Discussion
Using a common garden experiment we show that differences in gut microbial diversity across Cyprinodon pupfish species largely reflect phylogenetic distance among generalist populations in support of phylosymbiosis (Bordenstein and Theis 2015), rather than novel trophic specializations as predicted by adaptive radiation theory. Our study is highly consistent with Ren et al. (2016) which also found limited microbiome divergence and minimal associations with ecomorph in an adaptive radiation of Puerto Rican Anolis lizards, even within wild lizards. Gut microbiome diversity has also been found to associate more strongly with geography than phylogeny (Godoy-Vitorino et al., 2012) or a combination of geography, diet, and host phylogeny (Antonopoulou et al., 2019). These emerging studies of microbiome divergence within adaptive radiations of hosts provide an important counterpoint to the classic expectation of rapid phenotypic diversification and speciation during adaptive radiation (Schluter 2000; Stroud and Losos 2016; Martin and Richards 2019; Gillespie et al. 2020).
A major caveat is that we did not examine the microbiota of wild-collected animals feeding on their diverse natural resources of macroalgae, scales, and snails. Scales form up to 50% of the diet in scale-eaters (Martin and Wainwright 2013) and wild gut microbiome samples surely would have revealed more substantial differences in microbiome diversity and composition among generalist and specialist species on San Salvador Island. However, our goal with this common garden study using lab-reared animals fed an identical generalist-type diet for one month was to uncover any genetically based microbiome differences in these taxa by eliminating environmental effects as much as possible. Pupfishes exhibit no parental care and deposit external eggs on the substrate so vertical transmission also appears highly unlikely (but see Satoh et al. 2019 for a potential example of vertical transmission in a scale-eating cichlid). Furthermore, by including two lab-reared colonies of each generalist and specialist species on San Salvador from genetically differentiated and ecologically divergent lake populations (Martin et al. 2016; Richards and Martin 2017), we aimed to connect significant differences in microbiome composition observed in our specialist species to their specialized diets, rather than their lake environment or genetic background. This provides strong evidence of genetic divergence in the host associated with trophic specialization. These results are all the more surprising because trophic specialists show very little genetic differentiation from generalists (Fst = 0.1 – 0.3; Martin and Feinstein 2014; Richards et al. 2020). Indeed, there are only a few thousand nearly fixed or fixed SNPs (Fst > 0.95) between scale-eaters and molluscivores out of over 10 million segregating SNPs and as few as 157 fixed SNPs and 87 deletions in scale-eaters (McGirr and Martin 2020). However, this minimal set of genetic differences may be driving differences in gut microbiome composition. Intriguingly, the only fixed coding indel uncovered so far in this system is a fixed deletion in all scale-eater populations of the fifth exon of the gene gpa33 (McGirr and Martin 2020). This gene is expressed exclusively in the intestinal epithelium and mice knockouts display a range of inflammatory intestinal pathologies in mice (Williams et al. 2015), suggesting it may play a role in shifting the gut microbiota of scale-eaters that we observed in this study. Overall, metabolic processes were the single most enriched category among all differentially expressed genes between these trophic specialists at the 8 dpf larval stage, accounting for 20% of differential expression (McGirr and Martin 2018).