Core intestinal microbiomes of planktivorous and algae-farming coral reef damselshes (Actinopterygii: Pomacentridae) reects feeding behaviour

Background: Fish harbour diverse microbiomes within their gastro-intestinal system that effect the host’s digestion, nutrition and immunity. Despite the great taxonomic diversity of sh, little is understood about sh microbiome diversity and the factors that determine its structure and composition. Damselsh are important coral reef sh species that play a strong role in determining algae and coral structure of reefs. Broadly, damselsh belong to either of two trophic guilds based on whether they are planktivorous or algae-farming. In this study, we use 16s rRNA sequencing to interrogate the intestinal microbiome of 10 damselsh species (Pomacentridae) from the Great Barrier Reef to compare the composition of their intestinal bacterial assemblages across the planktivorous and algae-farming trophic guilds. Results: We identify core intestinal bacterial taxa for each host sh species. Gammaproteobacteria, belonging to the genus Actinobacillus, were detected in 80 % of sampled individuals and suggests a possible core member of pomacentrid microbiomes. Core microbiomes of algae-farming species were more diverse than planktivorous species with farming species sharing 35 ± 22 ASVs and planktivorous sharing 7 ± 3 ASVs. We also provide evidence for signicant shifts in bacterial community composition along the intestines. We show that Bacteroidia, Clostridia and Mollicutes bacteria are more abundant in the anterior intestinal regions while Gammaproteobacteria are generally highest in the stomach. Finally, we highlight differences in microbiomes associated with both trophic guilds. Algae-farming and planktivorous damselsh host species signicantly differed in their composition of bacteria belonging to Vibrionaceae, Lachnospiraceae and Pasteurellaceae. Conclusions: Our results demonstrate that core intestinal bacterial communities of damselsh reect host species diet and feeding behaviour, whereby algae-farming hosts have larger and more diverse core microbiomes than planktivorous hosts. We suggest that the trophic guild of a host sh species is a strong determinant of microbiome structure.

There are many factors that affect the structure of sh gastrointestinal microbiomes [3,4]. These include host-related factors such as genetic attributes, size, age, sex [12][13][14], host phylogeny [15][16][17] and environmental factors such as water quality and diet [15][16][17][18][19]. Studies that investigate intestinal microbiome changes have concentrated on the impact of sh foods on species of aquaculture importance [20,21], although a few studies have investigated wild sh populations [15,22]. Bacterial symbiont diversi cation in wild herbivorous surgeon sh intestines is thought to be an important driver of host niche-partitioning [23,24], suggesting that intestinal microbiomes can in uence the trophic ecology of coral reefs.
There is increasing evidence that herbivorous shes have distinct microbiomes as compared to omnivorous and carnivorous shes [25]. Herbivorous and carnivorous sh diets are known to cause shifts in intestinal microbiomes; shes with plant-based diets have intestinal microbiomes dominated by Firmicutes, such as Clostridium, while shes with fat-based diets have microbiomes dominated by protease producing Proteobacteria [26][27][28][29]. In addition, the diversity of herbivorous sh intestinal microbiomes is higher than omnivorous and carnivorous host species under similar environmental conditions [30], suggesting that host feeding behaviour has a signi cant effect on sh intestinal microbiomes.
Damsel shes (Pomacentridae) are a diverse and abundant group of coral reef shes [31,32], and they are among the most widely studied family of reef shes [33,34]. Broadly, damsel shes are grouped into either planktivorous or herbivorous trophic guilds, although some herbivorous species may also feed on zooplankton [35]. Many herbivorous damsel shes that inhabit reef crest environments rare territorial, and they cultivate palatable algae within their territories, which they aggressively defend from other species.
In this study, we investigate the intestinal microbial diversity of ten species of planktivorous and algaefarming damsel shes, two guilds of damsel shes that signi cantly impact coral reef trophic dynamics. Planktivorous damsel shes play a key role transferring energy from the plankton to higher tiers of food chain, while algae-farming damsel shes in uence sediment and algae dynamics on coral reefs as well as increase the presence of coral disease associated pathogens within their territories [34,39,48,[51][52][53]. Thus, we hypothesise that differences in intestinal microbial communities will re ect the differences between these two feeding guilds. Speci cally, across the different host species and feeding behaviours, we examined (1) the phylogenetic differences in microbial communities, (2) the core microbial members, and (3) the changes in microbial community structure along the length of the intestinal tract.

Species collections and dissections
Fishes were collected from the Heron Island lagoon in the southern Great Barrier Reef, Australia (23°26'53"S, 151°56'52"E) in January and February of 2015. Collections occurred at a depth of 1-8 m within the research zone adjacent to the Heron Island Research Station. Three individuals of 10 damsel sh species were collected across the two feeding guilds which represent different trophic levels ( Table 1). Collections were conducted on SCUBA, and the planktivorous species were collected using a barrier net, while the algae-farming species were collected by speargun. Following collections, the shes were immediately placed on ice and transported to Heron Island Research Station. In the laboratory under sterile conditions, shes were weighed, measured and photographed, then the gastrointestinal tract was removed, and the gut length was recorded and photographed. The entire gut was xed in 4% DNA/RNA free paraformaldehyde and sterile phosphate-buffered saline for 12 hours, then it was stored in DNA/RNA free water.

DNA extraction and 16S rRNA MiSeq Illumina Sequencing
Samples were transported to James Cook University for subsampling along each intestinal tract and DNA extraction. Under sterile conditions, standardized biopsy cores were taken and from four locations along the intestinal tract: the stomach, the anterior intestine, the mid-intestine, and the posterior intestine. DNA was extracted from tissue biopsies using a QIAamp DNA Micro Kit (Qiagen, Hilden, Germany) following the manufactures guidelines. A nanodrop was used to record the quality (260/280 ratio) and quantity (ng/μL) of DNA from each extraction.
Ampli cation of the 16S rRNA gene using the primers 27F (5'-AGRGTTTGATCMTGGCTCAG-3') and 519R (5'-GTNTTACNGCGGCKGCTG-3') with barcodes on the forward primer. These genes were ampli ed in a 30 cycle PCR (5 cycle used on PCR products) using the HotStarTaq Plus Master Mix Kit (Qiagen, USA) under the following conditions: 94°C for 3 minutes, followed by 28 cycles of 94°C for 30 seconds, 53°C for 40 seconds and 72°C for 1 minute, after which a nal elongation step at 72°C for 5 minutes was performed. After ampli cation, PCR products were checked in 2% agarose gel to determine the success of ampli cation and the relative intensity of bands. Multiple samples were pooled together (e.g., 100 samples) in equal proportions based on their molecular weight and DNA concentrations. Pooled samples were puri ed using calibrated Ampure XP beads. Then the pooled and puri ed PCR products were used to prepare a DNA library by following Illumina TruSeq DNA library preparation protocol. Sequencing was performed at the Molecular Research LP (MR DNA; Texas, USA) on a MiSeq™ System following the manufacturer's guidelines.
Amplicon sequence data were sorted by the sample and demultiplexed using demux for QIIME 2 (version 2018.11; Bolyen et al. 2018). Sequences were screened for quality, trimmed at 450 bp after removal of primer sequences and assigned as amplicon sequence variants (ASVs) [55] using DADA2 [56]. Taxonomy of the ASVs was determined using a pre-trained, naїve Bayes classi er [57] and the q2-feature-classi er plugin [58]. The classi er was trained on the target 480 bp region of sequences in the Greengenes 13_8 99% database. ASV clusters were arranged in a phylogenetic tree using FastTree [59,60] and visualised using Interactive Tree of Life [61]. The feature table, metadata and taxonomic classi cations were exported from QIIME 2 in.biom format [62], and the rooted phylogenetic tree was exported in.nwk format.

Statistical analysis
The exported feature table and phylogenetic tree were imported into R version 3.5.2 (R Core Team 2019) and stored as a phyloseq object [63] for downstream analyses. All ASVs not assigned to phylum were ltered from the data, and those designated as chloroplasts or cyanobacteria were removed and stored as a separate object for further analysis. Samples were rare ed to minimum sampling depth for diversity analyses; however, non-rare ed data were used for multivariate modelling [64,65]. Multivariate generalised linear models were used to test for signi cant differences in bacterial communities among host sh species, feeding behaviour and location along intestines using mvabund in R [66,67]. Bacterial taxa were grouped by class when examining microbiome changes along the length of the intestinal tract.
Bacterial community data were tted to negative binomial distributions and tested using log-likelihood ratios (LRT) via 999 simulations. A nested analysis of variance (ANOVA) used to test the role of gut location when accounting for species variation. Traditional distance-based ordination methods to visualise variation across communities, such as non-metric multidimensional scaling (NMDS) and principal coordinate analysis (PCoA) may confound trends [68]. To avoid these issues, the R package boral [69] was developed to explain the bacterial community composition of each sample through a set of latent variables. Bacterial community data were tted to a negative binomial distribution, and the model was run with two latent variables to account for residual variation for each of the major phyla detected in the samples. Venn diagrams were produced using the VennDiagram package [70], and the core microbiome scatterplots were produced following [71].

Results
A total of 1,254,909 sequences were detected in 119 samples after denoising and trimming of all chloroplast, mitochondria sequences and host DNA. Among these sequences, 3,776 ASVs were detected; 39.4% of which belonged to the Phyla Proteobacteria, 26.2% to Bacteroidetes, 13.4% to Firmicutes and 12.6% to Planctomycetes. The 20 most abundant ASVs accounted for 41% of the total number of detected sequences. The most common ASV belonged to the genus Actinobacillus and accounted for 9.9% of the total detected sequences ( Table 2). A further two unknown members of Mollicutes and Pasteurellacea accounted for 9.9 and 3.8% of sequences, respectively.
An ordination analysis revealed that most host sh species have distinct Proteobacteria, Bacteroidetes and Firmicutes communities ( Figure 1). Abudefduf sexfasciatus and Abudefduf whitleyi displayed high variation in Proteobacteria communities while the two feeding behaviours have similar community composition ( Figure 1a). Bacteroidetes are distinct for A. sexfasciatus and Stegastes apicalis, with no discernible patterns between the two feeding behaviours (Figure 1b). Communities of Firmicutes were the most distinct between host species, although some host species, such as S. apicalis, Chromis atripectoralis and A. whitleyi,are variable in composition ( Figure 1c). However, there was reasonable separation of the two feeding behaviours in terms of Firmicutes community composition ( Figure 1f).
Different levels of ASV richness were detected for each host sh species. Dischistodus perspicillatus had the greatest mean richness of ASVs, with a total of 322 ± 17 ASVs per individual. The species with the lowest ASV richness were C. atripectoralis and A. sexfasciatus with 47 ± 21 and 30 ± 8 ASVs per individual, respectively (Supplementary Figure 1). Shannon diversity was greatest for three algae-farming species D. perspicillatus, Stegastes nigricans and S. apicalis and lowest for the planktivorous species C. atripectoralis, A. sexfasciatus and Pomacentrus moluccensis.

Core Microbiomes
Most ASVs occurred in less than 30% of sampled individuals (Figure 2a). 13 bacterial ASVs were found to occur in more than 30% of sampled individuals; therefore, they may represent core members of pomacentrid microbiomes ( Table 3). The most common ASV in this study belongs to the genus Actinobacillus, which occurred in more than 80% of sampled individuals, albeit at a low abundance in many individuals, with the highest abundances in the planktivorous damsel shes Acanthochromis polyacanthus and P. moluccensis.
Core bacterial taxa for each sh species were also detected ( Figure 2b), which were de ned as ASVs that were shared between all sampled individuals for each species. There were 70 bacterial ASVs shared between the three sampled individuals of D. perspicillatus and only two ASVs shared between the three A. sexfasciatus individuals. Core microbiomes within host species were higher in algae-farming species than planktivorous species, with algae-farming host species sharing 35 ± 22 ASVs and planktivorous species sharing only 7 ± 3 ASVs.
Core ASVs that occurred in all three sampled individuals of a host sh species were detected for the bacterial phyla Bacteroidetes, Firmicutes, Tenericutes, Spirochaetes, Planctomycetes, Proteobacteria and Verrucomicrobia. The core ASVs Coraliomargarita sp. and Verruco-5 (Verrucomicrobia), Pirellulaceae (Planctomycetes) and Desulfovibrionaceae (Deltaproteobacteria) occurred in all three sampled D. perspicillatus individuals (Figure 3). There was high richness of core Pasteurellaceae and Vibrionales ASVs, with 10 and 21 core members, respectively. High diversity of an unknown clade of Gammproteobacteria were also detected in the two host species belonging to the genus Pomacentrus, P. moluccensis and Pomacentrus wardi.
There were 61 core ASVs belonging to the Bacteroidetes, 28 of which occurred in S. apicalis and 38 in P. perspicillatus ( Figure 4). An unknown clade of Flavobacteriales and a diverse consortium of Rikenellaceae were core members of S. apicalis, while P. perpicillatus had a diverse core assemblage of ASVs belonging to the family Flavobacteriaceae. One ASV belonging to Spirochaetes, Brevinema andersonii, was a core member of S. nigricans and C. atripectoralis,while a Tenericutes ASV belonging to Mollicutes was a core member of all host species except the planktivorous damsel shes A. polyacanthus and A. sexfasciatus ( Figure 5). A rich consortium of core Firmicutes ASVs were detected for S. apicales and S. nigricans, which included members of the Erysipelotrichaceae, Ruminococcaceae and Lachnospiraceae families.

Bacterial shifts along the intestinal tract
Bacterial communities signi cantly shifted along the intestinal tract (LRT = 1263, P = 0.001; Supplementary Table 1). Nine classes of bacteria had signi cant changes in abundance across the different sh species and locations along the intestinal tract (P < 0.05; Figure 6). Bacteroidia, Clostridia and Mollicutes all displayed strong increases in abundance from the stomach to the posterior intestine. Gammaproteobacteria were highest in the stomach but were generally found in high abundance throughout the intestinal tract. The stomach had 286 unique bacterial ASVs, the anterior intestine 753, while 1139 and 656 ASVs were only found in the mid and posterior intestines, respectively. Only 19 ASVs were common to the stomach and posterior intestine while 152 ASVs were found throughout the intestine (Figure 7).

Feeding behaviour effect on microbiomes
We detected a signi cant difference between feeding behaviours and microbiomes (LRT = -0.021, P = 0.001). Most bacterial ASVs were unique to either of the feeding behaviours of the host sh, with only 124 ASVs common to both feeding behaviours (Figure 7). 78 bacterial ASVs, belonging to 20 families, were important drivers of this relationship. There were marked differences in abundances of ASVs belonging to Vibrionaceae, Lachnospiraceae and Pasteurellaceae. Two Vibrio sp. (Vibrionaceae) were more common in planktivorous host species, and ve members of Actinobacillus were more abundant in algae-farming host species. However, none of the ASVs exclusively occurred in the planktivorous or algaefarming damsel shes, suggesting that this relationship was driven by host species rather than feeding behaviour.

Discussion
This study reveals that algae-farming damsel sh species have larger core microbiomes than planktivorous species. This result is likely attributable to the specialised feeding behaviour of these species where they largely consume a narrow range of turf algae species [36,38,48], unlike planktivorous species which are adapted to a more opportunistic feeding strategy. We also provide evidence that algaefarming damsel sh tend to have more diverse intestinal microbiomes than planktivorous species. These results show that microbiome structure of host sh species that have specialised feeding behaviour have acquired specialised intestinal bacteria that play an important role in digestion.
Like surgeon sh intestinal microbiomes from the Red Sea [15], the damsel sh microbiomes presented here were dominated by members of Proteobacteria, Bacteroidetes, Firmicutes and Planctomycetes. Another dominant ASV in the damsel sh microbiome belonging to Mollicutes (Tenericutes) resembled bacteria detected in rabbit sh intestines [22]. The number of highly similar bacterial ASVs shared among pomacentrids, acanthurids and siganids may re ect the similar feeding behaviours of these coral reef shes. For instance, algae-farming damsel shes may also ingest prey items other than algae, such as zooplankton [35] or other invertebrates [72]. Our data support this notion given the dominance of Proteobacteria that are important in the digestion of proteins [29], and may re ect the similarities between pomacentrid, acanthurids and siganids diets.
Damsel sh microbiomes were largely dominated by Gammaproteobacteria of the Pasteurellaceae, with one ASV occurring in more than 80% of sampled shes and representing almost 10% of total detected sequences. Although this ASV currently represents an unknown species of the Actinobacillus genus, a 98% similar sequence has been collected from the intestines of surgeon shes in Saudi Arabia [24], suggesting these taxa are important components of reef sh microbiomes. Members of Pasteurellaceae have also been recorded in high abundances in adult damsel shes and cardinal shes collected around Lizard Island, Australia [73], and they are deemed important components of tropical planktivorous sh gut microbiomes [74]. Gammaproteobacteria are also very abundant on the skin of many coral reef shes [75]. The prevalence of Pasteurellaceae amongst the damsel shes in this study, as well as other reef shes, provides additional evidence that Pasteurellaceae are important members of coral reef associated sh microbiomes.
We provide evidence that algae-farming damsel shes have more specialised microbiomes than the planktivorous species. Algae-farming damsel shes had larger core microbiomes than the planktivorous damsel shes, and these core microbiomes were speci c to each host species. For example, P. wardi and P. moluccensis had diverse, but different strains of Gammaproteobacteria, while D. perspicillatus and S. apicalis had large Bacteroidetes core communities but were dominated by Flavobacteriaceae and Rikenellaceae, respectively. Different species of territorial damsel shes farm and consume different species of algae [36,48], and the large differences in their specialized microbiomes may re ect these narrow dietary preferences. Conversely, the small core microbiomes of the planktivorous damsel shes may re ect the high variation in consumed plankton of each species, suggesting these shes have opportunistic feeding behaviours. These results, however, do not support the notion that sh with greater diet variability have more diverse microbiomes [25]. In fact, the damsel sh with narrow, algae-farming feeding behaviours tended to have the greatest diversity of intestinal bacterial, suggesting that the host may select microbial populations that include specialised bacteria that enhance the digestion and absorption of nutrients from speci c algal diets.
Recent evidence suggests a high degree of resource partitioning in sh communities which is a key mechanism that facilitates the high diversity of coral reefs [76,77]. The largely different microbiomes of each species presented in this study may re ect this resource partitioning, where different species of planktivorous damsel sh may be consuming different size classes of zooplankton [76] or trophic niche [77]. The similarity between closely related host species and microbiomes, such as P. wardi and P. moluccensis, also demonstrates that phylogeny may in uence core microbiomes [15][16][17]75].
Interestingly, Photobacterium damselae, Vibrio harveyi, Vibrio ponticus and other Vibrio sp. were prevalent amongst the damsel shes sampled in this study. These bacteria represent potential pathogenic members of Vibrionacaea and have been detected in many shes of aquaculture importance, including Chromis punctipinnis [78], Lutjanus argentimaculatus [79], Seriola dumerili [80], Scophthalmus maximus [81,82], Sparus aurata [83], Solea quinqueradiata [84] and Solea senegalensis [85]. Although identi ed as Vibrio harveyi in the GreenGenes database, GenBank revealed there was high similarity of these sequences to other members of the Harveyi clade, such as Vibrio owensii [80]. There are thought to be up to 11 species of Vibrio belonging to this clade [86], most of which are pathogens of sh, shrimp and coral [87][88][89]. Given the apparent healthy state of the sampled shes and the high abundances of potentially pathogenic Vibrionacaea in the sh guts, we provide support to the idea that these organisms are natural components of healthy sh microbiomes and are opportunistic pathogens in shes only under speci c conditions [79,90].
The facultative anaerobic bacterial classes Bacteroidia, Clostridia and Mollicutes were generally in higher abundance in the mid and posterior intestinal regions than the stomach. Differences in microbiomes along the intestinal tract have been recorded in the rabbit sh Siganus fuscescens [91], with midgut communities more representative of the environmental sources and hindguts hosting a microbiome more specialised to anaerobic conditions and fermentation [92]. The increase in Bacteroidia, Clostridia and Mollicutes along the intestines may be due some members of the class being mutualistic components of the sh gastrointestinal ora. Some members of Bacteroidetes are known to breakdown polysaccharides and metabolise the derived sugars [93], while members of Clostridium are known to metabolise cellulose [29]. Our results con rm the increased prevalence of anaerobic bacteria in the hindgut of damsel shes, which probably consists of taxa responsible for the fermentation and metabolism of complex molecules before being absorbed by the host [3].

Conclusions
In this study, we demonstrate that damsel shes have diverse intestinal microbial communities whereby the core members of a species re ect diet and feeding behaviour. We show that algae-farming damsel shes have larger core microbiomes, which may re ect the more specialised diets of these species. We also provide evidence that damsel sh mid and posterior intestines have higher abundances of facultative anaerobic bacteria that are known to play important roles in fermentation and cellulose breakdown. These ndings add to a growing body of literature that suggests that host sh feeding behaviour has a strong in uence on the composition of intestinal microbiomes.

Declarations
Ethics approval and consent to participate

Competing interests
The authors declare that they have no competing interests.

Funding Not applicable
Authors' contributions CRJK analysed and interpreted the amplicon sequence data and was the major contributor in writing the manuscript and preparing gures and tables. JMC undertook the eldwork and collected all specimens, performed gut dissections, tissue biopsies and provided feedback on the manuscript. JHC was involved with the initial synthesis and design of this study and provided feedback on the manuscript. WL and TDA were involved with the initial synthesis and design of this study, provided the facilities to undertake laboratory work and provided feedback on the manuscript. All authors read and approved the nal manuscript.    Table 3: Taxonomic composition of core ASVs occurring in more than 80% of sampled individuals. Accession numbers for closest GenBank sequences (similarity given in brackets) are supplied. Occurrence and relative abundances were generated from rare ed data.    Phylogenetic tree of core Bacteroidetes ASVs. Coloured host species re ects whether the ASV was present in all three sampled individuals of each host sh species.
Page 25/27 Figure 5 Phylogenetic tree of core Firmicutes, Spirochaetes and Tennericutes ASVs. Coloured host species re ects whether the ASV was present in all three sampled individuals of each host sh species.