Numerous studies on diverse vertebrate species have convincingly demonstrated that the composition of bacterial communities of the gut is affected by diet [13, 51-54]. What remains largely unknown are gut microbiota dynamics during the host’s adaptation to novel food sources, particular during early stages of species divergence [but see 14, 15]. To address this question, we studied trophic ecology and the gut microbiota of repeated Nicaraguan Midas cichlid crater lake radiations, a model system for rapid ecological diversification and speciation [19, 20, 29]. We asked whether the parallel evolution of trophic diversification is associated with respective changes of the gut microbiota among sympatric crater lake species (i.e., are the gut microbiotas of ecologically similar species that independently evolved in two crater lakes more similar to each other than they are to their closest relatives of the same lake?). Our results suggest that among individuals of the same crater lake differentiation in trophic ecology and the gut microbiota are, to some extent, associated, hinting at the importance of diet in shaping the gut microbiota. However, only a small proportion of gut microbiota variance is explained by differences in diet. Hence, future studies need to address the contribution of other factors to obtain a more comprehensive picture of gut microbiota dynamics in this system. Moreover, we found that interspecific variation in trophic ecology (measured as stable isotope ratios of carbon and nitrogen) is significantly higher than intraspecific variation, a pattern that was much less pronounced for the gut microbiota (Fig. 4). We did not find strong evidence for parallel changes of the gut microbiota across crater lakes, suggesting that diet affects Midas cichlids’ gut microbiota differently in these lakes. We want to emphasize that our analyses are restricted to bacterial communities harbored within the midgut of these fish and should be interpreted in that context. Signatures of differentiation might be different if another part of the gut (e.g., hind gut) is analyzed, and future studies will need to address this question.
Gut microbiota differentiation across lakes
Comparing bacterial communities from water of their natural habitats with those harbored in fish guts clearly revealed that, although some bacterial taxa are shared, the gut microbiota not merely represents the bacterial community of the natural environment. Water and gut samples were obtained at different time points and sequenced independently on different sequencing platforms (see Methods section for more details). Temporal variation and technical aspects might have contributed to the observed differences in microbial communities (Fig. 2). Thus, caution should be taken when drawing conclusions about the extent to which the gut microbiota of Midas cichlids is derived from bacteria present in the water based on our data. Future studies need to systematically investigate this question in more detail. However, the differences in microbial community composition (Fig. 2) and abundance of major bacterial (Fig. 3) are so strong between water and gut samples, suggesting that they represent biological signals and that the gut microbiota of Midas cichlids is largelycontrolled by the host, as has been found for other fishes [14, 15, 55]. In accordance with a previous study on this system [17], Proteobacteria, Firmicutes, Fusobacteria and Bacteroidetes were the dominant bacterial phyla of Midas cichlids’ gut microbiota. These bacteria are also found in many other freshwater fishes [14, 15, 56]. Albeit the bacterial diversity of environmental samples strongly differed among lakes (Fig. 3B), the gut microbiota of Midas cichlids, except for A. citrinellus from Lake Managua [but see 17], showed constant levels for this measure. This provides further evidence that the diversity of bacterial species in the gut might be constrained by the host and stabilized at a given level, as predicted by the holobiont theory [57]. In A. citrinellus from Lake Managua, bacterial diversity was by far the lowest among all populations (Fig. 3B) and was also lower in water from Lake Managua compared to Lake Nicaragua. The city of Managua, Nicaragua’s capital with a population of more than 2 million, is located on the shore of Lake Managua and for decades, domestic and industrial waste water has been disposed into the lake [58]. As a result, concentrations of mercury and other toxic substances are extremely high in the lake and are also enriched in fishes [58, 59]. Mercury levels have been shown to be correlated with δ15N values [60, 61] and Midas cichlids from Lake Managua showed the highest δ15N values (Fig. 1), in agreement with the observation that mercury accumulates in these fishes. Further, aquatic pollutants such as heavy metals or pesticides have been shown to alter community composition and reduce diversity of the gut microbiota in aquatic organisms [reviewed in 62]. Albeit speculative at this point, high levels of contamination might have decreased the bacterial diversity of Lake Managua as well as the gut microbiota diversity of Midas cichlids inhabiting this lake, pointing to the combined influence of host and environment in shaping the gut microbiota.
Association between trophic ecology and gut microbiota in crater lake adaptive radiations
Midas cichlids from crater lakes Apoyo and Xiloá represent an excellent model to study the dynamics of gut microbiota changes during early stages of ecological diversification and speciation. Species repeatedly diverged only very recently and show differentiation in trophic ecology (Fig. 1), albeit diet overlaps to varying degrees among species [24]. Therefore, we tested whether gut microbiota differentiation is associated with trophic ecology of crater lake Midas cichlids.
In general, we detected significant differences of gut bacterial community composition among sympatric species in both crater lakes, which is in contrast with previous results [17]. These discrepancies could have been the result of several differences between the two studies: (i) sample sizes were larger in our study, (ii) we included one additional species in crater lake Apoyo (A. globosus), which could affect the PERMANOVA test statistics, (iii) sequencing depth per individual was higher in our study after rarefaction (20,000 vs. 15,000 reads), which might increase the number of rare ASVs in our study as exemplified by the rarefaction curve (Fig. S1) and (iv) different regions of the gut might have been sampled. We found some evidence for positive correlations between differentiation of the gut microbiota with carbon and nitrogen isotope signatures in both crater lakes (Fig. 4). However, we want to emphasize that results varied depending on distance metric (weighted UniFrac, unweighted UniFrac, Bray-Curtis dissimilarity) and statistical test (Mantel test or Pearson correlation). Overall, it appears that divergence of trophic ecology and the gut microbiota are to a certain degree associated, suggesting that adaptation to different food sources necessitated changes of the gut microbiota. But, these patterns appear not to be produced by shifts in abundance of similar bacterial taxa as we detected no evidence for parallelism in gut microbiota changes across the two crater lakes. It should also be noted that a substantial amount of gut microbiota variation is not explained by trophic divergence. Differences in stable isotope ratios, reflecting the host’s trophic ecology, were considerably higher across species compared to within species in both crater lakes (Fig. 5A & B). In contrast, such differences were much less pronounced for the taxonomic composition of the gut microbiota and the predicted functional bacterial metagenome (albeit statistically significant in some cases; Fig. 5C & D). This could mean that occupation of novel trophic niches might be achieved without drastically changing the overall composition of the gut microbiota, both taxonomically and functionally. Rather, subtle changes in some functionally important bacterial taxa might suffice to exploit new food sources and to allow ecological and evolutionary diversification of the hosts. Alternatively, the very recent divergence of crater lake Midas cichlids might impede clear differentiation of the gut microbiota, as discussed in more detail in the following paragraph.
Parallelism and non-parallelism of the gut microbiota in fishes
Parallel changes of the gut microbiota associated with differentiation in trophic ecology have been reported for older fish species [13], whereas other studies found no evidence for parallelism among more recently diverged populations [14, 15, but see 16]. In the very recent Midas cichlid adaptive radiations from Apoyo and Xiloá that diverged less than 1,700 and 1,300 generations ago, respectively [20], parallel changes in diet led us to expect that a similar pattern could also be found in the gut microbiota. However, we did not detect evidence for parallel changes of the gut microbiota. The only exception to this was a significant association of the predicted functional metagenome with trophic position (measured as normalized δ15N). These results indicate that functional rather than taxonomic characteristics of the gut microbiota might be important during early stages of trophic divergence. Taken together, studies of multiple groups of fishes suggest that parallel changes of the gut microbiota might only be expected on longer time scales [13]. These observations can be explained by the fact that during early stages of divergence, species might occupy novel niches but diet, to varying degrees, still overlaps among young species or ecotypes.
In crater lake Midas cichlids, stable isotope analyses showed that species largely occupy distinct niches with varying levels of overlap among species (Fig. 1). Combining these results with previous stomach content analyses suggest that these species are generally omnivorous and mainly feed on similar food items but their relative proportions differ among species [24]. Note that the study by Elmer et al. [24] classified Midas cichlids only as benthic or limnetic, thus, variation among benthic species had not been investigated to date in this system. The results of the stable isotope analysis suggest that young crater lake species might be in the process of adapting to specialized ecological niches but currently they are still opportunistic generalists with a varying diet. Although Midas cichlids from the two crater lakes diverged in trophic ecology in parallel, we did not detect evidence for parallel changes of the gut microbiota. One possible explanation for the lack of microbiota parallelism is that there might be hidden variation in prey items that are not captured by stable isotope data. Stable isotope data are suitable for showing general differences in trophic ecology, but they don’t provide detailed information on the exact prey items an organism feeds on. While it has been shown that prey items are largely similar between species and also across crater lakes [24], future studies that incorporate data on the gut microbiota, stable isotopes and stomach contents are needed to investigate this possibility in more detail. Further, short term changes of an individual’s diet are not reflected in the stable isotope signature of muscle tissue as this represents an average of this individual’s diet over a period of approximately three months [63, 64]. In contrast, the composition of the gut microbiota is highly variable and changes rapidly with diet [65, 66]. Hence, the gut microbiota rather represents a snapshot of an individual’s most recently acquired food items, generating high levels of intraspecific variation. This could explain why intra- and interspecific variation of the gut microbiota is much more equal compared to stable isotope data (Fig. 5). Accordingly, high levels of intraspecific dietary variation might mask interspecific differences in trophic ecology, thereby blurring any signal of gut microbiota parallelism in recently diverged ecotypes or species. This is what we can also see in other fishes like whitefish and guppies, where the main change between ecotypes is in the relative proportion of food items [14, 15]. In contrast, benthic-limnetic species pairs of threespine stickleback show little overlap in diet [67, 68], which might explain the strong and parallel changes of the gut microbiota, despite the young age of these species [16]. Only after species sufficiently diverged to become trophic specialists that do not overlap in food items, one would expect persistent and parallel patterns of gut microbiota divergence, as seen in African cichlids or stickleback [13, 16].