본문 바로가기
읽기 · 학술 · 공개 라이선스 원문

Drivers of Microbiome Biodiversity: A Review of General Rules, Feces, and Ignorance

Reese, Dunn · 2018

공개 학술지 mBio 에 실린 총설 논문입니다(2018년). CC BY 4.0 으로 공개된 글이고, 그림 · 표 · 참고 문헌 표시를 뺀 본문을 실었습니다.

출처 Reese & Dunn (2018), mBio · 라이선스 CC BY 4.0 · 원문

장 2개 올림 이 장 3293낱말
다음 장 →

이 장은 원문만 있습니다. 문장별 해석은 앞 장부터 차례로 붙입니다.

1장 · Drivers of Microbiome Biodiversity: A Review of General Rules, Feces, and Ignorance

The alpha diversity of ecologic communities is affected by many biotic and abiotic drivers and, in turn, affects ecosystem functioning. Yet, patterns of alpha diversity in host-associated microbial communities (microbiomes) are poorly studied and the appropriateness of general theory is untested. Expanding diversity theory to include microbiomes is essential as diversity is a frequently cited metric of their status. Here, we review and newly analyze reports of alpha diversity for animal gut microbiomes. We demonstrate that both diet and body size affect diversity in the gut but that gut physiology (fermenter versus simple) is the most important driver. We also assess the advantages of various diversity metrics. The importance of diversity in microbiomes is often assumed but has not been tested outright. Therefore, we close by discussing how to integrate microbiomes into the field of biodiversity-ecosystem functioning to more clearly understand when and why a host supports diverse microbial communities.

PERSPECTIVE

Diversity, one of the most widely studied ecologic properties, is fundamental to the field of ecology itself. It is widely reported and often promoted as an indicator of ecosystem state due to its relationships with productivity, functioning, and stability. However, diversity is not an inherent feature of communities, nor is it a universal good. Specifically, more diversity is not always better. Therefore, diversity cannot be assessed in a universal framework but instead must be contextualized or relativized within the ecosystem of interest. The difficulties of interpreting diversity, both its drivers and consequences, are particularly acute when considering microbial systems.

To date, the majority of research into patterns of diversity and biodiversity-ecosystem function has been carried out in plant communities or with vertebrates at larger spatial grains. In these macroscopic communities, diversity has long been documented to vary predictably with geographic area, latitude, altitude, time, and disturbance, among other factors. Major research efforts to build diversity-function relationships from this literature began in the early 1990s and have continued unabated since then. While the diversity of some microbial communities has also been shown to vary predictably along gradients or in response to disturbance, only a small proportion of the diversity-function literature addresses microbial eukaryotic or bacterial communities (but see references,, and) even though most species are microbes. An interesting exception is work on the importance of evolution to the dynamics of biodiversity. Because experimental evolution is more tractable for analyses of microbes, they are more often used for experiments studying diversity and evolution conjointly (see, for example, references,, and), demonstrating that such experiments are possible and even uniquely valuable for testing ecologic theories. Relatively little work has been done to bridge the divide between large and small (see, for example, reference). This dearth may in part reflect differences in biology and even physics between macroscopic and microbial systems, but we suspect that it just as often reflects differences in the scientific cultures of those who study “macrobes” and those who study microbes.

The lack of relevant microbial theory is particularly apparent when considering host-associated microbial systems, i.e., microbiomes. Diversity metrics are frequently reported in amplicon sequencing-based assessments of communities, and variations in diversity are calculated between hosts or between case and control states. Without a theory of diversity applicable to microbiomes, however, it is hard, if not impossible, to interpret these results. Even in parasites, for which there is a much longer tradition of research, there has been only limited work on patterns of diversity inside hosts (see, for example, references and) rather than globally irrespective of individual animals (see, for example, reference). The patterns of microbiome diversity, pathogenic or commensal, among individuals, within host species, or between host species have not been systematically described. Without such knowledge, the obstacles preventing our understanding when changes in diversity are important remain insurmountable.

Based on what has been found in plant, animal, and free-living microbial communities, we would expect diversity in microbiomes to be impacted by a number of factors. Factors which increase niche availability, such as diverse resources or increased area, or which limit growth, such as frequent disturbance or chronically extreme conditions, are expected to drive diversity in opposite directions. The specific forms of these factors likely differ depending on the body site considered—for example, disturbances to the skin could include bathing or variation in environmental temperature, whereas the gut would be more susceptible to ingested pharmaceuticals or changes in diet. Similarly, the levels of importance of microbial community diversity to the host are also likely to differ between body sites. Here, we focus on differences among species in the diversity of the gut microbiota, which is arguably the most highly studied host-associated microbial community type.

Drawing on previous reports, we suggest that gut microbial community diversity is positively associated with diet diversity and the proportion of calories derived from plant material and that increases in both should increase the number of available nutritional niches. We also suggest that gut microbial community diversity tracks body size, which increases available gut space (akin to increases in area in other ecosystems) and transit time. Physiological features such as stomach pH, which influences the ability of microbes to colonize the gut, and gut morphology, which alters available habitat and residence time, may also contribute. To test these hypotheses, we reviewed the literature and conducted new analyses of data drawn from previously published studies. We then discuss our limited current understanding of diversity-ecosystem functioning in the gut and propose future work to address this limitation.

VARIATION BETWEEN ANIMALS

Few studies have explicitly tested the drivers of variation in bacterial (or, more generally, microbial) community diversity in guts among animal species. In general, these studies have tended to suggest an effect of diet on gut microbial community diversity, with herbivory associated with higher diversity. Among New World bats, the phylogenetic diversity of gut bacteria was found to be associated with dietary strategy; phylogenetic diversity was greatest in fruit-feeding bats and lowest in blood-feeding bats. More recently, Godon et al. found that diversity was generally higher in herbivores but that the effect of diet was mediated by body size, with an overall positive relationship between body mass and gut bacterial community diversity (measured using the Simpson index) across a cohort of distantly related animals in captivity (including birds, mammals, and reptiles). Captivity itself has also been found to affect gut diversity in some mammals, but the relationship did not have a consistent signal.

Heuristic comparisons among taxa and contexts reveal patterns that support our hypothesis that diet diversity and gut morphology are important drivers of gut microbial community diversity. Small insects with simple diets often have low-diversity communities. Caterpillars, which have short gastrointestinal transit times and short guts, harbor very low microbial community diversity, driven by species sourced from the food plant material itself. Fruit flies and honeybees similarly carry only a few species of bacteria consistently in their guts. In the case of Drosophila, individuals can lose their microbiota in the absence of replenishment from bacterially colonized food although there is also evidence that some bacteria can persist and grow when added to an otherwise axenic gut. At the other end of the spectrum, humans living more traditional lifestyles, which are associated with higher-fiber diets and longer transit time, have been found to have higher-diversity gut microbiota than do those living Westernized lifestyles. Interestingly, there is not a significant difference in most diversity metrics between Inuit individuals who consume traditional but low-fiber diets and Canadians living in a city. This result suggests that it is not traditional lifestyles per se that necessarily lead to more diverse gut microbes but instead certain kinds of diets.

To test specific hypotheses of a general model of diversity in host-associated communities, we conducted a review of relevant literature on the gut microbiota. Gut microbiota are important for nutrient and waste processing, signaling, and immune function. Gut microbes are also, in most cases, among the most diverse host-associated communities. The field of gut microbiome research has grown exponentially over recent decades, including, but especially since, the decade in which the NIH Human Microbiome Project began. Therefore, we expected it to have the largest data sets to draw from. We conducted a literature search in Google Scholar for “gut AND diversity AND microb*” and also read papers cited in the work found through the search. We focused on 16S rRNA gene amplicon data and on healthy individuals found both in the wild and in captivity. We then extracted data from the published text and figures of 36 studies (see Table 1; see also Data Set S1 in the supplemental material). Data found only in figures were estimated using GraphClick, which is graph digitizer software. We used the authors' own definition of a taxonomic group (typically, species or operational taxonomic unit [OTU]), which in most studies, as in most of the literature, was 97% similarity of the 16S rRNA gene, although other cutoff values were also used (see, e.g., references and) or not specified (see, e.g., reference). In cases where host species were represented in multiple publications from the same population, we included only data from the study with greater sequencing depth (see, for example, references and). If averages of sequencing depth data or rarefaction depth data were not reported, we used the lower end of the range reported. In some studies, no information was reported on sequencing depth. In two studies (and), data were available only at the family level. We then collected metadata on animal diets and body size for a subset of animals from a comprehensive database for mammals and birds. Gut structure (ruminant, hindgut fermenter, simple gut) data were available in many publications that we included but were also corroborated for additional species through targeted literature searches.

We first tested for individual effects of taxonomic class (e.g., mammal, bird, or insect; n = 259), diet class (carnivore, omnivore, or herbivore; n = 259), and gut physiology (foregut ruminant, hindgut fermenter, or simple gut; n = 176) on diversity using a combination of Kruskal-Wallis and Mann-Whitney U tests. Multiple-hypothesis correction (Bonferroni correction) was applied to any comparison with more than three levels. To explore the interaction between diet and physiology, we analyzed a subset of mammalian fecal samples for which data on diet and physiology were available (n = 63 to 67) using linear mixed-effects models that included diet or physiology or diet and physiology as a fixed effect(s). All models included sequencing depth as a fixed effect and publication identity (i.e., which publication the data came from) as a random effect. Likelihood tests were used for comparisons of the models to one another and to a null model that included only sequencing depth. Similarly, we tested for an effect of body mass on diversity in mammalian fecal samples (n = 52 to 58) by comparing a linear mixed-effects model that included body mass and physiology to one that included only physiology. Again, sequencing depth and publication identity were included in both models. These analyses were carried out in the “lmer” package in R. Finally, we were able to extract more-detailed physiological data on gut length and stomach pH for a subset of animals. Gut length data were included in cases where data for members of the same genus were present in both data sets. Gut length (n = 15) and pH (n = 14 to 16) were analyzed with partial Spearman correlations which corrected for variations in sequencing depth. We included publication as a random effect or sequencing depth as a covariate in our analyses in order to correct for differences in sequencing approaches between studies. Not all studies were included in all analyses, because not all data were available for each parameter and, notably, because not all studies reported sequencing depth.

Shannon diversity and richness

Shannon diversity, a metric that weights the numbers of species by their relative evenness data, was the most commonly reported diversity metric. We found that Shannon diversity data varied among some taxonomic classes, diet groups, and gut physiologies (P < 0.05 [Mann-Whitney U tests]) but that there were no overall effects of any of these variables (P > 0.05 [Kruskal Wallis tests]; Fig. 1A to C). In mammalian fecal samples, Shannon diversity was associated with host diet but more strongly with gut physiology. Linear mixed-effects models that included diet class or gut physiology outperformed a null model that included only sequencing depth (P < 0.001 [likelihood tests]), but the model that included both diet class and physiology did not outperform the model that included only physiology and sequencing depth (P = 0.80). Fermenters, both foregut and hindgut, had higher gut microbial richness values than did animals with simple guts; the effect of the inclusion of ruminants led to a pattern wherein herbivores (a subset of which are ruminants) seemingly had higher diversity overall than omnivores or carnivores (none of which are ruminants; Fig. 2A). The diet effects disappeared in comparisons performed just within animals with simple guts. Shannon diversity was similarly associated with variations in body mass. A model that included body mass outperformed the null model (P = 0.03 [likelihood test]), but this effect did not result in greater performance than that seen with a model that simply included physiology (P = 0.90 [likelihood test]). The explanation for this result is likely that the relationships between body mass and diversity had different directions depending on host physiology (Fig. 2B)—as ruminants increased in body mass, their Shannon diversity increased, whereas animals with simple guts showed the opposite trend wherein the guts of larger animals with simple guts tended to be lower in microbial community diversity. This may be in part a consequence of the fact that large animals with simple gut physiology tend to be omnivores and carnivores.

Richness, i.e., the total number of OTUs or species recorded, was the second most commonly reported diversity metric and showed patterns similar to those seen with Shannon diversity. The richness data differed between some taxa, dietary groups, and physiological groups but did not show overall effects of these variables (P > 0.05 [Kruskal Wallis tests]; Fig. 1D to F). In mammals, richness was impacted by diet within animals with simple guts, but across the board, herbivorous hindgut or foregut fermenters had the highest diversity (Fig. 2D). Linear mixed-effects models that included diet class or gut physiology outperformed the null model (P < 0.001 [likelihood tests]), but the interaction model did not outperform the physiology model (P = 0.18). A body mass model outperformed the null model (P < 0.001 [likelihood test]), but, as with Shannon diversity, inclusion of the body mass term did not improve model performance relative to one that included physiology alone (P = 0.20 [likelihood test]; Fig. 2E). To the extent, then, that body mass matters, it seems to matter with respect to effects on the relative abundance and evenness of species rather than their total count.

Gut length measurements, which were available for a subset of taxa, were not associated with diversity (richness or Shannon diversity), even correcting for study sequencing differences (P > 0.05 [partial Spearman correlations]). Similarly, the pH of the stomach, which can act like a filter, preventing, in proportion to its acidity, the colonization of the gut by microbes in the years after birth, was not associated with richness (P = 0.31 [partial Spearman correlation]; Fig. 2F) and was associated with Shannon diversity only marginally (P = 0.05 [partial Spearman correlation]) (Fig. 2C). As gut length data were available for only 19 genera and pH data for 23 species, such relationships may require greater power to compensate for study-to-study variability.

Other metrics

Fewer than a third of animals had other estimates of diversity reported. These included phylogenetic diversity, Chao1, and, less commonly, Simpson’s diversity. Perhaps because of lower sample sizes, these metrics were not consistently associated with diet or physiology and were more often driven by large variations in sequencing depth.

Because different research projects differ with respect to sequencing technology, target primers, and, perhaps most importantly, sequencing depth, diversity values can vary greatly for reasons decoupled from biology. Diversity is highly correlated with sequencing depth, i.e., the number of reads sequenced per sample, similarly to how diversity is correlated with sampling effort in studies of plants or animals. But the debate over whether or not to rarefy microbial sequencing data to a set diversity depth is ongoing. Chao1 is an estimator designed to specifically account for differences in sampling effort between studies and should in theory be more robust in comparisons between microbial sequencing studies (but see reference). In practice, the sensitivity to sequencing depth in our analysis was most apparent for Chao1 (Fig. 3), although this may also reflect differences in sequencing technologies. While it is possible that these metrics of diversity would yield greater insight if available from a broader swath of the literature, variations between samples with respect to such values are currently extremely difficult to interpret. Overall, our results, in conjunction with other reports on the difficulties in estimating microbial community diversity, indicate that the data from comparisons of Shannon diversity may be the most robust and informative, particularly given the frequency with which this metric is already reported in the field.

How to improve these analyses

In general, we find a coarse relationship between gut architecture and host diet with respect to gut microbial community diversity. There are, of course, many other factors which, we predict, could impact gut alpha diversity, but such data are harder to find consistently in the literature. Immune system complexity and functioning can together serve as a filter for specific taxa or microbes in general, but studies that consider the influence of immune complexity and function would require generally agreed-upon definitions and measurement protocols for immune function. A varied diet is expected to create more metabolic niches for the microbiota and thus to increase microbial community diversity. Estimates of diet diversity, specifically, of plant diversity, are difficult to calculate without observational or genomic measurements of diet, though. In our survey, sequencing-inferred diet data were available for only one study, which included just two host species, but such data could easily be collected in future studies through inclusion of plant DNA sequencing in addition to the microbial sequencing. Finally, a short retention time in the gut is expected to lower diversity as it selects for only fast-growing taxa—similarly to how high disturbance rates are commonly associated with low diversity in plant communities. This trend has been shown for select host taxa but has not been tested with a diverse group of hosts. Because body size is not necessarily a strong predictor of gastrointestinal transit time, actual estimates of transit time will be necessary to truly test this hypothesis more generally.

Overall, we found that the best predictor of gut microbial community alpha diversity variation among animals was the presence or absence of a simple gut. Ruminants and hindgut fermenters consistently supported higher levels of diversity than did species with simple guts. While herbivores are typically thought to have more diverse gut microbial communities, we found that herbivores with simple guts, such as some bears, tended to have diversity levels similar to those of predators. We hypothesize that this effect is due specifically to the gut architecture and, as an extension, the short transit times.

Such variations between herbivores on the basis of gut architecture lead to an important outstanding question in the study of microbiomes—does it matter for the host whether its microbial community is diverse? If diversity is important for function, herbivores with simple guts, such as giant pandas or spectacled bears, would be predicted to receive less benefit from their microbiomes and as such might experience evolutionary pressure for behavioral or physiological compensation. Alternatively, it may be that a diverse community is not necessary for digestion of plants and that the large, complex communities found in herbivorous foregut and hindgut fermenters primarily reflect the possibility of higher diversity in large, complex environments rather than the diversity’s adaptiveness for the host per se. Addressing this issue requires moving beyond measuring diversity to experimentally manipulating it, as discussed below.

이 글은 위 출처의 원문을 그 라이선스(CC BY 4.0)에 따라 옮겼습니다. 참고문헌 목록과 인용 번호, 그림 · 표 · 각주 · 수식 · 부록은 빼고 절별로 나누어 실었으며, 그 밖의 문장은 원문 그대로입니다. 원 저작자와 발행 기관은 이 화면을 만들거나 검토하지 않았고, 이곳의 해석과 낱말 풀이는 공식 번역이 아닙니다.