Introduction

The “microbiome” has emerged as a key determinant of many aspects of organismal biology, capable of shaping developmental, physiological, and reproductive phenotypes1,2,3,4,5. Yet, the contribution of the microbiome to host adaptation remains an evolutionary puzzle6,7,8,9. Microbiomes are traditionally viewed as nongenetic, environmental factors that influence host phenotypes. However, unlike abiotic environmental conditions, the effects of microbial variation have a genetic basis and can evolve10, but are not always inherited in the same way as host genes6,7,9. While microbiomes have substantial phenotypic effects on their hosts, these effects strongly depend on the ecological context2,3,8. Despite their clear importance, the complex effects of microbial inheritance and genetics on host phenotypes remain underappreciated, limiting our ability to understand host-microbiome evolution.

In this perspective piece, we explore how the microbiome influences host phenotypic variation. If the microbiome expands the host genetic repertoire and influences trait heritability, then the microbiome may have substantial impacts on host phenotypic evolution. We propose a path forward by integrating the microbiome into quantitative genetic models. Models that account for patterns of microbial inheritance and phenotypic effects will allow researchers to make predictions into the microbial impact on host evolutionary trajectories. Using this perspective, we detail two common scenarios found in the literature. First, microbial variation may shift the mean phenotype of the population, as would be expected when hosts leverage the microbiome to become locally adapted. Second, microbial variation may change host phenotypic variance, which could reduce variance by buffering against host variation or increase variability within the population. These two patterns may occur together (e.g., shift in mean and decrease in variance) and will change how hosts explore their fitness landscape. We review approaches to measure host-microbiome evolution, addressing current theoretical and technical limitations. We conclude by suggesting key avenues of research powered by the integration of the microbiome into common evolutionary tools, like quantitative genetics and experimental evolution. Through this perspective piece, we highlight how considering shifts in phenotypic mean and variance will help elucidate how the microbiome influences host evolution.

Phenotypic effects: extending the host genetic repertoire

Dawkins’ “Extended Phenotype” recognized how organisms modify surrounding environments and ecological communities11,12. Through environmental modification, an organism’s phenotypic effects are extended beyond its own genome, suggesting evolution is influenced through interacting ecological communities. This theory, developed for free-living ecosystems, also applies to host-microbiome interactions12,13. The microbiome, with its consortium of genomes, extends the genetic repertoire of the host to form what some are now calling the “Extended Genotype” because the host integrates the extended effects of the microbiome into its phenotype3,14,15,16,17,18,19,20,21,22. This extended genetic repertoire may shape the distribution of host phenotypes within a population, and consequently, shape the evolutionary potential of the host.

To formalize these verbal arguments, there is a genuine need for quantitative genetic models that incorporate the contribution of the microbiome to host phenotype, and subsequently, response to selection (Box 1). Explicitly incorporating the genetic variance encoded by microbes (VG-MICRO) as one would for the other components of phenotypic variance (e.g., VG-HOST or VE) should be a useful starting point. Host phenotypic variance can then be decomposed as:

$${V}_{P}={V}_{{{{{{\rm{G-HOST}}}}}}}+{V}_{{{{{{\rm{G-MICRO}}}}}}}+{V}_{{{{{{\rm{E}}}}}}}$$
(1)

When VG-MICRO contributes to host phenotypic variance, the microbiome may shape the evolutionary potential for a population. For illustrative purposes, imagine a scenario where a microbe turns a host blue (i.e., VG-MICRO contributes substantially to VP), and this blue phenotype increases host fitness in a hypothetical environment. We would predict that these beneficial blue-inducing microbes will increase in frequency in the host population, shifting the mean host color. We consider this scenario as hosts leveraging locally adaptive microbes (Fig. 1). The evolutionary benefits in this scenario will depend on the match between phenotype and selective pressure. Second, microbial genetic variation may alter variance in host phenotypes within a population (Fig. 1). The microbiome may effectively act as a buffer against environmental perturbation, thus decreasing phenotypic variance. Under this scenario, modifying the degree of host phenotypic robustness may be advantageous for specialization, but may constrain future adaptation if the environment is not constant23,24. On the other hand, the microbiome may also increase host phenotypic variance. This may enable the simultaneous exploration of novel regions of fitness landscapes, potentially enhancing host evolutionary responses to rapidly changing environments25,26,27,28. We note that these scenarios are not mutually exclusive, and the microbiome may shape both mean and variance in a population. Finally, it is important to consider that not all host phenotypes will be influenced by the microbiome, and not all microbes will influence host phenotypes (e.g., VG-MICRO = 0). Naturally, many host-microbiome interactions are also pathogenic, but such relationships have been studied theoretically and empirically in disease ecology8,29 and could also be incorporated into our quantitative genetics framework. Here, however, we are primarily focusing on adaptive microbiomes.

Fig. 1: Microbial influence on host phenotypic variation.
figure 1

The microbiome encodes many more genes than the host genome alone. Interactions with variation in the microbiome may alter the host genotype-phenotype map, shaping host phenotypic variance within populations. A First, some host phenotypes will not be affected by the microbiome (visualized in gray) and it is worth noting that not all microbes will influence host phenotypes. B The microbiome may shift the mean host phenotypes. The blue distribution is suggestive of when hosts leverage locally adaptive microbiomes to match local selective pressures. C Alternatively, the microbiome may also alter phenotypic variance (conceptualized in purple). D Finally, both the phenotypic mean and variance may be affected by the microbiome (in green). We note here that these scenarios are not mutually exclusive. Expanding host phenotypic variation through the microbiome may allow hosts to explore novel regions of fitness landscapes. These are conceptualized phenotypic distributions, and more experimental work is necessary to confirm how the microbiome affects host phenotypic distributions.

Using this simple quantitative genetics framework enables the partitioning of the genetic contribution of the microbiome to host phenotypic variance, and subsequently, host evolution. We explore how the two broad scenarios of the microbiome on host phenotypic effects described above (shifting mean phenotype and changing variance) can affect host evolutionary potential (Box 1). Overall, when the microbiome contributes to host phenotypes, the response to selection is amplified and shifts host evolutionary trajectory. However, the magnitude of microbial effects on host evolution will also strongly depend on the complexities of microbial inheritance.

A complex inheritance

Integrating microbial genetic variation into host evolutionary processes builds on ‘hologenome’ theory. The hologenome is defined as an evolutionary unit combining the eukaryotic host and all associated microbes15,30,31. Under this concept, evolution operates on this single unit because eukaryotic hosts are never isolated from microbes in the natural world15,22. Thus, the evolutionary fate of both hosts and microbes is intertwined. However, the intertwined fate of a single host-microbiome evolutionary unit is the focus of much criticism of the hologenome. One key criticism is that the host and microbiome rarely operate as one selective unit because transmission of the microbiome between host generations is rarely strictly vertical6,7,9. The microbiome can be transmitted via strict vertical transmission through embryos, but many microbes are also transmitted through quasi-vertical and environmental modes, like vaginal birth in mammals, regurgitation in birds, environmental inoculation in insects, or coprophagy in many taxa32,33. Environmental factors, like diet or climate, substantially influence the reservoir of possible microbial partners, leading to changes in the microbiome independent of any host evolution34,35,36. Host phenotypes also may be more strongly shaped by distinct microbial variation early in life, but because microbial dynamics operate at shorter timescales than host generations, those relevant microbes may not be faithfully transmitted across host generations8,37,38,39,40. Such environmental and demographic complexities muddle the inheritance of the microbiome, challenging our understanding of how and when the microbiome could influence host evolution6,7,8,9,31.

From a quantitative genetics perspective, heritability of the microbiome can be described as the proportion of microbial variation attributable to host genetic variance, just like any other complex trait41. Thus, heritable microbes are those where the relative abundance or community structure is associated with particular host genotypes41,42. Traditionally, only vertically transmitted microbes have been considered to be heritable, but the conflation of ‘microbial inheritance’ (i.e., transmission mode) and ‘microbial heritability’ (i.e., host genetic contribution to microbial variation) overlooks the many host systems that acquire their microbes through diverse transmission modes43. Indeed, in hosts dominated by environmentally acquired microbes, the contribution of host genetic variation to the relative abundance of particular microbes is as high as 42% in humans44,45, 39% in Drosophila46, and 25% in maize47. Interestingly, these and other studies show that not all components of the microbiome are heritable, with estimates ranging from 8 to b56% of microbes being transmitted faithfully45,47,48,49.

One might expect microbes with beneficial phenotypic effects to be more faithfully inherited. If a microbe contributes to a trait that promotes fitness in particular environments, then selection may reinforce faithful transmission to maintain the beneficial association, as observed in the many complex behaviors in host-microbe associations like aphid-Buchnera50, squid-Vibrio51, or beewolf-Streptomyces associations52. However, for many hosts, more complex communities are transmitted53,54,55. The drivers of this relationship between inheritance and phenotypic effects is not always clear. For example, in a UK twin study investigating the influence of host genetics on the microbiome, Methanobrevibacter species with the highest heritability had a strong association with low body-mass index44,45. However, another study in humans found low microbiome heritability, but microbial variation still explained 22-36% of metabolic traits35. Similarly, microbial variation explained 33% of weight gain in pigs48 and 13% of methane emissions in cows49 but also occurred largely independently of host genetic control of the microbiome. In other words, we still do not know whether the most heritable microbes explain the most significant variance in host traits.

Variation in transmission and the environment can impact the detection of heritable microbes. Controlling for transmission and environment are essential to identifying heritable microbes, and the disagreement between studies may result from differences in experimental design. The UK twin studies44,45 compared dizygotic and monozygotic twins to untangle maternal transmission and genetic contributions, while Rothschild et al.35 used large cohorts of unrelated individuals combined with metadata on diet and lifestyle to determine the environmental contribution. Animal and plant studies often use common garden experiments to expose different host genotypes to similar pools of available microbes46,47,48,49. The differences in these approaches may bias the ability to detect the contribution of heritable microbes to host trait variation, especially for hosts with largely environmentally acquired microbes. Despite these experimental limitations, these studies suggest that for a range of host traits, the microbiome contributes almost as much to phenotypic variance as do host genetics, even with the complicated inheritance of the microbiome.

Overall, these studies suggest that variation in how faithfully the microbiome is inherited may also influence host phenotypic variance. A quantitative genetics framework provides a path forward to partition the microbial contribution (VG-MICRO) to host phenotypes and allows us to further dissect the evolutionary consequences of this variation in transmission fidelity (Box 1). Through scaling the microbial effects by transmission fidelity, the microbiome could modulate the host response to selection. Microbes with phenotypic effects that are faithfully transmitted are likely to enhance the host phenotypic response to selection.

Our conceptualization suggests that the microbiome can modulate the host evolutionary potential, but there is currently limited empirical data to validate these conclusions. However, within the last 10 years, researchers have started to identify connections between the microbiome and host adaptation in diverse taxa and environments. We synthesized the literature and indeed found evidence suggestive of two common patterns in how the microbiome may shape host evolution connected to the processes suggested by our quantitative genetic framework: (1) hosts leverage locally adaptive microbes, shifting phenotypic means between host populations, and (2) the microbiome exposes novel host variation, changing host phenotypic variance within populations. These scenarios are not mutually exclusive, but as we explain below, may lead to different evolutionary trajectories for hosts.

Hosts leverage locally adaptive microbes

Hosts may leverage microbes to acquire new traits that are adaptive in the local environment. Locally adaptive microbes may facilitate the ability of hosts to explore the fitness landscape to match local environmental stressors by shifting host phenotypic means (Fig. 1). In particular, microbes with larger effective population sizes, rapid generation time, and pangenomes may evolve novel functions faster than their hosts56,57,58,59, and hosts may also acquire these adaptive microbes from the standing microbial variation in the local environment. If hosts can leverage locally adaptive microbes, then hosts can increase survivorship and/or reproduction to rapidly adapt to novel environments.

There are a number of remarkable studies that illustrate the substantial impact of locally adaptive microbes on host phenotypes and fitness. For example, bean bugs can gain pesticide resistance by acquiring a pesticide degrading bacterium, Burkholderia60,61,62 that is already present in the soil environment. Many other hosts utilize their microbiome to detoxify harmful chemicals. In habitats with toxic creosote plants, woodrats have a gut microbiome that can degrade phenolic toxins; exposure to creosote resin predictably structures this microbiome, enriching for Actinobacteria that degrade phenols63 and enabling the woodrat to occupy a specialized dietary niche63,64,65,66,67. Woodrat populations naïve to creosote have distinct microbiomes that do not degrade phenolic toxins even after creosote resin exposure68; however, transfer of the toxin-adapted microbiome to these woodrats increases woodrat survival on a creosote diet64. The microbiome can also facilitate survival in other kinds of stressful environments. Plants on geothermal soils are associated with the thermotolerant endophyte Curvularia69; thermotolerant Curvularia can increase survival up to 40 °C in non-adapted tomatoes, while Curvularia isolated from non-geothermal soils did not70. Additionally, salt tolerant fungal endophytes can confer salt tolerance to non-adapted plant populations70. Overall, these examples show that hosts can utilize specific microbes with large phenotypic effects to specialize and persist in novel environments, likely adapting more rapidly than the host genome alone could.

Microbial variation exposes novel host variation

During assembly of the microbiome, stochasticity and priority effects may create microbial variation between hosts within a population8,38,71. This change in microbial variation may increase phenotypic variability within populations (Fig. 1), enabling individuals to explore more phenotypic space26,28. Microbial variation would then enable the simultaneous exploration of different regions of fitness landscapes, changing the evolutionary trajectory of the host. We next review several examples that show how the microbiome increases phenotypic variance, but note that in natural populations, a combination of increases and decreases in phenotypic variance along with shifts in the mean likely occur.

Microbial variation increases host phenotypic variance in many organisms. For example, in Drosophila, different microbial communities increase the variance in larval development time, pupal weight, and adult weight compared between flies with different microbes and flies without microbes72,73. More so, different combinations of microbes shape life history traits in Drosophila nonlinearly, suggesting that microbial variation increases host phenotypic variance in complex ways74. Similarly, in Daphnia, different bacteria increase variance in body size and hatching success compared between sterile and among non-sterile treatments75. Microbiomes evolved for rapid flowering time in Arabidopsis increased variance in biomass compared to microbiomes evolved for slow flowering time76. In zebrafish, different bacteria species can either stimulate or suppress neutrophil responses, suggesting that variation in the microbiome is associated with varying immune responses to other bacteria77. In these cases, hosts are permissive to colonization by many microbes, and in turn, phenotypes respond to microbial variation. A major question for host-microbiome evolution is why some host phenotypes are responsive to diverse microbiomes while others maintain more specialized interactions?

The ability of the host to tune specific traits using diverse microbes may allow hosts to better match phenotypes to environmental stressors16,17, especially those stressors that may vary within populations. First, hosts that experience extreme variation in environment over their lifespan may need flexibility in their microbiome16. For example, coral microbiomes shift to protect against pathogens and extreme heat78,79,80. Bear microbiomes are enriched for microbes that harvest more energy during summer, most likely to ensure abundant fat accumulation before hibernation81. Second, the adaptive benefit of variable microbiomes may be especially important for organisms where high spatiotemporal variation in the environment occurs between parent and offspring. When offspring occupy different environments than parents, microbes may provide critical cues for local environments. For example, in Daphnia, microbes acquired from their local environment increased fecundity compared to those acquired via maternal transmission before diapause75. To best match phenotype to the local environment, hosts may frequently use microbial variation to alter developmental timing. Development may be especially responsive to microbial variation in many taxa, from insects to plants to crustaceans82. Developmental plasticity, through microbial variation, may expose novel phenotypic variation83, and this may in turn generate phenotypes that allow rapid adaptation to novel environments.

Particularly for organisms that live in fluctuating environments, the microbiome may enable a form of evolutionary bet-hedging. Bet-hedging occurs when phenotypic variation is maintained within a population to maximize long-term fitness of the lineage, despite shorter term fitness costs84. In the microbiome, bet-hedging could occur when hosts harbor associations with diverse microbes and responsive phenotypes. For some host systems, microbial variation may increase variability among individuals within a population, like when a diverse reservoir of microbes stochastically infects individuals, leading to increased inter-individual variation. Bet-hedging may also occur at the intra-individual scale, like when microbes have different phenotypic effects during different life stages. For example, studies have shown that Lactobacillus typically slows development in flies, suggesting different fitness costs in different environments72,73. However, on a nutrient poor diet, Lactobacillus increased in abundance and promoted larval growth85. These studies suggest the microbiome may enable a form of bet-hedging in Drosophila and potentially other hosts, but more work is necessary to explicitly test the role of the microbiome in bet-hedging86.

Long-term effects of the microbiome on host evolution

So far, we have focused on how variation in the microbiome shapes both the mean and variance of host phenotypic traits. When the phenotypic distribution changes, then the response to selection in the host also changes (Box 1). Over time, the adaptive potential of the microbiome will also leave signatures of selection in the host genome as well. The role of the microbiome in long-term host evolution is poorly understood, but by drawing on studies in evolutionary ecology and traditional symbioses, we can draw basic predictions for the long-term effects of microbiome-induced phenotypic variation on host evolution.

Relying on locally adaptive microbes may facilitate host trait evolution in the short term—but what are the long-term evolutionary consequences of host reliance on locally adaptive microbes? If beneficial, we may expect that the frequency of those microbes will increase within the host population, and in turn, this evolutionary benefit may allow hosts to rapidly cross the fitness landscape. Thus, hosts should evolve to maintain locally adaptive microbes and their phenotypic effects to either host traits or mitigation of the environmental stressors, much like genetic accommodation or niche construction83. Host taxa then specialize in these unique niches, as observed for taxa like bark beetles87, pine weevils88, and coffee berry borers89 that use their microbiomes to detoxify plant secondary compounds in specialized niches.

When host taxa experience similar ecological processes, the microbiome may converge to provide similar functions, like conifer pests that use microbes to degrade otherwise toxic terpenes87,88,90. In all of these cases, selection may favor specific functions to ensure consistent benefits to these specialized host–microbe associations. However, it is unclear whether selection operates first on the standing pool of microbes in the environment (that are then acquired by hosts), or if selection on the host exerts selective pressure on the microbiome, which may in turn evolve novel traits, benefiting the host. In some cases, like for bean bugs, pesticide application first increases pesticide resistant Burkholderia in the soil, and then bean bugs acquire the pesticide resistance through Burkholderia62.

It is worth considering that when microbial variation exposes novel host phenotypic variation, hosts must deal with inconsistent benefits of the microbiome in the short-term. The benefits and stability of a given host-microbiome association will ultimately be a function of the environmental context, with different expectations for hosts that inhabit stable versus more rapidly fluctuating environments (as discussed in Bruijning et al.86).

However, relying on environmentally acquired microbes increases the probability of random microbial acquisition, like pathogens and non-adaptive microbes. If specific microbes are consistently beneficial, hosts should evolve mechanisms to ensure faithful microbial transmission, similar to traditional symbioses91. The quantitative genetic framework introduced above is focused on a specific functional trait, but this framework may also be extended to selection on transmission mechanisms as well. For example, the bean bug acquires pesticide-resistant Burkholderia even at only 0.04% of the total soil microbiome62. With a unique physical sorting structure in the midgut that imposes competition between different potential microbes, the bean bug ensures the establishment of only beneficial Burkholderia species92,93. Hosts can evolve mechanisms to either screen potential symbionts, like in the bean bug-Burkholderia example, or other behaviors that promote the acquisition of beneficial microbes, like coprophagy32,33. While it is still contentious if most microbiomes behave like traditional symbioses, ultimately, strict vertical transmission and extreme reliance on specialized microbes is likely an evolutionary dead end, constraining hosts and microbes to the ecological conditions that generated the symbiosis94. Furthermore, when hosts leverage locally adaptive microbes, different populations will have different locally adaptive associations, which will likely lead to increased genetic differentiation between host populations. Overall, the challenge is to determine how microbial genetic variation is linked with and influences host genetic change during adaptation. Integrating the microbiome into quantitative genetics provides a framework to generate formal predictions. However, as discussed next, accurately quantifying microbial variation and its effects on host phenotypic variation by taking advantage of clever experimental design are necessary to test these hypotheses.

Experimental approaches to study the influence of the microbiome on host evolution

Above, we identified common scenarios of how the microbiome may influence host adaptation, with a focus on how host phenotypic distributions across individuals are shaped by microbial variation. With recent advances in sequencing technologies, microbial variation can be quantified at many different scales, from strains to whole community composition to complete ‘metagenomes’ (see Knight et al.95 for a recent review). A major challenge is to identify which scale of microbial variation (i.e., strain, community, metagenome) best measures microbial adaptation and influences host phenotypes8,95,96. Overcoming the technical hurdles in accurately quantifying microbial variation will be key to detecting how the microbiome influences the host evolutionary response.

The majority of host–microbiome interactions are characterized through “marker gene” studies (i.e., 16S rRNA for bacteria, ITS for fungi). Marker genes can classify microbial communities at broad taxonomic levels, are relatively inexpensive to sequence, and have established bioinformatic and analytical pipelines95. This approach demonstrates that microbial communities respond to many different kinds of selective pressures, like drought97,98, antibiotics99,100, different diets101, or warming environments102. Some studies show that microbial community diversity decreases in response to stressful environments98,99,100, while others show only shifts in community composition97,101,102. Marker gene studies are well-suited for taxonomically defining communities (though see103), but changes in microbial community composition do not always correlate to changes in host phenotype. For example, as measured by 16S rRNA marker genes, the macroalgae Ulva is colonized by different bacteria in different environments, but metagenomic data show the same core functional genes encoded by different bacterial taxa104. Similar patterns are observed in human microbiomes105 and the tree phyllosphere106,107. It remains unknown how taxonomic and functional identity provide different adaptive value for hosts. However, this conclusion may be strongly impacted by how microbial variation is characterized. Therefore, techniques beyond marker gene classification are needed to determine if and how microbial variation influences the host response to selection.

Strain-level analyses of microbial variation (i.e., polymorphisms and other genetic variants beyond marker genes) may provide more functional insight into the microbiome response to selection. For example, strain-level variation in honey bee microbes influences the response to pathogens or different diets for the hosts108. Strain-level analyses also provide insights into the stability of the microbiome across the lifetime of the host54,109,110 as well as transmission dynamics111,112. Metagenomes, which fully characterize the genes encoded by microbes, will provide crucial insights into functional variation of the microbiome95,108,113,114, like how bacterial metabolism varies across different niches within the same host115. This approach however remains costly, computationally intensive, and the results can be difficult to interpret. The emerging field of metatranscriptomics aims to study gene expression in the complex microbial communities95,116,117. Linking the metagenome to the metatranscriptome will be particularly insightful to understand how microbial variation influence host phenotypic variation95,118. For example, by comparing metagenomes to metatranscriptomes in the human gut, it was discovered that only a few specialized bacteria express unique biosynthesis pathways that may reflect the microbiome response to changing diets or other environmental perturbations in hosts118,119.

Experimental evolution can provide a powerful approach to investigate how microbial change interacts with host evolution13,120,121. In Drosophila, “Evolve and Resequence” (E&R) experiments, have uncovered many aspects of the genetic basis of adaptation to a diverse range of selective pressures122. E&R experiments use outbred populations, exert a selective pressure, and use sequencing to identify the genomic signatures of selection (Fig. 2). However, missing from this approach is whether and how the microbiome shapes the adaptive response of the host. As a first step in answering this question, the re-analysis of ten E&R experiments in D. melanogaster123 focusing on characterizing the microbial response to selection has been particularly insightful (Fig. 2). First, not all selective regimes had an effect on the microbiome. This observation is consistent with the idea that some phenotypes are more responsive to the influence of the microbiome than other phenotypes. Across the ten experiments, microbial diversity significantly decreased in four of the evolved populations (compared to the control populations), while it increased in one population. The decrease in diversity arises from a substantial increase in relative abundance for only a few bacteria, suggesting that flies are leveraging locally adaptive microbes. For example, for starvation resistance, Wolbachia increased in frequency following selection.

Fig. 2: Microbiome responds to host experimental evolution.
figure 2

Experimental evolution is a powerful tool to understand the genetic basis of adaptation. A Hypothesized schematic of the microbial response during adaptation in flies. B Microbial diversity is frequently reduced in evolved populations during experimental evolution in flies. Data is reproduced from Henry & Ayroles123. Here, each point represents Shannon diversity of metagenomes from a pool of sequenced flies. Asterisks denote the experiments where microbial diversity is significantly different between control and evolved populations.

While these results are correlative, this E&R survey suggests that the microbiome also responded to various selective regimes, and that microbiomes evolved in tandem with their hosts. However, additional work is necessary to understand if and how microbial evolution in these experiments also influenced host evolution. Like many other studies, we can only observe the endpoint of the host-microbiome evolutionary trajectory. Ultimately, understanding how the microbiome alters the evolutionary trajectory in hosts is the critical missing link13,15,16,18,19,120,124,125,126. Monitoring the temporal dynamics as populations respond to selection should provide the missing link—specifically, does the microbiome adapt faster than the host? And how frequently does the host utilize this rapid microbial evolution? A recent study, where the temporal evolutionary dynamics were monitored, suggests that environmentally acquired Lactobacillus bacteria in Drosophila evolve rapidly to nutrient-poor diets, and Drosophila can leverage rapid microbial adaptation independent of their own evolution121. We propose that experimental evolution, combined with statistical tools used in quantitative genetics, will be an important tool to understand how the microbiome shapes host phenotypic distributions, patterns of inheritance, and how these together influence host evolutionary trajectories (Fig. 3).

Fig. 3: Partitioning microbial effects on host adaptation.
figure 3

Experimental approaches developed for variance partitioning in quantitative genetics can be a powerful way to assess the influence of the microbiome on host evolution. A Experimental evolution will provide critical insights into how hosts and microbiomes respond to stressful environments. By including the microbiome (visualized as different colored circles) in experimental evolution, then genetic responses in both host and microbiome can be measured following selection. At the end of experimental evolution, we expect both host and microbiome to adapt - visualized as blue fly and blue/purple microbes. Microbial evolution may occur at the strain level (e.g., when beneficial mutations in particular genes drive adaptation to the novel selective pressure). Alternatively, selection may increase the frequency of a particular microbial taxon, leading to loss of microbial taxa in the adapted microbiome. To test how the microbiome interacts with the host genome to influence host phenotypes, one can perform fully factorial, reciprocal transplants between host, microbiome, and environment. B Key insights will be gained from examining the evolutionary trajectory of alleles that emerge or change in frequency during experimental evolution. C To test how microbial variation influences host phenotype, hosts can be inoculated with different levels of microbial variation. Removing the microbiome through antibiotics (or other manipulations) will show how hosts respond to perturbation to their microbiomes. D Finally, diallel crosses can be used to show how host and microbial genetic variation interact. Diallel crosses are performed by crossing all possible combinations between inbred lines to each other in a common environment (represented by fly colors). Rearing F1s in different microbial environments will enable partitioning of the additive and nonlinear, epistatic components between host and microbial genetic variation.

A practical outlook

Natural variation in the microbiome may also provide new tools to combat a range of challenges with practical applications. Indeed, there are several promising avenues of research leveraging natural microbial variation in fields ranging from conservation biology to agriculture. For example, variants of Pseudomonas bacteria isolated from caves help protect vulnerable bats from the devastating fungal infections of white-nose syndrome127,128. A similar strategy is also in development to protect amphibians from chytrid pathogens129. From a public health perspective, manipulating the microbiome in disease vectors may reduce vector competence130. The success of cancer management can be influenced by the presence of particular gut microbes in humans and finding novel probiotics to improve the efficacy of chemotherapy is under development131. Isolating host-associated microbiomes from organisms in extreme environments, like geothermal soils69,70, will serve as reservoir of adaptive microbes for agriculture in changing environments13,132. Natural variation in the microbiome represents a largely untapped resource with many novel solutions to address applied challenges.

Our quantitative genetics framework provides a useful starting point that can be further developed to leverage the microbiome in applied challenges. For conservation efforts, ensuring that microbial diversity is buffered from anthropogenic effects may help endangered organisms persist and restore both free-living and within-host ecosystems133,134,135. The effect may be similar to the microbiome increasing phenotypic variation to buffer fluctuating environments. Second, probiotics could be used to increase host resistance to specific challenges136, much like when the microbiome shifts host phenotypic distributions. The evolutionary persistence of microbiome engineering will depend on how selection operates on both host and microbiome13. For this to work, our framework suggests that maintaining transmission fidelity will enhance host response to selection (Box 1). Finally, many of the studies cited here utilize highly controlled environment to test the microbial contribution to host phenotypes. Identifying the processes that shape host-microbiome interactions in the wild is a major missing link135,137. Understanding the match between host phenotypic effects with the environment will be key to leveraging the microbiome for applied challenges.

Conclusions

In this perspective piece, we have identified scenarios of how the microbiome has the greatest potential to influence host evolution. Nevertheless, several key evolutionary questions remain to be addressed (see also Box 2): How does the microbiome change the host response to selection? What is the genetic basis of host responsiveness to microbial variation? Do the results of lab studies apply to microbiomes in the wild? Answering these questions will require a diversity of approaches using tools from quantitative genetics, community ecology, and genomics. Box 3

Several key analytical and theoretical gaps limit our current understanding of microbial influence on host evolution. More advanced bioinformatic tools are necessary to evaluate microbial diversity and its impact on host phenotypic diversity95. On the theoretical side, many aspects of the microbiome may poorly match the assumptions of theory developed for free-living organisms. Can we, for example, apply indirect selection12 or multilevel selection models (i.e., Price equation)96 to ask how independent or synergistic host and microbiome evolution are? One novel approach to detect the signatures of microbial variation shaping host evolution is ‘interspecies linkage disequilibrium’138, where selection may result in non-random allelic assortment of beneficial microbes and host genetic variation. The challenge is to identify how frequently variation in the microbiome increases host fitness, and whether this increases the probability of transmission across host generations.

Developing meaningful null models to describe host-microbiome evolution is an important and nontrivial challenge. A typical comparison would be between germ-free and conventionally reared hosts, where the null model would assume no difference. If hosts with a microbiome differed in their response to a selective pressure compared to the germ-free group, then the microbiome would be implicated in shaping host evolution. However, this may not be the most informative null, as how often have host organisms experienced a microbe-free world in their evolutionary history? Is the complete lack of a microbiome relevant to the evolutionary context in which organisms adapt to selective pressures? For some host taxa, recent work suggests that the modification and removal of the microbiome do not change host physiology139,140, and thus variation in the microbiome may not be evolutionary relevant for the host. First, we need more empirical data to calibrate the null model to understand how often the microbiome does not influence host phenotypes. Second, as outlined in Fig. 3, comparing evolved and ancestral microbiomes (and not only against germ-free conditions) across environments could greatly improve our understanding of host-microbiome evolution. Third, the null model should consider the balance in evolutionary interests between host and microbiome. We cannot assume that host-microbe associations evolve together to mutually benefit each partner6. Hosts and microbiomes will rarely have perfectly aligned evolutionary interests, suggesting a balance between cooperation and conflict6,8,37. Intergenomic conflict may maintain variation in microbial associations, but how often this influences host-microbiome evolution is not well understood141. Ultimately, the null hypothesis should drive appropriate experimental design to understand the balance between cooperation, conflict, and indifference between the host and microbiome during adaptation.

In conclusion, the contribution of the complex interplay between host and microbial genetic variation to host evolution is surprisingly understudied. Until recently, most quantitative and evolutionary genetic models addressing this topic have largely neglected the contribution of the microbiome to host phenotypic variance and evolution. Yet work from community ecology, quantitative genetics, and evolutionary biology suggest that the microbiome frequently shapes host phenotypic distributions across taxa and environments and may play a critical role in host evolution. Here, we have proposed a simple framework based on quantitative genetics that explicitly considers how genetic variation in the microbiome can extend the genetic repertoire of the host genome, influence host trait heritability, and subsequently impact host phenotypic evolution. Many challenges remain in this burgeoning field, but continuing advances in sequencing technology will facilitate the necessary characterization of host-microbiome evolutionary dynamics. Incorporating microbial variation into quantitative genetic models will provide fundamental novel insights into how selection operates across ecological and evolutionary scales.