Results

Environmental differences among HP and LP environments varied mostly as expected based on what is already known about Trinidadian streams. Namely, lower elevation, high-predation (HP) sites had higher stream water temperature (HP: 24.87°C, ±0.08 vs. LP: 22.82°C, ±0.03) and pH (HP: 7.5 ±0.03, LP: 6.7 ±0.03) compared to low-predation sites (p<0.01 for both comparisons), which include LP native, old, and new introductions (see Fig. S1 for site-level stream data). Fish morphology also fit expectations: fish in LP environments had longer gut lengths relative to body size (Fig. 2A) and were larger overall (Fig. 2B). Site-specific variation overwhelmed significant effects of site category on gut content (as measured by percent invertebrates, Fig. 2C), and fish condition (weight/length, Fig. 2D). If anything, at the time of sampling, LP fish were eating more invertebrates (Fig. S1D), contrary to expectations. HP fish had diets richer in detritus, as well as higher relative abundance of cyanobacteria (according to 16S) in their guts (p<0.001, Fig. S2, Table S4).
After pre-filtering steps (Table S1), we identified 12,068 bacterial OTUs in 485 guppy gut microbiome samples. In general, the posterior portion of the gut harbored bacterial communities that were more diverse (Fig. S3) and compositionally distinct from the anterior portion (PerMANOVA p<0.001). Across site categories, however, gut position explained less variation than other factors (Table 1), and differences in diversity and composition across site categories were similar in anterior and posterior communities (e.g. Fig. S2, S4, S5). Thus, we focused remaining analyses on the posterior gut microbiome, which may be less transient than the anterior gut community.
Site category had a significant effect on gut microbiome community composition (Fig. 3A; PerMANOVA p=0.001). The presence or absence of bacterial taxa in the gut was largely driven by stream water temperature (which correlated closely to stream pH) and gut length (Fig. 3 inset, Table S2). When HP source populations were removed and only LP environments were considered (LP native, old, and recent introduction sites), diet was a larger contributor to microbiome composition (Fig. 3B), but not in the dimension that corresponded with site category (arrows in Fig. 3B), for which water temperature was more important. Diet also had elevated importance when changes in relative abundances of taxa were accounted for (Fig. S6), compared to presence-absence alone, as also found by Sullam et al. (2015).
There were larger differences between gut microbiomes of HP source populations and those in LP environments (i.e. LP native and introduction sites) than among LP sites. When we compared microbiomes in LP environments to microbiomes of their respective HP source populations, microbiomes of introductions (e.g. Intro 2008, 2009, 1957, etc.) more closely resembled each other, and LP native populations, than their respective HP source site (Fig. 4, distance from 0 in y-axis). Gut microbiomes in introduced populations sometimes, but not always became more similar microbiomes of reference LP native populations (Fig. 4 does not convey this, but see Fig. S7). In two of four drainages, gut microbiome composition of introduced populations remained significantly different from those of LP native sites (Fig. S7, Table S3), with no effect of time since introduction.
Variation in microbiome composition across individuals was highest in HP sites (Fig. S8, betadispersion test p<0.01), and lower in the introductions than native sites, even though we might expect it to be higher in more genetically varied populations. Higher betadispersion was not driven by a greater among-individual diet diversity (p>0.05 for correlation, data not shown). While betadispersion differences can influence betadiversity tests (i.e. PerMANOVA), these differences were small so not likely to be the sole reason we observed significant differences in composition (Anderson & Walsh 2013).
Mean number of species within individual guts also differed across site categories. Gut bacterial richness was significantly lower in native populations (HP source and LP native) and higher in introductions (Fig. 5A), in both anterior and posterior gut positions (Fig. S4), while community evenness and alpha diversity were less responsive (Fig. S4). Richness was positively correlated to relative gut length (r2=0.43, p<0.001), which may explain previously-observed positive correlations between fish length and microbiome diversity (Bolnick et al. 2014a; Forberg et al.2016), and was not explained by higher diet diversity (Bolnick et al. 2014a).
Gut communities were dominated by Proteobacteria, Actinomycetes, and Firmicutes (Fig. S2), similar to other fish microbiomes (Sullam et al. 2012), and previous dominant phyla in guppies at this site (Sullamet al. 2015). Changes in the relative abundances of the most dominant taxa (Table S4) largely drove trends in weighted Unifrac betadiversity. Five of the 10 most abundant taxa across the entire dataset were from the order Rhizobiales, a genus able to fix nitrogen to create nutrients. While we did not measure fitness, we did observe patterns that suggest this additional source of nutrients could have increased fitness. For instance, the significantly more abundant Rhizobiales species in El Cedro (Fig. S9) could explain why these fish were in better condition than their LP native counterparts, despite a lower quality diet (Fig. S1D,E). The total Rhizobiales abundance (Fig. 5B, S9), as well as the abundances of these highly abundant taxa (Fig. 5C), tended to be highest in the old introduction streams, and lowest in HP source streams.