Divergence across diet, time and populations rules out parallel evolution in the gut microbiomes of Trinidadian guppies

Abstract

Diverse microbial consortia profoundly influence animal biology, necessitating an understanding of microbiome variation in studies of animal adaptation. Yet, little is known about such variability among fish, in spite of their importance in aquatic ecosystems. The Trinidadian guppy, Poecilia reticulata, is an intriguing candidate to test microbiome-related hypotheses on the drivers and consequences of animal adaptation, given the recent parallel origins of a similar ecotype across streams. To assess the relationships between the microbiome and host adaptation, we used 16S rRNA amplicon sequencing to characterize gut bacteria of two guppy ecotypes with known divergence in diet, life history, physiology and morphology collected from low-predation (LP) and high-predation (HP) habitats in four Trinidadian streams. Guts were populated by several recurring, core bacteria that are related to other fish associates and rarely detected in the environment. Although gut communities of lab-reared guppies differed from those in the wild, microbiome divergence between ecotypes from the same stream was evident under identical rearing conditions, suggesting host genetic divergence can affect associations with gut bacteria. In the field, gut communities varied over time, across streams and between ecotypes in a stream-specific manner. This latter finding, along with PICRUSt predictions of metagenome function, argues against strong parallelism of the gut microbiome in association with LP ecotype evolution. Thus, bacteria cannot be invoked in facilitating the heightened reliance of LP guppies on lower-quality diets. We argue that the macroevolutionary microbiome convergence seen across animals with similar diets may be a signature of secondary microbial shifts arising some time after host-driven adaptation.

Figure 1

Taxonomic composition of gut bacterial communities. Results are based on 454 pyrosequencing across 55 Trinidadian guppies from the wild and 14 included in a dietary manipulation experiment. Stream, ecotype and year of collection are listed on the x-axis for wild-caught fish. Communities in the central box represent gut bacteria of guppies from three streams on the southern slope of the Northern range, and those in the leftmost box show gut bacteria of guppies from a stream on the northern slope. Columns in the rightmost box represent samples taken after the dietary manipulation experiment in the lab, for which fish collected in 2011 from the Aripo were used, and HP fish and LP fish were fed either an invertebrate (I) or spinach (S)-based diet. Bacteria are classified to order by color, with phyla indicated to the right of ordinal labels.

Figure 2

Bacterial community composition and enterotype groupings in relation to 97% OTU distributions. A heatmap is shown for all OTUs with over 500 reads in the whole data set. The phylum associated with each OTU is listed above, and the asterisks indicate dominant OTUs that appear to be driving enterotype groupings (Supplementary Table 5). The shade of blue in the heatmap indicates the proportion of reads that come from the given OTU out of all reads for that sample. The color bar on the right of the heatmap indicates experiment/sample type—light gray: samples from the dietary manipulation experiment; dark gray: non-fish environmental samples; remaining colors: 2010 and 2011 field collections (yellow: Guanapo; red: Quare; green: Marianne; and blue: Aripo). Names on the left aid in further classifying samples, with ‘11' denoting field samples obtained in 2011 (2010 samples have no year abbreviation), and LP/HP/HP2 abbreviations are used as described for Figure 1. The sample names on the left are also shaded in green according to their enterotype grouping (displayed to the left of the sample label). The dendrogram on the right shows community similarity based on hierarchical clustering using Ward's method.

Figure 3

Effects of time, diet and population background on guppy gut microbiomes. (a) Similarity of gut communities among wild guppies from the Aripo and Guanapo Rivers during the 2011 and 2010 surveys. Communities from the two years and streams were significantly different from each other (Supplementary Table 7), although the stream effect was weak. Principal coordinates analyses (PCoAs) shown were generated from unweighted UniFrac distances, and PCoA 3 shows the strongest correlation with yearly variation. (b) Principal coordinates of guppy gut bacterial community similarity from experimental dietary treatments based on Bray–Curtis distances. Shape indicates treatment diet (invertebrate- or spinach-based diet) and color indicates guppy ecotype (HP or LP). Note that there are two overlapping LP spinach data points at (0.36, −0.17). Analyses of Bray–Curtis distances indicate a clear effect of ecotype on gut community composition. Subtle, yet significant, effects included both experimental diet and an ecotype by diet interaction (Table 1) because of a dietary difference in bacterial communities within HP but not LP guppies.

Figure 4

Relationships between guppy gut microbes and those from other habitats. (a) Closest relatives of core bacterial OTUs. Core OTUs were separately identified for LP fish (14 OTUs), HP fish (15 OTUs) and environmental (28 OTUs) samples when they occurred in >50% of the samples from one of these categories. Each OTU was then searched using BLASTn against the NCBI database and its three closest hits were categorized based on their environment of isolation, and proportions of these categories are shown. (b) Network analysis of all gut microbes derived from guppies and environmental samples. Square nodes correspond to 97% OTUs, with colors indicating whether these were found in one or more samples. Purple reveals the OTUs found in both gut and environmental samples. Black symbols correspond to guppy gut (lab vs field) vs environmental (water and sediment) samples.

Figure 5

Results of PICRUSt analysis showing predicted relative abundance of KEGG ortholog groups for guppy (a) ecotypes and (b) enterotypes. Asterisks on the right of each panel indicate significant differences among groups. No significant differences were found between ecotypes, whereas the majority of KEGG orthogroups differed among Enterotypes. The samples from the dietary manipulation (and their two associated enterotypes: 3 and 5) are not shown because an insufficient number of reads per sample (<800) clustered with the closed reference OTUs in approximately one third of the dietary manipulation samples.