­

The paper presents a study of ungulate nematodes in an East African savannah ecosystem. The authors use faecal metabarcoding to identify host-parasite interactions and subsequently test existing hypotheses about divers of parasite richness, host specificity, and host centrality in parasite sharing networks. I found this to be a very interesting paper to read, and was impressed by the thorough execution of the analyses. In fact, many of my initial gut-reaction concerns were subsequently addressed by additional analyses, or clarifications. Overall I find this to be a very well conducted study, presented in a clear and compelling manuscript.

I do find some room for improvement in the outlining of hypotheses, and justification of methodological choices. In some cases, I suggest alternative analyses which will improve the paper. Nice work :)

Line 30: You say here “evolutionary distinctiveness”, but in the phylogenetic literature this refers to a particular metric which you did not calculate (see the function evol.distinct in picante). This is a bit pedantic, but it might be best to use a different term here.

Line 71: Host phylogeny does not necessarily act to only preserve deep co-evolutionary relationships – recent divergences may in fact be more likely to determine rates of parasite host switching, or co-inheritance as divergence from a common ancestors is much more recent.

Line 74: Why is studying these networks at the individual level difficult? Is this only due to limited sampling, or is there a conceptual / methodological reason for this difficulty?

• Nice simple hypothesis on line 77 – if identity of hosts is phylogenetically conserved, there may not be a positive relationship between parasite richness ad parasite PD.

Lines 78 – 80: I agree that host body size, geographic range size, and population density regularly come up as important predictors of parasite species richness. However, I’m not sure that the respective allegories to habitat size, transmission opportunities, and parasite density are perfectly aligned. Host body size is also a proxy for accumulation of parasites (via positive relation with longevity – see Cooper et al. 2012 DOI: 10.1371/journal.pone.0042190), and larger hosts are sometimes referred to as a more “stable” resource. As for geographic range size, I agree that this is a proxy for transmission opportunities, but I would clarify that these are inter-species transmission (i.e. larger range size means potentially greater amounts of overlap among species – see Farrell et al 2015 DOI: 10.1111/1365-2656.12342 for a discussion of this in the context of ungulate declines). Similarly, in the infectious disease modelling literature, population density is often a proxy for intra-species transmission potential (for density-dependent components of transmission). In this case, parasite density would be considered an outcome of this increased transmission.

Lines 85 – 86: I am surprised there are no previous studies of ungulates that investigated feeding strategy or gut morphology (didn’t the Walker 2017 paper investigate feeding type and digestive system?). If existing studies for ungulates are not copmarable, it would be good to acknowledge previous research that investigated these effect for other taxonomic groups (i.e. a quick search revealed some studies in primates, rodents, birds, and lizards)

Lines 108 & 112: Because this is likely to substantially impact your richness estimates, and in turn the structure of your networks, please briefly mention your process for mOTU delineation here, including and outline of sensitivity analyses altering the similarity threshold.

Line 121: Here you introduce social-group size as a predictor for the first time. It would be good to provide rationale for this, either here or above when discussing the potential effects of other life history predictors on parasite species richness.

Line 124-125: You predict that host-parasite cophylogeny drives the importance of host identify. This hypothesis should be expanded upon as the expectation from highly mirroried cophylogenies is that closely related hosts should harbour more similar parasite communities (and these parasites should be closely related than expected by chance). I can see how host identity would come out as important here, but I don’t think the hypothesis as you have it is linked specifically enough to the expectations from coupled host-parasite co-evolution.

Line 130: Again, I think you can be more specific about your hypothesis here. E.g. What do you expect the pattern to look like? How are host phylogenetic relationships influencing the host-host network?

Line 142: How many samples were analyzed in total?

Line 153: “Ct” should be defined, and a citation provided for choice of threshold for positive / negative sample delineation. As an aside, if Ct is negatively correlated with egg-counts (line 158), why do you need this first step? Is it only to save money on further amplification? Would these samples with Ct > 35 be considered false positives / contamination?

Line 167: Please provide a justification for rarifying your data. In the microbiome world, some consider rarification of metabardocing-based count data “inadmissible” (see McMurdie & Holmes 2014 DOI:10.1371/journal.pcbi.1003531), and advocate for a more model-based analyses of raw counts.

Line 168: Why did you remove this sample from the hybrid plains x Grevy’s zebra?

Line 170-171: I understand removing sheep, goat, and waterbuck if all samples were negative for helminths, however part of your rationale was to the importance of increasing cross-transmission with domestic species. If cattle parasite communities are altered by anthelminthics, I understand they may not be representative in analyses of more life history predictors. However, they may still be important in terms of network centrality for key parasites that are present despite this treatment.

Line 188: Do you mean binomial for prevalence rather than Bernoulli?

Line 193: Why did you not account for phylogeny in this main model? In my opinion, the base model should include a phylogenetically informed species-level effect.

Line 198-200: Should the local data not be used as primary, then supplemented with PanTHERIA when not available? Also, how do you measure if something “differed dramatically”?

Line 227: Just a suggestion, but there are alternative methods to calculate parasite host specificity, such as those that include measures of phylogenetic distance askin to your sesPD metrics for parasite diversity (e.g. Park et al. 2018, DOI: doi.org/10.1098/rspb.2017.2613). These measures of specificity would be better aligned to your hypotheses of cophylogeny than measures counting the numbers of hosts per parasite.

Line 242-243: I dont understand why you didn’t conduct a PGLS controlling for other variables when including phylogenetic distinctiveness. This could be included in the PGLS analyses in lines 239-242. A Spearman’s P does not take into account phylogenetic non-independence.

Lines 256-260: As you say that PD is often highly related to species richness, it might be easier to understand these results is presented in the context of observed parasite richness, and sesPD. Therefore you could say something like, “X had low parasite richness (Z), but these parasites were significantly more phylogenetically diverse than expected by chance (sesPD=Q).”

Lines 261 – 266: Very interesting that these patterns are lost when host phylogeny is included.

Lines 657 – 662 (Figure 1): I think the PD analyses should be removed as they are not independent of richness. Further – there should be clarification as to what the means and error bars represent (which model are these estimates from?)

Data & Code Accessibilty – Data and code archiving is not mentioned in the manuscript. While online I see there is a plan to submit to some things to Dryad it looks like only the data and code to conduct the ecological analyses will be available. I cannot easily find mention of where the raw sequence reads are archived, or how they are processed into mOTUs, but this should be required for full reproducibility, and inclusion of these sequences in future work.