Habitat and Client Diversity Inuence the Skin Microbiome of the Caribbean Cleaner Goby Elacatinus Evelynae

Fish associated microorganisms are known to be affected by the environment and other external factors, such as microbial transfer between interacting partners. One of the most iconic mutualistic interactions on coral reefs are the cleaning interactions between cleanershes and their clients, during which direct physical contact occurs. Here, we characterized the skin bacteria of the Caribbean cleaner sharknose goby, Elacatinus evelynae, in four coral reefs of the US Virgin Islands using sequencing of the V4 region of the 16S rRNA gene. We specically tested the relationship between gobies’ level of interaction with clients and skin microbiota diversity and composition. Our results showed differences in microbial alpha-and beta-diversity in the skin of gobies from different reef habitats and high inter-individual variation in microbiota diversity and structure. Overall, the results showed that sh-to-sh direct contact and specically, access to a diverse clientele, inuences the bacterial diversity and structure of cleaner gobies’ skin. Because of their frequent contact with clients, and therefore, high potential for microbial exchange, cleanersh may serve as models in future studies aiming to understand the role of social microbial transfer in reef sh communities.


Introduction
Coral reefs are highly complex marine ecosystems that have been the focus of numerous studies examining the drivers of biodiversity and community dynamics [1]. As in other ecosystems, coral reef microorganisms are emerging as key members to maintaining reef health and resilience in the face of large-scale degradation due to climate change, human impacts and emerging diseases of corals (reviewed by [2]). This has resulted in increased efforts to characterize diseases of reef microbial communities, and to identify which organisms in uence resilience and recovery (e.g. [3]). Recent studies have shown the importance of interactions between the coral microbiome and the larger reef community (e.g. between reef-building coral and benthic algae, [4]). However, relatively few studies have examined the microbiome associated with the most mobile members of the reef community, such as shes (e.g. [5,6]).
Fish microbial communities are known to be affected by multiple biotic and abiotic variables. There is evidence that sh-associated bacteria are organ-speci c, species-speci c and individual speci c, thus comprising highly diverse communities [7][8][9]. While host factors seem to be the major drivers of sh gut bacterial diversity [10,11], the physicochemical properties of the water exert a considerable effect on the diversity of sh skin-associated microbes (e.g. temperature and salinity, [12]). Despite the impact of the surrounding environment, the contribution of the water microbiota to the sh skin microbiota composition seems negligible, with sh mucosae being a highly selective environment (e.g. [13]). However, other external factors, such as direct transfer of microorganisms between shes, might also play a major and still unexplored role [14][15][16]. Microbial transfer between interacting partners has been shown to be common in nature, shaping microbial consortia in humans and other animal groups (reviewed in [17,18]), including sh [14]. Although social microbial transmission could ultimately increase microbiome complexity, which may reduce the abundance of opportunistic and/or pathogenic taxa, as seen in bees [19] or chimpanzees [18], social interactions may also facilitate pathogen transmission and consequently increase levels of infection and disease (reviewed by [20]).
One of the most iconic mutualistic interactions on coral reefs is the relationship between cleaner shes and clients. Cleaners attract individuals from multiple species (clients) to their "cleaning stations", which are usually xed territories where they inspect the body of multiple client shes per day to remove parasites, dead tissue and mucus [21]. Although many small sh species clean opportunistically, members of two genera are obligate or "dedicated" cleaners [22]. These include the cleaner wrasses (Labroides spp.) in the Indo-Paci c and the cleaner gobies (Elacatinus spp.) in the Caribbean region.
Cleaner gobies of the genus Elacatinus reside on benthic substrate, moving only to make contact with client shes [21,23], which travel and interrupt other activities to visit cleaner gobies [24,25]. Visits to cleaner goby stations can be in uenced by multiple factors including location relative to territorial client sh [24][25][26], local sh abundance [27], structural complexity [28], and parasite activity and abundance [26,29]. Consequently, the abundance and diversity of client shes can vary widely among cleaning stations. Because of their frequent contact with heterospeci cs, and therefore, high potential for microbial exchange, cleaner sh may serve as a useful animal model system to understand the role of social microbial transfer in ecological communities [16]. Indeed, a recent study comparing "cleaner" vs "noncleaner" ecotypes of E. prochilos from Barbados found that bacterial diversity was signi cantly increased in "cleaner" ecotypes [15].
Here, we characterized the skin bacteria of the most ubiquitous Caribbean cleaner goby species, the sharknose goby Elacatinus evelynae, in several reefs within the US Virgin Islands, using 16S rRNA gene (V4 hypervariable region) amplicon sequencing. We speci cally studied the relationship between gobies' level of interaction with clients, and skin microbiota diversity and composition. We expected to nd a relationship between microbial diversity and client diversity and geographical differences in the skin microbiota among reefs due to putative socio-environmental differences.  (Fig. 1). Donkey Bight is dominated by a mix of rocky reef, live and dead coral, sand, and seagrass, while West Lameshur is highly degraded with almost no live coral, and it is located near a mangrove swamp [30].

Study species, sites, and behavioral observations
We focused on the sharknose goby Elacatinus evelynae, which are small sh (1.2-3.5 cm total length) with a prominent lateral blue and yellow stripe running from the snout to the base of the tail. This species is common across the study reef sites, inhabiting the surface of living coral, usually Siderastrea spp., Orbicella spp., Montastrea spp., Diploria spp., and Pseudodiploria spp. [23]. Cleaning stations were identi ed, and a subset was randomly selected for this study. Observations of cleaning interactions were made by two snorkelers positioned as far from the station as possible while still being able to see cleaning interactions (at least 2 m). Individual gobies were observed for 30 min at each location three times a day (at dawn, midday and dusk, totaling 1h30min for each cleaner) encompassing the hours during which E. evelynae are more active [25]. Observations were registered after a 2-5 min delay to allow the sh to become accustomed to the presence of the observer. During each observation period, the number of clients visiting each cleaning station, number of clients inspected, number of client genera visiting, number of client genera inspected, and average inspection time were recorded.
Following the nal observation, cleaner gobies were captured using individual hand nets. Nets were submerged in a 30% bleach solution and rinsed with fresh water prior to each use. Upon capture, both sides of each sh were swabbed with tubed sterile cotton swabs (MedicalWire), which were stored at -80°C until DNA extraction.

Laboratory procedures
Total DNA was extracted using the PowerSoil extraction kit and following manufacturer's protocol.
Extracted DNA was shipped in dry ice to the Centre for Microbial Systems at the University of Michigan Medical School (USA) where the V4 hypervariable region of the 16S rRNA gene (~ 250 bp) was ampli ed for each sample and controls (i.e. extraction and PCR blanks) with the primers 515F/806R [31] and sequenced using a dual indexing sequencing strategy [32]. Libraries were pooled and sequenced in a single Illumina MiSeq sequencing run.

Data analysis
Raw FASTQ les were analyzed using the DADA2 pipeline [33] for merging paired end reads, ltering and sequencing error correction using the following parameters: trimLeft = 20, truncLen = c(220, 200), maxN = 0, maxEE = c(2,2), truncQ = 2. Singletons were discarded and reads were collapsed into amplicon sequence variants (ASVs). Taxonomy was determined against the SILVA reference database (release 132) [34]. ASVs present in PCR and extraction blanks, that remained unclassi ed or were classi ed as Mitochondria (identi ed as Family) and Chloroplast (identi ed as Class) were removed from the dataset. Archaea were also excluded since the primers used are known to discriminate against this group in the marine environment [35]. An ASV frequency table was constructed with the R package phyloseq [36] and read normalized counts were obtained via the negative binomial distribution implemented in DESeq2 [37]. Raw sequence reads were deposited into NCBI's Short Read Archive under accession PRJNA756005.
Alpha-diversity (intra-sample) was estimated using Shannon, Fisher and Faith's Phylogenetic Diversity (PD) indices using the R package phyloseq [36]. Alpha-diversity differences among localities were tested using Generalized Linear Models (GLMs) and post-hoc comparisons were evaluated with Tukey's HSD test. To test the effect of each cleaning behavior on microbial alpha-diversity, Linear Mixed Models (LME) were performed using the number of clients visiting each cleaning station, number of clients inspected, number of client genera visiting, number of client genera inspected and average inspection time for each model as xed factors (predictors). Locality was included as a random factor and models were built using the R package lmer. The signi cance of GLM and LME models was estimated using ANOVA of type III with Satterthwaite approximation for degrees of freedom. Beta-diversity (inter-samples) was estimated using phylogenetic informed weighted and unweighted Unifrac [38] and Bray Curtis (BC) indices. Dissimilarity in microbial structure among reef locations was visualized using principal coordinates analysis (PCoA). The homogeneity of beta diversity dispersion among localities was also assessed by rst calculating the average distance to the sample group centroid using the betadisper function of the vegan R package [39] and then compared using a permutation test. Tukey's HSD test was used for posthoc comparisons. Differences in microbial structure among reefs were then tested with a PERMANOVA with the strata option for locality and 9999 permutations, as implemented in the adonis function of the vegan package. Post-hoc comparisons were performed using a pairwise PERMANOVA with the Bonferroni p-value correction for multiple testing. Additionally, the differences in the abundances of phyla or genera represented by ≥ 3% on average of all sequences were also assessed among reef locations and cleaning activity using the same GLM and LME structure described above. A Venn diagram was used to depict the number of ASVs shared among reef locations.
A dissimilarity matrix with the client genera inspected by each goby was constructed and differences among localities were also tested with a PERMANOVA using the BC index. To test the correlation between cleaning activity and the microbial community of each locality, Mantel statistical correlations based on Spearman's rank were performed between the number of clients visiting the cleaning station, number of clients inspected, number of client genera visiting and inspected and the beta-diversity distance matrices using the vegan R package. For all tests, differences were considered signi cant when P < 0.05.
Variation of the skin microbiome of cleaner gobies with cleaning activity Alpha diversity of the bacterial communities associated with the cleaner goby skin was positively (i.e., increasingly) correlated to the number of clients and client genera visiting the cleaning station and inspected for all reefs except for West Lameshur, which showed an inverted pattern, i.e., higher number and diversity of clients corresponded to lower values of alpha diversity (Fig. 3) Fig. 4). No differences were observed in the alpha diversities of individuals from the remaining locations.
Bacterial community structure (beta-diversity) was signi cantly different amongst reef localities with all beta diversity indices (p < 0.003), with signi cant pairwise differences between Donkey Bight and each of the other reef sites (p < 0.02; Fig. 5a,b,c and Supplementary Table S5). Beta dispersion was also signi cantly different among locations considering the Bray Curtis (F = 5.94, p = 0.002; Fig. 5d) and weighted Unifrac (F = 6.97, p = 0.001; Fig. 5e) indices. Pairwise comparisons of beta dispersion for the Bray Curtis index showed differences between Donkey Bight and West Lameshur, as well as between Donkey Bight and Hull Bay (TukeyHSD, p < 0.003; Fig. 5d). For the weighted Unifrac distance, differences were found between Donkey Bight and all remaining localities (Tukey HSD, p < 0.03; Fig. 5e).

Discussion
Cleaning stations have been shown to attract a wide diversity of sh species and thus, enhance local reef sh biodiversity and abundance (e.g. [40]). Because of the direct physical contact between cleaners and clients, there is the potential for cleaning stations to act as hubs for microbial exchange between sh. In this study, we used a 16S rRNA gene amplicon sequencing approach to test whether clientele diversity was associated with differences in the skin microbiome of the cleaner goby E. evelynae in the US Virgin Islands. Our results showed differences in alpha-and beta-diversity amongst gobies from different sampled reefs with few shared ASVs among them and high inter-individual variation in microbiota diversity and structure. Overall, the results showed increasing bacterial alpha diversity with the number of clients and client genera inspected (except in West Lameshur), as well as a positive correlation between beta-diversity and clientele diversity.

Goby cleaning activity impacts skin microbial diversity
Access by gobies to different reef sh species might be shaped by the level of reef degradation and therefore in uence cleaning interactions and consequently, goby microbiome. Our results showed a positive correlation between goby skin bacterial beta-diversity and the diversity of clientele inspected (i.e., number of client genera) in all sampled reefs. Moreover, differences in clientele diversity visiting cleaning stations were also observed in our study. For example, although diversity of client species was high at West Lameshur, the most common clients were Stegastes damsel sh, which are territorial sh only visiting cleaning stations within their territories [24], but usually less parasitized [29]. The territorial behavior of these sh can also in uence which other potential client species gain access to cleaning stations. By contrast, the remaining sites showed higher visitation rates of larger sh species, such as striped parrot sh (Scarus iseri) in Brewers Bay, yellow goat sh (Mulloidichthys martinicus) in Hull Bay and ocean surgeon sh (Acanthurus bahianus) in Donkey Bight, all of which are "preferred" clients, likely due to larger body size and thus higher parasite burden [23,41]. Larger clients, therefore, engage in longer cleaning interactions [41], which could increase bacterial transfer. However, in our study only client diversity was positively correlated with bacterial structure in the skin of cleaner gobies, indicating a stronger effect of client diversity compared to duration of inspection.
Recent studies have shown microbial changes between sh participants of symbiotic relationships. For example, microbial composition of clown sh mucus changes with contact with its anemone host [42]. Similarly, a "cleaner" ecotype of the Barbadian broadstripe cleaner goby Elacatinus prochilos harbored higher skin microbiota diversity than "non-cleaner" ecotypes [15]. Microbial interhost dispersal in zebra sh has also shown to in uence diversity and composition of microbial communities affecting host immune system [14]. However, the mechanisms involved in those changes are not well understood. Despite the high diversity of clients visiting cleaning stations in West Lameshur, not only did cleaners from that location have similar bacterial alpha-diversity levels to the ones from Brewers Bay and Hull Bay, but they also showed a contrasting relation between microbial diversity and cleaning activity when compared to the gobies in the other reefs. A possible explanation for the differences found in West Lameshur might be related to the greater habitat degradation at this sampling locality [30], which could have altered local reef community dynamics. Although reef animal communities harbor some of the most diverse microbial communities of the marine environment [42], it is estimated that the changes in host communities in a given location may impact microbial diversity in the entire reef [43]. Cleaner gobies inspect multiple client sh per day and engage in direct sh-to-sh contact [21,44]. Although this creates the opportunity for exchange of microbes, given the interspeci c nature of the interactions, actual exchange and persistence of microbes cannot be assumed. Our data, while correlative, supports this hypothesis (i.e., microbial exchange increases with cleaning activity). The alternative explanation, that cleaners with more diverse microbiota attract a more diverse array of clients, seems less likely but cannot be ruled out without experimental manipulation.
Cleaner sh have been shown to remove signi cant numbers of ectoparasites from hosts [45], which could otherwise compromise client welfare by causing skin damage, feeding on blood, and acting as vectors for diseases (reviewed by [46]). Nonetheless, the gain of a seemingly easy meal for cleaners (client-gleaned ectoparasites and mucous) may come with a price: while obviously predatory clients may eat the cleaners [23], less obvious is the fact that clients may also be vectors of parasites, bacterial contamination, and consequently disease to cleaners [47]. Although frequent contact with other reef sh seems to potentiate chances of increased microbial diversity and diversity in cleaners, which may protect against infections [48], further experiments on cleaner sh microbial exchange should be performed to understand the potential risks of cleaning activity and their impacts on reef communities. Goby skin microbiota varies across reef locations Spatial differences in skin microbiota have been reported for vertebrates, such as bats [49], amphibians [50], and marine species [51,52], including reef shes [5,6]. Although our main goal was to examine the relationship between client diversity and microbial diversity, we also observed differences in goby skin microbial composition among reef sites. Interestingly, those differences include contrasting results from sh captured from different reef habitats within less than 500 m of each other. Even though they are located within the same bay, Donkey Bight and West Lameshur differ in coral cover [30]. Additionally, a mangrove swamp empties out near West Lameshur site, and therefore water quality parameters and reef communities are likely to vary among our sites [53], leading to differences in sh microbial consortia. In fact, several studies have shown that sh skin microbiota respond to changes in the physicochemical composition of the water [11,13].
Despite host taxonomy being considered one of the main factors in uencing sh microbiome [5], a high skin microbial variability among individuals was found alongside the absence of a core microbiome in the study area for E. evelynae. Indeed, only a small proportion of bacterial taxa (3.5%) was common to all localities (Fig. 2), suggesting that local environmental differences might have a signi cant impact on the structure of goby skin bacteria. Other studies examining microbial composition of marine species between different localities found a core microbiome even between distant locations (e.g., whale blow between Paci c and Atlantic humpback whales, [54]). The present data shows that cleaner gobies share a considerably low proportion of bacteria even in a small geographic context, suggesting a substantial effect of environment, which may also be responsible for differences in clientele diversity and abundance at each sampled site.

Conclusions
This study suggests that sh-to-sh direct contact and speci cally, access to a diverse clientele, in uences the bacterial diversity and structure of cleaner gobies' skin. However, our study did not control for environmental factors and therefore, the extent to which microbial diversity of cleaner gobies can be in uenced by the surrounding environment and social behavior needs to be further explored in controlled experimental conditions. Nonetheless, this study sets the stage for future research using cleaner gobies as models to understand microbial dynamics in coral reefs. Besides the cleaner gobies studied herein, the microbiome of other obligate cleaners such as wrasses in Indo-Paci c reefs and the less studied but highly diverse group of cleaner shrimps [22], may also be in uenced by cleaning behavior, and speci cally by client diversity. Given current concerns over reef degradation worldwide and the importance of microbial commensals towards reef resilience, holistic studies examining microbial transfer to and from cleaner sh and other reef sh and the potential cascading effects deriving from such interactions are warranted. Additionally, microbial communities residing in areas surrounding cleaning stations, where sh largely congregate, should also be investigated due to their potential effects to the entire reef holobiont.

Declarations Funding
Funding was provided by the National Science Foundation awards OCE-2023420 to PCS and OCE-2022955 to AA, and by the European Regional Development Fund (ERDF) through the COMPETE program and by National Funds through Foundation for Science and Technology (project PTDC/BIA-MIC/27995/2017 POCI-01-0145-FEDER-027995) to RX. RX was also supported by Foundation for Science and Technology (FCT) under the Programa Operacional Potencial Humano-Quadro de Referência Estratégico Nacional funds from the European Social Fund and Portuguese Ministério da Educação e Ciência (IF/00359/2015, and 2020.00854.CEECIND). Field data were collected with support from NSF OCE-1536794 to PCS.

Con icts of interest/Competing interests
The authors have no con icts of interest to declare that are relevant to the content of this article.  Boxplots of the alpha-diversity measures for each locality with Tukey's HSD signi cance for pairwise differences. * indicates signi cant differences Linear regression plots depicting alpha diversity measures versus each of the observed cleaning variables Figure 5 a), b), c) PCoA plots with beta-diversity distances grouped by locality with 95% con dence interval ellipse, d), e), f) beta dispersion represented by distance to centroid for each beta diversity measure

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.