Shared Evolutionary Path in Social Microbiomes

Abstract

Social networks can influence the ecology of gut bacteria, shaping the species composition of the gut microbiome in humans and other animals. Gut commensals evolve and can adapt at a rapid pace when colonizing healthy hosts. Here, we aimed at assessing the impact of host-to-host bacterial transmission on Escherichia coli evolution in the mammalian gut. Using an in vivo experimental evolution approach in mice, we found a transmission rate of 7% (±3% 2× standard error [2SE]) of E. coli cells per day between hosts inhabiting the same household. Consistent with the predictions of a simple population genetics model of mutation–selection–migration, the level of shared events resulting from within host evolution is greatly enhanced in cohoused mice, showing that hosts undergoing the same diet and habit are not only expected to have similar microbiome species compositions but also similar microbiome evolutionary dynamics. Furthermore, we estimated the rate of mutation accumulation of E. coli to be 3.0 × 10⁻³ (±0.8 × 10⁻³ 2SE) mutations/genome/generation, irrespective of the social context of the regime. Our results reveal the impact of bacterial migration across hosts in shaping the adaptive evolution of new strains colonizing gut microbiomes.

Introduction

Individuals that share the same space are known to harbor more similar microbiota compositions than individuals that do not. In humans, substantial strain sharing was found among cohabiting persons, with 12% and 32% strain-sharing rates observed for the gut and oral microbiomes, respectively (Valles-Colomer et al. 2023). This indicates that microbial migration between hosts, which is increased when living in the same household, is an important factor structuring species diversity in the microbiome (Johnson and Clabots 2006; Johnson et al. 2008; Siranosian et al. 2021). Indeed, it has been shown that cohousing of mice, which are coprophagic, can reduce the diversity of the microbiota species composition among hosts. Cohousing is also a common practice aimed at reducing microbiota variation in many studies of host immune phenotypes (Ericsson and Franklin 2015).

Emerging data in both mice and humans show that significant evolutionary change can occur within strains of each microbiota species (Garud and Pollard 2020). In germ-free mice colonized with either a single or multiple strains of commensal bacteria (Li et al. 2015; Barroso-Batista et al. 2020; Yilmaz et al. 2021) or in mice with a native microbiota (Barroso-Batista et al. 2014; Lescat et al. 2017; Frazão et al. 2019, 2022), evolutionary change has been observed within days, weeks, or months. Time series data of human metagenomes also show that several adaptive evolutionary events can occur within months (Garud et al. 2019; Zhao et al. 2019).

To begin unraveling how bacterial transmission across hosts may affect patterns of molecular evolution of their gut microbes, population genetics theory of metapopulations, where a deme mimics a host, can be useful (Pannell and Charlesworth 1999; Booker et al. 2021). Theoretical models of adaptation in the context of metapopulations predict that the rate of adaptive evolution of a clonal population (i.e., of bacteria) resulting from the accumulation of new beneficial mutations should be affected by the amount of migration/transmission between demes (i.e., individual hosts). In particular, for a given level of migration, the rate of adaptation should be increased relative to when no migration occurs (Gordo and Campos 2006; Yeaman and Whitlock 2011). The models typically consider a single microbial species and therefore ignore the multispecies complexity characteristic of the gut ecosystem. However, their predictions should be robust for ecosystem level complexities under conditions where the strength of intraspecific competition is much higher than that of interspecies competition. Such conditions are observed when ecological models aiming at explaining the diversity and stability of microbiomes, like the generalized Lotka–Volterra, are fitted to 16s RNA data (Coyte et al. 2015).

Here, we study how the evolutionary path of a new lineage colonizing the mouse gut is influenced by its host social environment. We performed both transmission and evolution experiments. The former were designed to estimate Escherichia coli transmission rates among cohoused mice, while in the latter we use in vivo experimental evolution to test the following hypotheses: 1) that cohousing increases the rate of evolution of commensal bacteria and 2) that high migration rates lead to an increase in the number of shared evolutionary events accumulating across hosts. We test this hypothesis for the two key processes of evolution: horizontal gene transfer (HGT) and mutation accumulation. Under HGT, we expect that adaptive events occurring in only one mouse will transmit rapidly to the entire mouse metapopulation. Under mutation accumulation, we expect higher allele sharing in a social regime if adaptive evolution within the host is marked by intense competition between clones carrying different adaptive mutations. In fact, when clonal interference is pervasive within hosts, which live in the same environment (e.g., same diet), migration/transmission may help the bacterial clones which carry the highest combination of beneficial alleles to spread more rapidly across hosts.

Results

Model of Mutation–Selection–Migration and Genetic Drift

To quantify the conditions under which between-host transmission is expected to affect the spectrum of mutations shared across hosts, we used a model of mutation–selection–migration and genetic drift (see Materials and Methods). The simulations predict that when the number of migrants is ≫1, the metapopulation behaves as a single population and all beneficial mutations are shared across demes (fig. 1A and supplementary fig. S1A, Supplementary Material online). At the other extreme, when the number of migrants is extremely small, demes behave as close to independent populations, and the level of shared adaptive polymorphisms is expected to be very low.

Fig. 1.

Escherichia coli transmission between mice. (A) Model of bacterial transmission. The relation between the level of shared evolutionary events and migration in a model where all mutations have the same effect (type singleS) and in one where mutations are exponentially distributed (type exp). Parameter values N = 10000, U_d = 0.01; U_a = 0.00001; S_d = 0.1; S_a = 0.1. Populations accumulate mutations during 400 generations, in a time scale of E. coli evolution in the gut of 27 days. (B) Experimental design to estimate migration rates. Mice were colonized with an E. coli clone expressing either a cyan (CFP) or a yellow (YFP) fluorescent protein (10⁷ CFUs) while in an asocial environment (individually caging). In Experiment I, mice (n = 2) were moved to a social environment (cohousing) at Day 8, while in Experiment II (n = 2), mice were moved to a social environment at Day 15. (C) Fecal CFP and YFP E. coli loads were assessed under asocial and social contexts. Mice and time periods in a social context (cohousing) are indicated. Error bars represent 2SE, and the dashed line indicates the limit of detection (330 CFU/g of feces). Escherichia coli transmission rate (%): percentage of E. coli cells in the mouse gut that resulted from transmission events, each day, from the remaining mouse inside the cage. Error represent 2SE. See supplementary table S2, Supplementary Material online.

The simulations also predict that the rate of evolution, M(t), estimated from the sum of frequencies of the beneficial mutations accumulated until a given time, should increase as migration rates increase (supplementary fig. S1B, Supplementary Material online), although the power to detect differences in M(t) requires a high number of replicates given that large variances are observed for this statistic. Concerning the population mean fitness, the simulations predict that it should be similar across different transmission rates (supplementary fig. S1C, Supplementary Material online). As for the fraction of high-frequency mutations (>50%) in a population, the simulations predict that these should be higher in regimes of high migration rates (supplementary fig. S1D, Supplementary Material online).

High Levels of Strain Transmission Under Cohousing

The first set of experiments in the present study (transmission experiments) was performed to estimate E. coli transmission rates between cohoused animals. The use of two isogenic E. coli strains (supplementary tables S1 and S2, Supplementary Material online), each carrying a different neutral marker: yellow fluorescent protein (YFP) or cyan fluorescent protein (CFP), allowed us to quantify the extent of migration across hosts after they are cohoused. We estimated the rate of transmission by sampling the mice at only one timepoint, 18 h after cohousing, given that once one of the fluorescently labeled strains is transmitted to another animal, subsequent transmission episodes cannot be distinguished from the first (supplementary table S2 andfig. 1C, Supplementary material online). The CFP and YFP were found to behave as neutral markers in previous competitions in the mouse gut (Sousa et al. 2017). The experimental design is described in figure 1B: each mouse was colonized with one of the labeled lineages, CFP or YFP during a period of T days (T = 7 or 14 days, fig. 1B—Experiments I and II, respectively). After this period, the mice were cohoused (social regime) to allow for microbiome transmission between hosts and to quantify the rate of transmission of E. coli between hosts. The cohousing after 2 weeks of colonization (Experiment II) was performed to test if transmission would still occur after a period when within-host adaptation could already have occurred. Both the CFP and the YFP E. coli marked strains used in the transmission experiments were able to colonize another mouse gut already colonized with a marked strain, demonstrating their invasion potential (fig. 1C). We found transmission of the E. coli across mice to be pervasive and extremely fast (fig. 1C and supplementary table S2, Supplementary Material online), with cocolonization with both YFP and CFP lineages occurring in less than 24 h of cohousing both animals. On average, 1.5 × 10⁷E. coli cells (±1 × 10⁷, 2× standard error [2SE]) transmit to a new mouse per day, assuming that migration happens at a constant rate (supplementary table S2, Supplementary Material online). We thus estimate a rate of transmission of 7% (±3% 2SE) per day for this E. coli strain (fig. 2C and supplementary table S2, Supplementary Material online). This rate is similar to that observed in germ-free mice (Vasquez et al. 2021). This establishes that cohousing constitutes a regime of high migration and thus theoretically expected to result in high levels of allele sharing (fig. 1A and supplementary fig. S1A, Supplementary Material online).

Fig. 2.

Escherichia coli evolution experiment in a social and asocial context. (A) Loads of invader E. coli clone introduced by gavage into the mice in an asocial or social regimes. Error bars represent 2SE. The limit of detection (dashed line) is 330 CFU/g of feces. (B) Transmission examples of E. coli with specific mutations (e.g., psuK/fruA, ybi/fiu, yjfL/yjfM, frlR, and rnd) from donor to recipient mice living in a social regime. See supplementary table S6 and figures S5 and S6, Supplementary Material online, for more detailed information.

Rate of Molecular Evolution Under Cohousing

Having established that bacterial migration is common in cohoused mice with the transmission experiments described above, we performed a new set of experiments—evolution experiments—with a different experimental setting (supplementary fig. S2, Supplementary Material online) to test the influence of no transmission versus high transmission on the patterns of molecular evolution in the mouse gut. For these evolution experiments, all mice (no transmission regime: A2, B2 G2, H2, and I2; high transmission regime: A1, B1, C1, and G1, H1, I1) were colonized with a single E. coli strain expressing YFP, and the transmission rate was assumed to be identical to that observed in the aforementioned transmission experiments. In five out of six mice living asocially successful colonization with the invader E. coli was observed, whereas under social conditions, all mice were successfully colonized (fig. 2A and supplementary table S3, Supplementary Material online). Provided that colonization was successful for at least 1 month (27 days), we did not find significant differences in the loads of the invader E. coli in the asocial versus social environments, stabilizing around log10 6.3 (±0.58, 2SE) CFU/g feces (fig. 2A and supplementary fig. S3, Supplementary Material online).

We then sequenced pools of clones that evolved in the mouse gut for either a month (27 days) or more than 3 months (104 days) using Illumina technology and mapped the reads to the genome of their ancestor (supplementary table S4, Supplementary Material online). Parallel mutations, a sign of adaptive evolution, were noticed as previously (Barroso-Batista et al. 2014; Frazão et al. 2019; Barroso-Batista et al. 2020; Frazão et al. 2022), being present in both the E. coli populations isolated from animals in social and asocial settings (supplementary table S5, Supplementary Material online). The transmission of mutations such as psuK/fruA, frlR, rnd, yfjL/yfjM, and ybiX/fiu could be traced back by sequencing populations from previous sampling points. This allowed us to identify the mice where mutations were selected originally and then transmitted to the remaining mice in the same cage (fig. 2B and supplementary figs. S5 and S6 and table S6 Supplementary Material online). The intergenic mutation psuK/fruA was parallel and thus considered adaptive, being associated to the production of pyrimidine nucleotides from pseudouridine. The frlR mutation was found to be a common and adaptive mutation target in the gut. The corresponding gene plays a role in fructoselysine consumption and was very recently implied in E. coli chemotaxis and AI-2 signaling in the mouse gut (Laganenka et al. 2023). The rnd mutation is expected to affect the processing of precursor tRNA, and yfjL/yfjM is associated to the E. coli defective prophage CP4-57, while the ybiX/fiu mutation potentially affects the promoter region of the hydrolase gene (ybiX) (fig. 2B and supplementary figs. S5 and S6 and table S6 Supplementary Material online). No synonymous mutations were detected in clones sampled from the mice in a social setting, leading to a ratio of nonsynonymous to synonymous mutations dN/dS = infinity in all the animals. In populations isolated from individually caged mice, the dN/dS ratio was >1 in only 3/5 animals (supplementary table S5, Supplementary Material online). While the dN/dS statistic suggests that adaptive evolution may be stronger when bacterial transmission is occurring, it is also important to note that dN/dS may fail to detect adaptive evolution even when it is occurring (Kryazhimskiy and Plotkin 2008).

We then quantified how the dynamics of mutations accumulation in each host are affected by its social environment. The evolution rate of E. coli, when colonizing in a social or asocial setting (fig. 3A and B and supplementary table S5, Supplementary Material online), was assessed by dividing M(t) by the number of generations at a given sampling point. M(t) is the total allele frequency for all mutations in the population at a given sampling time, which approximates the expected number of mutations in a randomly sampled individual clone (Good et al. 2017). Mutations accumulated at an average rate of 3.8 × 10⁻³ (±1.0 × 10⁻³ 2SE) per genome per generation in the social regime, a rate higher but not significantly different from that observed in the asocial regime: 3.1 × 10⁻³ (±0.8 × 10⁻³ 2SE) (fig. 3B and supplementary table S5, Supplementary Material online), suggesting that the social regime did not affect the rate of molecular evolution in the number of generations followed in these experiments.

Fig. 3.

Molecular E. coli evolution in a social and asocial context. (A) Escherichia coli mutation accumulation during in vivo evolution. Mutations M(t) correspond to the sum of allele frequencies at each sampling point along time (generations). (B) Escherichia coli evolution rate based on the M(t). The E. coli evolution rate per genome per generation was calculated by dividing the M(t) by the number of generations for each mouse of each biological replicate of the social and asocial regimes. See supplementary table S5, Supplementary Material online. Biological Replicate 1 includes mice: A1, B1, and C1 (social regime) and A2 and B2 (asocial regime). Biological Replicate 2 includes mice: G1, H1, and I1 (social regime) and G2, H2, and I2 (asocial regime). Bars represent the average rate of evolution in social or asocial regimes. (C) Proportion of shared or adaptive evolutionary events observed in the invader E. coli populations isolated from mice living in an asocial or social regime at Day 27 or 104 (D). Proportion of shared or adaptive high-frequency evolutionary events (>50%) observed in the invader E. coli populations at Day 27 or 104. Statistics correspond to binomial test for proportions.

A Social Regime Boosts Sharing of Evolutionary Events

The number of mutations accumulated per mouse was marginally lower in the social environment (median 3.5 in the social vs. 6 in the asocial, Mann–Whitney two tailed test U = 5, P = 0.08; supplementary table S4, Supplementary Material online). In the first cohort of cohoused mice, 9 out of 12 (75%) mutations were common across all 3 mice, whereas in the second cohort, 6 out of 8 (75%) were common. In the asocial context, only 4 out of 16 (25%) mutations were common across all mice in the first cohort, while none was common in the second (supplementary table S4, Supplementary Material online). Consistent with the presence of a resident E. coli colonizing the mouse gut (Frazão et al. 2019) (supplementary fig. S4, Supplementary Material online), 15 phage-driven HGT events were detected during the same time period (27 days: ∼400 generations). Of these, 11 occurred in cohoused animals (social regime, n = 6 mice), while only four events were observed in individually caged animals (asocial regime, n = 5 mice), as the resident strain was present only in two out of the five animals (supplementary fig. S4, Supplementary Material online). In the first and second cohorts of cohoused mice, 100% of phage-driven HGT events were common across all animals, while in the asocial context, phage-driven HGT events were observed only in two animals where the resident was cocolonizing (supplementary table S7, Supplementary Material online).

We next evaluated how many evolutionary events (mutations and/or phage-driven HGT) are shared between individuals. The proportion of evolutionary events (mutation and phage-driven HGT) shared among hosts at Day 27 (∼400 generations) postcolonization is significantly higher under a social environment, 77.4% (24 shared out of 31 evolutionary events), than in an asocial regime, 11.8% (4 out of 34 events) (binomial test for proportions, P < 0.00001; fig. 3C and supplementary tables S4, S7, and S8, Supplementary Material online). We next asked whether the level of allele sharing in cohoused mice would still be higher after long-term evolution and sequenced pools of E. coli clones evolved for 104 days (∼1,500 generations) in the first cohort of mice. The sharing of evolutionary events was still significantly higher under cohousing (68.2%) than in the asocial environment (15%), even after more than a thousand generations of evolution (binomial test for proportions, P < 0.00001; fig. 3C and supplementary tables S4, S7, and S8, Supplementary Material online). Importantly, when focusing on bona fide adaptive evolutionary events, that is, those that were observed in more than one independent mouse, among the two social regimes, the same trend toward more genetic parallelism in the social regime was observed (fig. 3C). The proportion of adaptive events (mutations and phage-driven HGT) that were shared among the hosts at Day 27 (∼400 generations) was significantly higher under a social environment, 83.2% (15 out of 18 evolutionary events) than in an asocial regime, 12.5% (2 out of 16 events) (binomial test for proportions, P < 0.00001; fig. 3C and supplementary tables S7, S9, and S10, Supplementary Material online). Additionally, even after 104 days (∼1,500 generations), the proportion of adaptive events was significantly higher, 85.7%, in animals under a social regime than in asocial conditions, 35.3% (binomial test for proportions, P = 0.02444; fig. 3C and supplementary tables S7, S9, and S10, Supplementary Material online).

A Social Regime Increases the Proportion of High Frequency Evolutionary Events

To understand if the social regime affects the frequency trajectory of the evolutionary events (as predicted by the model; supplementary fig. S1D, Supplementary Material online), we started by analyzing the proportion of high-frequency events (>50%) among the total events observed in E. coli populations isolated from mice living in the social or asocial regimes. The proportion of total mutations or HGT events that reached high frequency (fig. 3D, frequencies >50%) was significantly higher at Day 27 (∼400 generations) when E. coli colonized mice living in a social regime, 41.9%, than in animals independently caged, 17.6% (binomial test for proportions, P = 0.03156; fig. 3D and supplementary table S8, Supplementary Material online). At Day 104 (∼1,500 generations), the proportion of high-frequency events (fig. 3D) remained high in the cohoused animals, 77.3%, while in mice living in independent cages, it was only 50% (binomial test for proportions, P = 0.03662; fig. 3D and supplementary table S8, Supplementary Material online). When the same analysis was performed focusing only on bona fide adaptive events (mutations and/or phage-driven HGT events observed in more than one independent animal), the proportion of high-frequency events (fig. 3D and supplementary tables S9 and S10, Supplementary Material online) was not significantly different between E. coli populations isolated from mice in social or asocial conditions. This could be due to lack of statistical power as we observed a limited number of mutational events when compared with the total of events. Nevertheless, adaptive events show the same tendency as when considering all evolutionary events, that is, a higher number of high-frequency events in the social regime (fig. 3D and supplementary tables S9 and S10, Supplementary Material online).

Discussion

Bacterial evolution in the mammalian intestine has been studied mainly under asocial conditions, where bacterial transmission among mammalian healthy hosts does not take place (Barroso-Batista et al. 2014, 2015; Lourenço et al. 2016; Lescat et al. 2017; Sousa et al. 2017; Frazão et al. 2019, 2022; Ghalayini et al. 2019; Ramiro et al. 2020). To our knowledge, the evolution of E. coli in the presence versus absence of bacterial transmission in mice bearing a microbiota is compared here for the first time. We established a mouse model of bacterial transmission to compare E. coli evolution when colonizing the gut of mice living in a social or asocial regime. We found that in mice inhabiting the same cage 7% (±3% 2SE) of the E. coli colonizing each mouse gut results from daily migration events. It is known that E. coli can adapt to better colonize the mouse intestine in less than a week of colonization, with the mutation and phage-driven HGT events occurring concomitantly (Frazão et al. 2019). Here we observed that E. coli transmission between hosts can still occur even when an adapted E. coli is already colonizing the gut. Interestingly, the first clone to colonize a given mouse was not always the one ending up as dominant after a month of evolution, suggesting that priority effects (Sprockett et al. 2018) should not be dominant in this species.

The E. coli population size, a key aspect in determining the level of variability in a population and the effectiveness of selection relative to drift (Charlesworth 2009), was not affected by transmission, with populations isolated from mice living socially or asocially presenting similar loads. The E. coli cells transmitting among microbiome-bearing mice was comparable to the 10% migration inferred in a study using germ-free animals (Vasquez et al. 2021), suggesting that E. coli transmits between hosts at a very similar rate independently of the presence or absence of a microbiome. Interestingly, we show invasion and coexistence of the same strain of E. coli (expressing different fluorescent markers) in mice living in a social regime, suggesting that strain colonization resistance, observed for other important species in the gut such as Bacteroides fragilis (Lee et al. 2013), is absent from the E. coli species.

Importantly, whole-genome sequencing of populations revealed that when in a social regime, where host-to-host bacterial transmission is pervasive, the evolution of E. coli is always characterized by an elevated ratio of nonsynonymous to synonymous mutations, indicative of evolution driven by strong selection of adaptive haplotypes, a phenomenon we found to be less common in an asocial regime.

Our theoretical model of bacterial transmission predicted that the outcome of bacterial migration between hosts would affect the evolutionary process in terms of the number of shared evolutionary events between hosts. The model simulations predicted that when the number of migrant clones between mice is large enough (≫1), the metapopulation of cohoused mice acts as a single population and genetic changes are shared across hosts. Alternatively, when transmitting a very low number of bacterial migrants, each host behaves as an independent population, with the level of sharing of genetic evolutionary events being extremely low. Indeed, a significantly higher proportion of shared evolutionary events (total and adaptive) was found in E. coli populations from the social condition. These observations relate not only to the mutational process but also to the evolutionary phage-driven events. We were able to study the latter process because our mouse cohort was colonized with a resident E. coli carrying several active prophages that can be transferred to the invader strain, thus conferring an adaptive potential to consume sugars that are common in the gut (Frazão et al. 2019). The adaptive mutational process targeted mostly the same biological processes in both invader E. coli populations, isolated from the social or asocial regime, namely, production of pyrimidine nucleotides from pseudouridine (psuK/fruA mutation) and fructoselysine consumption (frlR mutation).

Phages are known key players in shaping the composition and diversity of bacterial communities in many environments, namely, in the gut of mice and humans (Kim and Bae 2018; Frazão et al. 2019; Sutton and Hill 2019). Here, we observed that phage-driven HGT events, namely, the phages Nef and KingRac, were transferred to the invader E. coli genome as observed previously (Frazão et al. 2019). This process was independent of the social or asocial condition, though limited to when cocolonization with both the resident and invader lineages occurred. Nevertheless, under a social regime, bacterial transmission across hosts spread these adaptive phage-driven events across all cohoused animals.

We found that under the social regime, total evolutionary events reached significantly higher frequencies in the invader population, with adaptive events following the same tendency. This was indeed predicted in the simulation modeling, where the highest frequency mutations were more common at higher levels of bacterial transmission. This suggests that unlike in asocial contexts, in a social regime, a beneficial mutation reaching high frequency should be more prone to transmit to another host and avoid the drift barrier.

Invader populations evolving during transmission did not show any signs of diversifying selection. On the contrary, these populations exhibited genetic homogeneity and an elevated dN/dS ratio indicative of adaptive evolution driven by strong selection. The present study also reveals that the evolution rate of E. coli is unchanged in both social and asocial settings, being on average 3.0 × 10⁻³ (±0.8 × 10⁻³ 2SE) mutations/genome/generation. A possible explanation is that in a social regime, due to E. coli transmission, the number of adaptive mutations in each animal is expected to be higher, given that these can originate not only de novo in each animal but also in other hosts being subsequently transmitted. By sweeping in the population, transmitted adaptive mutations may reduce genetic diversity and the number of nonadaptive mutations in the receiving mouse, thus leading to higher adaptive evolution (high dN/dS ratio) while preserving the same molecular evolution rate as in the asocial context.

Interestingly, the evolution rate found is similar to those reported in previous studies in vitro (Good et al. 2017) and in vivo (Frazão et al. 2022), where bacterial host-to-host transmission was not taking place. Thus, as proposed previously (Frazão et al. 2022), E. coli appears to follow a clock-like rate of evolution, which the present data show is independent of the host social context where E. coli evolves.

We conclude that selective pressures acting on bacteria depend on the host's social status and that the effect of migration should be taken into account when analyzing the evolutionary history of a bacterium colonizing the mammalian gut. Moreover, in contrast to a social context, social isolation can lead to psychoemotional stress, impair normal development of organs and tissues (Grigoryan et al. 2022), and can alter the gut microbiota (Donovan et al. 2020). These differences related to the host's social status could constitute an additional layer of selection pressure besides bacterial transmission, also contributing differential bacterial evolution in the gut.

Our findings demonstrate that the evolution of new bacterial strains invading the gut microbiota can be strongly affected by the host's social environment, which calls for further experimental evolution studies comparing social versus asocial conditions.

Materials and Methods

Escherichia coli Clones

The clone used in the present study is described in supplementary table S1, Supplementary Material online. Ancestral and evolved E. coli clones were grown at 37 °C under aeration in brain heart infusion (BHI) or Luria broth (LB). Media were supplemented with antibiotics when specified. Serial plating of 1× phosphate buffered saline (PBS) dilutions of feces in LB agar plates supplemented with the appropriate antibiotics was incubated overnight, and YFP-labeled bacterial numbers were assessed by counting the fluorescent colonies using a fluorescent stereoscope (SteREO Lumar, Carl Zeiss) with the detection limit for bacterial plating being 330 CFU/g of feces.

In Vivo E. coli Transmission and Evolution Experiments

Mice drank water with streptomycin (5 g/L) for 24 h before a 4-h starvation period of food and water. The animals were then inoculated by gavage with 100 µL of an invader E. coli bacterial suspension of ∼10⁸ colony forming units (CFUs). For the transmission experiments, invader E. coli clones expressing either CFP or YFP fluorescence were used to assess bacterial transmission between animals. In the evolution experiments, mice littermates A1/A2, B1/B2, C1/C2, G1/G2, H1/H2, and I1/I2 were all gavaged and successfully gut colonized with a single invader E. coli clone (YFP), except mouse C2. The same letter describes animals that are littermates, while number 1 or 2 means the mice are in a social or asocial regime, respectively. Five (A2, B2, G2, H2, and I2) out of the 11 mice colonized with E. coli were already followed, in previous studies (Frazão et al. 2019; Frazão et al. 2022). Six- to 8-week-old C57BL/6J female mice were kept in ventilated cages under specified pathogen free (SPF) barrier conditions at the Instituto Gulbenkian de Ciência (IGC) animal facility. The gut microbiota of mice used in the experiments is natural to the animals and not the result of the introduction of a defined microbiota into germ-free animals. Fecal pellets were collected and stored in 15% glycerol at −80 °C for later analysis. This research project was ethically reviewed and approved by the Ethics Committee of the ICG (license reference: A009.2010) and by the Portuguese National Entity that regulates the use of laboratory animals (Direção Geral de Alimentação e Veterinária [DGAV]; license reference: 008958). All experiments conducted on animals followed the Portuguese (Decreto-Lei n° 113/2013) and European (Directive 2010/63/EU) legislations concerning housing, husbandry, and animal welfare.

Whole-Genome Sequencing of Evolved E. coli Populations

DNA was extracted (40) from E. coli populations (mixture of >1,000 clones) growing in LB plates supplemented with antibiotic to avoid contamination. DNA concentration and purity were quantified using Qubit and NanoDrop, respectively. DNA library construction and sequencing were carried out by the IGC Genomics facility using the Illumina platform. The mean coverage per population was 200 × and processing of raw reads and variants analysis was performed as previously described (Frazão et al. 2019; Frazão et al. 2022).

Model of Bacterial Transmission

A model of mutation–selection–migration and genetic drift was used to predict the effect of bacterial transmission on the landscape of mutations shared among populations of E. coli colonizing the gut of mice. For simplicity, we assume that there are two demes of equal carrying capacity (N). Individuals reproduce clonally within each deme and acquire deleterious mutations at a rate U_d and fitness increasing mutations (adaptive mutations) at a rate U_a per generation (typically 1/100 of U_d; Perfeito et al. 2007). Under the simplest model, the effect of a deleterious mutation is assumed to be constant (S_d) as well as that of an advantageous mutation (S_a). A distribution of fitness effects, where the effects of both mutations are taken from exponential distributions of mean S_d or S_a, is also modeled to establish the generality of the qualitative expectations. Selection acts within each deme, and N individuals are chosen according to their fitness. The model ignores any trade-offs of mutation effects outside the host environment, implicitly assuming transmission by direct contact. Migration follows selection, where a Poisson number of migrants (Nm) is exchanged between two demes according to the migration rate m. To measure the level of shared adaptations across demes, each individual adaptive portion of the genome is explicitly modeled with a finite size G and a biallelic model for adaptive mutations. Thus, the mutation rate toward new advantageous alleles is U_a/G per site per generation. Deleterious mutations follow an infinite site model, as these are in principle much more numerous than beneficial mutations.

Statistical Analysis

A linear mixed-effects model was used to do the analysis of temporal load differences of the invader E. coli from Days 9 to 27, with time as repeated measures and housing as treatment factor. An analysis of variance (ANOVA) test was used to compare the evolution rate in social and asocial groups. A binomial test was used to compare proportions in the two regimes (social and asocial). A P < 0.05 was considered for statistical significance.

Supplementary Material

Supplementary data are available at Molecular Biology and Evolution online.

Author Contributions

I.G. and N.F. designed and coordinated the study. N.F. performed the experiments. I.G. performed the simulations. N.F. and I.G. analyzed the results and wrote the manuscript giving final approval for publication.

Data Availability

Raw sequencing reads were deposited in the sequence read archive under the name of bioproject PRJNA930727. The code for the model of mutation–selection–migration and genetic drift is available at the GitHub platform (https://github.com/isabelgordo/SharedEvolutionSocialMicrobiomes.git).