it possible that an entire tooth replacement cycle has occurred within the 18-day span, or is tooth replacement sufficiently slow that you would not expect teeth to be "missed" by this analysis?

The image label in the figure seems like it should be "G" as per the figure legend, rather than "H".

If Wnt and BMP are thought to have an oppositional effect on tooth replacement/regeneration, why are these two genes expressed in such similar patterns in early regenerating teeth?

Perhaps the early expression of BMP has the effect of repressing tooth formation in nearby tissue? What is the expression of BMP receptors within Wnt-expressing cells? It seems potentially notable that the Wnt-expressing cells appear to be a subset of the BMP-expressing cells, although a lack of bicolor in situ prevents any solid conclusions about coexpression.

It would be nice, if available, to show images of the zebrafish teeth with corresponding staining to in panels B and C.

Is there a more quantitative summary of the categories of teeth (early, mid, late, erupted, etc.) observed for each condition you could point to as evidence of this?

This result seems to depend on the inclusion of nascent teeth in the total number of teeth; the number of tooth families marked by erupted teeth is unchanged.

Is there a difference in the number of nascent teeth vs. erupted teeth in the stickleback case between the heat shock and no-heat shock conditions? Might that also affect the significance of the total tooth number result?

Probably refers to Fig. 6C?

Might be interesting to note that BMP OE seems not to be able to eliminate nascent teeth that have already begun the process of formation, as evidenced by the stalled stain-free teeth that are inferred to have arisen just after the pulse-chase treatment.

Maybe could be related to the fact that Bmp6 and Wnt seem to be expressed simultaneously in early teeth - perhaps those cells aren't receptive to Bmp6 signalling and can't be eliminated once they get going, but new tooth germs can be inhibited before they start being formed.

I appreciated the thorough phenotyping analyses of these experiments and the inclusion of multiple species in this paper! I imagine it was also a great deal of work to perform these transgenic experiments in the stickleback, especially for multiple genes and across large numbers of individuals for several different conditions. I had some questions about the implications of the results, as well as about some of the more detailed phenotypes that were discussed but I wasn't able to find figures for in the manuscript.

More of a stylistic choice, but some kind of summary figure would be great to help encapsulate the findings and the comparison of results between zebrafish and stickleback. Perhaps a table of some kind indicating + vs - effects for each phenotype, for each species?

It would be nice to indicate this in the figure legends, even if it's the same in each figure, for the sake of the reader.

Where possible, it would be great to include the exact number of individuals phenotyped for each condition plotted (basically, state the number of data points for each plot).

Would the authors be able to provide images of the no-probe controls, perhaps as a supplement? Given the high levels of background in the epifluorescence images, it is indeed challenging to distinguish what is true signal from autofluorescence or debris.

OTU 180?

The images for this figure in both the web-based full text and .PDF document appear to have compression artifacts. This might be a problem with how BioRxiv image compression works. However, having clearer images would make it much easier to determine if the co-localization indicated by the blue arrows is genuine. Perhaps the authors could also add cropped, zoomed versions of the specific colocalization spots to make the images easier to interpret.

For ease of comprehension, it would also be helpful to have text labels for each channel, perhaps overlaid on the bottom left corner of each image. This change could be made to each of the figures in the manuscript.

This manuscript appears to be a follow-up study of previous metagenomic analysis. The authors use FISH to assess whether members of the genus Rickettsia might be found in tardigrades. The authors use three probes: a Rickettsia-specific probe, a probe that generally labels the 16S ribosomal RNA of most bacteria, and a nonsense probe to assess whether RickB1 signal is genuine. The authors report a single sample with apparent Rickettsia staining.

This is intriguing, but additional images and information about the number of samples examined and image processing methods would be helpful to evaluate the strength of the evidence presented in this manuscript. I've annotated particular places where additional information might be helpful.

Would the authors be able to provide some statistics about the number of samples of each species examined? Perhaps a table of the number of individual animals imaged for each species, and a categorical description of how many animals appeared to have "true positive" puncta.

As the images in Figs 2 and 3 were taken using a confocal microscope, could the authors provide a rendering of the Z-axis of each image for each channel? Could the authors also describe how the images were processed? Are the images maximum intensity projections, or a single slice from a Z-stack? If the images are maximum intensity projections, for example, it is possible that some foci that appear to be fluorescent in the RickB1 channel but not the NONEUB channel could be obscured.

It would be helpful to explicitly state the values shared in Table 4 within the text to provide readers some context about the efficiency of recovering a phenotype.

It might be nice to indicate which species have the features/ methods of interest already developed. For example, you could have a number of columns to the right of the cladogram with a colored circle for:

a) specific biological features, such as life history (parthenogenesis vs. hermaphrodism); b) whether or not RNAi methods have been demonstrated; c) whether or not CRISPR/Cas9 methods have been demonstrated.

This could be a nice way to put the findings in this manuscript into a broader context with a figure.

If available, it might be useful for others familiar with nematode injection if an image of this needle compared to a standard needle could be included in the supplement.

Would the authors be able to place the efficiency in the context of existing literature for C. elegans or P. pacificus, for example?

Would this optimal timing potentially vary depending on what the specific phenotype of interest might be?

While the reported phenotypes (motility and survival) are convincing, if the authors performed such experiments, it might be helpful to know whether any sham injections (buffer only, Cas9 only, guide RNA only) are able to produce any offspring with the uncoordinated phenotype.

The authors indicate in Table 2 that some experiments were summarized together. It seems that the specific comparison being made in this section could benefit from some additional repeated trials to see whether the differences in annealing temperatures reproducibly result in a difference in phenotype recovery.

The authors of this study demonstrate CRISPR/Cas9 mutagenesis and knock-in for the non-model nematodes Panagrolaimus sp. PS1159 and Propanagrolaimus sp. JU765 using multiple orthogonal approaches, including behavioral analysis, chemical perturbation, nucleic acid biochemistry (T7, restriction digest), and Sanger sequencing. The results presented are well-supported and enable functional genetic perturbations in these organisms.

There are a few cases where additional data might strengthen the authors' claims, which I've marked. In some locations, the authors could provide a bit more context for their results for the ease of reader comprehension. I find that the explanation of the authors' more anecdotal observations is particularly helpful, especially for others who might try similar experiments in diverse nematodes.

This seems to be a pretty generous cutoff for log2 fold change - might this result in some false positive DE gene assignment?

At what point in the ortholog assignment approach were these genes "lost"? If the DIOPT cutoff were set to 7 instead of 8, for example, would these 4 remaining "positive control" genes be recovered?

The authors used differential expression between soma and periphery in mouse/rat data to try to identify genes localized to the periphery in Drosophila. Were the authors able to confirm enrichment of these predicted genes in the periphery? It's a bit difficult to tell based on the images, but it seems more that these genes are simply present at the periphery.

I think it would be helpful to more clearly state the number of transcripts evaluated for localization using smFISH. The "70%" value could be misinterpreted to be with respect to the 1700 mRNAs predicted to be localized to Drosophila glia.

I'm curious if more mappings of mouse to fly genes could be recovered if a newer approach were used. There are a lot of papers now using OrthoFinder (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1832-y) to identify orthologs between species, for example.

It might be helpful to include this result in as a fourth column in Table 5.

This manuscript leverages publicly available data from vertebrate glial periphery transcriptomes, as well as Drosophila single-cell RNA-Seq data, to try to predict the localization of mRNAs in the periphery of glia in the fruit fly Drosophila melanogaster. The authors predict around 1700 transcripts they expect to be localized to glial projections in Drosophila and test their predictions using a subset of genes previously known to be expressed in Drosophila PNS glia. The authors also functionally perturb genes they identified as localized to glia and show that these genes play a role in glial function.

This manuscript does a great job of leveraging publicly available datasets and databases of multiple types, such as Gene Ontology and Disease Ontology, as well as drawing insights from the biology of multiple organisms.

I think the authors were able to demonstrate that some genes in their analysis were present in the periphery of Drosophila glia, and that those genes function in the nervous system. It's unclear to me whether these genes are truly differentially expressed in the periphery relative to the soma, as might be predicted based on the starting mouse and rat data. It will be exciting to see how well these predictions perform in other glial cell types in Drosophila or other protostomes, and how this kind of analysis might be extended to explore additional invertebrates.

For genes primarily expressed in the mesoderm, would it be possible to display an orthogonal view of the confocal Z-stack? Perhaps as a supplement? It appears that some of these images are of a single slice or projection of a subslice of a full sample.

Most of trunk and posterior marker genes used for HCR in this study were Hox genes. Hox expression also appears to be restricted to mesoderm in a variety of other echinoderms, so perhaps it would be expected that these genes would not be expressed in the ectoderm of this organism. Are there other non-Hox trunk ectodermal markers – perhaps homologs of chordate or hemichordate genes – which would be strong markers for trunk ectoderm?

The claim that AP patterning is restricted to the ambulacral region seems to depend strongly on wnt3 being a terminal or posterior marker. Moreover, pax2/5/8 appears to be expressed more laterally than wnt3, which would correspond with a more "trunk-like" identity based on the corresponding expression of chordate/hemichordate homologs. Additional ectoderm-expressed marker genes showing a similar boundary to wnt3 would strengthen the argument that there is no ectoderm equivalent to a trunk region.

For ease of interpretation and consistency with other figures, it might be useful to replace "Left-right" with "Medial-lateral".

It might be helpful to have this figure also elaborated as a schematic with corresponding colors, such as in Supp. Fig. 3. As someone without familiarity with the anatomy, it's a little difficult to wrap my head around this image. It would also be helpful to indicate where the ambulacral boundary is with respect to the other features, to be able to more easily interpret panels u-b'. Adding schematics of the classified expression pattern to the left of each row of images could help ease interpretation.

It would be very interesting to look at the oral-aboral tomography data to look for genes that are more highly expressed on the aboral side - perhaps those genes could help clarify the axial identity of this unplaced tissue.

This manuscript explores how the AP axis is patterned in the derived body plan of echinoderms. The authors use RNA tomography in multiple sequential sections of juvenile starfish limbs to evaluate the expression of canonical groups of AP genes with respect to the proximal-distal, oral-aboral, and medial-lateral axes. The authors visualize the gene expression patterns of canonical AP marker genes using HCR in starfish juveniles.

Based on these data, the authors build a model of the starfish body plan that places anterior genes in the ambulacral region of the ectoderm (the ambulacral-anterior hypothesis). The authors show that a large portion of the starfish ectoderm – the interambulacral region – does not appear to express canonical ectodermal markers of the AP axis. Most of the trunk markers used in this study were Hox genes, which in this study and other echinoderm studies appear to be primarily restricted to the mesoderm. HCR of additional ectodermal trunk markers would help clarify the identity of the interambulacral region.

It will be very interesting to understand what genes are expressed in the interambulacral region to determine whether starfish are indeed "mostly head-like animals."

This manuscript explores how the AP axis is patterned in the derived body plan of echinoderms. The authors use RNA tomography in multiple sequential sections of juvenile starfish limbs to evaluate the expression of canonical groups of AP genes with respect to the proximal-distal, oral-aboral, and medial-lateral axes. The authors visualize the gene expression patterns of canonical AP marker genes using HCR in starfish juveniles.

Based on these data, the authors build a model of the starfish body plan that places anterior genes in the ambulacral region of the ectoderm (the ambulacral-anterior hypothesis). The authors show that a large portion of the starfish ectoderm – the interambulacral region – does not appear to express canonical ectodermal markers of the AP axis. Most of the trunk markers used in this study were Hox genes, which in this study and other echinoderm studies appear to be primarily restricted to the mesoderm. HCR of additional ectodermal trunk markers would help clarify the identity of the interambulacral region.

It will be very interesting to understand what genes are expressed in the interambulacral region to determine whether starfish are indeed "mostly head-like animals."

It would be very interesting to look at the oral-aboral tomography data to look for genes that are more highly expressed on the aboral side - perhaps those genes could help clarify the axial identity of this unplaced tissue.

Most of trunk and posterior marker genes used for HCR in this study were Hox genes. Hox expression also appears to be restricted to mesoderm in a variety of other echinoderms, so perhaps it would be expected that these genes would not be expressed in the ectoderm of this organism. Are there other non-Hox trunk ectodermal markers – perhaps homologs of chordate or hemichordate genes – which would be strong markers for trunk ectoderm?

The claim that AP patterning is restricted to the ambulacral region seems to depend strongly on wnt3 being a terminal or posterior marker. Moreover, pax2/5/8 appears to be expressed more laterally than wnt3, which would correspond with a more "trunk-like" identity based on the corresponding expression of chordate/hemichordate homologs. Additional ectoderm-expressed marker genes showing a similar boundary to wnt3 would strengthen the argument that there is no ectoderm equivalent to a trunk region.

For genes primarily expressed in the mesoderm, would it be possible to display an orthogonal view of the confocal Z-stack? Perhaps as a supplement? It appears that some of these images are of a single slice or projection of a subslice of a full sample.

It might be helpful to have this figure also elaborated as a schematic with corresponding colors, such as in Supp. Fig. 3. As someone without familiarity with the anatomy, it's a little difficult to wrap my head around this image. It would also be helpful to indicate where the ambulacral boundary is with respect to the other features, to be able to more easily interpret panels u-b'. Adding schematics of the classified expression pattern to the left of each row of images could help ease interpretation.

For ease of interpretation and consistency with other figures, it might be useful to replace "Left-right" with "Medial-lateral".

This manuscript leverages publicly available data from vertebrate glial periphery transcriptomes, as well as Drosophila single-cell RNA-Seq data, to try to predict the localization of mRNAs in the periphery of glia in the fruit fly Drosophila melanogaster. The authors predict around 1700 transcripts they expect to be localized to glial projections in Drosophila and test their predictions using a subset of genes previously known to be expressed in Drosophila PNS glia. The authors also functionally perturb genes they identified as localized to glia and show that these genes play a role in glial function.

This manuscript does a great job of leveraging publicly available datasets and databases of multiple types, such as Gene Ontology and Disease Ontology, as well as drawing insights from the biology of multiple organisms.

I think the authors were able to demonstrate that some genes in their analysis were present in the periphery of Drosophila glia, and that those genes function in the nervous system. It's unclear to me whether these genes are truly differentially expressed in the periphery relative to the soma, as might be predicted based on the starting mouse and rat data. It will be exciting to see how well these predictions perform in other glial cell types in Drosophila or other protostomes, and how this kind of analysis might be extended to explore additional invertebrates.

The authors used differential expression between soma and periphery in mouse/rat data to try to identify genes localized to the periphery in Drosophila. Were the authors able to confirm enrichment of these predicted genes in the periphery? It's a bit difficult to tell based on the images, but it seems more that these genes are simply present at the periphery.

It might be helpful to include this result in as a fourth column in Table 5.

At what point in the ortholog assignment approach were these genes "lost"? If the DIOPT cutoff were set to 7 instead of 8, for example, would these 4 remaining "positive control" genes be recovered?

I'm curious if more mappings of mouse to fly genes could be recovered if a newer approach were used. There are a lot of papers now using OrthoFinder (https://genomebiology.biomedcentral.com/articles/10.1186/s13059-019-1832-y) to identify orthologs between species, for example.

This seems to be a pretty generous cutoff for log2 fold change - might this result in some false positive DE gene assignment?

I think it would be helpful to more clearly state the number of transcripts evaluated for localization using smFISH. The "70%" value could be misinterpreted to be with respect to the 1700 mRNAs predicted to be localized to Drosophila glia.

The authors of this study demonstrate CRISPR/Cas9 mutagenesis and knock-in for the non-model nematodes Panagrolaimus sp. PS1159 and Propanagrolaimus sp. JU765 using multiple orthogonal approaches, including behavioral analysis, chemical perturbation, nucleic acid biochemistry (T7, restriction digest), and Sanger sequencing. The results presented are well-supported and enable functional genetic perturbations in these organisms.

There are a few cases where additional data might strengthen the authors' claims, which I've marked. In some locations, the authors could provide a bit more context for their results for the ease of reader comprehension. I find that the explanation of the authors' more anecdotal observations is particularly helpful, especially for others who might try similar experiments in diverse nematodes.

Would this optimal timing potentially vary depending on what the specific phenotype of interest might be?

The authors indicate in Table 2 that some experiments were summarized together. It seems that the specific comparison being made in this section could benefit from some additional repeated trials to see whether the differences in annealing temperatures reproducibly result in a difference in phenotype recovery.

Would the authors be able to place the efficiency in the context of existing literature for C. elegans or P. pacificus, for example?

While the reported phenotypes (motility and survival) are convincing, if the authors performed such experiments, it might be helpful to know whether any sham injections (buffer only, Cas9 only, guide RNA only) are able to produce any offspring with the uncoordinated phenotype.

It would be helpful to explicitly state the values shared in Table 4 within the text to provide readers some context about the efficiency of recovering a phenotype.

If available, it might be useful for others familiar with nematode injection if an image of this needle compared to a standard needle could be included in the supplement.

It might be nice to indicate which species have the features/ methods of interest already developed. For example, you could have a number of columns to the right of the cladogram with a colored circle for:

a) specific biological features, such as life history (parthenogenesis vs. hermaphrodism); b) whether or not RNAi methods have been demonstrated; c) whether or not CRISPR/Cas9 methods have been demonstrated.

This could be a nice way to put the findings in this manuscript into a broader context with a figure.

OTU 180?

Would the authors be able to provide some statistics about the number of samples of each species examined? Perhaps a table of the number of individual animals imaged for each species, and a categorical description of how many animals appeared to have "true positive" puncta.

As the images in Figs 2 and 3 were taken using a confocal microscope, could the authors provide a rendering of the Z-axis of each image for each channel? Could the authors also describe how the images were processed? Are the images maximum intensity projections, or a single slice from a Z-stack? If the images are maximum intensity projections, for example, it is possible that some foci that appear to be fluorescent in the RickB1 channel but not the NONEUB channel could be obscured.

The images for this figure in both the web-based full text and .PDF document appear to have compression artifacts. This might be a problem with how BioRxiv image compression works. However, having clearer images would make it much easier to determine if the co-localization indicated by the blue arrows is genuine. Perhaps the authors could also add cropped, zoomed versions of the specific colocalization spots to make the images easier to interpret.

For ease of comprehension, it would also be helpful to have text labels for each channel, perhaps overlaid on the bottom left corner of each image. This change could be made to each of the figures in the manuscript.

Would the authors be able to provide images of the no-probe controls, perhaps as a supplement? Given the high levels of background in the epifluorescence images, it is indeed challenging to distinguish what is true signal from autofluorescence or debris.

This manuscript appears to be a follow-up study of previous metagenomic analysis. The authors use FISH to assess whether members of the genus Rickettsia might be found in tardigrades. The authors use three probes: a Rickettsia-specific probe, a probe that generally labels the 16S ribosomal RNA of most bacteria, and a nonsense probe to assess whether RickB1 signal is genuine. The authors report a single sample with apparent Rickettsia staining.

This is intriguing, but additional images and information about the number of samples examined and image processing methods would be helpful to evaluate the strength of the evidence presented in this manuscript. I've annotated particular places where additional information might be helpful.

Where possible, it would be great to include the exact number of individuals phenotyped for each condition plotted (basically, state the number of data points for each plot).

I appreciated the thorough phenotyping analyses of these experiments and the inclusion of multiple species in this paper! I imagine it was also a great deal of work to perform these transgenic experiments in the stickleback, especially for multiple genes and across large numbers of individuals for several different conditions. I had some questions about the implications of the results, as well as about some of the more detailed phenotypes that were discussed but I wasn't able to find figures for in the manuscript.

It would be nice to indicate this in the figure legends, even if it's the same in each figure, for the sake of the reader.

More of a stylistic choice, but some kind of summary figure would be great to help encapsulate the findings and the comparison of results between zebrafish and stickleback. Perhaps a table of some kind indicating + vs - effects for each phenotype, for each species?

Probably refers to Fig. 6C?

Might be interesting to note that BMP OE seems not to be able to eliminate nascent teeth that have already begun the process of formation, as evidenced by the stalled stain-free teeth that are inferred to have arisen just after the pulse-chase treatment.

Maybe could be related to the fact that Bmp6 and Wnt seem to be expressed simultaneously in early teeth - perhaps those cells aren't receptive to Bmp6 signalling and can't be eliminated once they get going, but new tooth germs can be inhibited before they start being formed.

Is there a more quantitative summary of the categories of teeth (early, mid, late, erupted, etc.) observed for each condition you could point to as evidence of this?

This result seems to depend on the inclusion of nascent teeth in the total number of teeth; the number of tooth families marked by erupted teeth is unchanged.

Is there a difference in the number of nascent teeth vs. erupted teeth in the stickleback case between the heat shock and no-heat shock conditions? Might that also affect the significance of the total tooth number result?

It would be nice, if available, to show images of the zebrafish teeth with corresponding staining to in panels B and C.

Might there be physical constraints on the tissue itself that prevent an increased number of tooth fields from forming? Are there examples of total tooth number increasing without a concurrent increase in the pharyngeal tooth plate surface?

Additionally, as this experiment only marks the number of teeth at two timepoints – the start and the end of the experiment – is it possible that an entire tooth replacement cycle has occurred within the 18-day span, or is tooth replacement sufficiently slow that you would not expect teeth to be "missed" by this analysis?

The image label in the figure seems like it should be "G" as per the figure legend, rather than "H".

If Wnt and BMP are thought to have an oppositional effect on tooth replacement/regeneration, why are these two genes expressed in such similar patterns in early regenerating teeth?

Perhaps the early expression of BMP has the effect of repressing tooth formation in nearby tissue? What is the expression of BMP receptors within Wnt-expressing cells? It seems potentially notable that the Wnt-expressing cells appear to be a subset of the BMP-expressing cells, although a lack of bicolor in situ prevents any solid conclusions about coexpression.