Microbiomes as sources of emergent host phenotypes

See allHide authors and affiliations

Science  27 Sep 2019:
Vol. 365, Issue 6460, pp. 1405-1409
DOI: 10.1126/science.aay0240


Microbial communities associated with animals exert powerful influences on host physiology, regulating metabolism and immune function, as well as complex host behaviors. The importance of host–microbiome interactions for maintaining homeostasis and promoting health raises evolutionarily complicated questions about how animals and their microbiomes have coevolved, and how these relationships affect the ways that animals interact with their environment. Here, we review the literature on the contributions of host factors to microbial community structure and corresponding influences of microbiomes on emergent host phenotypes. We focus in particular on animal behaviors as a basis for understanding potential roles for the microbiome in shaping host neurobiology.

Animal microbiomes are powerful modifiers of host biology that influence many phenotypes characteristic of a species. The highly integrated physiologies of hosts and their associated microbes—their microbiota— have led to the increasing popularity of the concept of the holobiont; i.e., the sum of a metazoan host organism and its closely associated microbial communities, evolutionarily functioning as an inextricably intertwined single unit (1). The resulting multispecies collective (which we will distinguish from single-species collectives known as superorganisms) faces natural selection, leading to coevolution of the member organisms. This concept is hotly debated, with discussions raising issues concerned with ecological theory, strictly obligate versus facultative associations, and microbial transmission and/or heredity (2, 3). Regardless, microbiomes contribute substantially to the biology of their eukaryotic partners in associations that often expand the physiological capacity of the partner organisms.

The vast biochemical potential of the microbiome, often distinct from that of the host, can lead to the rise of emergent phenotypes, where host organisms gain new physiological abilities as a result of contributions from their microbial partners. Beyond nutritional and metabolic couplings, microbial communities are becoming recognized as modulators of complex animal phenotypes, including behavior. In certain cases, these connections have an experimentally identified biochemical basis, but in others, these relationships are poorly defined. This is especially true when considering the human microbiome, in which incredible amounts of diversity and complexity are observed, but ethical considerations limit experimental potential. Nevertheless, correlations emerging from human studies, as well as experimental evidence from animal models, point to substantive functional interactions between the microbiome, host neurobiology, and behavior. These associations present a series of interesting questions: (i) What are the selective processes that shape complex microbiomes? (ii) What distinctive physiologies can microbiomes convey to their hosts? (iii) How do neurobiological processes affected by the microbiome influence host fitness?

Here, we review recent findings on how animals shape their autochthonous microbial communities and how these communities influence emergent phenotypes in their hosts. We will focus our discussion on the connections between the microbiome and the nervous system and their potential to inform our understanding of the evolution of animal behavior.

Host determinants of symbiotic partners

Animals shape their microbiomes in several ways. Many of these mechanisms are enforced at the genetic level of the host and can be revealed experimentally in genetically tractable model organisms, or through large-scale comparisons that leverage natural variation between individuals. Using crosses between inbred mouse strains, researchers have identified host single-nucleotide polymorphisms (SNPs) that are associated with the relative abundances of specific gut microbes. For example, variations in genes required for the immunomodulatory Irak3–TLR2 (Toll-like receptor 2) signaling pathway shape the relative abundances of Lactococcus and Coriobacteriaceae, likely by altering the host’s ability to immunologically respond to peptidoglycan (4). Similarly, Toll-like receptor 5 (Tlr5) gene expression early in life has been associated with altered gut microbiota in mice, possibly by enhancing immune responses to flagellated bacteria (5). Results like these suggest that broad innate immune activation mediates the host’s control over its microbiota. However, there is evidence that these effects are at least partially caused by microbial transmission or by environmental factors, as differences between wild-type and Tlr5-deficient mice are absent when the mice are not raised in separate colonies (6, 7).

In the adaptive arm of immune responses, pattern recognition and humoral responses can also influence the composition of the microbiota. Immunoglobulin A (IgA) is the primary isotype that recognizes microbiota species in the intestinal tract. The host secretes large quantities of microbiota-binding IgA into the gut lumen that is either broadly reactive with low microbial affinity (8) or relatively specific with higher affinity (9), and binds to motifs common across several bacterial (and in some cases, viral) lineages. This has led some researchers to consider such broad IgA responses as a form of innate immunity. These broadly cross-reactive antibodies show increased reactivity to bacteria in vivo, displaying synergy between environmental context and host response. This effect could arise because of increased expression of microbial antigens in the natural context, or improved antibody–antigen interactions because of host environment–specific factors, potentially leading to niche-specific sanctioning of microbiota members (8).

In humans, twin studies leveraging genetic variation between pairs of identical and fraternal twins have identified connections between host genes and gut microbiota composition. In a study of fecal microbiotas from >1000 twin pairs, several bacterial taxa, including Christensenellaceae and Turicibacteraceae, showed significant correlation with host genetic variation (10). Comparing these abundance measurements with SNPs identified several genetic loci as having the potential to shape microbiota composition. Select polymorphisms in the lactase gene, LCT, were associated with reduced host lactase activity and increased representation of Bifidobacterium species, which can metabolize lactose. Twin studies such as this have also revealed several correlations between microbiota makeup and genes associated with taste and olfaction (10), which together point to another avenue of host control of the microbiota: diet.

Diet alters the composition and activity of the microbiota by introducing new species and providing specific nutrients, selecting for enrichment or depletion of certain microbes through nutrient surplus and starvation, while also shifting the expression profiles of members of the microbiota to degrade a dynamic bounty (11). In humans, this is exemplified by human milk oligosaccharides, which are complex carbohydrates found in breast milk that are inaccessible to infant digestion. However, these oligosaccharides support specific members of the neonate microbiota, including specific Bifidobacterium species, which promote mucosal immunity while shaping the maturing infant microbiota (12).

In several examples, there is evidence for diet exerting a greater impact on the microbiota than host genotype. Various mouse studies have shown that skewing the nutrient composition of the diet overshadows the effects of host factors such as immune signaling (13). Food composition can alter the interactions between host and microbiome, mediating microbial dependence on host-provided nutrients. For example, genetically varying the fucosylation of host mucins alters the composition and transcriptional activity of the gut microbiota (14). This effect was only seen when the microbiota was dependent on mucus glycans as a nutrient source and could be diminished when mice were fed a diet rich in varied carbohydrate sources, pointing to diet as a master regulator of microbial communities in the intestine. These observations offer promise for “correcting” microbiota imbalances or deficiencies through dietary intervention, potentially by nutritional supplementation with substrates predicted to select for or against specific microorganisms (prebiotics).

By surveying host genotype and microbiota composition across taxonomic lineages, it is possible to gain insights into potential relationships between evolutionary history and microbial colonization. Brooks et al. described the concept of “phylosymbiosis”—i.e., the relationship between host taxonomy and microbiota structure—and showed that this concept is valid across different animal groups (15). That is, animal speciation events are typically mirrored by microbiota composition, with more closely related host individuals wielding more similar microbiomes. This pattern holds true within and across species, as members of the same species exhibit more similar gut microbiomes than members of different species.

Notably, these coevolved microbiomes have impacts on host health. For example, specific bacteria, such as Lactobacillus plantarum, promote a variety of fitness phenotypes at various stages of fruit fly development. In this case, Drosophila melanogaster larvae are attracted to L. plantarum–derived volatiles (16) and promote replication of L. plantarum on their foodstuff, increasing bacterial abundance and promoting fly larval growth (17). Moreover, L. plantarum must be repeatedly ingested (i.e., it does not stably colonize the gut tract), indicating that external support of a microbiota reservoir may be valuable to host physiology. The benefits conveyed by specific microbiota compositions are often sufficient to select for mechanisms that animals use to shape their microbiomes over evolutionary time scales.

Emergent phenotypes regulated by microbiomes

Beyond obligate mutualisms, animals and their evolutionary relatives rely on specific microbes for important parts of their life cycles and activities (Fig. 1). In one of the closest evolutionary relatives to animals, the choanoflagellate Salpingoeca rosetta, the transition from unicellularity to multicellularity is mediated by a series of lipid species produced by a naturally cohabitating bacterium, Algoriphagus machipongonensis. The bacterial partner produces at least two sulfanolipids that stimulate multicellular rosette formation in S. rosetta, while also producing at least two lysophosphotidylethanolamines that promote rosette stability and one capnine lipid that inhibits rosette formation (18). Multicellularity is predicted to reduce motility while improving feeding, so rather than being constitutively required, bacterial stimulation of rosette formation likely has context-specific fitness outcomes by providing S. rosetta with different lifestyle strategies (Fig. 1A). Thus, bacterial association provides the host with advantages in adaptability and fitness during periods of environmental change.

Fig. 1 Emergent microbiome-derived phenotypes.

(A to D) Examples of bacterial members of the microbiota providing new capacities to their animal hosts.

Credit: Kellie Holoski/Science

In some cases, these burgeoning relationships can provide sufficient mutual benefit to lead to coevolution between partners. For example, the mutualism between the Hawaiian bobtail squid Euprymna scolopes and the bioluminescent gamma-proteobacterium Vibrio fischeri is tightly regulated at the interface of a specifically evolved symbiotic organ in the squid (19). Although both partners can be raised separately without substantial impairment in laboratory conditions, the bacterium provides the squid with a singular capacity for predator evasion by counterilluminating bioluminescent camouflage, an ability that squids themselves have not evolved (Fig. 1B). In return, the squid provides V. fischeri with a privileged niche inaccessible to other bacteria and abundant nutrients to promote high bacterial densities.

Similarly, functions endowed by the microbiota can confer the ability to consume otherwise intractable food sources, opening new ecological possibilities to the host. In wood-feeding higher termites, the ability to metabolize wood carbohydrates is facilitated by the synergistic activities of a bacterial consortium in the midgut and hindgut. The termite Nasutitermes takasagoensis utilizes xylan-degrading glycoside hydrolases, provided primarily by a hindgut Treponema bacterial symbiont, to render wood celluloses bioavailable to other members of the microbiota and the host (20). Through phylogenetic analysis, these genes were discovered to most likely originate from bacteria of the Firmicutes phylum. Reintroduction of these xylanases by horizontal gene transfer among bacterial partners allows these termites to eat wood, just as lower termites do with the aid of eukaryotic symbionts (Fig. 1C, top).

The potential for similar microbiota-guided metabolic shifts has also been seen in the human microbiota, where a gut microbe, Bacteroides plebeius, is genomically enriched with genes for porphyranase and agarase enzymes, allowing it to digest algal carbohydrates (21). B. plebeius is believed to have acquired these genes from related marine bacteria that degrade seaweeds. Analysis of metagenomic datasets shows that B. plebeius and its glycoside hydrolases are enriched in the gut microbiomes of individuals from Japan, where seaweed is a common component of the diet. These enzymes are less abundant in individuals from North America, suggesting environmental acquisition and/or dietary maintenance of this capacity (22) and the potential for dietary shifts enabled by microbiome composition (Fig. 1C, bottom).

Beyond affecting host dietary abilities and activities, members of the endogenous microbiota can alter host immune status, having broad impacts on host health. For instance, comparing the gut microbiomes of inbred mice has revealed the wide-ranging effects of segmented filamentous bacteria (SFB), an epithelium-associated Clostridium relative. The presence of SFB skews host immune responses, such that helper T cells are polarized toward a proinflammatory/anti-extracellular pathogen state (promoted by T helper 17 cells and interleukin-17) that protects mice from bacterial infection (23). This same SFB-induced T cell profile has also been found to be detrimental in mouse models of some autoimmune (24) and neurological disorders (25). For example, in an inflammation-based mouse model for autism spectrum disorder, colonization of dams with SFB promoted abnormal social behavior and stereotypies in the offspring (25). These findings highlight the potential for the microbiota to influence aspects of host biology once thought to be independent of gut physiology, such as brain development and behavior.

Microbiome influences on behavioral and neurological phenotypes

The connections between microbiome state and host behavior complicate our understanding of how microbial communities shape hosts. Alterations in the microbiota have been correlated with or explained as a cause of a wide range of brain and behavioral phenotypes in a wide range of organisms (26). Some of these studies have been controversial, raising questions surrounding proper study design, data interpretation, and the physiological relevance of animal behaviors to human behavioral disorders. Other observations have been reproducible within and across various model organisms, providing proof of concept that manipulating the microbiome can modify animal behaviors and neurological endophenotypes of disease.

Some of the first examples discovered of microbial influences on behavior involve olfactory communication between organisms. Volatile chemicals produced directly by microbes can shape host odorant profiles that signal social interactions. For example, depleting the microbiome of the eusocial leaf-cutting ant Acromyrmex echinatior alters the expression of odorant cuticular hydrocarbons of the treated ants and triggers aggressive behavior from untreated nestmates (Fig. 1D, top) (27). Recolonization with microbiota from conspecifics abrogates the effect, whereas colonization with microbes from different ant colonies incites particularly aggressive behavior from nestmates. Similar phenomena have been shown to affect mate choice in particular lineages of D. melanogaster (28).

The microbiota can modify the odorant profiles of vertebrates, too. Although microbial odorant profiles can be used to distinguish age, sex, and group differences in mammals (26), the degree to which they truly influence social behavior remains unclear. This is particularly the case for primates and other mammals with more complex social motivation and who rely less heavily on olfactory communication.

Aside from regulating chemical cues that inform behavioral responses, the microbiome is increasingly recognized for its ability to shape the development and function of endogenous neurobiological pathways that control animal behaviors (Fig. 1D, bottom, and Fig. 2). Laboratory rodents, flies, and zebrafish reared without their native microbiomes exhibit deficient sociability and social preference in conventional behavioral assays (26). These assays typically use automated tracking software to quantify the duration of time that an animal spends interacting with a familiar or unfamiliar conspecific, similar to metrics used to assess human social disorders. Changes in the gut microbiome are reported to contribute to social behavioral abnormalities seen in offspring exposed to maternal immune challenge and maternal high-fat diet (25, 29), whereas postnatal treatment with select bacteria is reported to promote sociability and communication in mice across etiologically distinct genetic and environmental models of autism spectrum disorder (30, 31). Although the exact mechanisms are unclear, there is evidence that microbes modify or produce biochemicals that could alter the activity of neural circuits in the brain that regulate social motivation (31). Advances in identifying microbial metabolites and their cognate receptors, and in tools for selectively activating and inhibiting subtypes of neurons, have enabled studies that highlight direct and indirect signaling between indigenous microbes and neurons (26). In some cases, molecules produced by microbes, including classical microbial-associated molecular patterns and select secondary metabolites, can enter host tissues and are sensed directly by neurons (Fig. 2). In other cases, microbial modulation of the immune system or host core metabolism confers downstream influences on the host nervous system. Overall, such findings support the notion that the microbiome can alter host physiologies that ultimately influence the nervous system and behavior.

Fig. 2 Impacts of the microbiome on aspects of neurobehavior.

Experimentally proposed examples of mechanisms by which the microbiota affect behavioral phenotypes.

Credit: Kellie Holoski/Science

In addition to social communication–related behaviors, many responses to stress, such as risk avoidance and decreased exploration, have been linked to altered microbiomes. Several studies across different model organisms consistently report that depletion or absence of the microbiota reduces stress-induced behaviors, such as freezing in motion and thigmotaxis (26). For example, germ-free mice and rats consistently display more exploratory behavior compared to conventionally colonized controls in paradigms that measure the time spent or number of entries into an environment typically considered aversive (26). Although the exact mechanisms underlying this behavioral phenomenon remain unsettled it is thought to be mediated by microbiome-dependent changes in levels of circulating stress hormones, such as corticosterone and adrenocorticotropic hormone, that feed into the hypothalamic-pituitary-adrenal axis (32).

Since the first observations were made, the list of microbially derived and microbially modulated hormones, neuropeptides, and neurotransmitters has grown substantially. Several hundred circulating metabolites are regulated by the gut microbiome, and there is evidence that some, like p-cresol, 4-ethylphenyl sulfate, and indoxyl sulfate, can modify behavior through endocrine signaling. Precursors of these small molecules are reported to cross the gut epithelium, become modified in host tissues, enter the systemic circulation, and penetrate the blood-brain barrier (33). Some microbial metabolites, like indoles and their derivatives, are reported to activate the vagus nerve directly, which contains a subset of neurons with receptive fields in intestinal villi that innervate the brainstem. Rats colonized with indole-producing Escherichia coli exhibit higher stress-induced immobility, reduced exploratory behavior, and increased expression of an activation marker in vagal neurons (34). Both endocrine and sensory neuronal pathways are also thought to contribute to the influence of the gut microbiota on feeding behavior in animals. Microbial metabolism regulates circulating nutrients and hormones, such as leptin, insulin, and glucagon-like peptide 1, that affect neural circuits underlying hunger and satiety (35), and vagal neurons detect peripheral energy status and relay that information directly to homeostatic centers in the hypothalamus.

As more studies reveal the effect of members of the microbiota on animal behaviors, a key question is whether the composition of the human microbiota has any relevance to symptoms of human behavioral disorders. There are an increasing number of reports that alterations in the gut microbiome are associated with various neurodevelopmental, neuropsychiatric, and neurodegenerative disorders (26). However, cause and effect are unclear, and even across studies of individuals with the same diagnosis, there is little consensus on whether there is a definitive microbial signature for these disorders. This could be attributed to multiple factors, including inherent heterogeneity within each disorder, variations in lifestyle and medical status, differences in study methodology, the difficulty of analyzing data from small sample sizes, or simply a lack of specific microbial association with the condition.

Although roles for the microbiota in the etiopathogenesis of neurological diseases remain controversial, there is some evidence that microbiome based interventions can be helpful in alleviating behavioral symptoms. For example, in an open-label study of a small group of 18 children with autism spectrum disorder and comorbid gastrointestinal disturbances, gastrointestinal and behavioral symptoms improved by 8 weeks and up to 2 years after fecal microbiota transplantation (36). Stronger evidence for a microbiota association with mood comes from randomized double-blind placebo-controlled studies reporting positive effects of select probiotics, consisting of select Bifidobacteria and Lactobacilli, on reducing signs of emotional distress in healthy human volunteers, preterm infants, and individuals with chronic fatigue syndrome (3739). In addition, there is increasing evidence that host-associated microbes interact with common neurochemicals (40) and xenobiotics, such as atypical antipsychotics and antidepressants (41), which is inspiring interest in targeting these interactions to modify host behavior (42). Together, these findings lend support to the prospect of microbiome-based interventions for influencing mood-related symptoms. Further, they raise the interesting possibility that host conditions along the gut-brain axis could be products of emergent phenotypes stimulated by the microbiome.

Partnership between hosts and their microbiomes

Hosts and their microbiomes have coevolved to maintain the homeostasis of the holobiont. To answer the questions that we posed earlier, we can say that (i) the composition and functional activity of host-associated microbial communities are shaped not only by internal host factors but also by responses to external influences. Host immune responses, diet, and behavior can shape and be shaped by animals’ microbiomes. (ii) Microbiomes can endow their hosts with myriad phenotypes, including metabolic novelty, predator defense, and certain social responses. (iii) Microbiota-neurobiology interplay can alter normative development and animal behavior, broadly molding activities required for animal fitness. With increasing evidence that microbiomes can modify complex behaviors, ranging from feeding and mate choice to social communication and risk avoidance, microbial modulation of behavior is likely to affect the evolutionary fitness of the host, but further experiments will be needed to directly test this.

This final point pushes several new questions to the forefront. Have modern neurobehavioral activities, which are apparently influenced by host-associated microbiomes, also been outsourced to members of the microbiota, akin to the host dependency and associated genomic degradation seen in endosymbionts? Broadly, have critical portions of host neurobiology been fully delegated to microbes, or have these host physiologies only arisen because of these partnerships?

If the microbiota are responsible for normative neural activities, over evolutionary history, the incorporation of new species into the microbiota may have allowed a host species to take behavioral and social leaps often ascribed to factors like altered brain morphology. In this case, what are the selective advantages to both host and symbiont for sharing these processes? Are these relationships true mutualisms, or could some of our microbiota act as parasites? Or has the relationship between host and microbiota been shifted by modern environmental circumstances? Despite the enticing prospect of the potential coevolution of host behavior with microbiome activity, it is also possible that the neurobehavioral effects of the microbiome are an epiphenomenon of normal bacterial metabolism. The downstream effects may thus either be coincidental or be the result of incidental host sensitivity to these processes or products (42).

Animal microbiomes can quickly alter host physiology, and disruption to the microbiota has been hypothesized to contribute to a bevy of health and disease states. Without understanding connections between host and microbiome, attempts to correct disease-associated microbiomes could create unpredicted issues. Newly discovered connections between behavior and the microbiota highlight the risks and rewards of this potential double-edged sword.

References and Notes

Acknowledgments: Funding: J.B.L. is supported by fellowship F32-GM119238 from the National Institutes of Health. E.Y.H. is a New York Stem Cell Foundation—Robertson Investigator, supported by a Packard Fellowship in Science and Engineering and Chan Zuckerberg Initiative Ben Barres Career Acceleration Award. Competing interests: E.Y.H. is a former cofounder of Axial Biotherapeutics, cofounder of Bloom Science, and scientific adviser to Holobiome, Inc.
View Abstract

Stay Connected to Science

Navigate This Article