Monitoring Infection and Antibiotic Treatment in the Skin Microbiota of Farmed European Seabass (Dicentrarchus Labrax) Fingerlings

The microbiota of fish skin, the primary barrier against disease, is highly dynamic and modulated by several factors. In fish aquaculture, disease outbreaks occur mainly during early-life stages, with associated high economic losses. Antibiotic treatments sometimes remain the best option to control bacterial diseases, despite many reported negative impacts of its use on fish and associated microbiota. Notwithstanding, studies monitoring the effects of disease and antibiotic treatment on the microbiota of fingerlings are scarce. We sequenced the bacterial 16S rRNA V4 gene region using a metabarcoding approach to assess the impact of a mixed infection with Photobacterium damselae ssp. piscicida and Vibrio harveyi and subsequent antibiotic treatment with flumequine, on the skin microbiota of farmed seabass (Dicentrarchus labrax) fingerlings. Both infection and antibiotic treatment led to a significant increase in bacterial diversity and core microbial communities and impacted microbiome structure. Dysbiosis was confirmed by changes in the abundance of potential pathogenic and opportunistic bacterial taxa. Skin bacterial metabolic function was also significantly affected by flumequine administration, suggesting a detriment to fish skin health. Our results add to an increasing body of literature, showing how fish microbiome response to infection and antibiotics cannot be easily predicted.


Introduction
Commensal microbiota are an essential part of the immune response of animals, including fish, with a crucial role in disease prevention [1,2]. Perturbations to the homeostasis of commensal organisms are termed dysbiosis and can occur through three, not mutually exclusive, events linked to the occurrence of diseases [3]: (i) pathobiont proliferation, (ii) reduced diversity, and (iii) loss of beneficial microbes. Most of the microbiome research in vertebrates is focused on the gut microbiota, due to their recognized role in sustaining the gut-brain axis and in disease outcome [4]. However, in the case of fish, because pathogens are ubiquitous in the aquatic environment, fish skin and associated mucus are considered the primary barrier against diseases, with increasing numbers of studies focusing on the skin microbiota (e.g., [5][6][7]). Evidence shows that fish skin bacteria are highly dynamic, with composition and diversity being sensitive to both biotic (e.g., ontogeny, [8]) and abiotic factors (e.g., infection, [9]).
Although aquaculture is the fastest-growing food-producing industry, disease outbreaks due to pathogenic bacteria are one of the biggest constraints for its sustainability [10]. The impact of disease on the microbiota of several fish species has been assessed, albeit mainly regarding adult populations (e.g., skin and gill, [5][6][7]9]; gut and stomach, [11,12]). It is well established that disease often leads to dysbiosis, through a decrease in bacterial diversity and/or proliferation of pathogenic taxa other than the main etiological agent of disease (e.g., [5,13]). Frequently, dysbiosis also led to changes in predicted microbial function (e.g., [6,12,14]). Although disease and mortality incidences in aquaculture are higher for fry or young fingerlings, the impact of disease on the microbiota during these early life stages is still poorly understood [10]. Nevertheless, existing reports show that diversity, structure, and potential function of the microbiotas of early-life stages are also affected by disease (skin, [13]; gut, 15; 16]). The European seabass Dicentrarchus labrax is susceptible to several bacterial pathogens, two of the most concerning being Photobacterium damselae and Vibrio spp. [17]. Both bacterial genera are usually reported from the skin of healthy fish, and dysbiosis is typically accompanied by increases in their abundances (e.g., European seabass, [7]; perl gentian grouper, [18]). These pathogens cause photobacteriosis and vibriosis respectively, and infections can be systemic, affecting multiple organs [19,20].
Although vaccines are available against major diseases, procedures are costly and cause substantial stress and mortality, and typically only confer short-term immunity [21]. For this reason, antibiotics are still widely used to control disease in aquaculture. Several studies have reported the negative impacts of antibiotic use on fish, which include behavioral changes [22], microbial diversity decrease [7,22,23], increased susceptibility to secondary infections [24], changes in predicted microbial function [23], and mortality [25]. Flumequine is a fluoroquinolone antibiotic active against gram-negative bacteria and widely used in aquaculture [26]. Depending on water temperature, it can persist in fish skin and muscle up to 4-14 days after the last administration [27]. To the best of our knowledge, the impact of flumequine on the microbiota of fish has never been assessed.
Here, we used a 16S rRNA metabarcoding approach [i.e. metataxonomy, 28] to characterize the skin microbiota of the seabass Dicentrarchus labrax fingerlings before, during, and after a natural disease outbreak of Photobacterium damselae ssp. piscicida and Vibrio harveyi, and subsequent antibiotic treatment with flumequine. Our goal was to describe changes in composition, structure, and potential function of the skin microbiota caused by disease and subsequent antibiotic treatment. We predict that (i) disease will cause dysbiosis through a decrease in the microbial diversity and core microbiota, while antibiotic treatment will have the opposite effect; and (ii) both disease and antibiotic treatment will lead to changes in the most abundant bacterial taxa and differences in microbiome structure.

Experimental Design, Sampling, and Preparation
Fish were reared in an open water circulation system in a semi-intensive farming facility, where water is supplied from the Alvor Estuary (Southern Portugal). Sampled fish belonged to the same age cohort and were collected from the same rearing pond. Fish were 6 months old on the first sampling date, weighing on average 57 g, and 7 months old on the last sampling date, weighing on average 70 g. Individuals were randomly caught using a fishing pole and skin samples were taken using tubed sterile swabs (Medical Wire & Equipment, UK). We swabbed the right upper lateral part of the fish skin from head to tail. Swabs were immediately stored at − 20 °C until transportation on dry ice to the CIBIO-InBIO laboratory, where they were kept at − 80 °C prior to being processed.
Individuals were collected once a week between August 23 and September 13, 2016, encompassing four different sampling dates corresponding to four different fish health stages: healthy (N = 30), infected (N = 30), treatment (N = 30), and recovery (N = 15) ( Fig. 1). During the "healthy" state (sampled on August 23), all individuals were considered healthy due to lack of visible disease symptoms. The second sampling point occurred on August 30, and on August 31, fish began to die. Hence, samples collected on August 30 were categorized as "infected" (although still asymptomatic). Treatment with flumequine antibiotic was initiated on August 31, being administered at 35 g/ton of fish through commercial feed until September 6. Bacterial isolates from the liver, kidney, and spleen of infected fish were collected prior to the start of antibiotic treatment and pathogens were identified via PCR by a commercial company (Acuipharma, Spain). PCR amplification showed that the etiological agents of infection were Photobacterium damselae ssp. piscicida and Vibrio harveyi. Fish were again sampled on the last day of antibiotic administration (September 6) and these samples were categorized as "treatment." A final sampling point Total DNA from 105 fish samples and 6 controls (DNA extraction kit negative controls) was extracted using the PowerSoil DNA Isolation Kit (QIAGEN, Netherlands), following the manufacturer's protocol. We selected this kit based on its performance for extracting both gram-negative and gram-positive bacteria using mechanical cell lysis (bead-beating). Extraction kit negative controls were pooled into one single sample. DNA extractions were shipped in dry ice to the University of Michigan Medical School (USA) for amplification and sequencing on a single run of an Illumina MiSeq platform according to the protocol of [29]. Sequencing was done using a MiSeq Reagent Kit V2 500 cycles (Cat# MS-102-2003), according to the manufacturer's instructions. Each sample plus 4 PCR blanks and 4 identical mock communities (ZymoBIOMICS Microbial Community DNA Standard) were amplified and sequenced for the V4 hyper-variable region of the 16S rRNA gene (~ 250 bp).
In total, 3,389,081 partial 16S rRNA gene sequences were retrieved for the skin of the seabass fingerlings, with the number of sequences per sample ranging from 12,891 to 48,211. Amplicon sequence variants (ASVs) present in negative controls (extraction kit and PCR) were removed from downstream analysis. After removal of contaminants and non-bacterial sequences, 6,163 ASVs (3,300,989 sequences) were assigned to the skin microbiota of the seabass fingerlings. Diversity and bacterial abundances of the mock communities corresponded to those described by the manufacturer. Microbial taxa showing a mean relative proportion ≥ 1% were considered as part of the most abundant taxa in the microbiota.

Data and Statistical Analysis
Raw FASTQ files were denoised using the DADA2 pipeline in R [30]. We estimated a midpoint rooted tree of ASVs using the Quantitative Insights Into Microbial Ecology 2 package (QIIME2; release 2020.11). We constructed a table containing ASVs and made taxonomic inferences against the SILVA (138 release) reference database [31]. We normalized ASV abundances using the negative binomial distribution, which accounts for library size differences and biological variability [32]. The core microbiota was assessed at the ASV level for each health state (i.e., sampling date) separately. An ASV was considered as part of the core microbiota if present in 100% of the samples from each state.
Microbial alpha-diversity was calculated at the ASV level using Shannon and Faith's phylogenetic (PD) diversity indices as implemented in the R phyloseq package [33]. Additionally, Pielou's evenness was calculated at the ASV level as implemented in the R microbiome package [34]. Microbiome structure (beta-diversity) was also estimated at the ASV level using phylogenetic UniFrac (weighted and unweighted) and Bray-Curtis distances. We assessed variation in microbiome diversity and structure using the Kruskal-Wallis and PERMANOVA [adonis function of the R vegan package, [35]] tests, respectively. Dissimilarity between samples was visually assessed through a principal coordinates analysis (PCoA) and dendrograms. A heatmap was built to depict changes in the abundance of the most abundant phyla and genera (≥ 1% of all reads). All analyses were performed in R-studio v4.0.2.
Predicted bacterial metabolic functions were estimated using the metagenomic Phylogenetic Investigation of Communities by Reconstruction of Unobserved States software (PICRUSt2) embedded in QIIME2 [36] and applying a weighted nearest sequenced taxon index (NSTI) cutoff of 0.03. Predicted metagenomes were collapsed using the Kyoto Encyclopedia of Genes and Genomes (KEGG) Pathway metadata [37]. We used linear discriminant analysis (LDA) in LEfSe to identify differentially abundant metabolic pathways in the skin microbiota of the seabass fingerlings using state as a class, a P-value cut-off of 0.05 and a LDA effect size cutoff of 2 [38].

Microbial Diversity and Composition
Alpha-diversity estimates varied significantly between states (P ≤ 3 −3 , Table 1), with the exception of the PD index between treatment and recovery states, and Pielou's evenness between healthy and recovery states (P = 0.2, Table 1). There was a significant increase in diversity and evenness from healthy to infected states, as well as from infected Table 1 Microbial diversity comparisons for the skin of the seabass Dicentrarchus labrax across all samples and between the four different states. For each Kruskall-Walis test (alpha-diversity), we report the chi-squared value (overall) and significance (P value, overall and pairwise) and for each PERMANOVA test (beta-diversity), we report the R 2 statistics (overall) and significance (P value, overall, and pairwise). Significant differences are indicated in bold

Metric
Overall Healthy/ Infected to treatment states (Fig. 2). Between treatment and recovery states, a significant decrease in diversity and evenness occurred (Fig. 2). There were a total of 49 bacterial phyla and 926 bacterial genera identified across all samples. Of these, 6 phyla and 32 genera were found in high abundance in at least one of the health states (Fig. 3, Online Source 1). The abundance of these taxa varied with the onset of infection and antibiotic treatment (Fig. 3, Online Source 1). It is noteworthy that the abundance of Photobacterium (identified as P. damselae) in the skin remained stable between healthy and infected states and decreased with antibiotic treatment and then again during recovery (Figs. 3 and 4, Online Source 1). On the other hand, the abundance of Vibrio increased from healthy to the infected state, decreased with antibiotic treatment, and increased again in recovery (Figs. 3 and 4, Online Source 1).
Of the 74 ASVs recovered from the core microbiota across the four states, 6 were present in the healthy state, 20 in the infected state, 30 in the treatment state, and 67 in the recovery state (Online Source 2). These numbers corresponded to 0.4, 0.7, 1, and 4% of the total ASVs found in the healthy, infected, treatment, and recovery states, respectively. There were 2 unique core ASVs in the infected and treatment states each, while 38 were unique to the recovery state (Online Source 2). Of the 20 unique core ASVs recovered from the microbiota of infected fish, one was identified as Photobacterium damselae (Online Source 2). Additionally, there were 2 ASVs belonging to Vibrio that were part of the core microbiota during the infected state, but not during the healthy or treatment states. In the recovery state, those two and four other Vibrio ASVs were part of the core microbiota of the skin.
Beta-diversity estimates showed significant differences between states (p ≤ 6 −3 , Table 1), except for the UniFrac weighted distance between healthy and recovery states (P = 0.1, Table 1). On the other hand, only a low percentage of the variation between states was explained by our model (R < 0.3, Table 1). Analyses of the PCoA of Bray-Curtis distances showed that samples from the healthy, treatment, and recovery states clustered separately, while no apparent structural cluster of samples between states was observed using the UniFrac distances (PCoA, Fig. 2). Moreover, the dendrogram constructed from the UniFrac unweighted distance showed two main clusters separating healthy fish from the remaining samples. Within the latter cluster, treatment and recovery states were more closely related (Fig. 2). Dendrograms constructed from the UniFrac unweighted and Bray-Curtis distances showed no clustering of the health states (Online Source 3).

Microbial Predicted Function
There were a total of 478 predicted KEGG pathways in the skin microbiota of the seabass. Linear discriminant analysis of the metagenomic predictions performed in LEfSe showed that 3, 5, and 7 different pathways were significantly enriched in the healthy, treatment, and recovery states, respectively (Fig. 5). Interestingly, there were no significantly enriched pathways in the infected state. On a broad level, all enriched pathways were related to either biosynthesis (67% in healthy, 57% in recovery) or degradation/utilization/assimilation (33% in healthy, 100% in treatment, 43% in recovery) categories (Fig. 5). Specifically, enriched metabolic pathways in the healthy state were related to amino acid degradation and fatty acid and lipid biosynthesis; in the treatment state, enriched pathways were related to carbohydrate degradation and secondary metabolite degradation; and finally, in the recovery state, enriched pathways were related to carbohydrate degradation, vitamin biosynthesis, fatty acid biosynthesis, purine nucleotide biosynthesis, and sugar derivative degradation (Fig. 5).

Discussion
We characterized for the first time the effects of bacterial infection with Photobacterium damselae ssp. piscicida and Vibrio harveyi and treatment with flumequine on the skin microbiota of seabass fingerlings. Most of our predictions were confirmed with one important exception; both core microbiota and microbial diversity increased with the onset of infection. However, dysbiosis was accompanied by an increase in the abundance of potential pathogenic and opportunistic taxa.

Disease Effects on Skin Microbiota of Seabass Fingerlings: Healthy vs Infected States
Impacts on microbial diversity, richness, and evenness caused by infection by bacterial pathogens and parasites have been described in some fish, including the skin of rainbow trout infected with Ichthyophthirius multifilis [14]; the gut of Asian seabass infected with Tenacibaculum singaporense Fig. 3 Heatmaps of the most abundant (≥1%) phyla and genera recovered from the skin of the seabass Dicentrarchus labrax fingerlings for healthy, infected, treatment, and recovery states. Unknown genera are labeled as u.g Fig. 4 Barplots of relative frequency of key genera recovered from the skin of the seabass Dicentrarchus labrax fingerlings for healthy, infected, treatment, and recovery states. Unknown genera are labeled as u.g [16]; the skin of Atlantic salmon infected with Lepeophtheirus salmonis [6]; the gut of grass carp with enteric infection [12]; the gut of brown trout infected with Tetracapsuloides bryosalmonae [15]; the skin of orbicular batfish infected with Tenacibaculum maritimum [13]; the gut and stomach of rainbow trout infected with Caligus lacustri [11]; and the skin of adult seabass infected with Photobacterium damselae [7]. Dysbiosis was reported in the vast majority of these studies through decreases in fish microbial diversity and increases in pathobionts. Although an increase in diversity was observed in the present study, the direction of the changes in the abundance of key microbial taxa indicates dysbiosis occurred in the skin microbiome of seabass fingerlings. In the present study, Vibrio (that encompasses one of the etiological agents of infection), and two other unidentified genera belonging to families with opportunistic taxa, Flavobacteriaceae and Vibrionaceae [39], increased their abundance in the infected state. Another genus that increased in abundance in infected fish was Aureispira, previously found to be highly abundant in the intestinal microbiota of grouper juveniles after iridovirus infections [40]. Similar increases in bacterial diversity after infection have been already described in other fish (e.g., [6,12,14,15]), indicating that changes in microbial diversity cannot be readily anticipated, and growth or decline of specific taxa is easily predicted. On the other hand, the stage of the infection could be influencing the observed results, since different signs of dysbiosis in fish were previously reported for different infection stages [5].
Alterations to the core microbiota in response to infection were also previously reported in the skin of adult European seabass infected with Photobacterium damselae [7], and in yellowtail kingfish infected with enteritis [5]. In the present study, a Photobacterium damselae ASV had a prevalence of 100% in the infected state. However, its mean abundance remained unaltered between the healthy and infected states, suggesting the skin was only indirectly affected by this pathogen. These results are in line with our previous study showing that infection caused by P. damselae can lead to dysbiosis of skin microbiota of farmed seabass despite no increase in abundance [7]. Similar results were also obtained by Legrand et al. [5] when describing skin and gill dysbiosis in the yellowtail kingfish during enteritis gut disease.
Microbiome community structure of the skin microbiota was also significantly affected by infection. Although samples of diseased fish were collected on the day prior to disease onset, only a few samples from the infected state (7 out of 30) clustered within the healthy group, confirming that significant taxonomic changes had occurred in most individuals analyzed. These results suggest that some properties of the skin mucous that allowed certain phylogenetically related taxa to thrive in the skin may have been altered by infection, consequently affecting the resident microbiota. However, despite the increase in microbial diversity and changes in structure driven by infection, predicted microbial metabolic functions remained unaltered. This suggests that the increase in diversity observed between healthy and infected states was due to colonization by bacteria capable of performing similar metabolic functions [1].

Flumequine Effects on Skin Microbiota of Seabass Fingerlings: Infected vs Treatment States
In the present study, microbial diversity was observed to increase on the 8 th day of treatment with flumequine. However, as expected, administration of flumequine resulted in a decrease in abundance of both etiological agents of disease in this study. This is unsurprising given the reported sensitivity of both species to this antibiotic [41]. Importantly, this treatment led to an increase of potentially harmful Flavobacteriaceae [39]. Interestingly, the genus Alteromonas, which has been shown to exhibit antibacterial activity against fish pathogens, including Photobacterium damselae and several Vibrio spp. [42], and resistance against amoxicillin, erythromycin, and gentamicin [43], increased in abundance during antibiotic treatment. Microbial disruptions have been reported after oxytetracycline and rifampicin treatment in microbiota of adult fish (e.g., [7,[22][23][24]), and after streptomycin, ciprofloxacin, or oxytetracycline treatment on earlier fish life stages (e.g., zebrafish larvae, [25,44]). Although an increase in fish microbial diversity caused by antibiotic administration is less common, it has been also reported before (e.g., [7,44]). In fact, in the studies where diversity decreased after antibiotic exposure, there were no pre-existing health conditions. On the other hand, in this study as well as in [7], infection had occurred, and in the study [44], fish were immersed with the anti-nutritional compound saponin before antibiotic treatment. This suggests pre-existing infection (microbiota disruption) and antibiotics may have a cumulative effect on skin microbial diversity.
Significant changes in the predicted potential function of the skin microbiota were detected after antibiotic treatment. Specifically, the degradation of carbohydrates and secondary metabolites were significantly enriched during antibiotic treatment. However, the production of carbohydrates and secondary metabolites is linked to the protective role of the microbiota (e.g., [1,45]). For example, carbohydrates are directly related to specific cell-cell adhesion, modulating microbial binding to the mucus [46]. It has been suggested that carbohydrate synthesis by the human microbiota helps to establish symbiosis with microbial commensals and aids pathogenic evasion [45]. Furthermore, production of secondary metabolites is one of the mechanisms by which commensal microbiota fight against pathogens [1]. These results suggest a microbial response to antibiotics, which may ultimately have a negative effect on fish immunity. On the other hand, the extracellular matrix that covers the biofilms formed by bacteria are known to have carbohydrate-rich components [47,48]. The predicted enrichment of degradation of carbohydrates pathways seen here in the microbial communities of the fish skin may also be an effect of antibiotic administration.

Recovery of the Skin Microbiota in Seabass Fingerlings: Healthy vs Recovery States
Previous studies reported that short-term recovery of the microbiota of fish after antibiotic treatment does not lead to the diversity levels observed in the healthy state (e.g., 1-week recovery, [24]). In the present study, with the exception of Pielou's evenness, diversity significantly increased between healthy and recovery states. Importantly, the abundance of Vibrio increased in the recovery state, indicating microbial balance may not have been fully obtained.
Microbiome structure was also significantly different between the recovery and healthy states. However, closely related microbial structuring was found in fish from treatment and recovery states. Almost half of the enriched predicted metabolic pathways during the recovery state were related to the same categories of pathways enriched during the treatment state (carbohydrate and secondary metabolite degradation), with significant differences from the healthy state. To the best of our knowledge, only the study by Brumlow et al. [49] has effectively measured the effects of 3-day antibiotic treatments (tetracycline and rifampicin) in the biochemical profile of the skin of Gambusia affinis. In this study, the authors also report changes in microbial community composition relative to pre-treatment after an 8-day recovery. However, unlike the present results, where significant changes in microbial function were predicted, the results of Brumlow et al. [49] indicated that the biochemical functions of the microbiota were mostly reestablished after the 8-day recovery. Flumequine is a highly persistent antibiotic and can take several weeks to be fully depleted from the blood and tissues of fish [50,51]. This antibiotic has a slower depletion rate in the skin than in the muscle or liver and can be present in the skin 20 days after oral administration [52]. Although a longer time frame would be necessary to evaluate whether full functional recovery of the skin microbiota does occur, our results highlight the high susceptibility of skin microbiota to antibiotic exposure.

Conclusions
Homeostasis of the microbial communities in the mucosal surfaces of fish is central to control pathogen abundances [2]. Here, we describe a dysbiosis episode caused by an infection outbreak induced by Photobacterium damselae ssp. piscicida and Vibrio harveyi, and subsequent antibiotic treatment with flumequine. Although overall, antibiotic treatment appeared to have a greater impact on the skin microbiota when compared to infection, this could be a result of a cumulative effect of both infection and antibiotic treatment. Moreover, the microbial profile of the fish in the recovery state differed from that of the fish in the healthy state, as well as their predicted metabolic functions. The results of this study highlight that the physiological response of fish commensal bacteria to infection and antibiotics is complex, and not easily predicted.

Data Availability
The raw sequences are available at NCBI Sequence Read Archive (SRA) database within the BioProject ID PRJNA741392.

Declarations
Ethics Approval Animals in this study were reared in a commercial fish farm located in the estuarine environment of the Alvor Estuary (Portimão), southern Portugal. Fish were handled by the fish farm employees and samples were taken non-invasively. According to the Portuguese legislation DL Nº 113/2013, our work does not involve animal experimentation and therefore is exempted from the need of ethical approval.