Focal strain evolution was different in each community
We measured performance (competitive fitness) of the evolved populations relative to the ancestral population when both were grown in separate dialysis bags embedded within the community at the end of the experiment (see Supplementary Fig. 1). The focal strains (rows of Fig. 1) had wide variation in performance across the different communities (columns of Fig. 1). Some evolved populations adapted strongly to laboratory conditions and their population densities greatly increased compared to ancestral populations (e.g. Raoultella sp.2, Curtobacterium sp.1, Rhizobium sp.1), while others showed weak or non-significant adaptation (e.g. Serratia sp., Microbacterium sp.). Similarly, some communities (columns of Fig. 1) were permissive in allowing focal strains to adapt (e.g. communities 1, 3, 4, and 7), while other communities were restrictive (e.g. communities 2, 5, 6, and 8). Similarities in evolvability were not simply explained by genetic similarity: Raoultella sp.1 clearly had the capacity to adapt rapidly in some communities, but Raoultella sp.2 did not show the same pattern. Instead, the evolvability of most species depended on the interaction between focal strain and community, which explained 38% of the observed evolutionary change, while focal strain and community alone only explained 13% each (Fig. 1, pie plot, Supplementary Table 1).
Each row is a focal strain, and each column is the community in which it evolved. Performance was measured as the growth of the evolved focal strains divided by the growth of the ancestral strain in the evolved community over a period of 2 weeks. Blue squares indicate that the evolved focal strain grew significantly better than its ancestral counterpart, while red squares indicate that the evolved strain grew significantly worse. Non-significant differences are not shown. Colour intensity reflects the significance level (n = 4). The pie plot shows the overall amount of variance explained by species, communities, and their interaction. Details of the statistical analysis are given in the Supplementary Methods. Source data are provided as a Source Data file.
Extrinsic and intrinsic properties determine evolution
We quantified the impact of extrinsic factors on the evolvability of the focal strains by characterising the communities according to their diversity (Shannon’s Index) and their ecological ‘robustness’ in response to an environmental perturbation. Ecological robustness was quantified as the degree to which the instantaneous activity of each ancestral community was reduced by the low pH conditions (see Supplementary Methods, Supplementary Fig. 4); robust communities were little impacted by changes to pH. Our findings demonstrate that focal strains evolved higher performance in communities with low diversity (low Shannon’s index, Fig. 2a). The result implies that, while bacterial strains have great capacity to adapt, as has been shown in numerous simple laboratory systems1,2,3,7,16,33, that capacity is reduced as communities become more diverse. This does not appear to result simply from the richness (number of taxa) but is a consequence of an uneven distribution of abundance. Several factors could contribute to this pattern. For example, focal species might adapt more easily to a few dominant species (low Shannon’s index) than to multiple moderately abundant species. Diversity and community robustness both mainly negatively affected resource degradation (Fig. 2a), so competition for resources might be a factor that limits adaptation of the focal strain in diverse communities27. Alternatively, low-diversity communities themselves might adapt less rapidly to new conditions because a greater proportion of the species are at low abundance, giving the focal species a relative advantage. We also observed that evolved strains had a greater capacity to use chitin as a resource in communities with low robustness. This may indicate that chitin was more available in communities that were struggling with pH 5. While we cannot definitively separate these mechanisms using the available data, further experiments could test whether resource competition was driving these patterns, for example by creating environments where resources were not limiting to growth. Regardless of the underlying cause, the result shows that properties of the background community alter focal strain adaptation.
a We modelled how each response variable (different colours in the figure) was impacted by extrinsic factors and their interaction (rows in the figure), and b showing the effect of intrinsic factors and their interaction. The t-statistics associated with each factor are shown, with positive/negative t-statistics indicating a positive/negative association between the factor and the response variable. Only those factors that were retained in the best model are shown. Non-significant interactions that were not retained for any factor are not shown. A full explanation of the statistical modelling is given in the Supplementary Methods. Non-linear trends are indicated by the square of the variable (e.g. Shannons2); detailed definition of the extrinsic and intrinsic factors (Shannons, Robustness, Phylogeny, etc.) are given in the Supplementary Methods. Source data are provided as a Source Data file.
We also quantified four intrinsic properties of the ancestral focal strains: (i) their degree of maladaptation when encountering the low pH conditions in monoculture (i.e. instantaneous reduction in activity at pH 5.5 compared to pH 7, Supplementary Fig. 4b), (ii) their mean abundance across all communities, (iii) their inferred genome size, and (iv) their phylogenetic distance from one of the focal strains (see the “Methods” section). We found that strains with larger genomes were associated with a greater capacity to adapt to the laboratory environment (Fig. 2b). Larger genomes are more likely to contain pre-adaptations to new conditions42, have higher number of mutations per generation, and might contain multiple gene copies, allowing one copy to change with little fitness cost42. Focal strains with the highest evolvability were also relatively rare on average in their native communities and were more closely related phylogenetically to the strain with the highest average evolvability (Raoultella), thus potentially occupying underutilised niches. Finally, we found that maladapted focal strains evolved more, which might reflect that well-adapted species experience little selection pressure and hence respond slowly to selection7.
The extrinsic and intrinsic performance correlates point toward a key role of pre-adaptation in providing the opportunity for further adaptation. Under this model, ecological opportunity is granted by extrinsic factors (e.g. a low-diversity community), while the capacity to adapt is governed by pre-adaptation (degree of maladaptation) and the available toolkit (genome size). Our experimental design is unable to confirm a causal link between these factors and the focal populations, but we hypothesised that the basis for extrinsic and intrinsic effects revolved around pre-adaptation for resources within the microcosms, and whether the focal strains could capitalise on the resource environment left underexploited by the background community. We characterised the resource niches of the ancestral and evolved focal strains by quantifying their potential to degrade labile (xylose), recalcitrant (cellulose), and intermediate (chitin) substrates (Supplementary Table 2) found in tree-hole environments. We found that the intrinsic and extrinsic factors had significant impacts on the evolution of resource usage, particularly related to cellulose and chitin degradation (Fig. 2). The ability of the focal strains to exploit the more recalcitrant resources was constrained by extrinsic factors. Cellulose degradation was constrained in more diverse communities and chitin was constrained in more robust communities. Improved chitin degradation was also associated with some intrinsic factors, notably genome size. There was little evolution to use xylose (the most labile substrate), perhaps because rates of degradation could not be improved or because xylose was rare late in the experiment when we assume most of the labile substrates had been degraded.
Evolved focal strains used recalcitrant resources
A more detailed look at how substrate usage evolved revealed a remarkably consistent evolutionary trajectory (Fig. 3a). First, for most strain–community combinations there was little change in enzyme activity, indicating that the strains were either constrained to their ancestral niche by intrinsic or extrinsic factors, that the ancestral niche was a local optimum, or that there was insufficient time for significant evolutionary change (see the “Methods”section). For those focal strains that did respond to selection with a niche shift, there was a consistent change in substrate usage, with an increased ability to use cellulose mainly at the cost of chitin, and with less impact on xylose. Overall, the result is plausible, since the low replacement rate of fresh media would likely have led to the accumulation of recalcitrant substrates38. Finally, there was a tendency for the response to selection (length of the arrows in Fig. 3a) to be highest for those ancestral focal strains that started with the highest cellulose and chitin degradation ability (Fig. 3b, Supplementary Table 3). The result again implies an important role of pre-adaptation to the resource environment43,44.
a We measured the ability of each focal strain to degrade three substrates (xylose, chitin, cellulose). The base of the arrow indicates the mean ability of the ancestor to degrade the substrates, and the arrowhead indicates the mean ability of the evolved strain to degrade the substrate. Evolved and ancestral substrate degradation was averaged across replicates. Colours correspond to the focal strain. b We investigated how ancestral resource utilisation influences the response to selection. The x-axis is the resource utilisation of the ancestral focal strains. The y-axis is the response to selection, measured as the change in resource utilisation from the ancestral to the evolved for each focal strain in every community. These distance values correspond to the length of the arrows in Fig. 3a. Lines represent best fit lines and dotted lines represent 95% confidence intervals highlighting the steeper slope for cellulose and chitin (see Supplementary Table 3 for statistics). Source data are provided as a Source Data file.
Genetic variants underlying adaptation
We used genome sequencing to determine whether genetic changes underlie the observed phenotypic changes. In particular, we conducted shotgun sequencing of populations of evolved Raoultella sp.1 and sp.2 across the communities (Fig. 4, Supplementary Table 4) and compared these to the genomes sequenced from the ancestral populations. This approach revealed several genetic variants, including variants that were gained over the course of the experiment. In addition, several genetic variants arose during the initial growth of the originally clonal ancestral population and were either lost or changed in frequency during the experiment, as has been observed in comparable studies45,46. There was evidence that some of the variants were involved in carbon metabolism (kdgT gene; a 2-keto-3-deoxy-d-gluconate transporter that regulates pectin metabolism) and in inter-specific interactions (pipB2 gene, a secreted effector protein that can modify cell interactions), consistent with our observations that substrate degradation and community interactions could explain the evolutionary trajectory of the focal populations. However, most of the genes involved have not been annotated, so further work will be needed to confirm which of the variants are responsible for the changes in focal strain performance and metabolic phenotypes we observed. There was no clear impact of community diversity on distribution and identity of the genetic variants in evolved Raoultella strains, which may be because selection was targeting similar loci in different communities. Alternatively, some of the phenotypic changes we observed that were linked to changes in community diversity could have been due to behavioural (rather than genetic) modifications in the focal strains. Finally, the reduced statistical power to detect correlations in this smaller sample set could have obscured a signal, which could be remedied by more comprehensive sequencing across the focal strains or by exploring alternative statistical models.
a Genetic variants detected in Raoultella sp.1 focal strains. b Genetic variants found in Raoultella sp.2 focal strains. Points show significant changes in SNP frequency compared to the ancestral isolate, with colours indicating a significant gain (green), loss (yellow), or significant change in frequency (blue) of genetic variants. Gene names are given for SNPs that lie within known genes. Variants without annotation are hypothetical proteins. The size of the points is proportional to the number of evolved focal strains from the different communities showing the variant class. Source data are provided as a Source Data file.
Final communities were dominated by similar taxa
To see whether the biotic environment changed during the experiment, we determined the change in the background communities by 16S amplicon sequencing. Although all final communities were distinct and highly diverse, initial differences between the communities dissipated during the experiment, and final communities converged such that there was no significant difference among the communities at the conclusion of the experiment (Fig. 4a, PERMANOVA, d.f. = 7, F-value = 0.7679, p = 0.987). In particular, all of the final communities were dominated by Erwinia sp., Klebsiella sp., Serratia sp., Pseudomonas sp., or Pantoea sp. (Fig. 4b). A likely explanation of these dynamics is ecological sorting of species with environmental filtering47, but the phenotypic and genetic evolution observed in the focal strains suggest that evolutionary processes in the communities could also play a role. The result also implies that the variation in response to selection of the focal strains in the different communities (Fig. 1) likely resulted from differences in community composition early in the experiment.
Although different abiotic selective pressures would undoubtedly produce different results in terms of which strains and genes evolved, we suggest that these results illustrate general features of bacterial communities: that extrinsic and intrinsic impacts on adaptation cannot be viewed in isolation, since the interaction between these two components is considerably more important than each one alone (Fig. 1). The change in substrate degradation rates by the focal strains (Fig. 3) provides a likely mechanism by which this occurs. Communities where competition is high (diverse communities or communities that contain a high relative abundance of congeneric taxa) act to constrain adaptation, likely through pre-emption of the available niche space4,27,30,33,48,49. In this scenario, extrinsic factors provide the opportunity for adaptation, but focal strains can only capitalise on the opportunities if they have the capacity. The interaction between ecological opportunity and evolutionary capacity is likely to have a great influence on how populations will respond to changing environmental conditions, for example due to global change, or exposure to antibiotics. Our experiment provides a preliminary insight into how bacterial population navigate adaptive landscapes when they are embedded within complex communities, revealing that understanding the biotic environment is vital for understanding evolutionary trajectories. More work will be needed to disentangle mechanisms of the changes we observed, but future work on evolving communities will inevitably require the joint consideration of intrinsic species variables and external community variables.
Source: Ecology - nature.com
