Collectively, results from this experimental evolution model system provided support for some of the assumptions underlying the concept of hologenomic evolution. First, the selection on the host was effective, in that voles from the selected H lines exhibited improved ability to cope with low-quality diet compared to those from the control C lines. Additionally, we again confirmed that this selection yields changes in the caecal bacterial community. Importantly, although diet had a profound effect on microbiome composition, the microbiome differences due to selection were partially robust to dietary change. The selection differences were also robust to dispersal and environmental exposures, as cohousing had no effect on selection-related performance traits and only a weak effect on the bacterial community. Furthermore, the multivariate characteristics of the bacterial community and the abundances of several taxa were correlated with selection-related performance traits at the level of individual variation within selection and dietary groups. In the context of hologenomic evolution, we find that for our system, selection on a host performance trait leads to the ability of the host to maintain a distinct microbiome that correlated with host performance and is robust to horizontal bacterial transfer from individuals harboring a different microbiome.
Our experimental evolution model was designed to mimic the early stages of the evolution of herbivorous strategy in mammals (Sadowska et al. 2008), a transition requiring symbiosis with bacteria digesting cellulose and other fibrous compounds of plant cells. Therefore, it could be argued that the digestibility of a herbivorous diet would be the appropriate target of selection. However, from an organismal and evolutionary perspective, coping with a particular diet in terms of percent digested mass may be less important than ability to convert food into body growth or offspring. Therefore, we argue that the ability of juveniles to grow or maintain body mass during a short period of feeding on the low-quality diet (LQD) is an appropriate proxy for measuring "adaptation" to the herbivorous strategy. It is also consistent with the intended evolutionary scenario in which a non-strict herbivore may face a temporal shortage of typical food, and natural selection would favor those individuals that can immediately cope with the herbivorous diet (Sadowska et al. 2008, 2015). An important advantage of selection experiments is the potential to reveal the multi-level nature of phenotypic changes and to uncover proximate mechanisms that underlie the differences observed at the organismal level. For example, voles from the H lines tended to have a decreased basal metabolic rate, locomotor activity, and hormonal recovery after an acute stress (Sadowska et al. 2015; Maiti et al. 2019; Lipowska et al. 2020), but increased fat content (unpublished data). In this and the parallel experiment (Lipowska et al. 2024) we added microbiome as another level of investigation, and, in agreement with predictions based on comparative analyses, we showed that changes in microbiome are fundamental in evolutionary adaptation towards herbivory, and appear already at its initial stage.
Voles from the H lines were larger at the beginning of the feeding trial, grew faster during the trials on both diets than those from C lines (Fig. 5). They also had a higher rate of digestion of the LQD, and thus an increased metabolizable energy intake, due to an increased rate of food consumption rather than increased digestive efficiency. However, the ability to consume and process the low-quality food at a higher rate without compromising digestive efficiency indicates an improved capacity for herbivory, given that there is typically a tradeoff between digestion rate and digestive efficiency (Karasov and Martínez del Rio 2007). Such results could be due to increased alimentary tract size or performance, or improved efficiency of symbiotic digestion at the biochemical level. Additionally, the difference in body mass between the groups may contribute, as greater size is generally considered to be an adaptation to the herbivorous strategy to allow for greater food retention and lower relative energy requirements (Demment and Van Soest 1985), though greater body size also presents physiological challenges in the need to absorb and distribute nutrients through the body (Clauss and Hummel 2005). All these results were very similar to those obtained in the independent experiment on animals of the next generation (Lipowska et al. 2024), demonstrating integrity of the experimental protocols.
As expected, the vole gut microbiome was most strongly modulated by diet. The bacterial communities of animals fed the grass-diluted, fiber-rich diet, were more diverse than that of voles fed the standard diet, as shown by increased values of all three alpha-diversity traits we analyzed, altered community membership and structure, and altered abundances of most of the bacterial taxa. In general, many of these diet-induced changes reflect previous observations regarding feeding on fibrous diets (Reese and Dunn 2018). While these findings are useful for their confirmatory nature, our primary interest in this study is how microbiome traits are affected by the selection linetype of the focal individual (combined genetic and maternal effects) and the cohabitant (horizontal bacterial transfer).
The bacterial community membership and structure differed markedly between the H and C lines, and, while the general pattern revealed by PCoA was similar to that reported in the cross-fostering experiment (compare Fig. 3 in (Lipowska et al. 2024) with Fig. 3 in this study), both PCoA and PERMANOVA showed that the effect of selection was more profound in the current experiment. A plausible explanation is that in (Lipowska et al. 2024) all newborns were cross-fostered, and thus the difference attributed to selection was mainly due to genetic differences, whereas here the pups were reared by their biological mothers, which provided an opportunity for vertical bacterial transmission as well as physiological and behavioral early life maternal stimuli. Thus, the effect attributed to selection here included the maternal environmental effects. The importance of early life vertical transmission from mothers in shaping the microbiome of independent, post-weaning juvenile voles was documented by a highly significant effect of foster mother origin that was as large as the effect of biological mother origin, albeit associated with different bacteria than those dependent on genetic background (Lipowska et al. 2024). For example, the cross-fostering experiment showed that voles raised by mothers from the H lines, regardless of their biological mother, harbored lower abundances of Fournierella, a recently-characterized anaerobic genus first isolated from the human gut (Togo et al. 2017), while the abundance did not depend on the selection line of the biological mother. However, in agreement with a prediction based on those results, in the current experiment, in which the animals were reared by their biological mothers, the abundance of Fournierella was decreased significantly in the H lines. A similar pattern, though not so clear, appeared in seven other genera (Firmicutes: Lachnospiraceae_UCG-010, Oscillibacter, UCG-003, UCG-014; Campilobacterota: Helicobacter; and an uncultured and unclassified Verrucomicrobiota). Importantly, as will be argued in the next paragraph, the abundance of these genera did not depend on the linetype of cohabitant. Thus, the difference between the microbiome of post-weaning juveniles from the H and C lines is shaped both by the genetic and maternal environment effects.
On the other hand, in this experiment we showed that the microbiome community of juveniles, although changing in response to changes in diet and housing environment, is highly robust to horizontal transmission from other voles: the community membership was only marginally affected (p = 0.044) and community structure was not affected at all by the linetype of the cohabitant (p = 0.6; Table S6, Fig. 3). Both of the experiments showed that changes in abundances of the bacterial taxa that could be attributed to the selection linetype of the foster mother (vertical transmission) or of the cohabitant (horizontal transmission) occurred mostly in different genera than those attributed to the linetype of the focal individual. Moreover, with a few exceptions, the signal of horizontal transfer in juveniles appeared in different bacteria than those acquired by vertical transfer in early life. Thus, even if the presence or abundance of some bacteria is susceptible to the horizontal transfer, the post-weaning juveniles from the selected lines appear to protect the composition of bacteria beneficial in the context of our experimental evolution model system that are preferentially hosted due to both genetic and early life maternal effects. This picture is consistent with the observation that the linetype of the cohabitant had no effect on any of the selection-related performance traits (Fig. 5). Taken together, the results of the two experiments support the idea that microbiome community traits can be treated like any other organismal trait, determined early in life by genetic and maternal influences, and later subject to acclimation to specific dietary and habitat conditions, but largely robust to random influences from other conspecifics at later life stages. Consequently, the results can be treated as supporting some of the main assumptions of hologenomic evolution (Bordenstein and Theis 2015; Theis et al. 2016; Rosenberg and Zilber-Rosenberg 2018).
Lipowska et al. (Lipowska et al. 2024) reported that the effect of selection explained about 1% of the total variance in multivariate community characteristics, which might raise doubts about whether such a difference would be reproducible in an independent experiment. The results of this experiment confirm that the difference, although small, is real. Nevertheless, it could still be argued that such a small effect, even if real, has little biological significance. However, this effect concerns the difference averaged over four independently evolving replicate lines of both the selected and unselected control lines, and it appeared after only 22–23 generations of effective selection (Supplementary Methods, available online), i.e., on a very short evolutionary time scale. We also showed that the effect of selection is partially robust to disturbances such as changes in diet or housing conditions and to vertical bacterial transmission in early life, and highly robust to horizontal transmission in juveniles. Thus, consistent with other studies based on rodent selection experiments that have reported correlated changes in microbiome composition of comparable magnitude after more generations and with larger differences in the directly selected trait (McNamara et al. 2021, 2023), we believe that the small difference is still biologically meaningful. This conclusion is also supported by the results of analyses conducted to uncover plausible functional relationships between the microbial and selection-related performance traits.
The role of gut symbionts in cellulose digestion is widely known, but the gut microbiome can facilitate mammalian herbivory also by other mechanisms (Dearing and Kohl 2017). The gut microbiome is integrated with maintenance of host mass balance, especially through interactions with metabolic physiology (Chevalier et al. 2015; Sommer et al. 2016; Regan et al. 2022). These metabolic interactions might also occur through general interactions with body size, as aspects of gut microbiome are correlated with body size across species (Godon et al. 2016; Reese and Dunn 2018; Sherrill-Mix et al. 2018). Additionally, through the gut-brain axis, the gut microbiome can modulate aspects of feeding behavior and feeding rates (Bo et al. 2020; Shu et al. 2021; Trevelline and Kohl 2022). Thus, the microbial contributions to mammalian herbivory may extend beyond digestion of cellulose, to include other aspects of the animals’ energetics, physiology, and behavior that contribute to improved efficiency in converting consumed food to body growth. Indeed, results of comparisons of univariate microbial traits (alpha diversity metrics and abundances of specific bacterial taxa) between the selected and control lines, as well as results of analyses of within-group correlations between microbial and morpho-physiological traits at the individual level, indicate that different microbial traits are partially independently correlated with body mass balance and digestive efficiency.
Although both (Lipowska et al. 2024) and this study showed that the alpha-diversity metrics do not differ between the selected and control lines, in both studies we observed a significant positive correlation between the digestive efficiency and the caecal microbial species richness, measured as both the number of amplicon sequence variants and Shannon index (Table R1). Notably, (Lipowska et al. 2024) reported nearly the same partial regression slopes, which again reinforces credibility of the results. Generally, relationships between diversity and function are enigmatic to ecologists and evolutionary biologists, though complex to interpret given the many measures of diversity and of function (Shade 2017; Reese and Dunn 2018). In the context of herbivory, however, as shown also by our results, it is typical that a greater taxonomic diversity yields higher functional diversity, which is beneficial towards degrading the complex fibers present in plant material (Reese and Dunn 2018). Therefore, in this context, it seems counterintuitive that in both of the experiments, the correlations between diversity metrics and digestive efficiency were stronger in animals fed SD than in those fed LQD (Tables 1,2). However, it should be noted that we did not analyze the bacterial composition in animals fed LQD for a long period of time, which would presumably lead to the establishment of a community "optimized" to cope with such a diet, but rather, in accordance with the assumptions and objectives of this selection experiment (Sadowska et al. 2008), a response of the animals and their microbiota to a sudden change in diet lasting only five days. Thus, the microbial community was presumably only in the initial stages of rebuilding, and since the rate of this process could vary among individuals, the correlations were dissipated.
Results of analyses of correlation between the digestive efficiency and multivariate beta-diversity characteristics showed the same pattern, again, highly consistently in both (Lipowska et al. 2024) and in this study. Both the community membership and community structure were correlated with digestive efficiency, and the former also with the rate of food consumption and digestion, and all these correlations were stronger in animas fed the SD than LQD diet (Table 1, Table 1 in (Lipowska et al. 2024)). These results clearly document the significance of the bacterial community composition in the digestion, and support the notion that in the LQD groups the composition was presumably at a transition between the states stabilized for a given diet. We are aware that the experiments would be further strengthened by including groups fed both the SD and the LQD for a longer time and to an older age, enough to stabilize the microbial composition after the dietary change. However, doubling the effort would not be feasible due to logistical limitations and, more importantly, would not be acceptable for ethical reasons, since the negative body mass balance in the 5-day trial indicated that many individuals, especially those in the unselected C lines, would not be able to live on this low-quality diet for a long period of time.
On the other hand, however, the correlation between the beta-diversity metrics and body mass-change during the feeding trials (MDFT) was much stronger in animals fed the LQD than those fed SD, similarly as the difference in MDFT between the H and C lines is larger in animals fed LQD. The latter result is not surprising, because MDFT measured in the test with LQD was the trait directly selected for, whereas the increased growth rate in animals fed SD appeared only as a correlated response to selection. However, the first observation is not trivial, especially considering that the correlations with FC, ADE and FD were more profound in animals fed the standard diet. Taken together, the results provide an evidence that the bacterial composition is an important factor affecting growth rate of the voles independently of its effect on the food digestion, and imply that the effects of altered microbiome not related to food digestion contributed to the difference between H and C lines in the mass balance.
This multimodal functional character of the microbiome contribution to the evolution of “herbivory” in our experimental evolution model is revealed also by considering the effect of selection on abundances of particular bacterial taxa and correlations of the abundances with the selection-related performance traits (Fig. 4, 6; Table R1; Tables S7-S12).
The abundances of Desulfobacterota (mainly Desulfovibrio), Proteobacteria (uncultured genera of Rickettsiales and Paracaedibacteraceae), and Actinobacteriota (mainly Bifidobacterium) increased in the H lines, and were positively correlated with the body mass change in the feeding trial. While it is unclear how the Proteobacteria could be beneficial, Bifidobacterium, which was nearly absent in C lines but increased in H lines to considerable 0.12% of relative abundance, is well known for involvement (directly or through interactions with other bacteria) in degradation of fibrous carbohydrates and producing short-chain fatty acids (SCFA; mainly butyric and acetic) and lactate, which becomes the source of energy, and also in protection against pathogens, production of vitamin B, antioxidants, and stimulation of the immune system in humans (Rivière et al. 2016), and therefore are included in probiotic supplements. Desulfovibrio has both beneficial and adverse effects in humans (Singh et al. 2023), but it is a potent producer of acetic acid and other SCFA (Hong et al. 2021), and is also involved in hydrogen scavenging (Kohl et al. 2016a), which again may increase the efficiency of using the fibrous carbohydrates as a source of energy. The abundance of Firmicutes, the most abundant phylum, did not change systematically in response to selection, but several of its genera that increased abundance in the H lines can be also associated with plausible beneficial roles, not only associated with digestion. For example, Acetatifactor, in addition to producing SCFA, is involved in stress response (Liu et al. 2022; Ma et al. 2024), and its increased abundance suggests a lower stress level in voles from the H lines, which can contribute to the improved ability to maintain energy balance under the conditions of suddenly worsened diet. Ruminococcus is involved in fiber degradation (Christopherson et al. 2014), and Lactobacillus, which was positively correlated with the rate of effective food digestion, a proxy of metabolizable energy intake, is a the dominant genus in the foregut chambers of several herbivorous rodents (Kohl et al. 2014; Shinohara et al. 2016) and is associated with growth promotion in malnourished mice through interactions with hepatic growth hormone signaling (Schwarzer et al. 2016).
On the other hand, the abundance of Bacteroidota, the second most abundant phylum, was decreased in the selected H lines, and at the level of individual variation it was negatively correlated with both the body mass change in the trial with low-quality diet, i.e. with the value of the selected trait, and with the apparent digestive efficiency (this latter correlation concerned the standard rather than the low-quality diet, but, as we have explained above, this is not surprising). A similar pattern was revealed by analyses of abundance of two genera from this phylum (a genus of Muribaculaceae, the most abundant in this phylum, and a less abundant Rikenella), and, not so clearly, in analyses of abundance of two less common phyla: Campilobacterota (mainly Helicobacter), and Cyanobacteria (mainly Gastranaerophilales). These results can be taken as suggesting that high abundance of these bacteria compromise the efficiency of digestion, and consequently the energy balance and the growth rate, and therefore the selection favored voles that are able to control the abundance of these bacteria at a lower level. Such an interpretation is strengthened by known functional effects of some of these bacteria. The involvement of Helicobacter in compromised gastrointestinal function in humans is widely known, and such effects appear also in rodents (see references in: (Charles River 2011)). Thus, its decreased abundance may be associated with increased digestive performance and generally improved health. The abundance of Rikinella is deceased in gut microbiome of obese humans (Palmas et al. 2021), which implies a negative association of this genus abundance with energy balance or predisposition to store energy reserves, or both. A propensity to keep a positive energy imbalance and to accumulate fat are considered adverse in the context of contemporary human health, but these are exactly the characteristics favored in our experimental evolution model. Actually, according to the “thrifty genotype hypothesis”, which attempts to explain evolutionary basis of the human propensity to develop obesity when access to energy-rich food is unlimited and physical activity reduced, the same concerned also our human hunter-gatherer ancestors (Chakravarthy and Booth 2004).
To summarize, our results support the hypothesis that selection on a host performance trait leads to evolution of the preferential and stable maintenance of an altered microbiome composition, beneficial in the context the experimental evolution model. The microbiome community traits are determined early in life by genetic and maternal influences, and are later subject to acclimation to specific dietary and habitat conditions, but are largely robust to random influences from other conspecifics at later life stages. Thus, the microbiome traits can be treated like any other trait of the host. Consequently, the results can be regarded as supporting some of the main assumptions of hologenomic evolution. However, several questions remain opened, such as what specific genetic changes in the H lines confer the altered microbiome, in what specific way the beneficial bacteria interact with the host through the lower-level mechanisms (molecular, biochemical, neurobiological), and whether the microbiome in the selected lines evolved only by modifying the composition of bacteria from an available diversity, or whether specific genetically divergent bacteria evolved in response to the (hypothetical) hologenomic evolution. We believe continued work with our bank vole experimental evolution model system can answer such questions, and thus contribute to the understanding of hologenomic evolution of mammalian herbivory, and encourage the development of other similar experimental evolution approaches. promote