Background
Animals harbor communities of microbes, known as microbiomes, in their gut and other body sites [1–3]. Given that these microbiomes are capable of mediating animal physiology and ecology, a key challenge is to determine the functional consequences of microbiome variation within host populations [4–8]. Such inter-individual microbiome variation has been observed across many animal species and tends to be particularly strong in wild populations [9–11]. Because most animal species are not experimentally tractable, a common approach is to use a microbiome’s diversity and composition to infer its “healthiness”—i.e. its effects on animal performance and fitness. Researchers often use two rules of thumb to evaluate whether a given microbiome is healthy or unhealthy for a given host: (1) higher alpha diversity indicates a healthy microbiome and (2) changes in microbiome beta diversity (interindividual variation) or composition indicate an unhealthy microbiome. These principles are extrapolated from findings about the human microbiome or the most closely related laboratory model, with little known about their applicability to wild animal populations.
We argue that drawing conclusions about an animal microbiome’s healthiness from diversity metrics alone is problematic. First, even for the deeply studied human microbiome, what aspects indicate a healthy state remains controversial [12, 13]. Even among healthy humans, microbiome diversity and composition vary across individuals and populations, shaped by factors like diet and environment [13]. Second, host species differ in baseline alpha diversity, absolute microbial abundance, longitudinal dynamics, and ecological function [14, 15]. For example, gut microbiomes of humans and closely related primates show divergent temporal dynamics, inter-individual variation, and the degree of co-diversification with hosts [13, 16]. Thus, rules developed in humans or lab models to infer host health outcomes from diversity metrics are more tenuous than is often assumed and may not translate to other species. Here, we weigh the evidence for and against the use of these rules in wild animal populations and propose an alternate path to identify healthy microbiomes. Although much of the literature we draw on concerns gut microbiomes, our arguments are equally applicable to microbiomes of skin or other body sites.
Future directions
Identifying general rules about what makes a healthy microbiome is a key goal for the field. However, in most animal species, we currently do not know what compositions, levels of diversity, or degrees of change are within the sphere of a healthy microbiome. To accomplish this, we need to understand multiple key aspects of a species’ microbiome: its members, temporal dynamics, and its combined resulting function for the host.
Among closely related species, diversity-health rules might indeed exist and be generalizable. However, these relationships must be established and tested. For example, a series of studies could start by longitudinally monitoring both the microbiome and some aspect of host fitness or a relevant physiological indicator in a given species (Fig. 1). Data describing what both healthy and unhealthy animals’ microbiomes look like across time could enable a species-specific model of microbiome healthiness in terms of composition and variability. Using multi-omic data (e.g. shotgun metagenomics, transcriptomics, and metabolomics) can also help us understand, which levels of microbiome diversity, be they taxonomic or functional, may best predict host physiology and fitness. From there, studies could focus on testing whether this model holds in closely related species. Systematically comparing these patterns to those in more distantly related taxa could then determine the extent of their generalizability. Ultimately, manipulative experiments—although not possible for many species—are necessary to determine the causality of these relationships.
We argue that it is time to move away from extrapolating existing (and often equivocal) information from humans and lab models, and toward embracing the diversity of ways in which wild animals interact with their microbial partners. We know that the alpha and beta microbiome diversity metrics in current use do not consistently predict host health status. This is not particularly surprising, given that these metrics imperfectly capture what matters most to a host: microbiome function. Further, variation in microbiome dynamics across the tree of life means that truly general diversity–health relationships may not exist. For many wild populations, we may never achieve a causal understanding of how microbiome variation affects host health. Regardless, host health and physiology metrics can and should be directly measured alongside the microbiome. Determining which microbiome features predict host health within and (perhaps) across species will help us better understand and manage animal–microbe relationships.
