1. Ecology
  2. Evolutionary Biology
Download icon

Effects of domestication on the gut microbiota parallel those of human industrialization

  1. Aspen T Reese  Is a corresponding author
  2. Katia S Chadaideh
  3. Caroline E Diggins
  4. Laura D Schell
  5. Mark Beckel
  6. Peggy Callahan
  7. Roberta Ryan
  8. Melissa Emery Thompson
  9. Rachel N Carmody  Is a corresponding author
  1. Department of Human Evolutionary Biology, Harvard University, United States
  2. Society of Fellows, Harvard University, United States
  3. Wildlife Science Center, United States
  4. Department of Anthropology, University of New Mexico, United States
Research Article
  • Cited 1
  • Views 1,102
  • Annotations
Cite this article as: eLife 2021;10:e60197 doi: 10.7554/eLife.60197

Abstract

Domesticated animals experienced profound changes in diet, environment, and social interactions that likely shaped their gut microbiota and were potentially analogous to ecological changes experienced by humans during industrialization. Comparing the gut microbiota of wild and domesticated mammals plus chimpanzees and humans, we found a strong signal of domestication in overall gut microbial community composition and similar changes in composition with domestication and industrialization. Reciprocal diet switches within mouse and canid dyads demonstrated the critical role of diet in shaping the domesticated gut microbiota. Notably, we succeeded in recovering wild-like microbiota in domesticated mice through experimental colonization. Although fundamentally different processes, we conclude that domestication and industrialization have impacted the gut microbiota in related ways, likely through shared ecological change. Our findings highlight the utility, and limitations, of domesticated animal models for human research and the importance of studying wild animals and non-industrialized humans for interrogating signals of host–microbial coevolution.

eLife digest

Living inside our gastrointestinal tracts is a large and diverse community of bacteria called the gut microbiota that plays an active role in basic body processes like metabolism and immunity. Much of our current understanding of the gut microbiota has come from laboratory animals like mice, which have very different gut bacteria to mice living in the wild. However, it was unclear whether this difference in microbes was due to domestication, and if it could also be seen in other domesticated-wild pairs, like pigs and wild boars or dogs and wolves.

A few existing studies have compared the gut bacteria of two species in a domesticated-wild pair. But, studies of isolated pairs cannot distinguish which factors are responsible for altering the microbiota of domesticated animals. To overcome this barrier, Reese et al. sequenced microbial DNA taken from fecal samples of 18 species of wild and related domesticated mammals.

The results showed that while domesticated animals have different sets of bacteria in their guts, leaving the wild has changed the gut microbiota of these diverse animals in similar ways. To explore what causes these shared patterns, Reese et al. swapped the diets of two domesticated-wild pairs: laboratory and wild mice, and dogs and wolves. They found this change in diet shifted the gut bacteria of the domesticated species to be more similar to that of their wild counterparts, and vice versa. This suggests that altered eating habits helped drive the changes domestication has had on the gut microbiota.

To find out whether these differences also occur in humans, Reese et al. compared the gut microbes of chimpanzees with the microbiota of people living in different environments. The gut microbial communities of individuals from industrialized populations had more in common with those of domesticated animals than did the microbes found in chimpanzees or humans from non-industrialized populations. This suggests that industrialization and domestication have had similar effects on the gut microbiota, likely due to similar kinds of environmental change.

Domesticated animals are critical for the economy and health, and understanding the central role gut microbes play in their biology could help improve their well-being. Given the parallels between domestication and industrialization, knowledge gained from animal pairs could also shed light on the human gut microbiota. In the future, these insights could help identify new ways to alter the gut microbiota to improve animal or human health.

Introduction

Industrialized, agrarian, and foraging human populations differ along numerous ecological dimensions, including diet, physical activity, the size and nature of social networks, pathogen exposure, types and intensities of medical intervention, and reproductive patterns. Such changes have resulted in large shifts in the gut microbiota in industrialized populations relative to non-industrialized populations or closely related primates (De Filippo et al., 2010; Moeller et al., 2014; Moeller, 2017; Smits et al., 2017), including reductions in alpha-diversity and changes in composition that have been implicated in the rise of various metabolic and immunological diseases (Ley et al., 2006; Cox et al., 2014; Kamada et al., 2013). Many aspects of ecology that now differ between industrialized and non-industrialized human populations were similarly altered during the process of animal domestication (Zeder, 2012). For example, domestic animals often consume less diverse, more easily digestible diets than their wild relatives, expend less energy to achieve adequate (or excess) caloric intake, live in comparatively static and high-density groups, and can be subject to modern medical interventions including antibiotic treatment (McClure, 2013). Although industrialization and domestication are fundamentally different processes, the ecological parallels between human industrialization and animal domestication suggest that the gut microbiota of diverse domesticated animals may differ in consistent ways from those of their wild progenitors, and further, that their differences may resemble those observed between industrialized and non-industrialized human populations.

Many of the altered ecological features experienced by industrialized humans and domesticated animals have been independently observed to impact the gut microbiota, including diet (David et al., 2014; Carmody et al., 2015), physical activity (Allen et al., 2018; Lamoureux et al., 2017), the size and nature of social networks (Dill-McFarland et al., 2019; Antwis et al., 2018), antibiotic use (Bokulich et al., 2016; Cho et al., 2012), and changes in birthing and lactation practices (Bokulich et al., 2016; Li et al., 2018). The effects of these features on gut microbiota composition are often found to match or exceed the effects of genetic variation (Carmody et al., 2015Rothschild et al., 2018), which is also routinely modified by domestication. As such, ecological shifts under domestication might be expected to lead to gut microbial differentiation between domesticated animals and their wild counterparts. To this end, wild mice have been shown to differ from laboratory mice in gut microbial composition (Kreisinger et al., 2014; Rosshart et al., 2017). Similarly, a comparison of domesticated horses and wild Przewalski’s horses in adjacent Mongolian grasslands found that the wild animals harbored compositionally distinct, and overall more diverse, gut microbial communities (Metcalf et al., 2017). However, to date, no general survey has been conducted to characterize the global effects of domestication on the gut microbiota.

Apart from the pressures of ecological change that domestic animals experience in human environments, animal domestication has also entailed strong artificial selection for phenotypes desirable to humans, such as rapid growth and docility in agricultural animals, reliable reproduction and stress resistance in laboratory animals, and unique physical and/or behavioral attributes in companion animals. Although targeted phenotypes differ based on the species under domestication, all domesticated mammals share the legacy of having been intentionally or indirectly selected for tameness (Wilkins et al., 2014). This selection has been argued to have resulted in convergent morphological and physiological features across domesticated mammals that are collectively referred to as ‘domestication syndrome’ – including, for instance, reductions in brain size and tooth size, depigmentation, altered production of hormones and neurotransmitters, and retention of juvenile behaviors into adulthood – with the pleiotropic nature of these effects thought to be mediated by changes in neural crest cells (Wilkins et al., 2014). Therefore, to the extent that gut microbiota is dependent on host biology, we might additionally expect domestication to have shaped the gut microbiome in similar, potentially convergent, ways across diverse mammalian lineages. Such microbiota-structuring contributions ascribable to evolutionary rather than ecological forces have the potential to be much greater in and/or unique to domesticated animals relative to industrialized human populations since the process of domestication has been advancing for much longer than industrialization.

Here, we assess the effects of domestication on the mammalian gut microbiota, perform controlled dietary experiments that attempt to distinguish between the relative roles of ecology and genetics in driving these patterns, and compare the effects of domestication to those of human industrialization. While we focus primarily on the impacts of domestication on the mammalian gut microbiota, we include analyses of industrialized and non-industrialized human populations because much is known about the effects of industrialization on the gut microbiota and as such it can serve as a benchmark ecological process for domestication. In addition, to explore the extent to which deeper evolutionary history affects these patterns, we also compare humans to chimpanzees (Pan troglodytes), one of our two closest living relatives and arguably the better referential model for the last common ancestor between Pan and Homo (Muller et al., 2017). Early Homo sapiens is thought to have undergone a form of self-domestication as a result of selection against aggression (Wrangham, 2018; Theofanopoulou et al., 2017), suggesting that there could likewise be parallels between the gut microbial signatures of animal domestication and Pan–Homo speciation.

We predict that (i) gut microbial communities will differ between domesticated animals and their wild counterparts, (ii) gut microbial communities of diverse domesticated animals may exhibit convergent characteristics in a microbial counterpart to the physiological domestication syndrome (Wilkins et al., 2014), and (iii) gut microbial changes observed with domestication may parallel contrasts observed between chimpanzees and humans. In addition, to the extent that domestication effects are driven by ecology rather than host phylogenetic distance, we should expect (iv) experimental manipulation of ecology to overcome differences in the gut microbiota between closely related hosts, and (v) the gut microbiota of domesticated animals will share more features with industrialized human populations than with non-industrialized human populations.

Identifying the factors shaping the gut microbiota of domesticated animals will provide insights into the ecology of host-associated microbial communities and their impact on health. Domesticated animals serve as reservoirs for zoonotic diseases (Morand et al., 2014; Wolfe et al., 2007; Cleaveland et al., 2001; Han et al., 2016) and carriers of antibiotic-resistant bacteria (Sayah et al., 2005; EFFORT Group et al., 2018). Furthermore, the ecological impacts of domestication on the gut microbiota could conceivably contribute to the unique health problems experienced by captive (Hosey et al., 2009) and domesticated animals (Timoney et al., 1988). Differences between domesticated and wild animal microbiota may also manifest in poor translatability between laboratory studies and the real world (Leung et al., 2018; Beura et al., 2016). Finally, the convergent nature of many ecological shifts experienced by domesticated animals and industrialized human populations suggests that domesticated animals may provide a uniquely useful model for studying the microbially mediated health impacts of rapid environmental change that results in mismatch between host, microbiota, and/or environment, a situation thought to apply to humans in industrialized settings (Sonnenburg and Sonnenburg, 2019). Understanding what shapes the domesticated microbiota may therefore identify routes to improve experimental models, animal condition, and human health.

Results

Cross-species comparison of gut microbial composition

First, we characterized the fecal microbiota of wild and domesticated populations of nine pairs of artiodactyl, carnivore, lagomorph, and rodent species (Figure 1A) using 16S rRNA gene amplicon sequencing and quantitative PCR (qPCR). Despite observing no single convergent ‘domesticated microbiota’ profile, our analysis detected a global signal of domestication status on gut microbiota composition. Across the combined dataset, the factor that explained the largest proportion of variation was the host dyad (e.g., pig/boar; p<0.001, R2 = 0.39, F = 17.086, permutational multivariate analysis of variance [PERMANOVA]; Figure 1B). However, correcting for host dyad, domestication status also contributed significantly to variation in microbial communities (p<0.001, R2 = 0.15, F = 6.081), and these results were robust to the distance metric analyzed (Supplementary file 1). Furthermore, analyses of individual dyads found a significant effect of domestication status for all groups except canids (p<0.05, R2 = 0.18–0.41, PERMANOVAs; Supplementary file 1). Diet and digestive physiology were also primary determinants of the gut microbiota (diet: p<0.001, R2 = 0.12, F = 21.216; physiology: p<0.001, R2 = 0.14, F = 23.938; Figure 1—figure supplement 1), as seen in other surveys of mammals (Muegge et al., 2011), with effect sizes comparable to that of domestication status. Consistent with the idea that higher ecological homogeneity in domesticates may beget greater gut microbial homogeneity, we found that there was greater between-conspecific variability in wild gut communities than in domesticated gut communities (p=0.002, F = 8.838; permutation test for F).

Figure 1 with 4 supplements see all
The mammalian gut microbiota carries a global signature of domestication.

(A) Sampling scheme for cross-species study. (B) Nonmetric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarities illustrates a significant signal (p<0.001, R2 = 0.15, F = 6.081, permutational MANOVA) of domestication (closed versus open circles; N = 82 domesticated and 99 wild) and clustering by host dyad (color; N = 5–20 individuals per species). (C) Distance to dyad (color) mean along Bray–Curtis ordination NMDS axis 1 differs by domestication status (p=0.006, Mann–Whitney U test). (D) Bray–Curtis dissimilarity between individuals is lowest among conspecifics, but wild–domesticated pairs also have lower dissimilarity than unrelated pairs (p<0.001, bootstrapped Mann-Whitney U tests). * indicates p<0.05. Large circles are means; bars show standard deviations.

To determine whether there was a consistent change in microbial composition with domestication, we calculated the difference between an individual’s ordination coordinates and the average ordination coordinates of its host dyad along the first nonmetric multidimensional scaling (NMDS) axis. Quantifying this ordination shift allowed us to consider overall changes in composition while correcting for host dyad and retaining information on the directionality of changes. We found that domesticated individuals were typically further right relative to the average of their host dyad (p=0.006, Mann–Whitney U test; Figure 1C) and that this difference was significant regardless of the distance metric analyzed (Supplementary file 1). Most domesticated species displayed similar trends in these ordination shifts (Figure 1—figure supplement 2) with laboratory and companion animals showing significant differences when analyzed collectively (p<0.05, Mann–Whitney U tests; Figure 1—figure supplement 2).

Free-ranging wild animal populations representing the progenitor species were not sampled for all pairs, potentially limiting the scope of our analysis. To assess whether the patterns described held for more stringent groupings, we analyzed the subset of wild–domesticated dyads for which the wild member was from a free-ranging population (i.e., ‘truly wild’) as well as the subset for which the wild member was the known progenitor (i.e., ‘perfect pair’) (Supplementary file 2). In both cases, we still found that domestication status explained a meaningful portion of variation in gut microbial community composition, regardless of the distance metric used (all p<0.001, all R2 > 0.11, PERMANOVA; Supplementary file 1).

Domestication of mammalian species occurred at different times, so the evolutionary relationships between members of a host dyad are not all equal, even in cases where we sampled the known progenitors. Supporting an underlying influence of host evolutionary history on the gut microbiota, we found that host species that were more closely related (i.e., had a shorter time since divergence) had more similar microbial community compositions (p<0.001, r = 0.157, Mantel test). Similarly, the magnitude of the ordination shifts along NMDS axis 1 were smaller for animals from host dyads that were more closely related (p=0.012, rho = 0.19, Spearman correlation; Figure 1—figure supplement 3). Nevertheless, supporting the idea that ecology plays a dominant role in shaping the gut microbiota, average dissimilarity between members of a dyad was not lower in species pairs with more recent dates of domestication (p=0.854) or more recent divergence times (p=0.380; Figure 1—figure supplement 3). Moreover, differences in the ordination shifts along NMDS axis 1 associated with domestication remained significant even when correcting for host dyad and divergence time (p<0.001, likelihood test linear mixed effects models). Overall, dissimilarity between conspecifics was lowest, but dissimilarity between wild–domesticated dyads was significantly lower than for unrelated pairs (p<0.001, bootstrapped Kruskal–Wallis tests; Figure 1D).

We also tested for differences in specific features of the gut microbiota between domesticated and wild mammals. Domestication status did not affect microbial density quantified as copies of the 16S rRNA gene per gram of feces (p=0.089, Mann–Whitney U test), Shannon index (p=0.2017), or operational taxonomic unit (OTU) richness (p=0.3506; Figure 1—figure supplement 4), indicating that the domestication signal overall was not primarily driven by microbial species loss. Consistent with experiencing heightened environmental exposure, wild animals generally harbored potential pathogen communities that were more diverse (p=0.001, Mann–Whitney U test) and marginally more abundant (p=0.092; Figure 1—figure supplement 4). Among laboratory animals specifically, potential pathogen abundance (p<0.001) and pathogen richness (p<0.001) were substantially lower than among wild relatives, while total microbial density was higher (p=0.006; Figure 1—figure supplement 2). Companion animals did not differ significantly by domestication status for microbial density, diversity, or pathogen metrics. By contrast, agricultural animals had higher Shannon index and richness values (p≤0.001, Mann–Whitney U tests) as well as marginally higher pathogen abundances (p=0.067; Figure 1—figure supplement 2) compared with their wild counterparts.

Diet versus host taxon effects on domesticated gut microbial composition in mice

Domestication has had profound effects on both ecology and host genetics. To begin to tease apart the relative roles of ecological change and genetic change in shaping the gut microbiota in domesticates, we performed a series of reciprocal diet experiments that tested the extent to which gut microbial signatures of wild–domesticated dyads could be recapitulated and reversed solely by the administration of ecologically relevant diets. We first conducted a fully factorial experiment in which wild-caught and laboratory mice (Mus musculus) were maintained for 28 days on wild or domesticate diets (Figure 2A, Supplementary file 3). Overall, we found that host taxon explained the largest amount of variation in composition (p<0.001, R2 = 0.173, F = 64.255, PERMANOVA), but that diet (p<0.001, R2 = 0.042, F = 15.427) and a host taxon by diet interaction term (p<0.001, R2 = 0.020, F = 7.557) were also significant (Figure 2B, Figure 2—figure supplement 1). Ordination shifts describing changes in the gut microbial community over the course of the experiment depended on the experimental group (axis 1: p<0.001, linear mixed effects model likelihood test; Figure 2D). Confirming prior reports that diet plays a dominant role in shaping the murine gut microbiota (Carmody et al., 2015), the gut microbiota of wild mice fed a domesticate diet (WildH/DomD) moved toward the average microbial community of domesticated mice fed a domesticate diet (DomH/DomD), the microbiota of domesticated mice fed a wild diet (DomH/WildD) moved in the opposite direction, and those of control wild or domesticated mice consuming their habitual diets (WildH/WildD and DomH/DomD) did not shift (Figure 2B). Over the course of the experiment, alpha-diversity as measured by Shannon index also changed significantly across treatment groups (p=0.025, Kruskal–Wallis test; Figure 2—figure supplement 2), with DomH/WildD mice becoming significantly more diverse (p=0.004, one-sample Wilcoxon test) despite lower baseline levels of alpha-diversity in domesticated versus wild mice (p=0.011, Mann–Whitney U test; Figure 2C).

Figure 2 with 2 supplements see all
Gut microbial differences between wild and domesticated mice can be partially overcome by diet swap.

(A) Design scheme for fully factorial host taxon by diet mouse experiment (N = 10 laboratory mice or three wild mice per diet group). (B) Nonmetric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarities showing changes for mice from day 0 (open circle) to day 28 (filled circle) by experimental groups (color). Composition varied by host taxon (p<0.001, R2 = 0.173, F = 64.255, permutational MANOVA), diet (p<0.001, R2 = 0.042, F = 15.427), and a host taxon by diet interaction (p<0.001, R2 = 0.020, F = 7.557). (C) Shannon index differed between host taxa on day 0 (p=0.011, Mann–Whitney U test). (D) Animals on reciprocal diets (DomH/WildD and WildH/DomD) but not control diets tended to move in opposite directions along Bray–Curtis ordination NMDS axis 1 from day 0 to day 28 (p=0.048 and p=0.25, respectively, one-sample Wilcoxon test). (E) At the end of the experiment, distance to the mean of the diet control at baseline (DomH/DomD and WildH/WildD) was lower for wild mice than for laboratory mice (p=0.048, Mann–Whitney U test). * indicates p<0.05, Mann–Whitney U test. Large circles are means; bars show standard deviations.

Neither host taxon nor diet was associated with differences in gut microbial density over the experiment (p=0.272, Kruskal–Wallis test; Figure 2—figure supplement 2), but it is notable that the total amount of feces produced, and thus likely the total number of bacteria, was lower in both host taxon groups when consuming the wild diet (p<0.001, Kruskal–Wallis test; Figure 2—figure supplement 2). Despite similar trends in fecal production between wild and domesticated mice in response to diet treatment, wild and domesticated mice differed markedly in their ability to harvest energy from experimental diets (p<0.001, Kruskal–Wallis test; Figure 2—figure supplement 2), as indexed by bomb calorimetry of feces. While wild mice were equally efficient digesters of the wild and domesticated diets, laboratory mice captured 15% fewer calories when consuming the wild versus domesticated diet.

Interestingly, asymmetries were also observed between wild and domesticated mice in their gut microbial responses to reciprocal diets. Whereas the microbial communities of WildH/DomD mice grew to resemble those of untreated DomH/DomD mice, the microbial communities of DomH/WildD mice remained distinct from untreated WildH/WildD mice throughout the experiment (p=0.042, Mann–Whitney U test; Figure 2B, E). It is possible that the asymmetry in energy harvest between wild and domesticated mice was rooted in differential microbial responses to reciprocal diets and the inability of DomH/WildD mice to harbor a wild-type microbiota.

Loss of wild gut microbiota in domesticated mice

Based on the lower alpha-diversity in domesticated versus wild mice (Figure 2C), we hypothesized that the asymmetries between domesticated and wild mouse responses to altered diets were due to past extinction of relevant strains from laboratory microbial communities and no dispersal source of replacement strains (Sonnenburg et al., 2016). Therefore, we next tested whether experimental dispersal from a wild microbial community in conjunction with feeding a wild diet could support a fully wild microbial community in laboratory mice (Figure 3A). A single colonization treatment with a wild mouse cecal community (via gavage) led to significant shifts in the gut microbial community (Figure 3B, Figure 3—figure supplement 1), resulting in closer resemblance to the wild donor (p<0.001, Mann–Whitney U test; Figure 3C). Shifts in NMDS axis 1 varied across experimental treatment groups (p<0.001, linear mixed effects model likelihood test; Figure 3D). While laboratory mice fed a wild diet but given a control gavage (PBS) also moved toward the donor along NMDS axis 1 (p=0.002, one-sample Wilcoxon test; Figure 3D), reflecting the influence of diet, we observed a substantially greater shift following experimental colonization (p<0.001, Kruskal–Wallis test). Surprisingly, among colonized mice, movement of the microbial community toward the wild donor profile was profound even without reinforcement from the wild diet (p=0.182, Mann–Whitney U test).

Figure 3 with 1 supplement see all
Laboratory mice can be re-wilded through colonization with a wild gut microbial community.

(A) Design scheme for colonization/diet mouse experiment (N = 9–10 mice per treatment group). (B) Nonmetric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarities showing changes for mice from day 0 (open circles) to day 8 (filled circles) by experimental groups (color). (C, D) At the end of the experiment (filled circle), distance to the wild community donor decreased most in animals colonized with wild communities (p=0.004 WildC/DomD and p=0.002 WildC/WildD, Mann–Whitney U test; C), but all experimental groups exhibited change along Bray–Curtis ordination NMDS axis 1 (p=0.002 PBSC/WildD, p=0.004 WildC/DomD, and p=0.002 WildC/WildD, one-sample Wilcoxon tests; D) during the course of the experiment. * in (C) indicates p<0.05, Mann–Whitney U test comparing day 0 to day 8 for each experimental group. Large circles are means; bars in (C) show standard deviations.

Although all mice exhibited an increase in microbial density over the course of the experiment (p<0.01, one-sample Wilcoxon tests), colonization with a wild community did not lead to higher microbial density overall (p=0.449, Kruskal–Wallis test; Figure 3—figure supplement 1) nor to an increase in alpha-diversity relative to baseline (p=0.258, one-sample Wilcoxon test). As in the original reciprocal diet experiment, wild diet treatment led to lesser fecal production (p<0.001, Kruskal–Wallis test; Figure 3—figure supplement 1). No differences in fecal production were observed between mice colonized with a wild community and PBS-treated controls (p=0.79; Figure 3—figure supplement 1), suggesting that lower fecal output on the wild diet was not a direct consequence of harboring a wild microbiota. Together, these results suggest that differences observed with experimental colonization reflected shifts in gut microbial community structure rather than simple augmentation of microbial load.

Diet versus host taxon effects on domesticated gut microbial composition in canids

To test if our findings were generalizable beyond mice, we conducted an analogous reciprocal diet experiment in captive sympatric populations of wolves and dogs (Figure 4A). We tracked gut microbial dynamics in these canids for 1 week on their standard diet (raw carcasses or commercial dog food, respectively) and 1 week on the reciprocal diet. As in the mouse experiment, we found that host taxon (wild or domesticated) explained the largest amount of variation in gut microbiota composition (p<0.001, R2 = 0.098, F = 13.730, PERMANOVA), but that diet (p<0.001, R2 = 0.058, F = 8.151) and a host taxon by diet interaction term (p<0.001, R2 = 0.028, F = 3.934) were also significant (Figure 4B, Figure 4—figure supplement 1). There were significant differences among experimental groups in the magnitude of their ordination shifts along the first NMDS axis over the experimental periods (p<0.001, linear mixed effects model likelihood test; Figure 4D). As in the mouse experiments, we observed that animals on reciprocal diet treatments (DomH/WildD; WildH/DomD) moved significantly toward the habitual gut microbial profile of the other species (p<0.05, one-sample Wilcoxon tests; Figure 4D), while the microbiota of animals consuming their habitual diet (DomH/DomD; WildH/WildD) did not shift predictably (p>0.100).

Figure 4 with 2 supplements see all
Microbial differences between wild and domesticated canids can be partially overcome by diet shifts.

(A) Design scheme for fully factorial host taxon by diet canid experiment (N = 9 dogs or N = 10 wolves per diet group). (B) Nonmetric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarities showing changes for canids from day 0 (open circle) to day 7 (filled circle) by experimental groups (color). Composition varied by host taxon (p<0.001, R2 = 0.098, F = 13.70, permutational MANOVA), diet (p<0.001, R2 = 0.058, F = 8.15), and a host taxon by diet interaction (p<0.001, R2 = 0.028, F = 3.93). (C) Shannon index differed between dogs and wolves on day 0 (p=0.003, Mann–Whitney U test). (D) Canids on reciprocal diets (DomH/WildD and WildH/DomD) but not control diets moved in opposite directions along Bray–Curtis ordination NMDS axis 1 over time (p=0.004 and 0.002, respectively, one-sample Wilcoxon tests). (E) At the end of the experiment, distance to the mean of diet controls at baseline (DomH/DomD and WildH/WildD) was lower for dogs than for wolves on reciprocal diets (p=0.001, Mann–Whitney U test). * indicates p<0.05, Mann–Whitney U test. Large circles are means; bars show standard deviations.

In addition, we again observed an asymmetry between domesticated and wild animals in the degree to which the gut microbiota responded to diet. On experimental diets, dogs and wolves differed significantly in their dissimilarity to diet controls (p<0.001, Kruskal–Wallis test; Figure 4E), with the gut microbial communities of dogs fed raw carcasses resembling those of wolves at baseline but the gut microbial communities of wolves fed dog food remaining distinct from those of dogs at baseline (p=0.001, Mann–Whitney U test).

The difference in the direction of asymmetry between the canid and mouse experiments may be explained by the different trends in dietary ecology between carnivores and omnivores during domestication. Carnivores, through the addition of extensive carbohydrates to their diet (Wolfe et al., 2007), likely encounter more diverse diets in captivity than in the wild, whereas captive herbivores and omnivores typically eat a lesser number of plant species or are maintained on a single feed mix. Supporting this, we found that dogs initially had significantly higher OTU richness (p<0.001, Figure 4—figure supplement 2) and Shannon index (p=0.003, Figure 4C) than wolves, but that reciprocal diets led to a switch in diversity (richness: p=0.002, Mann–Whitney U tests), with wolves becoming more diverse when fed dog food while dogs lost diversity when fed raw carcasses (Figure 4—figure supplement 2).

Analogous pressures in the human gut microbiota

We next explored the extent to which humans harbor gut microbial signatures analogous to those of domestication. Given evidence that the gut microbiota of domesticated animals is shaped by both ecology and speciation, we began by comparing humans to chimpanzees, one of our two closest living relatives. Humans may have undergone a form of self-domestication as a result of selection against aggression (Wrangham, 2018; Theofanopoulou et al., 2017) in addition to significant ecological change since our divergence from Pan, suggesting that the gut microbial signatures of animal domestication and Pan–Homo speciation could share features in common. We first compared samples that we collected from industrialized humans and wild chimpanzees, finding that the gut microbial communities of these humans and chimpanzees exhibited differences that paralleled those observed between domesticated animals and their wild counterparts when compared in the same ordination space (p<0.001, Mann–Whitney U test; Figure 5A, B). Microbial density (p=0.002, Mann–Whitney U test) and Shannon index (p=0.018; Figure 1—figure supplement 4) also differed between industrialized humans and wild chimpanzees, confirming prior reports that industrialized humans harbor microbial communities with substantially lower alpha-diversity (Smits et al., 2017). We found only a marginal difference in between-conspecific variability in the gut microbiota of industrialized humans and wild chimpanzees (p=0.092, F = 3.0987; permutation test for F). Including these human–chimpanzee comparisons in our analysis of the relationship between gut microbiota dissimilarity and the time since dyad divergence strengthened the observed relationship (p<0.001, r = 0.5251; Mantel test), with a conservative divergence time of 6.5 million years assumed for Pan–Homo in this analysis.

Figure 5 with 2 supplements see all
Differences in gut microbial communities between industrialized humans and wild chimpanzees parallel those observed between domesticated and wild mammals.

(A) Nonmetric multidimensional scaling (NMDS) ordination of Bray–Curtis dissimilarities in the gut microbiota illustrates that industrialized human populations (US and US Jha) exhibit similar trends relative to wild chimpanzees as domesticated animals do to wild animals, but that non-industrialized human populations (Hadza, Chepang, Raji, Raute, and Tharu) do not (N = 5–7 individuals per primate population and 5–20 individuals per other animal species). (B) Distance along the first Bray–Curtis ordination NMDS axis relative to group mean differs in the same direction for the two industrialized human populations relative to wild chimpanzees or non-industrialized human populations as for domesticated animals relative to wild animals (p<0.05, Mann–Whitney U tests, N = 7–99). (C) The gut microbial communities of wild animals are more dissimilar to those of industrialized humans than are those of domesticated animals (p<0.001, bootstrapped Mann–Whitney U test, N = 82 domesticated and 99 wild). * indicates p<0.05, Mann–Whitney U test. Large shapes are means; bars in (C) show standard deviations.

However, given the vast ecological differences between wild chimpanzees and industrialized humans, it remained unclear the extent to which these Pan–Homo differences reflected host phylogenetic distance as opposed to ecology. To better gauge the divergence attributable to phylogenetic distance versus ecology, we proceeded to compare the gut microbial communities of humans living in industrialized versus non-industrialized subsistence or agricultural societies, who are all equidistantly related to chimpanzees. Reanalysis of our cross-species comparison including published data on human populations in rural Nepal and Tanzania pursuing various non-industrialized lifestyles Jha et al., 2018 found that the gut microbial communities of these non-industrialized populations differed substantially from those of two independent U.S. samples, instead clustering more closely to those of chimpanzees in this ordination space (Figure 5A). Only the gut microbial communities of our industrialized populations show the rightward ordination shift along the BrayCurtis NMDS axis 1 that is also exhibited by the gut microbial communities of domesticated animals (p<0.05, MannWhitney U tests; Figure 5B). Moreover, bacterial taxa previously found to distinguish among human lifestyles (Smits et al., 2017) typically had relative abundances that varied in the same direction between wild and domesticated animals as among wild chimpanzees, non-industrialized human populations, and industrialized human populations (Figure 5—figure supplement 1). These trends were clearest in the bacterial family Bacteroidaceae, which exhibited a continuous increase from chimpanzees to non-industrialized populations to industrialized populations as well as an increase in domesticated relative to wild animals (p=0.008, MannWhitney U test).

Together, these data indicate that the human gut microbiota does not carry a global signal of domestication, as would be predicted under a hypothesis of human gut microbial self-domestication. Rather, the corresponding gut microbial responses to domestication and industrialization suggest that these responses are more likely driven by common ecological factors, a conclusion further supported by the observation that the gut microbial communities of domesticated animals were more similar to those of industrialized humans than were those of their wild animal counterparts (p<0.001, bootstrapped MannWhitney U test; Figure 5C). Notably, the gut microbial communities of domesticated animals and industrialized humans most closely resembled one another for companion animals (p<0.001, KruskalWallis test; Figure 1—figure supplement 2), presumably reflecting their greater ecological convergence and degree of physical contact (Song et al., 2013). Also supporting the role of ecology driving these trends, we found that an independent sample of captive chimpanzees did not cluster exclusively with the wild chimpanzee samples; indeed, their gut microbial communities were more similar to those of non-industrialized humans than to those of wild chimpanzees (p<0.001, Mann–Whitney U tests; Figure 5—figure supplement 2).

Discussion

Our data demonstrate that while domestication has not led to a convergent ‘domesticated microbiota,’ there is nevertheless a significant signal of domestication on gut community composition across diverse mammalian hosts. Furthermore, our experimental and cross-sectional analyses suggest that the domestication effect can, in large part, be ascribed to environmental rather than genetic changes. As in many comparative microbiota studies, host taxonomy was responsible for the largest component of variation in our cross-species analyses, but the contribution of domestication status was comparable to those of diet type and host physiology, factors previously identified as key drivers of the mammalian gut microbiota (Ley et al., 2008; McKenzie et al., 2017). Experimental diet intervention in wild/domesticated pairs reduced gut microbial dissimilarity and alpha-diversity differences between members of the dyad over short time scales. However, differences due to loss or gain of taxa during domestication could not be overcome by diet shifts alone, necessitating experimental recolonization. Together, these results indicate that domestication has played a large role in shaping the microbiota and, through husbandry practices, likely continues to do so today, suggesting that studying variations in animal husbandry practices may illuminate new levers for manipulating the mammalian gut microbiota (Velazquez et al., 2019; Villarino et al., 2016; Schmidt et al., 2019).

Although there are many shared features of contemporary ecology and historic artificial selection on domestic animals, it is perhaps unsurprising that domestication has not produced a single, convergent domesticated gut microbiota. The animals we characterized represent a diverse set of lineages in the mammalian clade, and thus their microbiota also differ due to variation in factors such as gut structure and size, passage rate, diet, and biogeography (Ley et al., 2008; Youngblut et al., 2019). While the domestication signal was comparable in magnitude to those of gut physiology and diet type, it does not mask those fundamental structuring forces. Furthermore, the particulars of a domesticated lineage can help clarify what aspects of ecology are most salient to the domestication effect. Cases where domestication effects are weaker in our comparative study generally consist of animals where the ecological change associated with domestication has been small – for example, sheep and pigs, whose diets may be quite similar to their wild progenitors, at least when kept in the non-industrialized agricultural settings that were sampled (McClure, 2013) – or where ecological changes are in the opposite direction from the domesticated norm – for example, canids, where the domesticate diet typically involves lower protein and higher carbohydrate levels than wild diets, instead of the higher protein levels seen in most laboratory or farm animals (Axelsson et al., 2013).

Our reciprocal diet experiments in mice and canids substantiate our claim that ecology plays a predominant role in shaping the domesticated gut microbiota. However, they do not pinpoint the mechanism(s) for these effects. Variability in diet or other aspects of ecology and their concomitant effects on host physiology (e.g., passage rate) can alter microbial composition or abundance through changes in the selective landscape that microbes experience and changes in environmental exposure (David et al., 2014; Carmody et al., 2019). Animals may experience altered bacterial colonization, leading directly to changes in composition, and/or viral colonization, which could then alter the bacterial community if new bacteriophages target gut bacteria or if eukaryotic viruses activate the host immune system, leading to transformations in the gut environment. That the gut microbial impacts of change in a single ecological variable like diet were sufficiently profound to balance those of host taxon identity suggests that suites of ecological variables changing together, such as during domestication or industrialization, may have collectively exerted an even larger influence (Jha et al., 2018).

Of course, microbiota changes were not the only pathway for animals undergoing domestication to respond to changing ecological factors. For example, genetic changes have enhanced the capacity for starch digestion in dogs (Axelsson et al., 2013; Reiter et al., 2016). Nevertheless, the increased microbial diversity and shifts in microbial composition that we observed in dogs may have additionally contributed to carbohydrate digestion. Indeed, dogs fed conventional diets have greater representation of carbohydrate metabolism genes in their gut metagenomes than do dogs fed meat-based diets (Alessandri et al., 2019). Notably, the microbiota has been found to supplement evolutionary responses during dietary niche expansion in wild animals that consume plants high in toxins (Kohl et al., 2014). As such, although hosts and their various gut microbial taxa are each expected to pursue their own fitness interests, gut microbial disparities observed between domesticated and wild animals, and more generally in other organisms under rapid environmental change, could potentially be adaptive for the host (Alberdi et al., 2016).

Beyond host support of a gut microbiota that can better digest a domesticate diet, humans may have selected for animals harboring a microbiota that helped them grow and reproduce well on such diets, thereby applying unconscious selection on the microbiota (Zohary et al., 1998). Changes in microbial function that enhanced host dietary energy harvest, survivorship, or reproduction may have been particularly important early in domestication, before host evolution occurred, although that hypothesis remains to be tested empirically. Regardless, specialization of microbial performance on domesticate diets could conceivably have come at the cost of broader digestive capacity, as seen in the laboratory mouse microbiota, which was better at harvesting energy from domesticate diets than from wild diets (Figure 2—figure supplement 2). It may also have impacted the immunological functions of the gut microbiota. The elevated pathogen abundances found in wild populations overall may largely be ascribed to low pathogen abundances in laboratory animals (Figure 1—figure supplement 2), which are maintained under specific pathogen-free conditions that minimize the likelihood of infection. Under natural conditions, though, the domesticated microbiota may exhibit reduced colonization resistance or immune system functioning (Rosshart et al., 2017; Beura et al., 2016; Rosshart et al., 2019), resulting in higher pathogen colonization, as observed here in agricultural animals. Future studies examining the trade-offs among microbially mediated functions, like digestive capacity, reproduction, and immunity, will help to illuminate the complex selection pressures shaping domesticated animals and their gut microbiota (Reese and Kearney, 2019).

We observed some correspondence between the gut microbial signatures of animal domestication and human industrialization that is most likely attributable to convergent ecological changes. The observation that gut microbial divergence among Pan and Homo primarily affects industrialized populations specifically implicates recent ecological changes as opposed to either ecological changes with deeper roots in human evolution or host evolutionary changes. Many recent human ecological changes involve accelerations of basic patterns established during the evolution of Homo, including increased proportion of calories from fat and protein, increased dependence on animal source foods, and extensive food processing involving both chemical and physical changes to food (Carmody, 2017). Other ecological changes are likely specific to industrialization, including reduced physical activity, high population density, and antibiotic use. These factors would be absent even in populations currently transitioning from subsistence to industrialized lifestyles (Jha et al., 2018), but may overlap with changes experienced by domesticated animals in their diets, habitats, and social milieu. While we limited our analysis to human–chimpanzee comparisons because Pan is the closest sister clade to Homo, recent work has indicated that the human gut microbiota is more similar to that of baboons (Gomez et al., 2019; Amato et al., 2019). Baboons are more distantly related to humans but have been argued to be closer in terms of diet and dietary physiology (Codron et al., 2008; Lambert, 1998), accentuating our finding of the importance of ecological factors in shaping the microbiota. Further work will be required to assess the specific combination of ecological factors driving similarities between domesticated and industrialized gut microbial signatures.

Because laboratory animals demonstrate some of the largest overall differences relative to their wild counterparts, they might be expected to have high translational potential as models for studying the gut microbiota of industrialized human populations. However, recent findings show that laboratory mice are poorer immunological models for humans in industrialized settings than are wild mice or laboratory mice harboring a wild microbiota (Beura et al., 2016; Rosshart et al., 2019). While the industrialized human gut microbiota exhibits parallels to those of domesticated animals, it may experience a broader array of environments and greater temporal variability; for example, greater ecological variability may explain the elevated gut microbial Shannon diversity seen in humans as compared to laboratory animals (Figure 1—figure supplement 2). Alternatively, it may be that domesticated laboratory animals are strong models for some aspects of host–microbe biology other than immunology. Certainly, studies of non-domesticated animals will be necessary to understand the natural history of host–microbe interactions (Reese and Kearney, 2019; Hird, 2017), as well as to determine the most appropriate models for translational research.

The fact that laboratory mice were permissive of recolonization by wild strains indicates that the local extinctions that occurred during domestication and/or generations in captivity can potentially be mitigated, thereby potentially improving the utility of these animals for research. Previous work has relied on laboratory mice colonized with a wild microbiota but fed standard laboratory chow (Rosshart et al., 2017; Rosshart et al., 2019) or on wild mice fed wild diets (Martínez-Mota et al., 2020). A combination of these approaches – adding wild gut microbial community members and feeding wild diet – would be expected to best support a wild gut microbiota in laboratory mice. A wild-microbiota laboratory-genotype model could be especially useful for studying infection challenges, disentangling host gene versus microbiota contributions to disease phenotypes, and testing for host–microbiota coevolution (Rosshart et al., 2019).

More generally, our data add to growing evidence that the gut microbiota is finely tuned to variations in the environment, affording at once expanded opportunities for biological mismatch to arise between the host and microbiota and for rapid microbiota-mediated host adaptation to novel environments. Further work to characterize the ecological significance of gut microbial plasticity will help reveal the fundamental nature of the host–microbial relationship, the conditions under which plasticity is beneficial versus detrimental, and the ecological conditions promoting cooperative, commensal, and competitive dynamics. The answers will have profound implications for our definition and pursuit of a healthy gut microbiome.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Biological sample (Bos taurus)FecesThis paperN = 9, sex unknown
Biological sample (Bison bison)FecesThis paperN = 20, sex unknown
Biological sample (Ovis aries)FecesThis paperN = 13, twelve females
Biological sample (Ovis canadensis)FecesThis paperN = 10, sex unknown
Biological sample (Sus scrofa domesticus)FecesThis paperN = 9, sex unknown
Biological sample (Sus scrofa)FecesThis paperN = 16, five females
Biological sample (Vicugna pacos)FecesThis paperN = 8, sex unknown
Biological sample (Vicugna vicugna)FecesThis paperN = 5, two females
Biological sample (Canis lupus familiaris)FecesThis paperComparative: N = 7, four females Experiment: N = 9, sex unknown
Biological sample (Canis lupus)FecesThis paperComparative: N = 9, sex unknown Experiment: N = 10, sex unknown
Biological sample (Oryctolagus cuniculus)FecesThis paperDomesticated: N = 11, four femalesWild: N = 12, sex unknown
Biological sample (Cavia porcellus)FecesThis paperN = 10, zero female
Biological sample (Cavia tschudii)FecesThis paperN = 11, sex unknown
Biological sample(Mus musculus)FecesThis paperComparative:
N = 9 (domesticated), zero female
N = 9 (wild), sex unknown
Experiments:
N = 49 (domesticated), zero female
N = 6 (wild), sex unknown
Biological sample (Rattus norvegicus)FecesThis paperDomesticated: N = 6, sex unknown
Biological sample (Rattus norvegicus)Intestinal sampleThis paperWild: N = 10, three females
Biological sample (Pan troglodytes)FecesThis paperWild: N = 7, seven females
Captive: N = 3, two females
Biological sample (Homo sapiens)FecesThis paperN = 7, five females
Sequence-based reagent515FCaporaso et al., 2011PCR primersGTGCCAGCMGCCGCGGTAA
Sequenced-based reagent806RCaporaso et al., 2012PCR primersGGACTACNVGGGTWTCTAAT
Software, algorithmRR Core TeamVersion 3.3
Software, algorithmQIIMECaporaso et al., 2010Version 1.8
Software, algorithmveganOksanen et al., 2017
Software, algorithmlme4Bates et al., 2015
Software, algorithmTimeTreeKumar et al., 2017http://timetree.org
Software, algorithmbootCanty and Ripley, 2020Version 1.3-25

Fecal sample collection

Distal gut microbiota samples from a range of non-human species were collected by authors or collaborators. Fecal samples from non-human mammals were collected from the ground within seconds to a few hours (<6) of production over the course of 2017 and 2018. In the case of artiodactyl, carnivore, lagomorph, and rodent feces, this approach precluded the need for institutional approval. Wild chimpanzee fecal samples were collected by field assistants under the approval of the University of New Mexico IACUC (protocol 18-200739-MC) and with permission of the Uganda Wildlife Authority and Uganda National Council for Science and Technology. Captive chimpanzee fecal samples were collected passively by keepers at Southwick’s Zoo, Mendon, MA. Human samples were self-collected by healthy study participants after providing written informed consent under the approval of the Harvard University IRB (protocol 17-1016) (Carmody et al., 2019). Samples were immediately frozen prior to permanent storage at –80°C. The only exceptions were wild vicuña and wild chimpanzee samples, which were preserved in RNAlater stabilization solution (Invitrogen) due to logistical issues in transportation from remote sampling locales. RNAlater was removed from these samples with centrifugation prior to further processing, and while sample preservation method was significantly associated with microbiota composition (p<0.001, PERMANOVA) it explained only a minor portion (R2 = 0.01) of the variation in beta-diversity. Sample sizes were chosen based on animal availability with a N > 5 for all species.

Domesticated animals

Request a detailed protocol

Domesticated sheep (Ovis aries; N = 11, ten females), cattle (Bos taurus; N = 9, sex unknown), and pig (Sus scrofa domesticus; N = 9, sex unknown) fecal samples were collected from a farm in Vershire, Vermont (VT). Domesticated alpaca (Vicugna pacos; N = 8, sex unknown) and domesticated sheep (O. aries; N = 2, two females) fecal samples were collected from a farm in Groton, Massachusetts (MA1). Mouse (Mus musculus, N = 9, zero female), rat (Rattus norvegicus; N = 6, sex unknown), and guinea pig (Cavia porcellus; N = 10, zero female) fecal samples were collected from animals in Harvard laboratory facilities (MA2). Domesticated rabbit (Oryctolagus cuniculus; N = 11, four females) fecal samples were collected from a shelter in Billerica, Massachusetts (MA3). Dog (Canis lupus familiaris; N = 7, four females) fecal samples were collected from personal pets in Stacy, Minnesota (MN). All samples were collected in summer or fall 2017. We have limited ability to distinguish between locale and species effects since all but one species (sheep) had samples collected from only one locale. Some locales had multiple species present, and we do find a significant effect of locale on overall microbial community composition even when correcting for host phylogeny effects (p<0.001, R2 = 0.16, F = 6.14, PERMANOVA). However, it is clear that locale does not necessarily lead to convergent microbiota across taxa as evidenced by the low clustering by site in NMDS ordination space (Figure 1—figure supplement 1). When analyzing just the sheep samples, we find a minimal effect of locale (p=0.023, R2 = 0.07, F = 2.08, PERMANOVA).

Time since domestication for each species pair was drawn from published data (Zeder, 2012; Morand et al., 2014; Driscoll et al., 2009; Goñalons and Yacobaccio, 2006). Time since divergence from wild progenitor or paired wild sample was taken from http://timetree.org (Kumar et al., 2017) because not all species in our sample set had existing genome assemblies of sufficient quality from which to infer a genome-based phylogeny. For animals with the same species name as their wild pair (e.g., Sus scrofa for both pigs and boars), which TimeTree treats as the same node, we used time since domestication estimates in lieu of time since divergence. In two cases (alpaca/vicuña and guinea pigs), the species have different names, enabling TreeTime to estimate time since divergence. Notably, these estimates are much larger than the time since domestication estimates for those dyads despite the fact that the wild species sampled are widely considered to be the progenitors of the domesticates. We include analyses with both time since divergence and time since domestication, considering the former to be more conservative estimates of relatedness. Gut physiology and diet classifications were assigned based on published literature (Stevens and Hume, 2004) and are listed in Supplementary file 2.

Wild animals

Request a detailed protocol

Wild boar (Sus scrofa; N = 16, five females) fecal samples were collected from adults and juveniles in southeastern Alabama (AL) during fall 2017. Rat (Rattus norvegicus; N = 10, three females) distal gut samples from adults and juveniles were collected directly from the colon shortly following trapping in New York City (NY) between February and May 2017 (Combs et al., 2018). Bison (Bison bison, N = 20, sex unknown) fecal samples were collected from a semi-free-ranging population in Elk Island National Park, Alberta, Canada (CAN) during 2013 (Weese et al., 2014). Wild house mouse (Mus musculus, N = 9, sex unknown) fecal samples were collected from live-trapped animals in the Cambridge, Massachusetts area (MA4) during winter 2018. Pursuant to Massachusetts state law, permits were not necessary to trap animals indoors. Wild European rabbit (Oryctolagus cuniculus; N = 12, sex unknown) fecal samples were collected in Mértola, Portugal (POR), during spring 2018. Bighorn sheep (Ovis canadensis; N = 10, sex unknown) fecal samples were collected during 2017 and 2018 in Wyoming (WY). Vicuña (Vicugna vicugna; N = 5, two females) fecal samples were collected during spring 2018 from a captive population in Santiago, Chile (CHL) that was free-grazing but supplemented with hay. Wild guinea pig (Cavia tschudii, N = 11, sex unknown) fecal samples were collected at a facility in Lima, Peru, (PER) during spring 2018. Wolf (Canis lupus; N = 9, sex unknown) fecal samples were collected during fall 2017 from captive packs fed an exclusively raw diet at the Wildlife Science Center (WSC) sanctuary in Stacy, Minnesota (MN). Wild chimpanzee (Pan troglodytes schweinfurthii, N = 7, seven females) fecal samples were collected between September 2015 and January 2016 from adult members of the Kanyawara community in the Kibale National Park, Uganda; samples were randomly selected from adults from a larger set initially prepared for a separate project (Reese et al., 2021). Captive chimpanzee (N = 3, two females) fecal samples were collected in May 2019 from adults at Southwick’s Zoo in Mendon, MA.

Humans

Request a detailed protocol

Fecal samples were collected from healthy adult humans (N = 7, five females) residing in the Cambridge, Massachusetts area. All participants were provided with sterile study kits and self-collected fecal samples during the same 3-day period in December 2017. During this period, participants freely consumed their habitual diets. Fecal samples were immediately stored at –20°C and were transferred within 24 hr to permanent storage at –80°C.

Animal experiments

Wild mouse capture

Request a detailed protocol

Mus musculus were introduced to North America from Western Europe and are now commonly found in commensal settings (Schwarz and Schwarz, 1943). We set out Sherman live traps in the evenings in buildings and barns during February 2018. Traps were baited with peanut butter and a small cube of apple and outfitted with sufficient bedding and food to sustain an adult mouse for at least 48 hr. They were checked the following morning to minimize time spent in the traps. Rodents were transferred from their traps to a plastic bag, and unwanted rodent species were released immediately. Mice that were identified as M. musculus (rather than Peromyscus spp., also common in Massachusetts) were transferred to temporary cages for transport to lab facilities. At time of capture, we collected fecal samples and body swabs for zoonoses testing by Charles River. All individuals were tested for fur mites; MAV 1 and 2; MHV; MPV/MVM; MRV; mousepox; POLY; REO; LCMV; LDV; TMEV/GDVII; SEND; PVM; Mycoplasma; Mycoplasma pulmonis; Filobacterium rodentium (formerly CAR Bacillus); Citrobacter rodentium; Clostridium piliforme; Corynebacterium kutscheri; Corynebacterium bovis; Streptobacillus moniliformis; and pinworm. The only agent of concern found was fur mites. Because animals were not treated for parasites or pathogens in order to preserve their wild gut microbiota signature, they were housed under non-specific pathogen free (SPF) conditions at Harvard’s Concord Field Station. Mice were allowed at least 3 days to adjust to laboratory conditions without handling and provided with a wild mouse diet (a mix of bird seed [Wagner's Eastern Regional Blend Deluxe Wild Bird Food] and Bio-Serv freeze-dried mealworms; Supplementary file 3) before the beginning of the experiment. All mice were housed singly from the time of arrival at the Concord Field Station and had access to water and food ad libitum.

Wild/laboratory mice reciprocal diet experiment

Request a detailed protocol

A total of 10 wild mice were captured for this experiment. Of these, two were deemed too young for inclusion in the study, one died before beginning the experiment, and one died during the course of the experiment. As a result, we collected six wild mice (WildH) that were included in the full study. In addition to the wild mice, male C57BL/6 mice 10–12 weeks of age with a conventional microbiota were purchased from Charles River Laboratories for inclusion in the study (DomH). All mouse experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals using protocols approved by the Harvard University Institutional Animal Care and Use Committee (protocol 17-11-315). The sample size for the laboratory animal group was chosen following power analyses to allow for β less than 0.1; the sample size for the wild animal group was chosen based on animal availability (N = 3 per diet treatment). The experiment was conducted once. All mice were housed singly from the time of arrival at the Concord Field Station and had access to water and food ad libitum. Mice were provided nesting material and plastic enrichment housing atop corncob bedding. The mice were maintained in a 20–22°C room with natural light cycles.

Mice, both wild and laboratory, were randomly assigned to one of two dietary treatment groups (N = 10 laboratory mice or three wild mice per group). The first group (domesticate diet: DomD) was provided ad libitum mouse chow (Prolab Isopro RMH 3000) in overhead food hoppers, as is standard in mouse studies. The second group (wild diet: WildD) was provided a mix of bird seed (Wagner's Eastern Regional Blend Deluxe Wild Bird Food) and freeze-dried mealworms (Supplementary file 3) in excess of predicted consumption. The food was placed in the corncob bedding to simulate foraging.

Before initiating the dietary interventions, all individuals were weighed and multiple fecal samples were collected. The mice were then returned to a new, clean cage with the treatment diet present. Over the next week, fecal samples and weights were collected daily for each mouse. The amount of food remaining was weighed and additional wild diet was added daily. One week after beginning the experiment, mice were weighed and fecal samples were collected, then mice were moved to clean cages. Weights and fecal samples were henceforth collected weekly (days 14, 21, and 28) until the end of the experiment, although additional food was added biweekly for individuals assigned to the wild diet treatment. Additional chow was added to hoppers for individuals assigned to the conventional diet treatment, and all water bottles were refilled as necessary. At the end of each week, food hoppers were weighed (DomD), and cage bedding was collected and sifted to quantify uneaten food (WildD), determine total weekly fecal production (all groups during week 3), as well as to provide fecal samples for bomb calorimetry (6050 Calorimeter, Parr). All calorimetry results were adjusted for the average weekly fecal production and average weekly food intake of each experimental group. At the end of the experiment (days 28–30), mice were humanely sacrificed via CO2 euthanasia.

Wild/laboratory mice gavage experiment

Request a detailed protocol

Thirty 10-week-old male C57BL/6 mice with a conventional microbiota were purchased from Charles River Laboratories for inclusion in the study. The sample size was chosen following power analyses to allow for β less than 0.1; the experiment was conducted once. Mice were cohoused in litter groups of 3–4 until beginning the study. Cage groups were spread across the treatment groups, with individuals randomly assigned to a diet and colonization treatment. There were three treatment groups: wild colonized/wild diet (WildC/WildD); wild colonized/domesticate diet (WildC/DomD); or phosphate-buffered saline (PBS) gavage/wild diet (PBSC/WildD). The latter served as a colonization control, emulating the DomH/WildD group from the reciprocal diet mouse experiment.

On the first day of study, fecal samples were collected from each mouse and the mice were weighed before colonization. Mice receiving a wild microbiota were colonized with cecal contents collected from one randomly selected WildH/WildD individual in the wild/laboratory experiment (see above). The cecal contents were prepared following Rosshart et al., 2017. In short, frozen cecal contents were resuspended in sterile-reduced PBS (1:1 g:ml) under anaerobic conditions then diluted 1:30. Each recipient mouse received a single dose of 100 to 150 µl cecal solution via oral gavage. PBS control mice received 100–150 µl sterile-reduced PBS via oral gavage.

Following gavage, mice were transferred to single housing in new, clean cages with the treatment diet present. Mice receiving domesticate diet were provided ad libitum mouse chow (Prolab Isopro RMH 3000) in overhead food hoppers. Wild mouse diet consisted of a mix of bird seed (Wagner's Eastern Regional Blend Deluxe Wild Bird Food) and freeze-dried mealworms (Supplementary file 3), which was provided in excess of predicted consumption and placed in the corncob bedding to simulate foraging. All mice were provided with nesting material and plastic enrichment housing atop corncob bedding.

Additional fecal samples and weights were collected on days 1, 2, and 8 following gavage. After weights and fecal samples were collected on day 8, mice were humanely sacrificed via CO2 euthanasia. During the course of the experiment, one mouse in the PBSC/WildD (control) treatment group died, resulting in a N = 9 for that group. At the end of the experiment, food hoppers were weighed (DomD), and cage bedding was collected and sifted to quantify uneaten food (WildD) and total weekly fecal production (all groups).

Wolf/dog reciprocal diet experiment

Request a detailed protocol

Ten wolves (Canis lupus) and nine dogs (Canis lupus familiaris) participated in the study. The sample size for the canid experiment laboratory animal group was chosen following power analyses to allow for β less than 0.1; the experiment was conducted once. Wild-caught or captive-born wolves lived in packs of 2–6 wolves at the WSC (Stacy, MN). They were exposed to natural light cycles and weather conditions, with access to shelters and wolf-dug dens in their enclosures. Wolves had ad libitum access to water. Dogs enrolled in this study were privately owned in Stacy, MN, and were recruited to participate through their owners. Dogs were kept in their typical environment throughout the experiment. All canid experimentation was approved by the WSC IACUC (protocol HAR-001). Wolves were enrolled in the study from December 5–20, 2018, and dogs were enrolled from December 24, 2018 to January 8, 2019.

On every day of the study, across both the control and reciprocal diet arms, wolves were given inert colored glass beads via treats (~15 g raw meatballs). The beads can be passed naturally without harm to the animal and allowed for source identification for fecal samples in cohoused animals. Fecal samples were collected daily in a sterile manner then moved to −20°C storage before long-term storage at −80°C. For the first week of the experiment, all animals received a control diet that matched their genetic background (Supplementary file 3) – raw chicken parts (4 lbs/animal) for wolves (WildH/WildD) and commercial dog food (Nutrisource Lamb Meal and Peas Recipe, Grain Free) for dogs (DomH/DomD). Fecal samples were collected at least once daily from wolf enclosures and the dogs’ home environments without handling the animals. On day 8, wolves were provided no new food, but were able to complete consumption of previously provided food. Fecal samples collected on this day were considered baseline samples for the next arm of the experiment. Beginning on day 8, a week of reciprocal diet feeding was commenced. During this period, wolves were fed commercial dog food (WildH/DomD) and dogs were fed raw chicken parts (DomH/WildD). Daily fecal samples were again collected. Following completion of the study, animals were returned to their standard diet.

Human sample meta-analysis

Request a detailed protocol

To compare the microbial differences observed between wild and domesticated animals and between humans and chimpanzees with differences linked specifically to industrialization, we also performed analyses including all of the samples outlined above plus a subset of published data from Jha and colleagues (Jha et al., 2018). To match sample sizes used in our human–chimpanzee contrast, we subsampled seven adults from their Chepang (Nepalese foragers), Raji (Nepalese foragers transitioning to subsistence farming), Raute (Nepalese foragers transitioning to subsistence farming), Tharu (Nepalese subsistence farmers), and American populations, as well as seven adults from the Hadza (Tanzanian hunter gatherers) population they analyze, which were originally described in another study (Smits et al., 2017). All data were downloaded from the European Nucleotide Archive. These populations represent extremes of industrialized and non-industrialized human lifestyles with the variation among the non-industrialized groups not covering the full breadth of intermediate lifestyles (e.g., modern agricultural or recent urban transplants). We believe that these extremes enable us to test how the human gut microbial communities respond to major ecological change of a magnitude that could be argued to approximate that experienced by gut microbial communities of animals undergoing domestication.

These samples were not necessarily collected or processed in an identical manner to each other or to the new data collected in this paper – namely, the Chepang, Raji, Raute, and Tharu samples were collected and preserved using OMNIgene kits while the American and Hadza samples were frozen, while all samples were extracted with the MoBio PowerSoil kit and sequenced on an Illumina MiSeq. Unfortunately, no existing published data on non-industrialized populations have been generated using exactly the same methods employed here. However, we reprocessed the sequences using the 16S rRNA gene amplicon QIIME pipeline described below and rarefied all samples to 10,000 reads depth to make the data as comparable as possible. Importantly, to the extent that these discrepancies introduce biases to our analyses, we expect they would do so in a manner agnostic to the comparison with our chimpanzee and American human samples. The high similarity between the US samples that we collected and those collected by Jha and colleagues supports this expectation. Furthermore, the fact that the non-industrialized Hadza samples were not stored with OMNIgene kits precludes conflating any non-industrialized signal with a sample-processing signal.

Specific taxa chosen for targeted analyses were identified from the human lifestyle analyses by Smits and colleagues (Smits et al., 2017); only taxa that had a non-zero abundance in wild chimpanzees were analyzed here. Time since Pan–Homo divergence was drawn from http://timetree.org (Kumar et al., 2017) to be consistent with domestication analyses.

16S rRNA gene analysis

Extraction

Request a detailed protocol

Following collection during observational or experimental animal work, fecal samples were temporarily stored at −20°C then moved to −80°C for long-term storage. Individual mouse pellets or approximately 50 mg feces were used for DNA extraction using the E.Z.N.A. Soil DNA Kit (Omega, Norcross, GA) following manufacturer’s instructions.

Sequencing

Request a detailed protocol

We performed 16S rRNA gene amplicon sequencing on fecal samples to determine gut microbial community structure. We used custom barcoded primers (Caporaso et al., 2011) targeting the 515F to 806Rb region of the 16S rRNA gene following published protocols (Caporaso et al., 2011; Caporaso et al., 2012; Maurice et al., 2013). Sequencing was conducted on an Illumina HiSeq 2500 with single-end 150 bp reads in the Bauer Core Facility at Harvard University. Sequence files were processed using QIIME version 1.8 (Caporaso et al., 2010). We demultiplexed the sequences using split_libraries_fastq.py, then used parallel_pick_otus_uclust_ref.py to carry out closed reference operational taxonomic unit (OTU) picking with 97% similarity using the default parameters under UCLUST. Microbial taxonomy for these OTUs was assigned in reference to the GreenGenes database (version 13.5) (DeSantis et al., 2006). We obtained 158,611 ± 109,567 assigned reads per sample.

qPCR

Request a detailed protocol

To estimate total bacterial density, qPCR was performed on fecal DNA using the same primers as used for sequencing. qPCR assays were run using PerfeCTa SYBR Green SuperMix Reaction Mix (QuantaBio, Beverly, MA) on a Bio-Rad CFX384 Touch (Bio-Rad, Hercules, CA) in the Bauer Core Facility at Harvard University. Cycle-threshold values were standardized against a dilution curve of known concentration and then adjusted for the weight of fecal matter extracted.

Statistical analyses

Request a detailed protocol

All statistical analyses were carried out in R (R Core Team, version 3.3). Alpha-diversity (Shannon index, OTU richness) were calculated for rarefied OTU tables (rarefaction limit of 17,500 for cross-species dataset; 27,000 for wild mouse study; 15,500 for the mouse colonization study; 7,500 for canid experiment). Beta-diversity (Bray–Curtis, Weighted UniFrac, Unweighted UniFrac) metrics were calculated using the vegan package (Oksanen et al., 2017) or QIIME on unrarefied data. All statistical tests performed were non-parametric. PERMANOVA was carried out with the adonis2 function in vegan with the domestication status variable nested within the species pair to correct for known relationships within dyads. To test how beta-diversity varied based on relatedness (within species, between wild–domesticated pairs, or among unrelated pairs), domestication type (relative to US human samples), or human/primate population (relative to zoo chimpanzee samples), we used a bootstrapping approach, thus correcting for the non-independence of dissimilarity measurements that include the same individuals in multiple comparisons. In short, we permuted the Mann-Whitney U test statistics and p values, resampling (25,000 permutations) with stratification specified by individual identity, using the boot package (Canty and Ripley, 2020). Variability in a species’ microbial community composition was calculated with the permutest and betadisper functions in vegan. For changes in family-level abundance, a Bonferroni correction for multiple hypothesis correction was then applied to all test results.

Potential human pathogens were identified following published methods (Kembel et al., 2012; Reese et al., 2016). In short, we obtained a list of potential human pathogens, compiled by Kembel and colleagues (Kembel et al., 2012), then manually compared that list to the taxa identified to the genus or species level in our analysis. A subset of the data containing only these species was then analyzed for diversity with the same methods used for the total dataset.

To determine the consistency of gut microbial differences across ordination space due to domestication, Pan–Homo divergence, or industrialization in the observational study, we calculated the average position of the host dyad (e.g., pig/boar) or all primates (humans and chimpanzees) for axis 1 of the NMDS, then measured the displacement along each axis for an individual sample relative to that mean position. We tested for differences in these ordination shifts by domestication status or primate host taxonomy (e.g., chimpanzee versus US human). To estimate the direction and magnitude of changes in beta-diversity during the experimental studies, we tested whether inclusion of a treatment group term improved the performance of a linear mixed effects model relative to a model with only time and animal ID terms for predicting the NMDS1 axis value for an individual. These analyses allowed us to consider the direction of beta-diversity changes in addition to the magnitude. We estimated the direction and magnitude of dissimilarity from the expected community composition (donor microbial community in gavage experiment; baseline species average for DomH/DomD or WildH/WildD in diet experiments) as the length of the vector through the first axis of ordination space. In analyzing the experimental diet study data, we used the lmer and anova functions in the package lme4 (Bates et al., 2015) to perform likelihood tests comparing a linear mixed effects model that included the variable of interest (i.e., treatment group) to a model that included only time variables. In both models, individual identity was included as random effects.

We explored the role of relatedness in structuring the cross-species dataset by (i) performing a Mantel test to compare divergence times and Bray–Curtis dissimilarities; (ii) testing for Spearman correlations between the NMDS shifts and the time since divergence and performing likelihood tests to compare a linear mixed effects model that included both domestication status and dyad as fixed effects and divergence time as a random effect with a model that only included the dyad and divergence time terms; and (iii) testing for Spearman correlations between the average dissimilarity within a wild–domesticated dyad (e.g., the average dissimilarity for all combinations of boar–pig pairs) and the time since domestication and time since divergence. We also used Mann–Whitney U tests to determine if dissimilarity between unrelated pairs was higher than for wild–domesticated dyads or within sets of conspecifics.

Data availability

All sequencing data included in this study are available in the European Nucleotide Archive under accession number PRJEB36262.

The following data sets were generated
    1. Reese AT
    2. Carmody RN
    (2020) European Nucleotide Archive
    ID PRJEB36262. Effects of domestication on the gut microbiota parallel those of human industrialization.
The following previously published data sets were used

References

    1. Bates D
    2. Mächler M
    3. Bolker B
    4. Walker S
    (2015)
    Fitting linear mixed-effects models using lme4
    Journal of Statistical Software 67:1–48.
  1. Book
    1. Carmody RN
    (2017)
    Evolution of the human dietary niche – quest for high-quality
    In: Muller MN, Wrangham RW, Pilbeam DR, editors. Chimpanzees & Human Evolution. Cambridge: Harvard University Press. pp. 311–338.
  2. Book
    1. Goñalons GLM
    2. Yacobaccio HD
    (2006) The domestication of South American camelids: a view from the south-central Andes
    In: Zeder M. A, Bradley D, Emshwiller E, Smith B. D, editors. Documenting Domestication. Berkeley, CA: University of California Press. pp. 1–375.
    https://doi.org/10.1525/j.ctt1pnvs1
  3. Book
    1. Hosey GR
    2. Melfi V
    3. Pankhurst S
    (2009)
    Zoo Animals: Behaviour, Management, and Welfare
    Oxford, UK: Oxford University Press.
  4. Book
    1. Muller MN
    2. Wrangham RW
    3. Pilbeam DR
    (2017)
    Chimpanzees and Human Evolution
    Cambridge, MA: Harvard University Press.
  5. Book
    1. Stevens CE
    2. Hume ID
    (2004)
    Comparative Physiology of the Vertebrate Digestive System (Second Edition)
    Cambridge, UK: Cambridge University Press.
  6. Book
    1. Timoney JF
    2. Gillespie JA
    3. Scott FW
    4. Barlough JE
    (1988)
    Hagan and Bruner's Infectious Diseases of Domestic Animals
    Ithaca, NY: Cornell University Press.

Decision letter

  1. María Mercedes Zambrano
    Reviewing Editor; CorpoGen, Colombia
  2. Detlef Weigel
    Senior Editor; Max Planck Institute for Developmental Biology, Germany

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This work examines the role of domestication and industrialization on the microbiome by looking at changes in the gut microbiota of humans and wild and domesticated mammals. Despite being fundamentally different processes, the authors conclude that domestication and industrialization have impacted the gut microbiota in related ways.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Effects of domestication on the gut microbiota parallel those of human industrialization" for consideration by eLife. Your article has been reviewed by a Senior Editor, a Reviewing Editor, and three reviewers. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that we cannot pursue the publication of your manuscript in its present form. However, we consider that the work is very interesting and therefore a new version that thoroughly addresses the concerns raised would likely be reviewed again. Although it would be treated as a new submission, we would aim to retain an overlapping set of reviewers.

The work provides exciting and valuable information on the possible effects of domestication and industrialization on the gut microbiome. However, there were several methodological issues raised, such as host genotype determination, control for genetic distance and, in particular, concerns regarding data analyses (diversity metrics, OTU picking, error rates, Permanova, and FDR correction, to name some) that can take a considerable amount of time to perform. There were also misgivings regarding the validity of some of the conclusions based on the data presented. These include group comparisons that do not necessarily agree with the idea that domestication and industrialization similarly impact the gut microbiota, and the effect of host genotype or genetic distance on the microbiota. Please also take into account the comments about differences with respect to previous publications regarding the claim that domestic animals may be useful as models.

Reviewer #1:

This study examines effects of domestication on the gut microbiome of wild animals to the effect of industrialization on the gut microbiota of humans. They report consistent shifts in composition of gut microbiota in domestic animals and in humans from industrialized (but not from traditional) societies. They also perform cross-feeding experiments of wild and domesticated animals (lab mice/wild mice; dogs/wolves) and report that apart from genetics, diet plays a dominant role in shaping (loss of diversity) of the domestic gut microbiota.

I have the following comments:

1) Introduction: "genetic changes under domestication" How did the authors control for differences in genetic distance among the individual domesticated/wild animal pairs? Are the shifts in composition of microbiota during domestication and industrialization still consistent if controlled for genetic distance?

2) Introduction: "Finally, the convergent nature of many ecological shifts experienced by domesticated animals and industrialized human populations suggests that domestic animals may provide a uniquely useful model for studying the microbially-mediated health impacts of rapid environmental change." and Discussion "their translational potential as models for studying the gut microbiota of industrialized populations may be greater than is currently appreciated." This statement is not clear – please clarify in the context of two publications (Nature. 2016;532(7600):512-6 and Science, 708 2019;365(6452):eaaw4361) that appear to state the opposite.

3) The authors describe that diet plays a major role in changing the microbiota of wild animals to those of domestic ones. Diet is a great source of viruses. To which extend is the introduction or loss of viruses (in particular phages) responsible for the shift in gut microbiota?

4) The authors state repeatedly that wild animals more diverse microbiota. Are there uniform changes in taxa? Are some taxa lost, and if so, is this observed in several wild / domestic pairs?

Reviewer #2:

This is an exciting paper with important implications for how diet and environment interact to shape the composition and diversity of the gut microbiome. The results are interesting – particularly the results of the robustly designed diet challenge experiments. Concluding with the host-microbe-environment mismatch puzzle is thought-provoking. I am slightly concerned about how the framing of the study is phrased. I additionally have some questions/suggestions/concerns regarding the methods.

1) The results of this study are super interesting, but the authors need to be sure to make it very clear throughout that they are examining how environmental and dietary shifts associated with domestication may parallel environmental and dietary changes in some human populations (not that some human populations are domesticated and some are wild). The authors are mindful to make this clear most of the time, but it would be good to make it explicit all of the time.

In addition, I would ask that the authors carefully consider how human populations are described – traditional is not the best term, unless it is how those populations self-identify. There are real and very important ecological differences that distinguish the human populations that were sampled. Using language that somehow indicates what those differences are might be more impactful than using industrialized vs. traditional. Or, at the very least, clearly defining those terms early on in the article is necessary. Industrialized vs. non-industrialized or traditional can be read as placing as elevating either group or could be read as saying the populations are "advanced" and "not advanced" (particularly important as this paper will likely generate some media attention).

2) I am wondering if it is better to categorize the genotype/diet experiment as a provenance/diet experiment or something similar. As the authors did not actually look at host genetics in the wild-caught mice, they don't know how genetically distinct they are and there is certainly variation in genetic distance from the lab mice within the group of wild-caught mice.

3) Genetic changes kind of come up unexpectedly and without context the Introduction, which I found unclear. It may be better to focus this paragraph solely on ecological/environmental shifts? I was also a little confused if the authors were indicating the known genetic changes caused by domestication would change something about host physiology that would impact the gut microbiome somehow, or if the effect of divergence in host genetics would cause a simultaneous divergence in gut microbiome composition, or both.

4) Gomez et al., 2019 and Amato et al., 2019 both found that the human gut microbiota is actually closer to that of baboons than chimpanzees. I don't think the authors necessarily need to add baboons to the analysis, but it would be relevant to acknowledge in the discussion that chimps may or may not be the best comparison for humans.

5) Introduction (and elsewhere): I don't think domestic can be used in place of domesticated – the meanings, to me at least, are distinct.

6) Methodological concerns:

- Samples collected in RNAlater are not necessarily comparable to freshly frozen – please note in the methods which species were preserved with each method and describe how you accounted for this difference in preservation.

- Why was closed-reference OTU-picking chosen? Open-reference OTU-picking is the recommended method, unless one is comparing amplicons from different regions of the 16S rRNA gene. I would suggest that analyzing the data using one of the ASV strategies (DADA2 or Deblur) is recommended, but also do not want to force the authors to reanalyze their entire dataset (and the newer ASV methods become less useful when including 454 data).

- Yatsunenko et al., 2012 used 454 sequencing – I am curious how the authors corrected for the differences in sequencing-related error rates between 454 and HiSeq? And why they did not choose to use available human datasets sequenced in a manner comparable to the newly produced dataset in this paper?

- Using the adonis2 function in vegan would allow the authors to use marginal sums of squares in the PERMANOVA analysis – this might allow them to better tease apart which factors are accounting for what proportion of the variance in the dataset.

- A Bonferroni correction is quite conservative for microbiome datasets – FDR correction could be used instead.

- I would like to see an explanation for the choice of method to measure the magnitude of change in β-diversity, as it is one I haven't seen before and measuring change along an axis that does not have an easily interpretable meaning might not be the most informative. Alternatively, comparing pairwise unweighted and weighted UniFrac between domesticated/wild and baseline/treatment and/or performing a Procrustes analysis may be preferred.

Reviewer #3:

Reese et al., compare the microbiota of domestic animals and their closest wild counterparts, including a comparison of humans and chimpanzee microbiotas. They report similar changes to the microbiota from domestication and industrialization. Overall, the data presented is fairly noisy and many of the conclusions seem overstated given slight differences between groups. Even if we set aside the issues with the data, which are not trivial, it is unclear how important the conclusions are. For example, the last sentence of the Abstract:

"We conclude that domestication and industrialization have similarly impacted the gut microbiota, emphasizing the utility of domestic animal models and diets for understanding host-microbial interactions in rapidly changing environments, and the importance of studying non-industrialized human populations for understanding aspects of human health dependent on host-microbial co-evolution."

Not so easy to unravel the point(s) the authors are trying to make. The last passage is already very clear to the field, non-industrialised populations are important to study. The first part suggests that domestic animals and diets are useful in understanding the microbiota in changing environments. It is not clear exactly what this statement is trying to convey and it requires some clarification.

In the Abstract the authors state that "domestication and industrialization have similarly impacted the gut microbiota". A major concern is the data presented in Figure 5B for two reasons. First, the difference between two industrialized human populations appears to be larger than that observed between domestic and wild animals. Second, the shift to the left from industrialized humans to traditional humans is larger than from industrialized humans to chimpanzees. Not only is this problematic from the standpoint of implications about the "wildness" of traditional populations, but also difficult to interpret given the greater similarity in genetics, physiology, lifestyle, and diet between human populations than chimpanzees and humans.

The authors report greater between species variability in wild gut communities than domesticated. However, it does not look like they did this comparison for the human and chimpanzee data. Given published data showing that the between individual variability in the microbiota of industrial individual is larger than that of traditional population microbiota, it would be interesting to see how these data compare to that of chimpanzees given that this is not the result you would expect given the data from the other animal pairs.

It is not clear how α diversity was calculated. Was the data rarefied and if so to how many reads and were the samples sequenced sufficiently deep to ensure an accurate measurement of diversity.

Subsection “Diet vs. genotype effects on domestic gut microbial composition in mice”. "Domestication has profound effects on both ecology and host genotype." Do the authors mean "has had", ie, there is evidence that animals, when domesticated, show genotypic changes, eg, new traits are selected for. Domestication over short time periods may have little effect on genotype.

Subsection “Diet vs. genotype effects on domestic gut microbial composition in mice”. "we found that host genotype explained the largest amount of variation" It is unclear what data the authors are examining to reach this conclusion. The species appears to be Mus musculus for these analyses. Are the authors performing a host genotype (eg, SNP) analysis? Please clarify how differences in host genotype are being determined.

Figure 2.

- It is very difficult to draw conclusions from Figure 2B. Suggest that the authors show centroids or find a better way to represent the data. Some of the colors are too similar as well, so difficult to differentiate. Why are DomG/DomD points moving on the PCA plot? Same with WildG/WildD? Perhaps this data could reveal drift of the microbiome composition in the absence of intervention, which may inform whether their diet shift in the other groups is meaningful.

Figure 4 has many of the same issues described for Figure 2. It's very difficult to interpret these panels with so many points going in different direction and minimal color differences between some of the points.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for submitting your article "Effects of domestication on the gut microbiota parallel those of human industrialization" for consideration by eLife. Your article has been reviewed by Detlef Weigel as the Senior Editor, a Reviewing Editor, and three reviewers. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

As the editors have judged that your manuscript is of interest, but as described below that additional experiments are required before it is published, we would like to draw your attention to changes in our revision policy that we have made in response to COVID-19 (https://elifesciences.org/articles/57162). First, because many researchers have temporarily lost access to the labs, we will give authors as much time as they need to submit revised manuscripts. We are also offering, if you choose, to post the manuscript to bioRxiv (if it is not already there) along with this decision letter and a formal designation that the manuscript is "in revision at eLife". Please let us know if you would like to pursue this option. (If your work is more suitable for medRxiv, you will need to post the preprint yourself, as the mechanisms for us to do so are still in development.)

Summary:

The manuscript by Reese et al., explores the effects of mammalian domestication and human industrialization on the gut microbiota and has important implications on how diet and environment interact to shape the composition and diversity of the gut microbiome. They characterize the microbiome via 16S rRNA gene sequence analysis in various mammalian species and show that the microbiome shifts with domestication. Using cross-feeding experiments in mice, wolves and dogs they are able to demonstrate that diet and microbial inoculation can reverse the effects of domestication. Finally, they compare chimpanzees and both industrialized and non-industrialized human populations and show that shifts in microbiome composition are larger when chimpanzees are compared with industrialized populations. Overall the work presents clever experiments aimed at characterizing the effects of domestication on the gut microbiota and comparing these effects with those of human industrialization.

Essential revisions:

1) Reviewers were concerned with the comparisons between domestication and industrialization and the subsequent conclusions. This aspect of the work needs to be improved for clarity and the claims toned down as they are not fully supported by the data presented.

a) The authors should note that domestication, which has taken a long time, and industrialization, a fairly recent change to our ecology, are different processes. Therefore, the direct comparisons in the manuscript do not seem entirely appropriate and should be more carefully addressed. In particular, the data does not provide strong evidence to support the claim that animal domestication and human industrialization result in similar effects on their hosts microbiome, even though this conclusion may be correct, since it makes sense given that many ecological processes are probably affected in similar ways. This conclusion should therefore be toned down to agree with their data.

b) Is there a way to incorporate data from populations that use subsistence strategies involving domestication, but are not Industrialized (the other populations in Jha et al., even)? It could be expected that the agricultural or pastoral but non-Industrialized countries would be somewhat intermediate in their microbiome composition, as they experience the factors of domestication without some of the extreme ecological consequences of Industrialization (antibiotics, highly processed foods, etc.). Is this the case?

c) A more nuanced discussion should occur at some point in the manuscript on the choice and caveats of using highly Industrialized populations in this comparison given that the process being compared is domestication and not industrialization.

2) The revised manuscript has improved but still lacks clarity in many places and uses language that is vague and often misleading, making it difficult to understand what the authors are trying to say. The entire text should therefore be checked and improved to make the language more precise.

a) In the Abstract, for example, it is not clear what shifts the authors refer to, what is meant by microbiomes to be impacted “'similarly”, and what “parallel ecological changes” are. It can be argued that the ecological changes are quite different in industrialized humans and domesticated animals (housing, hygiene, diet, etc.). However, the ecological processes that impacted their microbiomes, and the compositional alterations, might have been similar.

b) This vagueness is also found through the entire manuscript. What are ecological parallels (Introduction)? What is a "suite of shared ecological changes" (Introduction)? Which “evolutionary forces” were studied? What do the authors mean by "individual shifts"? (figure legend of Figure 1C). Compositional shifts in an individual? Was that even assessed?

c) The term “shifts” is used inappropriately throughout the manuscript. For example, what are "shifts between industrialized humans and wild chimpanzees" (Figure 5 legend)? The microbiome does not really shift from a human to a chimpanzee. Do the authors refer to differences between microbiomes in different hosts?

3) The authors should be careful with the way they present their results to avoid biased interpretation and make claims that are clearly supported by their results.

a) It sometimes seems as if the authors have interpreted the findings to fit a preconceived idea of the findings. For example, the authors conclude a "consistent effect of domestication status" (Results), but the samples cluster by host, which has the highest effect sizes. The conclusion is then mainly based on a statistical analysis that showed domesticated samples to be "further right" on an NMDs axis. This is not very convincing, and not very clear in Figure 1C either.

b) Another claim is that in Figure 2, differences between domesticated and wild mice can be overcome by a diet switch, but looking at Figure 2—figure supplement 2, that is simply not the case. It is difficult to see how the data in Figure 5 provides strong evidence that the effects of domestication and industrialization are similar.

4) More clarification is needed for wild and domestic microbiome results and subsequent conclusions

a) The results presented (Results and Figure 1) do not seem to support the conclusion that domestication is shifting all species to the right along NMDS1. The magnitude and direction of shift seems to differ based on host species. While the general trend of all species lumped together is to the right, sheep and pigs don't seem to follow the pattern (and some others don't seem to have a strong shift to the right). What are the effect sizes for the Mann-Whitney U tests here? Also, looking at Figure 1—figure supplement 2A, only the companion species are denoted as having a p<0.05, which seems at odds with the statement in the Results.

This species-dependent direction and strength of shift is not entirely unexpected based on previous work. Shifting host ecology (diet or captivity) has previously been shown to differentially effect host species: Amato et al., 2015 and McKenzie et al., 2017.

The inconsistency in the direction of the shift might not actually negate the broader point, that domestication at times has effects on the gut microbiome that are very similar to the shift we see between industrialized and non-industrialized humans. In fact, it might be instructive to point out what specific species might be good models for the shift we see in humans – what are the specific ecological shifts with domestication in those species and how does that mirror the ecological shift with industrialization in humans?

b) What does it look like when you put the results of the mouse and canid experiments in the same ordination space with your wild/domestic and chimp/human pairs? Is the shift in the expected direction? When looking at the results of the mouse experiment and the canid experiments on their own, we see a shift to the left with experimental domestication (ie, for the Wildh/Domd treatments), but this might be a function of the ordination space?

c) Were any of the animals, either wild or domestic, from the same family, field, pen, etc.? Cohousing results in convergent microbiome profiles across a number of species due to horizontal microbial exchange. If conspecifics were collected from the same living situation or were related, one might expect higher microbiome sharing on those grounds alone. This potential confounder could explain the high similarity between the conspecifics. These details should be added to the Materials and methods. If this is an issue, it should be corrected for in statistical comparisons (if possible).

5) Technical concerns and data presentation

a) Figure 2 and Figure 4 are a difficult to interpret, because the lines used to indicate moving points are obscuring the points themselves in some cases. Would ellipses around the treatment groups in the NMDS plots be more informative than the moving points?

b) For the adonis2 function, to get the marginal sums of squares you need to include “by = "margin"” in the function call. Using adonis2 without specifying “by” is equivalent to using the older adonis function. This should be relatively quick to rerun and will make the effect of host vs. ecology vs. diet easier to parse.

c) The OTU picking strategy can introduce biases when sampling microbes that are better represented in the reference taxonomy since more of the sequences will be classified in one sample versus another. Even though the authors seem to have chosen the best option for this dataset, there very well could be differences given that comparisons are explicitly between Industrialized vs Non-Industrialized populations (there tends to be lower read mapping to closed ref OTUs in non-industrial populations) as well as human-associated vs. wild animals (it would be expected that lab animals and livestock microbiomes have been better characterized back when that GreenGenes taxonomy was created).

Can a Supplementary file be added that lists the proportion of reads classified per sample? Are there differences in the number of reads that classify between the major comparisons in this paper (Industrialized vs. Non-Industrialized, Wild vs. Domestic, etc.)? If there are, then reprocessing of these reads either with an open OTU calling method or ASV method should be implemented.

d) How does microbial load/density vary based on gut passage rates, and could this be influencing your results?

https://doi.org/10.7554/eLife.60197.sa1

Author response

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that we cannot pursue the publication of your manuscript in its present form. However, we consider that the work is very interesting and therefore a new version that thoroughly addresses the concerns raised would likely be reviewed again. Although it would be treated as a new submission, we would aim to retain an overlapping set of reviewers.

The work provides exciting and valuable information on the possible effects of domestication and industrialization on the gut microbiome. However, there were several methodological issues raised, such as host genotype determination, control for genetic distance and, in particular, concerns regarding data analyses (diversity metrics, OTU picking, error rates, Permanova, and FDR correction, to name some) that can take a considerable amount of time to perform. There were also misgivings regarding the validity of some of the conclusions based on the data presented. These include group comparisons that do not necessarily agree with the idea that domestication and industrialization similarly impact the gut microbiota, and the effect of host genotype or genetic distance on the microbiota. Please also take into account the comments about differences with respect to previous publications regarding the claim that domestic animals may be useful as models.

The author and editorial comments were very helpful in improving our manuscript. We provide a point by point response below but have targeted our efforts towards those specified above including clarifying genotype grouping, adding new relatedness-informed analyses, and expanding/altering the methods to address the statistical and data processing concerns. We have also updated the human group analysis with a new set of publicly available data that better matches our sequencing methods, as well as new data on captive chimpanzees. Our updated results confirm or clarify our previous findings, and we think solidify our overall result that domestication and industrialization similarly influence the gut microbiota.

Reviewer #1:

This study examines effects of domestication on the gut microbiome of wild animals to the effect of industrialization on the gut microbiota of humans. They report consistent shifts in composition of gut microbiota in domestic animals and in humans from industrialized (but not from traditional) societies. They also perform cross-feeding experiments of wild and domesticated animals (lab mice/wild mice; dogs/wolves) and report that apart from genetics, diet plays a dominant role in shaping (loss of diversity) of the domestic gut microbiota.

I have the following comments:

1) Introduction: "genetic changes under domestication" How did the authors control for differences in genetic distance among the individual domesticated/wild animal pairs? Are the shifts in composition of microbiota during domestication and industrialization still consistent if controlled for genetic distance?

In response to this comment and similar queries from reviewer 2 and reviewer 3, we have added additional analyses of genetic effects (i.e. time since divergence and time since domestication) to the cross-species comparison, and have altered our discussion of genetics in the diet experiments. Because we do not have individual genotype data for the animals included in the cross-species comparison, we cannot explicitly include a correction for relatedness to our β-diversity analyses (i.e. the PERMANOVA tests and permutation test for F). However, we have added a Mantel test result describing the relationship between β-diversity and relatedness as well as additional analyses assessing the effects of relatedness to dyad-level trends. See the Results.

“Domestication of mammalian species occurred at different times, so the evolutionary relationships between members of a host dyad, even in cases where we sampled the known progenitors, are not all equal. […] Moreover, differences in position along NMDS axis 1 remained significant even when correcting for host dyad and divergence time (P<0.001, likelihood test linear mixed effects models).”

2) Introduction: "Finally, the convergent nature of many ecological shifts experienced by domesticated animals and industrialized human populations suggests that domestic animals may provide a uniquely useful model for studying the microbially-mediated health impacts of rapid environmental change." and Discussion "their translational potential as models for studying the gut microbiota of industrialized populations may be greater than is currently appreciated." This statement is not clear – please clarify in the context of two publications (Nature. 2016;532(7600):512-6 and Science, 708 2019;365(6452):eaaw4361) that appear to state the opposite.

We appreciate the confusion these statements may have caused. To remedy and clarify, we have adjusted our discussion of animal models in both the Introduction and the Discussion. In the Introduction, we now emphasize the relevance of animal models to studies of ecological change rather than human health/biology specifically.

“Finally, the convergent nature of many ecological shifts experienced by domesticated animals and industrialized human populations suggests that domesticated animals may provide a uniquely useful model for studying the microbially-mediated health impacts of rapid environmental change resulting in mismatch between host, microbiota and/or environment, a situation thought to apply to humans in industrialized settings (32).”

In the Discussion, we discuss more extensively why lab animals may or may not be good models for humans. We present the evidence that mice are poor models for humans, including the dirty mouse references the reviewer recommends, and speculate that they may be better models for other aspects of human biology.

“Because laboratory animals demonstrate some of the largest overall differences relative to their wild counterparts, they might be expected to have high translational potential as models for studying the gut microbiota of industrialized human populations. […] Certainly, studies of non-domesticated animals will be necessary to understand natural host-microbe interactions and their evolutionary history (48, 54), as well as to determine the most appropriate models for translational research.”

3) The authors describe that diet plays a major role in changing the microbiota of wild animals to those of domestic ones. Diet is a great source of viruses. To which extend is the introduction or loss of viruses (in particular phages) responsible for the shift in gut microbiota?

This is an important and interesting point. We have added a paragraph to the Discussion outlining potential mechanisms for diet effects on the microbiome, including virus impacts.

“Our reciprocal diet experiments in mice and canids confirm that ecology plays a predominant role in shaping the domesticated gut microbiota. […] That the gut microbial impacts of change in a single ecological variable like diet were sufficiently profound to outweigh those of host taxon identity suggests that suites of ecological variables changing together, such as during domestication or industrialization, may have collectively exerted an even larger influence (36).”

4) The authors state repeatedly that wild animals more diverse microbiota. Are there uniform changes in taxa? Are some taxa lost, and if so, is this observed in several wild / domestic pairs?

We found that there are not consistent effects of domestication on α-diversity in the gut microbiota (see Figure 1—figure supplement 4D, E). On a host level, some wild progenitors have higher diversity (e.g. mice; Figure 1—figure supplement 2C), whereas some domesticated species have higher diversity than their wild progenitors (e.g. wolves; Figure 1—figure supplement 4C). We discuss some of potential mechanisms underlying this in the text. In light of this complexity, we do not necessarily see a lot of consistent differences in microbial relative abundance between wild and domestic taxa in our cross-species comparison. Some of the microbial taxon responses are shown in Figure 5—figure supplement 1 and are discussed in light of trends observed in industrialized versus non-industrialized human populations.

Reviewer #2:

This is an exciting paper with important implications for how diet and environment interact to shape the composition and diversity of the gut microbiome. The results are interesting – particularly the results of the robustly designed diet challenge experiments. Concluding with the host-microbe-environment mismatch puzzle is thought-provoking. I am slightly concerned about how the framing of the study is phrased. I additionally have some questions/suggestions/concerns regarding the methods.

1) The results of this study are super interesting, but the authors need to be sure to make it very clear throughout that they are examining how environmental and dietary shifts associated with domestication may parallel environmental and dietary changes in some human populations (not that some human populations are domesticated and some are wild). The authors are mindful to make this clear most of the time, but it would be good to make it explicit all of the time.

In addition, I would ask that the authors carefully consider how human populations are described – traditional is not the best term, unless it is how those populations self-identify. There are real and very important ecological differences that distinguish the human populations that were sampled. Using language that somehow indicates what those differences are might be more impactful than using industrialized vs. traditional. Or, at the very least, clearly defining those terms early on in the article is necessary. Industrialized vs. non-industrialized or traditional can be read as placing as elevating either group or could be read as saying the populations are "advanced" and "not advanced" (particularly important as this paper will likely generate some media attention).

We appreciate the reviewer’s recommendation for nuance and clarity and have tried to improve treatment throughout the text. In particular we have altered our introductory presentation of different human populations (see the Introduction) and now refer to industrialized and non-industrialized throughout (rather than traditional) to keep in line with other recent work on variation in the human gut microbiota.

“Industrialized, agrarian, and foraging human populations differ along numerous ecological dimensions, including diet, physical activity, the size and nature of social networks, pathogen exposure, types and intensities of medical intervention, and altered reproductive patterns. Such changes have resulted in large shifts in the gut microbiota in industrialized populations relative to non-industrialized populations or closely related primates (1-4), including reductions in alpha-diversity and changes in composition that have been implicated in the rise of various metabolic and immunological diseases (5-7).”

2) I am wondering if it is better to categorize the genotype/diet experiment as a provenance/diet experiment or something similar. As the authors did not actually look at host genetics in the wild-caught mice, they don't know how genetically distinct they are and there is certainly variation in genetic distance from the lab mice within the group of wild-caught mice.

The reviewer is correct that we did not have individual genotype data for these individuals and thus the genotype/diet nomenclature could be confusing. As such, we have changed our presentation of these experiments to be tests of the effects of host taxon and diet. The abbreviations used in the text and figures now refer to WildH/WildD etc.

3) Genetic changes kind of come up unexpectedly and without context the Introduction, which I found unclear. It may be better to focus this paragraph solely on ecological/environmental shifts? I was also a little confused if the authors were indicating the known genetic changes caused by domestication would change something about host physiology that would impact the gut microbiome somehow, or if the effect of divergence in host genetics would cause a simultaneous divergence in gut microbiome composition, or both.

In an effort to clarify this section and reduce confusion, we have edited this text, it reads:

“Many of the altered ecological features experienced by industrialized humans and domesticated animals have been independently observed to impact the gut microbiota, including diet (9, 10), physical activity (11, 12), the size and nature of social networks (13, 14), antibiotic use (15, 16), and changes in birthing and lactation practices (15, 17). These effects are often found to match or exceed the effects of genetic variation—also introduced by domestication—on gut microbiota composition (10). As such, ecological shifts under domestication might be expected to lead to microbial differentiation between domesticated animals and their wild counterparts…”

4) Gomez et al., 2019 and Amato et al., 2019 both found that the human gut microbiota is actually closer to that of baboons than chimpanzees. I don't think the authors necessarily need to add baboons to the analysis, but it would be relevant to acknowledge in the discussion that chimps may or may not be the best comparison for humans.

We are unable to add baboons to the analysis at this time because we did not have access to such samples and sequences from baboons were not included in Jha et al., the reference we used for comparative analyses. However, we do now refer specifically to these findings: (Discussion).

“While we limited our analysis to human-chimpanzee comparisons because Pan is our closest sister clade to Homo, recent work has shown that the human gut microbiota is more similar to that of baboons (50, 51). Baboons are more distantly related to humans but have been argued to be more similar in terms of diet and dietary physiology (52, 53), accentuating our finding of the importance of ecological factors in shaping the microbiota. Further work will be required to illuminate the specific combination of ecological factors driving similarities between domesticated and industrialized gut microbial signatures.”

5) Introduction (and elsewhere): I don't think domestic can be used in place of domesticated – the meanings, to me at least, are distinct.

We appreciate the reviewer’s recommendation on this point and have replaced “domestic” with “domesticated” throughout.

6) Methodological concerns:

- Samples collected in RNAlater are not necessarily comparable to freshly frozen – please note in the methods which species were preserved with each method and describe how you accounted for this difference in preservation.

We have added text to the Materials and methods specifying which samples were preserved with RNAlater and (now in the case of the new human data) OMNIgene kits. We do not correct for differences in preservation in our analyses but have found that it does contribute substantially to the variation we observe. Whether analyzing all samples (including previously published human samples collected with the OMNIgene kits) or just the cross-species dataset, we find that preservative method is significant factor (PERMANOVA P<0.001) but with an R2=0.01. This result is outlined in the Materials and methods section describing preservation method.

- Why was closed-reference OTU-picking chosen? Open-reference OTU-picking is the recommended method, unless one is comparing amplicons from different regions of the 16S rRNA gene. I would suggest that analyzing the data using one of the ASV strategies (DADA2 or Deblur) is recommended, but also do not want to force the authors to reanalyze their entire dataset (and the newer ASV methods become less useful when including 454 data).

We are unfortunately not in a position to totally reanalyze the dataset at this point although we appreciate that there are potential benefits in using an ASV approach going forward. Here, we relied on closed-reference OTU-picking to limit batch effects between sequencing runs and preservative methods. Closed-reference OTU-picking is a common approach that is more conservative and thus unlikely to introduce spurious results when dealing with high variability datasets such as our cross-species comparison.

- Yatsunenko et al., 2012 used 454 sequencing – I am curious how the authors corrected for the differences in sequencing-related error rates between 454 and HiSeq? And why they did not choose to use available human datasets sequenced in a manner comparable to the newly produced dataset in this paper?

The reviewer is correct that the difference in sequencing platforms between the published data and our own is significant. With that in mind we have replaced our usage of data from Yatsunenko et al., with more recent data from Jha et al., 2018. This new data was produced with the same primers as we used for our data collection and was also sequenced on the Ilumina platform. To maximize the comparability of their data and ours, we reprocessed the raw sequences from Jha et al. using our bioinformatics pipeline. There are limitations to this dataset as well, of course—in particular some of the samples were collected with OMNIgene kits and stored in preservative before DNA extraction—but we include new text describing these.

- Using the adonis2 function in vegan would allow the authors to use marginal sums of squares in the PERMANOVA analysis – this might allow them to better tease apart which factors are accounting for what proportion of the variance in the dataset.

Thank you for the recommendation. We have replaced all PERMANOVA results with output from the adonis2 function rather than the adonis function.

- A Bonferroni correction is quite conservative for microbiome datasets – FDR correction could be used instead.

The reviewer is right that Bonferroni is quite conservative. However, in this case, FDR and Bonferroni corrections did not produce substantially different outcomes so we have erred on the side of conservatism by retaining the Bonferroni correction throughout.

- I would like to see an explanation for the choice of method to measure the magnitude of change in β-diversity, as it is one I haven't seen before and measuring change along an axis that does not have an easily interpretable meaning might not be the most informative. Alternatively, comparing pairwise unweighted and weighted UniFrac between domesticated/wild and baseline/treatment and/or performing a Procrustes analysis may be preferred.

We have added text to the methods and results describing the utility of this metric for analyzing change in microbial community composition. It is comparable to an analysis employed by Ang et al., 2020 highlighted in Figure 1C. We have retained this analysis, despite its relative novelty, because it allows us to consider the direction of change in a way that comparing pairwise dissimilarity measurements does not. However, additional analyses on such measurements have also been added for the cross-species comparison (i.e., Figure 1D).

Reviewer #3:

Reese et al., compare the microbiota of domestic animals and their closest wild counterparts, including a comparison of humans and chimpanzee microbiotas. They report similar changes to the microbiota from domestication and industrialization. Overall, the data presented is fairly noisy and many of the conclusions seem overstated given slight differences between groups. Even if we set aside the issues with the data, which are not trivial, it is unclear how important the conclusions are. For example, the last sentence of the Abstract:

"We conclude that domestication and industrialization have similarly impacted the gut microbiota, emphasizing the utility of domestic animal models and diets for understanding host-microbial interactions in rapidly changing environments, and the importance of studying non-industrialized human populations for understanding aspects of human health dependent on host-microbial co-evolution."

Not so easy to unravel the point(s) the authors are trying to make. The last passage is already very clear to the field, non-industrialised populations are important to study. The first part suggests that domestic animals and diets are useful in understanding the microbiota in changing environments. It is not clear exactly what this statement is trying to convey and it requires some clarification.

Throughout the revised manuscript, we have attempted to be more concrete about what can and cannot be concluded from our data. In addition, we have edited the final sentence of the Abstract to improve clarity, it reads:

“Our findings emphasize the utility of domesticated animal models for understanding hostmicrobial interactions in rapidly changing environments, while highlighting the limitations of animal models and the importance of studying non-industrialized human populations for understanding aspects of biology dependent on host-microbial co-evolution.”

In the Abstract the authors state that "domestication and industrialization have similarly impacted the gut microbiota". A major concern is the data presented in Figure 5B for two reasons. First, the difference between two industrialized human populations appears to be larger than that observed between domestic and wild animals. Second, the shift to the left from industrialized humans to traditional humans is larger than from industrialized humans to chimpanzees. Not only is this problematic from the standpoint of implications about the "wildness" of traditional populations, but also difficult to interpret given the greater similarity in genetics, physiology, lifestyle, and diet between human populations than chimpanzees and humans.

We sincerely thank the reviewer for their sensitivity to potential (though unintended) interpretations. In response to this reviewer’s concerns and those of reviewer 2, we have updated our analyses on different human populations. Using a more recent dataset, with samples prepared in a manner more in line with our own methods, we have found generally the same results, with trends between industrialized populations and chimps paralleling those between domesticated and wild animals (see updated Figure 5). We now find that the non-industrialized populations (now a group of foragers from Nepal and a group of hunter-gatherers from Tanzania) do not differ in the opposite direction but instead are just in line with the chimpanzees. Furthermore, we do not observe a large difference between the two industrialized human populations. This updated analyses should address the concerns the reviewer outlines above. In addition, to help guard against readers inferring any inappropriate distinctions between industrialized and non-industrialized populations, we have made revisions throughout the text to emphasize that these populations are equally human and that their differences are primarily rooted in different ecologies.

The authors report greater between species variability in wild gut communities than domesticated. However, it does not look like they did this comparison for the human and chimpanzee data. Given published data showing that the between individual variability in the microbiota of industrial individual is larger than that of traditional population microbiota, it would be interesting to see how these data compare to that of chimpanzees given that this is not the result you would expect given the data from the other animal pairs.

We appreciate this recommendation and have added a variability analysis for the human/chimpanzee comparison using the new data that we produced here (7 wild chimpanzees and 7 US human adults). For these samples at least, we do not find a significant difference in variability. The result is described in the text:

“We also found a marginal difference in between-conspecific variability in the gut microbiota of humans and chimpanzees (P=0.092, F=3.0987; permutation test for F). “

It is not clear how α diversity was calculated. Was the data rarefied and if so to how many reads and were the samples sequenced sufficiently deep to ensure an accurate measurement of diversity.

We apologize for any confusion around our methods. We have added greater detail on how α-diversity was calculated in the Materials and methods and now include rarefaction read depths for each dataset:

“Alpha-diversity (Shannon index, OTU richness) were calculated for rarefied OTU tables (rarefaction limit of 17,500 for cross-species dataset; 27,000 for wild mouse study; 15,500 for the mouse colonization study; 7,500 for canid experiment).”

Subsection “Diet vs. genotype effects on domestic gut microbial composition in mice”. "Domestication has profound effects on both ecology and host genotype." Do the authors mean "has had", ie, there is evidence that animals, when domesticated, show genotypic changes, eg, new traits are selected for. Domestication over short time periods may have little effect on genotype.

This is a fair point. We have edited the sentence to read: “Domestication has had profound effects on both ecology and host genotype.”

Subsection “Diet vs. genotype effects on domestic gut microbial composition in mice”. "we found that host genotype explained the largest amount of variation" It is unclear what data the authors are examining to reach this conclusion. The species appears to be Mus musculus for these analyses. Are the authors performing a host genotype (eg, SNP) analysis? Please clarify how differences in host genotype are being determined.

The reviewer is correct that we did not collect individual host genotyping for the mice or canids. In response to this comment and one from reviewer 2 (R2C2), we have altered our presentation of these experiments to be comparisons of the effects of host-taxon and diet rather than genotype and effect.

Figure 2.

- It is very difficult to draw conclusions from Figure 2B. Suggest that the authors show centroids or find a better way to represent the data. Some of the colors are too similar as well, so difficult to differentiate. Why are DomG/DomD points moving on the PCA plot? Same with WildG/WildD? Perhaps this data could reveal drift of the microbiome composition in the absence of intervention, which may inform whether their diet shift in the other groups is meaningful.

Figure 4 has many of the same issues described for Figure 2. It's very difficult to interpret these panels with so many points going in different direction and minimal color differences between some of the points.

We thank the reviewer for highlighting the issues with clarity. We have now reformatted the figures to better highlight the patterns of interest, e.g. to more clearly demonstrate the baseline and endpoint points. It is typical for control animals to show some drift over the course of an experiment due both to natural biological turnover and sequencing noise. This is likely what we are observing in the DomH/DomD and WildH/WildD animals. However, we do not find that this variation is directional (e.g. control individuals are not moving in a consistent direction and thus do not have a shift significantly different from 0), unlike the patterns observed in the reciprocal diet treatment groups. We have tried to clarify this finding in the results and have redrawn the plots to more clearly demonstrate the baseline and endpoint points for each individual.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Essential revisions:

1) Reviewers were concerned with the comparisons between domestication and industrialization and the subsequent conclusions. This aspect of the work needs to be improved for clarity and the claims toned down as they are not fully supported by the data presented.

a) The authors should note that domestication, which has taken a long time, and industrialization, a fairly recent change to our ecology, are different processes. Therefore, the direct comparisons in the manuscript do not seem entirely appropriate and should be more carefully addressed. In particular, the data does not provide strong evidence to support the claim that animal domestication and human industrialization result in similar effects on their hosts microbiome, even though this conclusion may be correct, since it makes sense given that many ecological processes are probably affected in similar ways. This conclusion should therefore be toned down to agree with their data.

We now explicitly state in the Abstract and first paragraph of the Introduction that domestication and industrialization are fundamentally distinct processes.

“Although industrialization and domestication are fundamentally different processes, the ecological parallels between human industrialization and animal domestication suggest that the gut microbiota of diverse domesticated animals may differ in consistent ways from those of their wild progenitors, and further, that their differences may resemble those observed between industrialized and non-industrialized human populations.”

In this manuscript, our dual consideration of domestication and industrialization is solely to test for analogous effects. We have now endeavored to clarify this rationale throughout the manuscript. For example, the title of the section on the comparisons is now “Analogous pressures in the human gut microbiota”. Some other modifications of the text that emphasize the distinction between the processes of domestication and industrialization include:

“While we focus primarily on the impacts of domestication on the mammalian gut microbiota, we include analyses of industrialized and non-industrialized human populations because much is known about the effects of industrialization on the gut microbiota and as such it can serve as a benchmark ecological process for domestication. In addition, to explore the extent to which deeper evolutionary history affects these patterns, we also compare humans to chimpanzees (Pan troglodytes), one of our two closest living relatives and arguably the better referential model for the last common ancestor between Pan and Homo (23).”

“Finally, the convergent nature of many ecological shifts experienced by domesticated animals and industrialized human populations suggests that domesticated animals may provide a uniquely useful model for studying the microbially-mediated health impacts of rapid environmental change that results in mismatch between host, microbiota and/or environment, a situation thought to apply to humans in industrialized settings (36). Understanding what shapes the domesticated microbiota may therefore identify routes to improve experimental models, animal condition, and human health.”

“We next explored the extent to which humans harbor gut microbial signatures analogous to those of domestication. […] We first compared samples that we collected from industrialized humans and wild chimpanzees, finding that the gut microbial communities of these humans and chimpanzees exhibited differences that paralleled those observed between domesticated animals and their wild counterparts when compared in the same ordination space (P<0.001, Mann-Whitney U test; Figure 5A, 5B).”

“We observed some correspondence between the gut microbial signatures of animal domestication and human industrialization that is most likely attributable to convergent ecological changes. The observation that gut microbial divergence among Pan and Homo primarily affects industrialized populations specifically implicates recent ecological changes as opposed to either ecological changes with deeper roots in human evolution or host evolutionary changes.”

b) Is there a way to incorporate data from populations that use subsistence strategies involving domestication, but are not Industrialized (the other populations in Jha et al., even)? It could be expected that the agricultural or pastoral but non-Industrialized countries would be somewhat intermediate in their microbiome composition, as they experience the factors of domestication without some of the extreme ecological consequences of Industrialization (antibiotics, highly processed foods, etc.). Is this the case?

We appreciate this recommendation to test for industrialization effects in a more nuanced way. Following the reviewers’ suggestion, we now include a subsample of individuals from all of the Jha et al. Nepalese populations in addition to their Hadza and US populations. Notably, we added samples from two populations transitioning to subsistence farming and a population that transitioned to subsistence farming within the past two centuries. We concur with the reviewers that these additional populations could in theory experience aspects of ecology more similar to the industrialized/domesticated environment, such as exposure to domesticated animals and diets higher in grains, and thus could be intermediate in their microbiota composition. However, we found that all three populations clustered with the traditional populations previously included (Hadza and Chepang from Nepal) and that they did not display the ordination shifts associated with domestication. The text and figures (Figure 5 and Figure 5—figure supplement 1) have been updated to reflect this finding and we further emphasize that extreme ecological shifts associated with industrialization and domestication are likely the drivers of the shared pattern.

“We observed some correspondence between the gut microbial signatures of animal domestication and human industrialization that is most likely attributable to convergent ecological changes. […] These factors would be absent even in populations currently undergoing the transition from subsistence to industrialized lifestyles (39), but may overlap with changes experienced by domesticated animals in their diets, habitats, and social milieu.”

c) A more nuanced discussion should occur at some point in the manuscript on the choice and caveats of using highly Industrialized populations in this comparison given that the process being compared is domestication and not industrialization.

Throughout the manuscript, we have made changes to emphasize that our primary focus is on domestication and not industrialization, including the following prominent statement in the Introduction:

“Here, we assess the effects of domestication on the mammalian gut microbiota, perform controlled dietary experiments that attempt to distinguish between the relative roles of ecology and genetics in driving these patterns, and compare the effects of domestication to those of human industrialization. While we focus primarily on the impacts of domestication on the mammalian gut microbiota, we include analyses of industrialized and non-industrialized human populations because much is known about the effects of industrialization on the gut microbiota and as such it can serve as a benchmark ecological process for domestication.”

We have also added text to the Materials and methods describing our choice of populations and our intent in analyzing industrialization as an ecological analog to domestication:

“These populations represent extremes of industrialized and nonindustrialized human lifestyles with the variation among the non-industrialized groups not covering the full breadth of intermediate lifestyles (e.g. modern agricultural or recent urban transplants). We believe these extremes enable us to test the how human gut microbial communities respond to major ecological change of a magnitude that could be argued to approximate that experienced by gut microbial communities of animals undergoing domestication.”

2) The revised manuscript has improved but still lacks clarity in many places and uses language that is vague and often misleading, making it difficult to understand what the authors are trying to say. The entire text should therefore be checked and improved to make the language more precise.

a) In the Abstract, for example, it is not clear what shifts the authors refer to, what is meant by microbiomes to be impacted “'similarly”, and what “parallel ecological changes” are. It can be argued that the ecological changes are quite different in industrialized humans and domesticated animals (housing, hygiene, diet, etc.). However, the ecological processes that impacted their microbiomes, and the compositional alterations, might have been similar.

We have attempted to clarify the language in the Abstract, now specifying some of the types of ecological changes considered and noting that there were analogous, not identical, changes between domestication and industrialization. “Domesticated animals experienced profound changes in diet, environment, and social interactions that likely shaped their gut microbiota and were potentially analogous to ecological changes experienced by humans during industrialization… Although fundamentally different processes, we conclude that domestication and industrialization have impacted the gut microbiota in related ways, likely through shared ecological change.”

b) This vagueness is also found through the entire manuscript. What are ecological parallels (Introduction)? What is a "suite of shared ecological changes" (Introduction)? Which “evolutionary forces” were studied? What do the authors mean by "individual shifts"? (figure legend of Figure 1C). Compositional shifts in an individual? Was that even assessed?

In addition to altering the abstract, we have made edits throughout the manuscript in an attempt to minimize vagueness and increase reader comprehension. For instance, we have removed the term “ecological parallels” referenced above, and we now instead specify the types of changes expected to be common to both industrialized and domesticated populations.

“Industrialized, agrarian, and foraging human populations differ along numerous ecological dimensions, including diet, physical activity, the size and nature of social networks, pathogen exposure, types and intensities of medical intervention, and reproductive patterns. […] Although industrialization and domestication are fundamentally different processes, the ecological parallels between human industrialization and animal domestication suggest that the gut microbiota of diverse domesticated animals may differ in consistent ways from those of their wild progenitors, and further, that their differences may resemble those observed between industrialized and non-industrialized human populations.”

We have removed “suite of shared ecological changes” referenced above. The sentence now reads:

“Apart from the pressures of ecological change that domestic animals experience in human environments, animal domestication has also entailed strong artificial selection for phenotypes desirable to humans, such as rapid growth and docility in agricultural animals, reliable reproduction and stress resistance in laboratory animals, and unique physical and/or behavioral attributes in companion animals.”

We have replaced “explore the ecological and evolutionary forces” referenced above. The sentence now reads:

“Here, we assess the effects of domestication on the mammalian gut microbiota, perform controlled dietary experiments that attempt to distinguish between the relative roles of ecology and genetics in driving these patterns, and compare the effects of domestication to those of human industrialization. While we focus primarily on the impacts of domestication on the mammalian gut microbiota, we include analyses of industrialized and non-industrialized human populations because much is known about the effects of industrialization on the gut microbiota and as such it can serve as a benchmark ecological process for domestication.”

We have also sought to more clearly define the ordination shift variable (both in the text and in the figure legends) and articulate why it was included. The point of first introduction of this variable in the text and the relevant section of the Figure 1 legend are copied below.

“To determine whether there was a consistent change in microbial composition with domestication, we calculated the difference between an individual’s ordination coordinates and the average ordination coordinates of its host dyad along the first nonmetric multidimensional scaling (NMDS) axis. Quantifying this ordination shift allowed us to consider overall changes in composition while correcting for host dyad and retaining information on the directionality of changes.”

Figure 1 legend: “Fig. 1. The mammalian gut microbiota carries a global signature of domestication… (C) Distance to dyad (color) mean along Bray-Curtis ordination NMDS axis 1 differs by domestication status (P=0.006, Mann-Whitney U test).”

c) The term “shifts” is used inappropriately throughout the manuscript. For example, what are "shifts between industrialized humans and wild chimpanzees" (Figure 5 legend)? The microbiome does not really shift from a human to a chimpanzee. Do the authors refer to differences between microbiomes in different hosts?

Throughout the manuscript, we have replaced the term “shift” with the phrase “ordination shift” to clarify that we are not considering ties in real world, just in analytical space. The term shift is frequently used to describe such ordination-based findings (e.g. see McKenzie et al., 2017 and Ang et al., 2020), but we believe that stipulating “ordination shift” will help clarify and alleviate reader confusion.

To further clarify, we have also added greater detail to the Materials and methods about how this metric is calculated and analyzed – specifically, where individuals are being tracked we include individual ID in our analyses and elsewhere we analyze by group:

“To determine the consistency of gut microbial differences across ordination space due to domestication, Pan-Homo divergence, or industrialization in the observational study, we calculated the average position of the host dyad (e.g., pig/boar) or all primates (humans and chimpanzees) for axis 1 of the NMDS then measured the displacement along each axis for an individual sample relative to that mean position. We tested for differences in these ordination shifts by domestication status or primate host taxonomy (e.g. chimpanzee versus US human). To estimate the direction and magnitude of changes in beta-diversity during the experimental studies, we tested whether inclusion of a treatment group term improved the performance of a linear mixed effects model relative to a model with only time and animal ID terms for predicting the NMDS1 axis value for an individual. These analyses allowed us to consider the direction of betadiversity changes in addition to the magnitude.”

References for similar ordination analyses: Ang et al., 2020; McKenzie et al., 2017.

3) The authors should be careful with the way they present their results to avoid biased interpretation and make claims that are clearly supported by their results.

a) It sometimes seems as if the authors have interpreted the findings to fit a preconceived idea of the findings. For example, the authors conclude a "consistent effect of domestication status" (Results), but the samples cluster by host, which has the highest effect sizes. The conclusion is then mainly based on a statistical analysis that showed domesticated samples to be "further right" on an NMDs axis. This is not very convincing, and not very clear in Figure 1C either.

We have attempted to address the reviewer concerns and remove the impression we are overinterpreting our findings. Specifically, we now start the presentation of our cross-species comparison results by stipulating we don’t find a convergent domesticated microbiota before noting we do see a global signal (not “consistent effects”) of domestication. We have also added PERMANOVA analyses of individual dyads (similar to analyses performed in McKenzie et al., 2017) that substantiate our claim that there are widespread (albeit not universal) effects of domestication.

“Despite observing no single convergent “domesticated microbiota” profile, our analysis detected a global signal of domestication status on gut microbiota composition. Across the combined dataset, the factor that explained the largest proportion of variation was the host dyad (e.g., pig/boar; P<0.001, R2=0.39, F=17.086, PERMANOVA; Figure 1B). However, correcting for host dyad, domestication status also contributed significantly to variation in microbial communities (P<0.001, R2=0.15, F=6.081), and these results were robust to the distance metric analyzed (Supplementary File 1). Furthermore, analyses of individual dyads found a significant effect of domestication status for all groups except canids (P<0.05, R2=0.18-0.41, PERMANOVAs; Supplementary File 1).”

We have moved the dyad-by-dyad breakdown of the ordination shift analysis to the supplement, and now present just the domesticated versus wild ordination shift analysis in the main text and figures to illustrate the overall effect. In addition, we have added discussion of some of the potential reasons for dyads not displaying identical trends when analyzed individually.

“Cases where domestication effects are weaker in our comparative study generally consist of animals where the ecological change associated with domestication has been small—e.g. sheep and pigs, whose diets may be quite similar to their wild progenitors, at least when kept in the non-industrialized agricultural settings that were sampled (9)—or where ecological changes are in the opposite direction from the domesticated norm—e.g. canids, where the domesticate diet typically involves lower protein and higher carbohydrate levels than wild diets, instead of the higher protein levels seen in most laboratory or farm animals (47).”

b) Another claim is that in Figure 2, differences between domesticated and wild mice can be overcome by a diet switch, but looking at Figure 2—figure supplement 2, that is simply not the case. It is difficult to see how the data in Figure 5 provides strong evidence that the effects of domestication and industrialization are similar.

We have addressed this overstatement by rephrasing the Figure 2 legend title to state that “Gut microbial differences between wild and domesticated mice can be partially overcome by diet swap.” We also made similar changes to the Figure 4 legend title, which now reads “Microbial differences between wild and domesticated canids can be partially overcome by diet shifts.” Additional explanation for changes made to address concerns that the effects of domestication and industrialization are overstated can be found in our response to 1a-c.

4) More clarification is needed for wild and domestic microbiome results and subsequent conclusions

a) The results presented (Results and Figure 1) do not seem to support the conclusion that domestication is shifting all species to the right along NMDS1. The magnitude and direction of shift seems to differ based on host species. While the general trend of all species lumped together is to the right, sheep and pigs don't seem to follow the pattern (and some others don't seem to have a strong shift to the right). What are the effect sizes for the Mann-Whitney U tests here? Also, looking at Figure 1—figure supplement 2A, only the companion species are denoted as having a p<0.05, which seems at odds with the statement in the Results.

This species-dependent direction and strength of shift is not entirely unexpected based on previous work. Shifting host ecology (diet or captivity) has previously been shown to differentially effect host species: Amato et al., 2015 and McKenzie et al., 2017.

The inconsistency in the direction of the shift might not actually negate the broader point, that domestication at times has effects on the gut microbiome that are very similar to the shift we see between industrialized and non-industrialized humans. In fact, it might be instructive to point out what specific species might be good models for the shift we see in humans – what are the specific ecological shifts with domestication in those species and how does that mirror the ecological shift with industrialization in humans?

As noted above, we have made language changes throughout the manuscript to tone down our claims and better clarify our conclusions. In particular, we added separate analyses of wild-domesticated dyads to support the finding of a signal of domestication in the microbiota of diverse lineages of mammals, and have added text discussing potential explanations for variation in the strength of this signal between lineages. For additional detail, please see response to item 3a, above.

The reviewers make a good point that leveraging differences in host species ecology could be a powerful way to understand which ecological aspects domestication and industrialization are most likely to have driven the parallel signatures observed. We will certainly consider such an analysis in the future, but we believe it is beyond the scope of the present manuscript, which serves to launch this line of inquiry by pointing out the existence of a global effect of domestication on the gut microbiota and its parallels with those of industrialization.

b) What does it look like when you put the results of the mouse and canid experiments in the same ordination space with your wild/domestic and chimp/human pairs? Is the shift in the expected direction? When looking at the results of the mouse experiment and the canid experiments on their own, we see a shift to the left with experimental domestication (ie, for the Wildh/Domd treatments), but this might be a function of the ordination space?

While we appreciate the suggestion for combining the comparative and experimental analyses to assess generalizability, we have chosen not to include this analysis in the revised document. A strong signal of species and sequencing run, typical in PERMANOVA analyses and ordination plots of microbiome data, is expected a priori to swamp the experimental domestication signal. The interpretation of left/right shifts is not inherently meaningful as these axes could be mirrored and still present the same data, as such the leftward shift in the experiments does not inherently conflict with the rightward shift observed in the comparative data. We retain these ordination shift analyses because they are inherently more informative than those of dissimilarity (which is directionless) and because the existence of changes along the first axis supports the claim that the effect is meaningful (since the lower the number of the axis, the greater the degree of variation captured). We continue to use the language of left/right in the text to help guide readers in viewing and interpreting the plots, and have provided greater detail concerning the analysis in the Materials and methods to help readers better understand the finding.

c) Were any of the animals, either wild or domestic, from the same family, field, pen, etc.? Cohousing results in convergent microbiome profiles across a number of species due to horizontal microbial exchange. If conspecifics were collected from the same living situation or were related, one might expect higher microbiome sharing on those grounds alone. This potential confounder could explain the high similarity between the conspecifics. These details should be added to the Materials and methods. If this is an issue, it should be corrected for in statistical comparisons (if possible).

Geographic closeness and, especially, cohousing are known to impact the gut microbiota and the reviewers are correct that this may be playing a role here. Unfortunately, due to limits on sample availability, we only have one domestic species where animals were sampled at more than one location. We do have multiple species sampled at the same location although these individuals were never cohoused (or kept in the same pen/field in the case of agricultural animals). To address the concerns of the reviewers, we have added an analysis of location and now include an ordination plot of the comparative data colored by collection locale (Figure S1C).

“We have limited ability to distinguish between locale and species effects since all but one species (sheep) had samples collected from only one locale. Some locales had multiple species present and we do find a significant effect of locale on overall microbial community composition even when correcting for host phylogeny effects (P<0.001, R2=0.16, F=6.14, PERMANOVA). However, it is clear that locale does not necessarily lead to convergent microbiota across taxa as evidenced by the low clustering by site in NMDS ordination space (Figure 1-figure supplement 1). When analyzing just the sheep samples, we find a minimal effect of locale (P=0.023, R2=0.07, F=2.08, PERMANOVA). ”

5) Technical concerns and data presentation

a) Figure 2 and Figure 4 are a difficult to interpret, because the lines used to indicate moving points are obscuring the points themselves in some cases. Would ellipses around the treatment groups in the NMDS plots be more informative than the moving points?

To improve the readability of Figure 2 and Figure 4, we have removed the lines indicating moving points and instead provide the ordination shift plots with individuals plotted separately. Hopefully this allows both to be read and interpreted more easily.

b) For the adonis2 function, to get the marginal sums of squares you need to include “by = "margin"” in the function call. Using adonis2 without specifying “by” is equivalent to using the older adonis function. This should be relatively quick to rerun and will make the effect of host vs. ecology vs. diet easier to parse.

We thank the reviewers for this clarification. We have updated these analyses and the results as reported in the manuscript. There were only minimal impacts of the altered function call, so no interpretations were affected by this update.

c) The OTU picking strategy can introduce biases when sampling microbes that are better represented in the reference taxonomy since more of the sequences will be classified in one sample versus another. Even though the authors seem to have chosen the best option for this dataset, there very well could be differences given that comparisons are explicitly between Industrialized vs Non-Industrialized populations (there tends to be lower read mapping to closed ref OTUs in non-industrial populations) as well as human-associated vs. wild animals (it would be expected that lab animals and livestock microbiomes have been better characterized back when that GreenGenes taxonomy was created).

Can a Supplementary file be added that lists the proportion of reads classified per sample? Are there differences in the number of reads that classify between the major comparisons in this paper (Industrialized vs. Non-Industrialized, Wild vs. Domestic, etc.)? If there are, then reprocessing of these reads either with an open OTU calling method or ASV method should be implemented.

This is a valuable point, and we thank the reviewers for raising it. As requested, we have added summary statistics for the proportion of reads classified per sample to Supplementary file 2. We have also tested whether variation in the proportion of reads classified is biased in a manner that might impact our results.

We do observe variation between wild and domesticated samples (P=0.045, Mann-Whitney U test). However, contrary to what reviewers (and we) initially expected, samples from wild animals had, on average, a larger proportion of reads mapped to the GreenGenes database compared with samples from domesticated animals (see Author response images 1 and 2). There was also variation observed among dyads (P<0.001, Kruskal-Wallis test) but, as far as we can tell, this variation was not patterned based on species ecology.

Author response image 1

Regarding the human analysis, we see variation in the proportion of reads classified across populations (P<0.001, Kruskal-Wallis test). However, industrialized and non-industrialized samples had similar proportions of reads mapped, overall. As expected, humans generally have more reads classified than animals, and notably, taxa present in the chimpanzee gut microbiota are substantially less well represented compared with those from all other groups.

Author response image 2

d) How does microbial load/density vary based on gut passage rates, and could this be influencing your results?

We don’t have the data to directly address passage rate here, although relationships between microbial density and transit time have been reported elsewhere in the literature (some citations included below). Broadly speaking, while transit time is likely tied to gut physiology and diet, it is not a static feature of the host. Thus, we did not feel it was appropriate to use published values to infer it and then test for a relationship here. However, to acknowledge this valid point, we now include reference to the possible effect of transit time on gut microbiota composition and load in the Discussion:

“Our reciprocal diet experiments in mice and canids substantiate our claim that ecology plays a predominant role in shaping the domesticated gut microbiota. However, they do not pinpoint the mechanism(s) for these effects. Variability in diet or other aspects of ecology and their concomitant effects on host physiology (e.g., passage rate) can alter microbial composition or abundance through changes in the selective landscape that microbes experience and changes in environmental exposure.”

References for microbial density and transit time: Kashyap et al., 2013 and Vandeputte et al., 2017.

https://doi.org/10.7554/eLife.60197.sa2

Article and author information

Author details

  1. Aspen T Reese

    1. Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, United States
    2. Society of Fellows, Harvard University, Cambridge, MA, United States
    Present address
    Ecology Behavior & Evolution Section, University of California San Diego, La Jolla, United States
    Contribution
    Conceptualization, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    For correspondence
    areese@ucsd.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-9004-9470
  2. Katia S Chadaideh

    Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, United States
    Contribution
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2251-1170
  3. Caroline E Diggins

    Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  4. Laura D Schell

    Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, United States
    Contribution
    Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9600-924X
  5. Mark Beckel

    Wildlife Science Center, Stacy, MN, United States
    Contribution
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  6. Peggy Callahan

    Wildlife Science Center, Stacy, MN, United States
    Contribution
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  7. Roberta Ryan

    Wildlife Science Center, Stacy, MN, United States
    Contribution
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared
  8. Melissa Emery Thompson

    Department of Anthropology, University of New Mexico, Albuquerque, NM, United States
    Contribution
    Funding acquisition, Investigation, Writing - review and editing
    Competing interests
    No competing interests declared
  9. Rachel N Carmody

    Department of Human Evolutionary Biology, Harvard University, Cambridge, MA, United States
    Contribution
    Conceptualization, Supervision, Funding acquisition, Visualization, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    carmody@fas.harvard.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7505-9646

Funding

National Institute on Aging (R01AG049395)

  • Melissa Emery Thompson
  • Rachel N Carmody

Harvard University (Dean's Competitive Fund for Promising Scholarship)

  • Rachel N Carmody

Harvard University (William F. Milton Fund)

  • Aspen T Reese

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank many collaborators for help in collecting wild and domesticated animal fecal samples, including Gwynne Durham and The Mountain School (cattle, pigs, sheep); Luina Greine Farm (alpaca); Steve Ditchkoff (wild boar); Jason Munshi-South and Matthew Combs (rats); Pedro Monterroso, Marisa Rodrigues, and Marketa Zimova (European rabbits); Margaret Gruen and Kyle Smith (dogs, wolves); Kevin Monteith (Bighorn sheep); J Scott Weese (bison); Cristián Bonacic (vicuña); Bridget Alex, Hopi Hoekstra, Nicholas Holowka, Irene Li, Daniel Lieberman, Mark Omura, and Antonia Prescott (wild mice). For help in collecting and processing wild chimpanzee samples, we thank Tony Goldberg, Zarin Machanda, Martin Muller, Emily Otali, Leah Owens, Sarah Phillips-Garcia, Richard Wrangham, and the staff of Kibale Chimpanzee Project. For assistance in collecting captive chimpanzee samples, we thank the staff at Southwick’s Zoo, Mendon, MA. For assistance in carrying out experiments, we thank Cary Allen-Blevins, Rachel Berg, Andy Biewener, Meg Callahan-Beckel, Brian Hare, Kathleen Pritchett-Corning, Pedro Ramirez, and Emily Venable. For helpful comments on the manuscript, we thank Daniel Lieberman, Richard Wrangham, and members of the Carmody lab. 

Ethics

Human subjects: Human samples were self-collected by healthy study participants after providing written informed consent under the approval of the Harvard University IRB (protocol 17-1016).

Animal experimentation: All experiments were conducted in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. All mouse experiments were performed with protocols approved by the Harvard University Institutional Animal Care & Use Committee (protocol 17-11-315). All canid experimentation was approved by the WSC IACUC (protocol HAR-001). Wild chimpanzee fecal samples were collected by field assistants under the approval of the UNM IACUC (protocol 18-200739-MC) and with permission of the Uganda Wildlife Authority and Uganda National Council for Science and Technology. Fecal samples from other non-human mammals were collected from the ground following natural production; this approach precluded the need for institutional approval and was non-invasive for the animals.

Senior Editor

  1. Detlef Weigel, Max Planck Institute for Developmental Biology, Germany

Reviewing Editor

  1. María Mercedes Zambrano, CorpoGen, Colombia

Publication history

  1. Received: June 19, 2020
  2. Accepted: February 12, 2021
  3. Version of Record published: March 23, 2021 (version 1)

Copyright

© 2021, Reese et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,102
    Page views
  • 143
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

  1. Further reading

Further reading

    1. Ecology
    Corey JA Bradshaw et al.
    Research Article Updated

    The causes of Sahul’s megafauna extinctions remain uncertain, although several interacting factors were likely responsible. To examine the relative support for hypotheses regarding plausible ecological mechanisms underlying these extinctions, we constructed the first stochastic, age-structured models for 13 extinct megafauna species from five functional/taxonomic groups, as well as 8 extant species within these groups for comparison. Perturbing specific demographic rates individually, we tested which species were more demographically susceptible to extinction, and then compared these relative sensitivities to the fossil-derived extinction chronology. Our models show that the macropodiformes were the least demographically susceptible to extinction, followed by carnivores, monotremes, vombatiform herbivores, and large birds. Five of the eight extant species were as or more susceptible than the extinct species. There was no clear relationship between extinction susceptibility and the extinction chronology for any perturbation scenario, while body mass and generation length explained much of the variation in relative risk. Our results reveal that the actual mechanisms leading to the observed extinction chronology were unlikely related to variation in demographic susceptibility per se, but were possibly driven instead by finer-scale variation in climate change and/or human prey choice and relative hunting success.

    1. Ecology
    Jakob Thyrring, Lloyd S Peck
    Research Article Updated

    Whether global latitudinal diversity gradients exist in rocky intertidal α-diversity and across functional groups remains unknown. Using literature data from 433 intertidal sites, we investigated α-diversity patterns across 155° of latitude, and whether local-scale or global-scale structuring processes control α-diversity. We, furthermore, investigated how the relative composition of functional groups changes with latitude. α-Diversity differed among hemispheres with a mid-latitudinal peak in the north, and a non-significant unimodal pattern in the south, but there was no support for a tropical-to-polar decrease in α-diversity. Although global-scale drivers had no discernible effect, the local-scale drivers significantly affected α-diversity, and our results reveal that latitudinal diversity gradients are outweighed by local processes. In contrast to α-diversity patterns, species richness of three functional groups (predators, grazers, and suspension feeders) declined with latitude, coinciding with an inverse gradient in algae. Polar and tropical intertidal data were sparse, and more sampling is required to improve knowledge of marine biodiversity.