The plasticity of interactions is neatly demonstrated here. I wonder if the authors have any intuition about whether this plasticity could be controlled or reduced in more complex communities. In other words, would the presence of more species stabilize the interactions in the community regarding environmental changes?

Thank you for this very interesting and clearly presented work! Looking forward to seeing what comes next for more complex communities.

I am curious to know if some of these random environments are similar to known/relevant ecological niches, and which one? It could be informative to add this information somewhere if available.

Have the authors investigated whether the distribution of competition and cooperation interactions shows any correlation with the phylogenetic distance of the species?

When both original environments shared similar compounds, did you add up the concentrations of these compounds, or did you keep the concentration from the original environment with the highest concentration?

As a general thought, I wonder if, while you don't see a profound effect of evolution on the species phenotype yet and potentially no specific mutant with enough fitness gain to have taken over a species population, the within-species genetic heterogeneity that's likely to be present for each species can influence the global set of interactions and the overall degradation phenotype/ability.

Thank you for this very interesting study. This represents a considerable amount of experimental work, and the strategy behind the directed evolution is clear and clever.

Typo

Typo: this should be mono- and co-cultures

Typo?

It is unclear here whether the authors are measuring the degradation activity of the species isolated from the best degrading communities after the 18 rounds or the original species.

Have these species (and any other species) in this work been sequenced?

Just a tiny detail, but there are inconsistencies in whether the authors use overrepresented, over-represented, underrepresented or under-represented

Typo?

Typo?

Is there a reason behind this choice? Did you want to keep experimenting until you reach a certain degradation score?

Are the species going extinct always the same? If this is the case, I wonder if this could highlight some potential specific physiological features or simply that this species may have been consistently inoculated at a lower concentration. Overall, I think it could be very interesting/informative to mention this information.

Typo: space missing

While the species have been normalized to the same OD before inoculation, do the authors know if this roughly represents similar cell concentration across species? Are all inoculations close to a 1:1:1:1 ratio?

Do the authors believe this would still be feasible for communities that need to be grown on solid medium?

Thank you for this great work. This seems like a game-changer for assembling microbial communities, and I hope many groups will take advantage of this. The versatility of the approach is really interesting, as the proof of concept with food coloring has illustrated it.

By 'highest biomass', do the authors mean highest cell density? I wonder if there could be interactions that lead to morphological changes in some strains that could then affect the absorbance reading. In other words, the OD is greater but not because there are more cells but because some cells have weird morphologies.

I wonder if any biofilm formed at the top or bottom of the wells? Have the cultures been homogenized before absorbance was measured?

typo

Can the authors give a little more context on the origin of the strains? And if/why a community with these strains could be ecologically relevant?

There seems to be a couple of typos in this section

This might just be a typo

Something seems to be missing in this sentence. Maybe ' Consequently, there has been an increasing number of studies on wine microbial ecology, ranging from investigation on the factors influencing yeast and bacterial diversity in grape must, such as climate and vine management (Bokulich et al., 2014; Grangeteau et al., 2017; Bagheri et al., 2018), to the mechanisms of interactions involved between the different species (Bordet et al., 2020) ' ?

Thank you for this very interesting work. The ability of bacteria to survive starvation and adapt to their environment is a very fascinating question.

While the authors mention in the introduction that the methyl compounds produced by the algae result from photosynthetic metabolism, I am wondering if there is anything known about other bacteria-algae interactions and whether algae benefit in any way from bacterial interactions.

As number of methyl groups seem to impact how the duration of the lag phase decreases, I am wondering if the authors have tried to mix/combined different compounds and see if the total number of methyl groups adds up and shortens the lag phase even more (or reach saturation)?

I am wondering if there is any knowledge about storage compound accumulated by P. inhibens before the starvation phase. It has been previously shown in other organisms that such compounds can help provide the energetic requirements for metabolic switch or exiting lag phase. For instance, glycogen appears to be important for *E. coli * to maintain ATP requirement when switching from a glycolytic to a gluconeogenic metabolism (https://doi.org/10.1128/mbio.01628-17) or primary source of glucose in lag phase (https://doi.org/10.1016/j.bbapap.2012.06.010)

I am wondering if the authors have performed statistical tests for panels A and B to test whether the lag phase is significantly decreased in the presence of the different methylated compounds compared to the control culture. Even if the results are quite clear, it might be appropriate to add some statistical validation.

I am wondering how genetically different are all the species. How much of the genomes of the species overlap? How much of each species' genome was recovered without cross-mapping?

Are the co-cultures inoculated with a 1:1 ratio for each species?

Thank you for this very interesting work. I agree that studying the roles of interactions in microbes' evolution and our ability to predict microbial community evolution is crucial. This works is a very nice contribution to this challenging goal.

FigS1 and FigS2 indeed show the diversity in growth. Is the diversity in degradation an expected phenotype as a consequence of the growth differences? Or is this statement about diversity in degradation associated with a measured trait?

With the sequencing data from the different transfers, is it possible to trace back when the acuC appeared?

The authors mentioned before that the experiment represents around 300 generations per species. Do they expect to see the fixation of many variations that provide a fitness advantage in 300 generations? Have the authors an idea of what the trajectories would be if they pursued the experiment for another 200 generations, would they see fewer variants with intermediate frequency? 300 generations are already a lot and it took almost a year, so I am just wondering if now that they see these trajectories, they could predict the trajectory for more generations.

Have you characterized the variant population earlier than transfer 11? While Ct accumulated more variations when evolving in the community, how wonder how this compares to the initial population, and how this could inform us about the evolution rate. Also, it could be interesting to see if all the species had a similar diverse population at the beginning and whether it may have influenced the outcome of the experiment.

This might be a typo. Is the number of the Table missing?

It would be interesting here to mention how easy or complex it is to obtain the GEMs for different species and what could be the current challenges or limitations.

How do these overlaps compare with genome similarity? Have the authors evaluated whether the species with the highest nutritional overlap are the ones with the highest genome similarity (or GEMs)?

While it is mentioned in the Methods section that the GEMs for these species have been obtained from the literature, it could be interesting to mention it here too and maybe give a little more detail about the these GEMs

Thank you for this very interesting study. The work is well explained and the high quality of the figures makes this work very enjoyable to read.

Is the total number of replicates n=6 as suggested in Figure 1c?

All the experiments have been carried out in liquid medium, while the natural environment of this microcosm might be closer to a solid environment. It could be interesting for the authors to briefly comment whether they would expect similar observations for growth on solid medium where microbial community structure might be important.

It is mentioned in the Methods section, that OD was measured at the end of each cycle. Do the authors observe any interesting patterns of OD depending on the community composition outcomes?

While this is briefly addressed in the Discussion section, it could be interesting to have a little more context about the choice of these species and their ecology. It could also be informative to know if (or how) the inoculation ratios used here would are representative of the natural community.

While these results have been observed for native human and mouse skin commensals, I wonder if people expect the same results if the damaged skin was treated with 'self-microbiome' (microbiome or single species formerly isolated from the wounded animal itself) ? Is there any precedent of such study? I wonder if somebody's immune system adapts to its own microbiome and vice-versa. Thus, introducing bacterial species, even if commensal but non from the host, could lead to specific immune response and /or bacterial adaptation to a new environment leading to inflammation and delayed healing.

I am curious to know if the authors had considered using combinations of different species in addition to single species. While single species don't seem to improve healing, do they expect that combination of species would yield a similar output? Because microbial communities phenotypes are rarely the sum of individual phenotypes thanks to microbial interactions, I wonder if healing would be improved or at least no delayed when using a mix of commensal species.

I am curious to know if the authors have measured, for all the tested strains, the growth and survival at wound site. Did E. coli grow well and did not delay the healing, or is it just because it did not grow, or not as much as the other species?

Thank you for this very interesting work that addresses fascinating abilities of bacteria (microbial interactions, phenotypic heterogeneity and antibiotic persistence).

This is a very nice and clear conclusion. Yet this has been only performed for 1 mutualistic 2-species community and for a single antibiotic. It would be interesting to know the authors' rational about choosing this antibiotic specifically, and why not extending these assays to different antibiotics and/or communities (yet I understand this is not that easy to find tractable mutualistic communities). Also, would the authors expect that this observation to be independent on the type of antibiotic?

This is a very interesting observation. Do the authors have ideas/suggestions about the molecular mechanism/response? Would antibiotic persistence be associated with a broad stress response (like stress response, stringent response, entry to stationary phase for instance?), or do they expect it to be a more specific mechanism? Would there be a way to use microscopy to see if such broad responses are activated in persistent cells?

How many time point and total replicates were used in the experiment?

Thank you for this elegant study. The text is clearly written and the figures are really nice. Looking forward to reading follow up work from this!

Does T.mu possess specific molecules that act like iron chelators (like siderophores?)

We can clearly see on Fig.1d that indeed the proportion of different phyla is altered in the presence of T. mu. I am wondering, if, at the species level, T. mu is affecting what species are present or if T. mu is not really affecting what species are present but mostly affecting their distribution/proportion?

Thanks a lot for this very interesting and exciting work. It is super motivating to see studies of understudied (and under appreciated) microbes and get to see how important they are in microbial communities. Also, the work is very clear and pleasant to read. Congratulations on this great work.

While the data strongly suggest that competition between specific bacterial species and the protist is associated with Tmu and Tc establishing in the gut, I am curious to know if the authors know if these two protist could actually graze on gut bacteria as well. Would that be possible to carry out some assays, especially now that the authors have developed a method to grow them in vitro , to test whether the Tmu and Tc actually graze on bacteria ; Assays comparing nutrient degradation rates of the protist species and the bacterial species could also be interesting to further confirm the hypothesis of nutrient competition, though it of course depends on the actual ability to growth these gut bacteria in vitro.

This is really exciting data. I wonder if, using these metagenomes data, the authors had the opportunity to investigate co-occurrence (or co-exclusion) of protist and bacterial or fungal species of the gut microbiome, and whether we can expect strong signal of species co-occurrence.

While this compound seems to be a good candidate for new antifungal therapy, I wonder how it could also impact the local microbiome (skin, gut, vaginal... depending on the infection + delivery method) - iron is crucial for (beneficial) microbes growth and is involved in microbial interactions. Do the author know if such antifungal therapy could interact with the local microbiome? It could be interesting/important to include such considerations in further studies and tests with this compound.

Looking at the Method section and the Figure legend, it seems that proliferation assays are OD-based. I am curious to know if the authors know (from this assay or complementary assay) if the the proliferation is reduced because cells are actually killed, or they are still alive but not dividing. Has a complementary dead/live assay been considered or performed? Or is it not necessary?

This is a very elegant demonstration.

This is great to have measured the functional performance. Do we know how this compare to the 'natural' performance in the ponds?

If the authors have any metagenomic information about the original or in situ sample (the one directly from the pound, that would be nice to compare with the domesticated community how this has changed.

Does this mean that 4 replicates have been grown for 7 days (without transfer) for each sample (and for a single round), or that each replicate has been grown for multiple consecutive rounds of 7 days to check stability of the community composition?

Looking at Figure 1, it seems that the authors have taken samples that were initially grown for 7 days before cryopreservation. I wonder how this growth period has affected the initial composition of the community, and how many species may have been lost in the process. Do the authors have that information? Maybe not for all samples as this would be a lot. Can the authors maybe elaborate about why they chose to carry out such initial growth rather than re-suspending the sample in buffer and cryopreserving the sample directly?

Is the microbial community of this environment uniquely bacteria? I wonder if other microorganisms such as fungi or protists are also parts of the microbial ecosystems and how much they interact with the bacterial community and the beech degradation function? What is the authors rational behind domesticating only bacteria?

Cryopreservation may have affected the community composition. Have the authors also looked at the community just after thawing the sample, to have an idea of the 'living' composition of the starting community? Would this be an important information to have to control or understand the process of reproducible domestication?

Very nice example of non additivity of interactions. Thanks!

I feel like the information of species viability in the different conditions could be mentioned at the beginning of the results and be an actual figure (not only a supplementary figure). I think it is an important level of information about interaction and gives more material and information to understand and interpret the RNASeq and Metabolomic data.

This is a very neat and clever system.

When multiple species are together, are their stationary phase synchronized? I am wondering whether one species would be in early stationary phase or still growth phase while the other species are already in stationary phase and how this would affect interpretation of the results.

Does that mean that each species would reach stationary phase after the same time independently of the presence of another species so the time point for RNASeq would match? In other, were the glucose concentrations set so that each species keep a similar growth rate in each condition?

The question of stationary phase is very interesting here. I have a couple of questions about how the authors define/characterize stationary phase in their system ? While stationary phase usually means no observation of active growth, do the authors know whether this is thanks to cell maintenance or a balanced ratio between cell death and cell division? - would one or the other explanation modify the interpretation of their results, as each scenario could be driven by very different metabolic states?

Have the authors looked at the interaction between both species in the absence of LA? I am wondering if other nutrients may actually be driving the competition, and to what extend the competition for LA would is actually the underlying mechanism for At growth inhibition.

Depending on the motility mechanism (flagellum, versus gliding etc...) of the bacteria, I am wondering if this is easier/faster to evolve resistance (for instance mutation of the phage receptor) versus improving motility through modification of the motility apparatus? Would it take more mutations do evolve better motility? Have the authors any thoughts on this, or idea whether this could be a potential explanation?

Have the authors considered doing whole genome sequencing of these 5 isolates to narrow down the mutations associated with these evolved lineages? Or based on the observed evolved phenotypes, can the authors thing of potential gene candidates to investigate mutations associated with the observed resistance. I am wondering if, even if you only see evolution of resistance and not of dispersal activity, this would be associated with similar mutations regardless of the experimental setup or whether we could expect a different mechanism of resistance between local and global phage distribution.

Very simple question: does the 5X on Figure 1 means each experiment has been performed in 5 replicates? Also, during the propagation, was each plate propagated to a single plate?

I believe 'emergent functional properties' can have multiple definitions. One definition is 'any functional property that is only observed in the community context and never in single organisms' and another, maybe broader definition, is 'any functional property observed in the community that cannot be predicted as the sum of individual behavior/function alone'. I assume this is the latest definition that is implied in this study, but it would be helpful to state it clearly.

Have the authors any ideas/thoughts about contexts that involved less metabolism related function like the community response to a perturbation (invader, predator, antimicrobial compound, temperature change, radiation...). If the function outcome that is measured is community biomass or individual species abundance, can we expect low-order approximations of landscape to work well as for the biomass prediction demonstrated in this work?

This is really exciting and really promising. Very interesting and clear work!

This is really exciting and really promising. Very interesting and clear work!

Have the authors any ideas/thoughts about contexts that involved less metabolism related function like the community response to a perturbation (invader, predator, antimicrobial compound, temperature change, radiation...). If the function outcome that is measured is community biomass or individual species abundance, can we expect low-order approximations of landscape to work well as for the biomass prediction demonstrated in this work?

I believe 'emergent functional properties' can have multiple definitions. One definition is 'any functional property that is only observed in the community context and never in single organisms' and another, maybe broader definition, is 'any functional property observed in the community that cannot be predicted as the sum of individual behavior/function alone'. I assume this is the latest definition that is implied in this study, but it would be helpful to state it clearly.

If the species of the community can be organized into niches (within the niche network mentioned in the first comment), it could also be informative to represent them organized as such, to eventually see where the species whose abundance is impacted by the presence/absence of Cs and Ch are directly or indirectly metabolically connected to them

If the gut microbiome can be divided into niches (or a network of niches), do the authors know how many ‘niches’ can thus be defined (in the model community used in the paper), how they are connected and how would the bile acid niche fit into such network?

Where do CA and DCA come from in the gut or in the synthetic community?

This work and these results are really interesting and exciting. This whole story and its conclusions could nicely be summarized into a diagram to make it even easier to navigate and understand.

How do the authors controlled for the species inoculation level? Inoculation level of a given species can also impact its final relative abundance. While in the Methods it seems that the inoculations are all done from the same species pools, aliquots have been frozen and thawed before inoculation, while the metagenomic analysis of the pool has been performed before freezing. So I am wondering if this could have introduced some inoculation differences and how the authors distinguish between potential inoculation effect versus interaction effect when comparing the relative abundance of the same species in the hCom1a experiment and the double dropout for instance? Would a comparison of abundance fold change (between Cecum and Inoculum) between 2 conditions lead to the same conclusions? For instance: comparing the abundance fold change of Ruminococcus obeum between Cecum and inoculum of the hCom1a condition to Ro abundance fold change (Cecum versus inoculum) in the double dropout condition.

What motivated the decision of FC > 100?

Have the authors considered looking at metatranscriptome modifications and untargeted metabolites profiles? That could provide an even broader insight into the modifications following the suppression of 1 or 2 strains, that may or may not lead to such drastic changes in community composition but rather in species physiology.

Are these species part of the same niches? Are they part of upstream or downstream niches? Replacing them in a metabolic network could be interesting to highlight potential interactions.

Since the authors are mentioning the species that degrade the chitin, reaction from which the rest of the trophic chain will cascade, shouldn't that be 'at the top' of the trophic hierarchy instead?

It is known that growth rate directly impacts gene expression, and that sometimes, gene expression changes between two conditions are simply connected to a decreased growth rate in the studied condition. I am wondering if there would be a way to sort out the genes that are associated with differential expression between genes whose expression change is likely related to growth rate effect versus genes whose expression changes are not (based on information that can be found in the litterature).

Have the authors considered or tried using spent supernatant from these 4 strains (instead of directly inoculating the strains)? It could be interesting to compare the effects of spent supernatant versus the presence of actual species to see if they mostly compare to the '6 hour prior to the target strain' or if they are pretty different.

Eventually, if the authors want to measure/map the interactions within microbial communities using uCI, have the authors tested whether the effects of 3 species inoculated in 3 different variable wells connected to the same target strain are the same effect on the target strain than if 3 species were inoculated together in the same variable well. I believe that having the 3-species mixed together and associated interactions would lead to a different growth outcome for the target strain (and might be more representative of the original natural ecosystem). It is quite nice that the uCI platform offers the opportunity to explore such questions (whether community effects are the sum of more simple effects) and help decipher higher-order interactions.

It could be informative to specify where the barcodes have been introduced in E. coli's genomes. While this information can be found in the paper mentioned as reference 24, this information is worth restating here to prevent confusion with barcoded gene disruption libraries (for instance RB-TNSeq libraries).

To help understand the different statements and conclusions about successful and unsuccessful colonization, the authors could define clearly what they mean 'colonization'. It is clear for the im cohort that E. coli doesn't survive, however what are the criteria to differentiate between a transient population versus an established population, especially in regards to the transit time in the mice gut.

Have the authors characterized the microbiome of all the mice at T0 before the E. coli gavage to identify any different starting points in terms of community composition that could potentially influence the outcome of the colonization (for instance the presence of a couple of specific species that could promote horizontal gene transfer)

• lineage ?

While the authors mention that the fraction of pairs exhibiting exclusion is consistent across communities independently of their richness, is there any pattern between coexistence scenarios highlighted in Figure 3 and community richness?

Maybe the authors could clarify why they chose to look at gene expression 4 days after milking.

Maybe the authors could clarify why they chose to look at gene expression 4 days after milking.

• lineage ?

Have the authors characterized the microbiome of all the mice at T0 before the E. coli gavage to identify any different starting points in terms of community composition that could potentially influence the outcome of the colonization (for instance the presence of a couple of specific species that could promote horizontal gene transfer)

To help understand the different statements and conclusions about successful and unsuccessful colonization, the authors could define clearly what they mean 'colonization'. It is clear for the im cohort that E. coli doesn't survive, however what are the criteria to differentiate between a transient population versus an established population, especially in regards to the transit time in the mice gut.

It could be informative to specify where the barcodes have been introduced in E. coli's genomes. While this information can be found in the paper mentioned as reference 24, this information is worth restating here to prevent confusion with barcoded gene disruption libraries (for instance RB-TNSeq libraries).

While the authors mention that the fraction of pairs exhibiting exclusion is consistent across communities independently of their richness, is there any pattern between coexistence scenarios highlighted in Figure 3 and community richness?

Eventually, if the authors want to measure/map the interactions within microbial communities using uCI, have the authors tested whether the effects of 3 species inoculated in 3 different variable wells connected to the same target strain are the same effect on the target strain than if 3 species were inoculated together in the same variable well. I believe that having the 3-species mixed together and associated interactions would lead to a different growth outcome for the target strain (and might be more representative of the original natural ecosystem). It is quite nice that the uCI platform offers the opportunity to explore such questions (whether community effects are the sum of more simple effects) and help decipher higher-order interactions.

Have the authors considered or tried using spent supernatant from these 4 strains (instead of directly inoculating the strains)? It could be interesting to compare the effects of spent supernatant versus the presence of actual species to see if they mostly compare to the '6 hour prior to the target strain' or if they are pretty different.

It is known that growth rate directly impacts gene expression, and that sometimes, gene expression changes between two conditions are simply connected to a decreased growth rate in the studied condition. I am wondering if there would be a way to sort out the genes that are associated with differential expression between genes whose expression change is likely related to growth rate effect versus genes whose expression changes are not (based on information that can be found in the litterature).

Since the authors are mentioning the species that degrade the chitin, reaction from which the rest of the trophic chain will cascade, shouldn't that be 'at the top' of the trophic hierarchy instead?

This work and these results are really interesting and exciting. This whole story and its conclusions could nicely be summarized into a diagram to make it even easier to navigate and understand.

Have the authors considered looking at metatranscriptome modifications and untargeted metabolites profiles? That could provide an even broader insight into the modifications following the suppression of 1 or 2 strains, that may or may not lead to such drastic changes in community composition but rather in species physiology.

Are these species part of the same niches? Are they part of upstream or downstream niches? Replacing them in a metabolic network could be interesting to highlight potential interactions.

What motivated the decision of FC > 100?

How do the authors controlled for the species inoculation level? Inoculation level of a given species can also impact its final relative abundance. While in the Methods it seems that the inoculations are all done from the same species pools, aliquots have been frozen and thawed before inoculation, while the metagenomic analysis of the pool has been performed before freezing. So I am wondering if this could have introduced some inoculation differences and how the authors distinguish between potential inoculation effect versus interaction effect when comparing the relative abundance of the same species in the hCom1a experiment and the double dropout for instance? Would a comparison of abundance fold change (between Cecum and Inoculum) between 2 conditions lead to the same conclusions? For instance: comparing the abundance fold change of Ruminococcus obeum between Cecum and inoculum of the hCom1a condition to Ro abundance fold change (Cecum versus inoculum) in the double dropout condition.

Where do CA and DCA come from in the gut or in the synthetic community?

If the species of the community can be organized into niches (within the niche network mentioned in the first comment), it could also be informative to represent them organized as such, to eventually see where the species whose abundance is impacted by the presence/absence of Cs and Ch are directly or indirectly metabolically connected to them

If the gut microbiome can be divided into niches (or a network of niches), do the authors know how many ‘niches’ can thus be defined (in the model community used in the paper), how they are connected and how would the bile acid niche fit into such network?