Fishing for the Bacteriome of Tropical Tuna

1 Background : Although tunas represent a significant part of the global fish economy 2 and a major nutritional resource worldwide, their consumption poses a risk of food 3 poisoning through the development of particular bacterial pathogens. However, their 4 microbiome still remains poorly documented. Here, we conducted a multi- 5 compartmental analysis of the taxonomic composition of the bacterial communities 6 inhabiting the gut, skin and liver of two most consumed tropical tuna species 7 (skipjack and yellowfin), from individuals caught in the Atlantic and Indian oceans. 8 Results : Our results revealed that the composition of the microbiome was 9 independent of fish sex, regardless of the species and ocean considered. Instead, 10 the main determinants were (i) tuna species for the gut and (ii) sampling site for the 11 skin mucus layer, and (iii) a combination of both parameters for the liver. 12 Interestingly, only 4.5% of all ASVs were shared by the three compartments, raising 13 numerous questions about the circulation of microorganisms within the tuna body. 14 Our results also revealed the presence of a unique and diversified bacterial 15 assemblage within the liver, comprising a substantial proportion of histamine- 16 producing bacteria, well known for their potential pathogenicity and their contribution 17 to fish poisoning cases. 18 Conclusions : These results indicate that the tuna liver is an unexplored microbial 19 niche whose role in the health of both the host and consumers remains to be 20 elucidated. 22 23 24


7
2X Phusion Mix (New England Biolabs ® , Ipswich, MA, USA), 1 µL of each primer at 1 10 µM (Eurofin ® , Luxembourg), 10 ng of DNA template and enough molecular-grade 2 H 2 O (Qiagen ® ) to reach a final volume of 25 µL. All samples were amplified in 3 triplicate to avoid PCR bias in the taxonomic diversity of the community [33]. 4 Triplicate PCR products were pooled and purified with a NucleoSpin Kit (Macherey-5 Nagel ® , Düren, Germany) following the manufacturer's instructions. Successfully 6 amplified samples (n=103) were sequenced on the Illumina platform (GenoToul ® , 7 Toulouse, France) using 2×250 bp MiSeq chemistry. Chimaeras were removed, and sequences were aligned to the SILVA 123 database 18 [35] to access their taxonomy. Analyses were performed on a random subsample of 19 6 847 sequences per sample, corresponding to the sample with the smaller number 20 of sequences, after trimming and quality processing. Using the phyloseq package 21 [36], final taxonomic and ASV tables were linked to sample metadata (tuna species, 22 sex, biological compartment and ocean). The relative abundances of ASVs in each 23 sample were assessed by phyloseq, and ASVs assigned to non-prokaryotes, 24 archaea, chloroplasts and mitochondria were removed. Using the phyloseq package 25 [36], taxonomic richness was calculated for each sample and tested for differences 1 between biological compartments (skin mucus, gut and liver), tuna species (yellowfin 2 and skipjack), oceans (Atlantic and Indian oceans) and sexes (female and male) 3 using the non-parametric Kruskal-Wallis ANOVA test. Statistical significance was 4 assumed when p < 0.05. Within phyloseq, the composition and diversity of bacterial 5 communities were represented at the class level, based on the relative abundances 6 of ASVs in each sample. Within each sample, the detection of histamine-producing 7 bacteria (HPB) was based on the presence/absence of bacterial species reported to 8 produce histamine in the literature. Dissimilarities between bacterial communities 9 were assessed using Bray-Curtis distances, which were calculated with the vegan bacterial communities between the three organs (i.e., skin, gut and liver), a Venn 16 diagram was constructed using the VennDiagram package [40]. From the Venn 17 calculations, the list of specific ASVs within each biological organ was sorted in RStudio. 18 The occurrence of each ASV, i.e., the frequency of its observation in the samples of 19 a dataset, was calculated. For each biological compartment, the five most frequent 20 ASVs were identified to the lowest taxonomic level available. The taxonomic richness of bacterial communities, defined as the number of amplicon 3 sequence variants (ASVs), showed important differences and similarities between 4 sexes, tuna species (skipjack and yellowfin), biological compartments (skin mucus, 5 gut and liver) and sampling sites (Atlantic and Indian oceans). 6 Variability between sexes. Regardless of the tuna species, ocean and biological 7 compartment considered, the taxonomic richness of the bacterial communities did 8 not show significant differences between male and female individuals ( Fig. 1; Tab. 1). 9 Variability between tuna species. In the gut and liver samples, bacterial richness was 10 significantly higher in yellowfin than in skipjack tuna in both oceans (Fig. 1). 11 Statistical analysis confirmed that bacterial alpha diversity differed significantly 12 between the two tuna species, while no effect of sampling site was observed (Tab. 13 1). 14 Variability between oceans. In the skin mucus samples, the opposite pattern was 15 observed for alpha taxonomic richness, which was significantly lower in tuna 16 captured in the Indian Ocean ( Fig. 1) but not significantly different between skipjack 17 and yellowfin (Tab. 1). 18 Variability between compartments. The skin mucus layer hosted a significantly higher 19 bacterial richness than the gut and liver of both tuna species, regardless of the 20 sampling site (Fig. 1). However, ASV richness did not differ significantly between the 21 gut and liver samples (Tab. 1). 22

Beta diversity. 1
As observed for alpha diversity, the composition of the bacterial communities (beta 2 diversity) did not show significant differences between sexes, regardless of the tuna 3 species, sampling site and biological compartment (PERMANOVA, p > 0.05). 4 Skin microbiome. Skin samples showed significant similarities between tuna species 5 but large dissimilarities between the two sampling sites ( Fig. 2A,D). In both the Indian 6 and Atlantic oceans, the skin bacterial communities greatly differed from those 7 examined in the surrounding seawater (Fig. 2D, Supplementary material Fig. 1). In 8 Atlantic yellowfin and skipjack tunas, the skin bacteriome was dominated by 9 Gammaproteobacteria, representing up to 83% of the sequences (Fig. 3A). Several 10 other bacterial classes, such as Actinobacteria, Alphaproteobacteria, Bacilli, 11 Bacteroidia and Mollicutes, were also present, together representing less than 50% 12 of the sequences in most samples. In Indian Ocean yellowfin and skipjack, the same 13 bacterial classes were observed in much greater proportions, representing more than 14 50% of the sequences in some samples (Fig. 3B). 15 Gut microbiome. By contrast with the skin microflora, the gut microflora included a 16 bacterial assemblage that was clearly distinct between the two tuna species, while 17 sampling site had no significant effect (Fig. 2B,E). In skipjack tunas, the gut 18 bacteriome was dominated by Mollicutes (Fig. 3C,D), whereas that of yellowfins 19 tunas was more diversified, with higher proportions of Gammaproteobacteria and, to 20 a lesser extent, Alphaproteobacteria and Actinobacteria (Fig. 3C,D). Although 21 Gammaproteobacteria were generally more abundant in the gut of tuna collected in 22 the Indian Ocean, no significant differences were observed between the two oceans 23 Bacilli were, on average, lower in skipjack than in yellowfin. Tuna from the Indian 6 Ocean hosted a liver microflora that was globally less diversified than that of their 7 Atlantic counterparts (Fig. 3F). However, no clear pattern was observed, and the 8 composition of hepatic bacterial communities in the liver seemed to be slightly more 9 influenced by the sampling site. 10 11 Shared taxa and specific ASVs among the three organs. 12 The Venn diagram revealed that a relatively small proportion of all ASVs (4.5%) were 13 common to the skin, gut and liver (Fig. 4). Among these 138 common ASVs, the five 14 most common (observed in 60% to 90% of the samples) corresponded to three 15 species of the genus Photobacterium (i.e., P. leiognathi, P. damselae and P. 16 angustum), which are histamine-producing bacteria (HPB); Mycoplasma sp.; and 17 Cutibacterium sp. In addition, each compartment hosted a specific and diversified 18 assemblage of taxa. The skin microflora, with 1661 specific ASVs, accounted for half 19 of the total microbiome diversity (i.e., 53.7%). The five most common taxa were 20 Flavobacterium frigidarium, Psychrobacter sp., Rothia muciloginosa, Streptococcus 21 sp. and Alkanindiges sp. Comparatively, the gut and liver hosted 560 and 440 22 specific ASVs, respectively. These relatively similar numbers were unexpected and 23 show that the liver harbours a unique bacterial assemblage that is almost as large as 24 that found in the digestive tract of tunas. In this organ, the five most common taxa were Photobacterium sp., Vibrio sp., Mycoplasma sp., Sulfitobacter pontiacus and 1 Corynebacterium-1 aurimucosum. 2 3

Diversity and location of HPB. 4
In the variety of samples analysed, the community of HPB comprised 7 known taxa, 5 namely, Aliivibrio fischeri, Klebsiella oxytoca, Photobacterium angustum, 6 Photobacterium damselae, Photobacterium leiognathi, Photobacterium phosphoreum 7 and Vibrio harveyi (Fig. 5). In general, HPB were largely dominated by species of the 8 genus Photobacterium, but their respective proportions greatly varied between the 9 biological compartments. The liver showed the greatest occurrence of HPB in both 10 tuna species and ocean, with a total relative abundance reaching up to 68%. 11 Photobacterium damselae was rather abundant in the liver of Atlantic Ocean tuna, 12 whereas P. angustum was more prevalent in the Indian Ocean, mainly in yellowfin. 13 Conversely, the gut generally hosted the lowest abundance of HPB, especially in 14 tuna from the Atlantic, which exhibited nearly undetectable levels of HPB (Fig. 5A). In 15 the skin mucus, the diversity of HPB varied between the two oceans, as 16 Photobacterium angustum and Photobacterium leiognathi were found in large 17 proportions in Atlantic Ocean tuna while Photobacterium angustum was rather 18 dominant in fishes from the Indian Ocean (Fig. 5A,B). microbiome of the threespine stickleback and Eurasian perch, which was explained 5 by a differential diet between males and females [6]. During reproduction, the levels 6 of sex hormones usually increase, and the production of gametes can lead to higher 7 energy expenditure, especially in females [42,46,47]. During this period, females are 8 likely to modify their diet [48], which could alter the composition of their gut 9 microflora. In our study, although all the yellowfin were smaller than 70 cm and 10 therefore sexually immature [48,49], the skipjack in their size class are considered 11 mature and with the ability to reproduce throughout the year [46]. Therefore, the 12 strong microbiological homogeneity between sexes for this species strongly suggests 13 that the composition of the tuna microbiome is likely not subject to the influence of 14 sex hormones. 15

16
The gut microbiome of tropical tuna is species-specific. Our results showed that 17 the composition of the gut microflora differed between the two tuna species but not 18 between the sampling sites (i.e., for a given species) (Fig. 2). Skipjack and small 19 yellowfin tunas (size classes sampled in our study) are very similar anatomically, 20 physiologically and behaviourally [50]. They also share the same habitat in the water 21 column [51] and usually feed on the same prey (i.e., mostly fish, crustaceans and 22 cephalopods) [52,53]. In addition, individuals in this study were caught around fish 23 aggregating devices (FADs), under which both tuna species tend to gather, feed on their diets, one could expect similar gut microbiota compositions. In our study, the 1 enteric flora of yellowfin tuna was dominated by Proteobacteria, which is often the 2 case with carnivorous fishes [1]. By contrast, the gut of skipjack tuna hosted a 3 majority of Mollicutes of the genus Mycoplasma sp. (Fig. 3), which also form a major 4 component of the gut microbiome of salmons, mackerels and gobies [4,11,12]. Such 5 species-specific composition of the gut bacteriome is also well known in vertebrates, 6 including birds, primates, reptiles, fishes and mammals, and is thought to be driven 7 by host genotype, physiology and diet [2]. Here, for the reasons cited above, the diet 8 and physiology hypotheses were discarded. Our results are in agreement with the 9 phylosymbiosis hypothesis, which assumes that the host phylogeny reflects the The skin microbiome is influenced by external conditions. The composition of 22 the skin microbiota showed completely different patterns and greatly varied between 23 the two oceans but not between the tuna species (Fig. 2). Proteobacteria, sampling sites (Fig. 3). These phyla typically dominate within the skin microbiome of 1 fish species [14,29,58- The strong microbial similarities found between skipjack and yellowfin tunas in both 9 oceans in this study are interesting and tend to minimize the role of parameters 10 related to the host (i.e., genetic, physiology, immune system, and diet) in shaping the 11 surface microbiome, unlike what was observed in the digestive tract. By contrast, 12 several other studies suggested that host species, as well as physiology or diet, 13 could be a major driver of skin microbiome composition in marine organisms [14,62]. 14 However, those studies compared species belonging to different families and orders, 15 with contrasting physiologies and feeding habits (omnivorous vs herbivorous), which 16 is not the case between skipjack and yellowfin tunas. 17

18
The liver microbiome: an unexpected niche of high bacterial diversity. The most 19 striking result in this study was the discovery of a highly diversified and unique 20 bacterial assemblage in the tuna liver (Fig. 4). Since the liver is a highly vascularized in several of our samples (Fig. 5), and the first three were among the top five taxa 17 present in the "common microbiome" comprising ASVs shared by the three organs 18 (Fig. 4). Altogether, these results raise the hypothesis of active circulation of HPB 19 between the different organs of tuna, which might be mediated by the bloodstream. 20 Our results thus provide new perspectives by describing the liver as another major 21 reservoir of HPB, where these bacteria may not only transit temporarily but also 22 proliferate. Our results also show the need to include this organ in animal microbiome 23 investigations in order to respond to health issues that might be posed by the 24 consumption of animals by humans.

The core and meta-microbiomes in tuna.
In our study, although endemic 1 microbiotas were detected in the skin, gut and liver of tuna, our results also 2 highlighted the existence of a common microbiome shared by the three 3 compartments. These shared taxa (mostly represented by the genera 4 Photobacterium, Mycoplasma and Cutibacterium) represented only less than 5% of 5 all ASVs (Fig. 4); however, their ubiquity raises various questions about the 6 circulation, establishment and connectivity of bacterial communities within the fish 7 body. It is now recognized that enteric or epibiotic bacterial communities can interact 8 with other organs, such as the liver, the brain and the lungs, via complex pathways 9 involving blood circulation, immune system components, hormones and various 10 metabolites [22,64,69]. Mono-and bidirectional communication pathways, such as 11 the gut-skin axis or the gut-liver axis, have been described in humans and are 12 thought to be strongly involved in the development of diseases [23,70,71]. For 13 example, the gut-liver axis is now the subject of much speculation in relation to 14 human health [18]. Recently, modification of the gut microbiota was shown to alter 15 the tightness of the epithelial barrier, allowing the transfer of microbes and various 16 other metabolites into the blood and triggering the inflammation of liver tissue [64,65]. 17 Similarly, changes in the intestinal microflora could have a direct effect on the 18 production of neurotransmitters, hormones and other bioactive molecules capable of 19 acting on cutaneous receptors, thus altering the skin structure and its functions 20 [19,72]. Ethics approval and consent to participate 6 Not applicable. 7

Consent for publication 8
Not applicable. 9

Availability of data and materials 10
All data generated or analyzed during this study are included in this published article 11 and its supplementary information files. 12

Competing interests 13
The authors declare that they have no competing interests.