Microbial Diversity of the GIT of Wild Marine Fish
Utilizing a whole genome shotgun sequencing approach followed by read filtering and taxonomic assignment via the Kraken2/Braken pipelines we were able to define the microbiome composition of seven fish/shark species and their seawater environment. We find that the gut microbiota of all species is predominantly composed of Proteobacteria (50.4%), Firmicutes (20.8%), and Bacteroidetes (10.0%) (Figure 2A).
This finding matches those of previous studies that have identified Proteobacteria and Firmicutes as being the major constituents of the gut microbiota of marine fish[4, 5, 17, 23, 46–49]. Within the phylum Proteobacteria, Photobacterium (5.6%), Vibrio (4.5%), Alivibrio (3.5%), and Edwardsiella (3.2%) were the most abundant genera found in the fish samples collected in the bay (Fig. 2B). Here we observe significantly increased levels of Proteobacteria in benthivorous/piscivorous (scup, summer flounder, smooth dogfish) species compared to a plankivorous species (butterfish) (Mann-Whitney U test, p < 0.0001). Previous studies have also identified an increased abundance in Proteobacteria in omnivorous and carnivorous organisms compared to herbivores[5, 19, 49, 50] suggesting trophic level and dietary guild play a role in the level of Proteobacteria present in the gut microbiota. Due to a high degree of variability within sample types, fish species did not group together significantly within a principal coordinate analysis of Bray-Curtis Dissimilarity (Fig. 2C). However, the microbiomes of species clustered more separately when samples were separated by site and sampling time, suggesting that there may be significant temporal and spatial variability within the microbiome of fish (Figure S1 B, D). Between the samples collected at the two Narragansett Bay sampling locations there were notable differences in the microbiome composition of the summer flounder at the genus level; Photobacterium was significantly increased in the Fox Island population (padj < 0.05) and six less prominent genera were significantly increased in the Whale Rock population (padj < 0.05) (Figure S1 A, B Table S2). Surprisingly, no significant differences in taxonomy were found between the two sites in the butterfish and scup populations. This suggests that the microbiota of summer flounder may have characteristics unique to either the upper or lower bay locations, while the butterfish and scup populations are more homogeneous.
Shark Spiral Valves Harbor Species-Specific Microbiota
The gut microbiota of sharks has only been characterized in a few reports[20, 21, 47, 51, 52], and represents an understudied area of shark physiology, which likely plays a major factor in host health. Here, we define the microbiota of four shark species, the mako shark, thresher shark, porbeagle, and smooth dogfish. Sharks have unique digestive architecture defined by the spiral valve, an organ that maximizes absorption and minimizes the length of GIT by increasing surface area through a corkscrew-like arrangement of intestinal tissue (Fig. 3A).
The spiral valve of all sharks was dominated by Proteobacteria (53.9%) and Firmicutes (18.0%), with Photobacterium (17.5%), Campylobacter (6.0%), and Dickeya (5.7%) the most prominent genera (Fig. 3B, 2B). Analysis of the Bray-Curtis Dissimilarity metric revealed a significant difference between the microbiota of each species (PERMANOVA, p = 0.001), defined by distinct clustering in a principal coordinate analysis (Fig. 3C). Previous studies of Elasmobranchii have also found an abundance of Photobacterium as well as Campylobacter in the spiral intestine of sharks[47, 52], but to date only one study utilizing 16s sequencing has examined the taxonomic differences between regions of the shark GIT[51].
Here, we compare the microbiota of the spiral valve (SV) to the distal intestine (DI) in smooth dogfish. The principal coordinate analysis plot of the Bray-Curtis Dissimilarity metric displays the significantly distinct clustering of the SV and DI microbial communities (PERMANOVA, p = 0.018) (Fig. 3E). These disparate communities are defined by a significantly greater abundance of Proteobacteria in the DI (63.3%) compared to the SV (39.1%) (p = 4.55E-05), and a significantly reduced abundance of Actinobacteria in the DI (1.8%) compared to the SV (8.0%) (p = 1.89E-08) (Fig. 3D, 3F, 3G). The most differentially abundant species between GIT sites was Photobacterium damselae, which was significantly more abundant in the DI compared to the SV (log2FC = 9.84, padj = 4.52E-74) (Fig. 3E). Such differences in microbial composition were not found in the GIT of the previously studied bonnethead shark (Sphyrna tiburo)[51], suggesting this may not be a universal phenomenon among sharks. Our findings suggest that the SV and DI represent unique ecological niches for commensal microbes, and that perhaps nutrient availability, host immunity, or oxygen levels may act as selective factors for bacterial colonization in these regions of the smooth dogfish GIT.
The GIT Microbiota of Marine Fish act as a Reservoir of ARGs which are Associated with Proteobacteria
Environmental reservoirs of antimicrobial resistance play an important role in the selection, proliferation, and transfer of resistance genes [53]. We used the computation tool DeepARG [54] to identify resistance genes and find that the gut microbiota of marine fish represent one such reservoir of ARGs. Across all fish GIT samples we detected 518 different resistance genes covering 27 antibiotic resistance classes (Table S3). The most abundant resistance gene classes were multidrug (34.3%), macrolide, lincosamide, streptogramin (MLS) (16.1%), tetracycline (16.0%), and beta-lactam (4.6%) (Fig. 4C). Recently, Collins et al. found multidrug and beta-lactam resistance genes in the microbiota of deep-sea fish [24], and similarly a study of ocean waters around the globe found tetracycline, beta-lactam, and multidrug resistance genes to be the most prevalent resistance gene types in seawater [55]. These findings suggest resistance mechanisms may be conserved across bacteria that inhabit the marine environment and fish GIT.
An increase in antibiotic resistance gene abundance was associated with certain fish/shark species, specifically in higher trophic level organisms (Figure 4A, top). In general, those fish that exhibited piscivorous feeding behavior, occupying a higher trophic level, had a greater burden of antibiotic resistance. Rowan-Nash et al. found a significant correlation between Gammaproteobacteria and ARGs in human gut microbiota samples suggesting that the presence of certain bacteria may be driving levels of resistance in host-associated microbial communities[34]. Expanding on this idea, we examined the relationship between ARGs and Proteobacteria in the GITs of fish and found that samples from piscivores with a higher relative abundance of Proteobacteria harbored an increased abundance of ARGs compared to planktivorous/benthivorous species with less Proteobacteria (Figure 4A, bottom, Figure S2). A correlation analysis between ARG abundance and Proteobacteria relative abundance in fish within Narragansett Bay showed a significant positive correlation (r = 0.7971, R2 = 0.6353, p < 0.0001, Pearson’s correlation) (Figure 4B). When we factored in the large offshore shark species, we find that this trend generally holds true with the exception of the thresher shark, which despite having high levels of Proteobacteria had relatively low levels of ARGs (Figure S3). These findings show that fish with high levels of Proteobacteria are likely to have an increased level of detectable ARGs. Furthermore, this may suggest that higher trophic level organisms with a more carnivorous diet and Proteobacteria rich gut microbiota will have a greater resistance gene burden.
In order to determine the bacterial hosts of these resistance genes, metagenomically assembled genomes (MAGs) were assembled using the MetaWRAP assembly pipeline [56] and subsequently queried for ARGs. From all metagenomic reads across fish and water samples, we assembled 267 MAGs covering 9 bacterial phyla (Figure S4). We found that the MAGs from Firmicutes (n = 8), Fusobacteria (n = 2), and Proteobacteria (n = 121) had the highest prevalence of ARGs, and had significantly more resistance genes than MAGs from Bacteroidetes, Verrucomicrobia, Spirochaetes, Planctomycetes, and Tenericutes (n = 104) (Mann-Whitney U test, p < 0.05) (Fig. 4D). Notably, the second most ARG-rich MAG was identified as Photobacterium damselae, which occurred at a high abundance in all the piscivorous fish and shark gut microbiota supporting the theory that higher trophic level organisms may harbor more ARGs (Fig. 2C).
Inferring Diet Through Metabarcoding of GIT Shotgun Metagenomic Data
The levels of Proteobacteria were highly correlated with the relative abundance of ARGs in the fish microbiome. This relationship may be driven in part by the dietary inputs. Dietary analysis provides insight into the trophic structure and predator/prey relationships within a community and is a driving factor in shaping the gut microbiome. Traditionally techniques to study diet in wild animals, such as direct observation or stomach content analysis, have been low throughput and time consuming and are unable to identify phenotypically indistinguishable or rapidly digested prey items [57]. The use of DNA-barcoding methods circumvents these issues by providing molecular level resolution that reduces the need for human identification of physical dietary components [57]. Additionally, molecular methods provide a high throughput alternative that can detect not only dietary items, but also potential parasites. Here, we utilize DNA-metabarcoding targeting the cytochrome c oxidase subunit I (COI) [58], elongation factor TU (tufA), and ribulose-1,5-bisphosphate carboxylase (rbcL) genes [59] to identify the diet and potential GIT parasites of seven marine species.
Of the seven species examined in this study, four occupy a shared demersal habitat in Narragansett Bay, RI providing an opportunity to detect interspecies predation and differential dietary preferences within a habitat. The planktivorous butterfish had a diet primarily consisting of diatoms (Bacillariophyta), algae (Chlorphyta, Ochrophya, Haptophyta), and to a lesser extent arthropods (Arthropoda), characteristic of an organism occupying a low trophic level (Figure 5A, C).
The benthivorous scup occupies a higher trophic level than the butterfish, characterized by dietary signatures of diatoms (Bacillariophyta), arthropods (Arthropoda), and segmented worms (Annelida) which were known to be a major prey source for this benthic species (Figure 5A, C) [60]. At the order level we find that the Metazoan portion of the scup diet is derived from amphipods (Figure 5D). Previous dietary studies of both the summer flounder and smooth dogfish in New England waters identified these species as high trophic level predators preying on fish, squid, and crabs [61, 62]. It is notable that due to the feeding patterns of these species they were sometimes captured with empty stomachs and intestinal tracts resulting in an absence of detectable DNA markers making diet identification impossible (Figure 5A). We find that these highly carnivorous species prey primarily on chordates in the class Actinopterygii (ray-finned fishes) as well as arthropods (Figure 5A, C, D). In summer flounder the Metazoan derived diet came from primarily Decapoda (crustaceans) and Clupeiformes (herring and anchovy family) (Figure 5D). The smooth dogfish DI contained Metazoan signatures of Stromatopoda (mantis shrimp) and fish across several orders (Figure 5E). DNA markers corresponding to butterfish and scup were found in the GIT of the high trophic level predators (summer flounder and smooth dogfish) suggesting that predation occurs within this benthic food web and represents a possible route of bacterial and ARG transfer from lower- to higher-trophic level organisms. We also obtained dietary signatures from three large migratory shark species that play an important role in the food web as apex predators. All three sharks exhibited piscivorous diets based on metabarcoding (Figure 5A, C, D). A closer look at order level taxonomy revealed that each shark had a fairly specialized diet with DNA from only one or two different prey species (Figure 5D). The COI dietary signatures for the thresher, mako, and porbeagle sharks were primary from Clupeiformes, Scombriformes, and Perciformes, respectively (Figure 5D). Using this metabarcoding approach for dietary contents we confirmed that the summer flounder, smooth dogfish, mako, thresher, and porbeagle sharks all had highly piscivorous diets compared to the butterfish and scup. Furthermore, these each species harbored a significantly distinct diet that was host specific (PERMANOVA, p = 0.006) (Figure 5B). These trends in prey preference likely influence the microbial communities inhabiting the GIT as diet is a strong modulator of the microbiome. From a metabarcoding analysis of wild marine fish GIT samples we were able to infer diet, trophic interactions, and gain insights into the role of host diet in shaping the microbiota through nutrient availability and potential bacterial transfer between diet and host.
Functional Differences in the Microbiota Linked to Host Diet and Trophic Level
The gut microbiota plays an important role in host digestion, increasing nutrient availability and uptake [4, 19, 63]. Host diet in turn plays a key role in determining the makeup of the gut microbiome, and evidence shows that dietary modulation and macronutrient availability can drastically alter the composition and function of the intestinal flora [12, 64, 65]. Through previous stomach content analyses [61, 62], and our own metabarcoding analysis (Figure 5), we are able to gain an understanding into the role of diet in shaping the gut microbiome of these marine fish. Investigation into the carbohydrate-active enzymes (CAZymes) known to play a role in metabolism of dietary polysaccharides revealed 120 differentially abundant CAZymes between the piscivorous and planktivorous/benthivorous species suggesting that the divergent diets of these groups may have an impact on the functional capacity of the microbiome (Figure 6A).
Glycosaminoglycans, including chondroitin, are a group of diverse polysaccharides that are components of a variety of tissues including cartilage derived from mammals, marine fish, squid, and other organisms [66–71]. The diet of piscivorous fish, such as those studied here, include a number of organisms known to contain chondroitin (Arthropoda and Chordata). Thus, the piscivorous fish and sharks occupying a higher trophic level would likely have greater dietary intake of this polysaccharide compared to the butterfish and scup, whose prey is less rich in chondroitin. We find that several CAZymes linked to chondroitin metabolism are significantly enriched in the piscivores compared to planktivores/benthivores (log2fc > 1.5, padj < 0.05) (Fig. 6A, B). Interestingly, these genes were predominantly detected in MAGs isolated from the piscivorous species, summer flounder, smooth dogfish, thresher, mako, and porbeagle sharks (Table S4). This data suggests that host diet, associated with trophic level and dietary guild, may select for bacteria with particular carbohydrate utilization patterns. In this case, piscivorous fish and sharks likely have a more chondroitin rich diet and the abundance of chondroitin could provide an ecological niche for bacteria with chondroitin lyase and hydrolase enzymes. Chitin is one of the most abundant polysaccharides in nature and makes up the exoskeletons of many arthropods [72, 73]. Several chitinases were detected across nearly all the gut microbiota samples collected (Fig. 6C), suggesting that the ability to utilize chitin may be a widespread trait among marine associated microbiomes likely due to the fact that chitin is ubiquitous in this environment. Overall, our evaluation of carbohydrate active enzymes within the fish gut microbiota suggests that the availability of dietary polysaccharides associated with different trophic levels may have a role in selecting for certain bacteria based on polysaccharide utilization. This finding has the potential to link host trophic level and related prey consumption with selection for specific microbes.