The gut microbiome of farmed Arctic char (Salvelinus alpinus) is shaped by feeding stage and nutrient presence

Abstract The gut microbiome plays an important role in maintaining health and productivity of farmed fish. However, the functional role of most gut microorganisms remains unknown. Identifying the stable members of the gut microbiota and understanding their functional roles could aid in the selection of positive traits or act as a proxy for fish health in aquaculture. Here, we analyse the gut microbial community of farmed juvenile Arctic char (Salvelinus alpinus) and reconstruct the metabolic potential of its main symbionts. The gut microbiota of Arctic char undergoes a succession in community composition during the first weeks post-hatch, with a decrease in Shannon diversity and the establishment of three dominant bacterial taxa. The genome of the most abundant bacterium, a Mycoplasma sp., shows adaptation to rapid growth in the nutrient-rich gut environment. The second most abundant taxon, a Brevinema sp., has versatile metabolic potential, including genes involved in host mucin degradation and utilization. However, during periods of absent gut content, a Ruminococcaceae bacterium becomes dominant, possibly outgrowing all other bacteria through the production of secondary metabolites involved in quorum sensing and cross-inhibition while benefiting the host through short-chain fatty acid production. Whereas Mycoplasma is often present as a symbiont in farmed salmonids, we show that the Ruminococcaceae species is also detected in wild Arctic char, suggesting a close evolutionary relationship between the host and this symbiotic bacterium.


Introduction
Symbiotic micr oor ganisms ar e essential for the health and wellbeing of all animals (Douglas 2018 ).In farm animals, microbiomes have been associated with different health c har acteristics whic h impact animal welfare and productivity (Jin Song et al. 2019, Chen et al. 2021, Wessels 2022 ).In aquaculture, fish are constantly exposed to high loads of micr oor ganisms in the rearing water and need to maintain a healthy pr otectiv e barrier to pr e v ent infection and disease (Sundh et al. 2010, Langlois et al. 2021 ).Commensal micr oor ganisms inhabiting fish m ucosal surfaces, suc h as the skin, gills, and gut, can aid in this function by outcompeting opportunistic pathogens for nutrients or activ el y pr e v enting their colonization through the production of antimicrobial compounds (Tarnecki et al. 2017, Perry et al. 2020 ).Apart from their role in disease prevention, the fish gut microbiome assists in nutrient uptake by breaking down complex carbohydrates and proteins in the gut, or through the production of vitamins and other essential nutrients (Clements et al. 2014, Yukgehnaish et al. 2020 ).T his , in turn, leads to more efficient feed utilization and impr ov ed gr owth.Despite the importance of microbiomes in maintaining the health and productivity of farmed fish, knowledge of their community composition and function across the div erse r ange of host species curr entl y being farmed r emains limited.
Within the group of salmonids, which includes se v er al highv alue aquacultur e species, the gut micr obiome of Atlantic salmon ( Salmo salar ) has so far r eceiv ed most attention (Rudi et al. 2018, Dv er gedal et al. 2020 ).Pr e vious studies hav e shown that the Atlantic salmon gut contains few autochthonous, or resident, gut micr oor ganisms (Gajardo et al. 2016, Karlsen et al. 2022 ), with Mycoplasma being one of onl y fe w r ecurring gut commensal (Lle well yn et al. 2015, Dehler et al. 2017, Fogarty et al. 2019, Rasmussen et al. 2021 ).Compared to Atlantic salmon, the microbiome of Arctic char ( Salvelinus alpinus ), a cold-water fish species, is less well-studied.First studies, using electr on micr oscopy and a cultiv ation-based a ppr oac h, demonstr ated substantial numbers of bacteria inhabiting the gut of Arctic char (Ringø et al. 2001(Ringø et al. , 2006 ) ). Research on wild Arctic char using high-throughput sequencing technology has shown a heterogeneity of the gut microbial diversity across geography, season and habitat (Hamilton et al. 2019, Element et al. 2020, 2021 ).Similar to Atlantic salmon, Mycoplasma was also among the dominant gut symbionts.In an aquacultur e envir onment, the gut micr obiota of Arctic c har has been studied to determine the impact of feeds and probiotics on the micr obial comm unity (Nyman et al. 2017, Knobloc h et al. 2022 ).Howe v er, it is not yet known which bacteria constitute the resident gut microbiota in farmed Arctic char or what role they might play in maintaining health, well-being and productivity of the fish.Understanding these inter actions, particularl y during the earl y life sta ges, could be v aluable for c har acterizing a healthy Arctic char gut micr obiome, form ulating pr ecision diets, or modulating less pr oductiv e gut micr obiota thr ough the selection and tr ansplantation of probiotic strains.
The objective of this study was to describe the microbiome of farmed juvenile Arctic char, identifying the stable members of the resident gut microbiota and analysing their putative role in the fish gut microbiome through 16S rRNA gene amplicon sequencing, fluorescence in situ hybridization (FISH) and metagenomic analysis.

Sample collection
Juv enile Arctic c har ( S. alpinus ) wer e collected fr om an ongoing feeding trial (Knobloch et al. 2022 ) at 104 days posthatch (dph) (T0, N = 22), 132 dph (T1, N = 15), and 157 dph (T2, N = 15).Before the beginning of the feeding trial (T0), one fish was collected fr om eac h of the 22 experimental tanks.At T1 and T2, five fish were collected from each of the three replicate control tanks.All fish had r eceiv ed the same control diet, consisting of fish meal, fish oil, gelatinized wheat, minerals, and vitamins as described in Knobloch et al. ( 2022 ), and were reared under identical conditions with a continuous fr eshwater exc hange and a water temper atur e of 8.6 ± 0.5 • C. The fish were fasted 12 h prior to weighing and sample collection.Fish were then euthanized with 500 ppm of phenoxyethanol and transported to the laboratory in sterile plastic bags on ice.Skin samples were collected by scraping along the lateral line using a sterile scalpel.The fish were then rinsed with 70% ethanol follo w ed b y sterile laboratory grade water to r emov e loosel y attac hed bacteria befor e dissecting and r emoving the mid-and hind-gut section.For histology, ∼5 mm sections were r emov ed fr om the hind-gut and fixed for 24 h in fr eshl y pr epar ed 4% paraformaldehyde solution in 1x PBS at 4 • C before being transferred to 70% ethanol for long-term storage.At time T0, T1, and T2, 1 l of tank water was collected and filtered on 0.2 μm cellulose filters (Advantec).In total, 11 feed samples were collected in sterile containers over the study period.Skin, gut, water filter, and feed samples were frozen at −80 • C until DNA extraction and sequencing.The experiment was performed according to European and Icelandic guidelines under the licence UST201707 from the Icelandic Environment Agency and FE-1134 from the Icelandic Food and Veterinarian Authority.
To compare the gut microbial community in relation to gut content 24 additional fish from time point T2, which had been collected and frozen at −80 • C at the end of the experiment, were dissected as mentioned abo ve .T he state of digestive content was described as "full" if digesta was present along the mid-and hindgut section, as "partially full" if digesta was present in parts of the mid-or hind-gut section, and as "empty" if no digesta was observable.
To compare the farmed Arctic char gut microbiota to those of wild specimens, 35 additional fish, ranging in weight from 1.8 to 15.5 g and collected from a fresh water spring in the south of Iceland (Kreiling et al. 2021 ), were dissected and processed as mentioned abo ve .

DN A extr action, PCR, and sequencing
DNA from the whole mid-and hind-gut with digesta, if present, was extracted as previously described in Leeper et al. ( 2022 ).In brief, defr osted guts wer e homogenized by bead-beating and then processed with the QIAamp Po w erFecal Pro DN A Kit (Qiagen).Tw o negativ e extr action contr ols wer e run alongside these samples.DN A from skin, w ater filter, and feed samples wer e extr acted using the MasterPure Complete DNA & RNA Purification Kit (Epicentre) following the manufacturer's instructions for DNA extraction.PCR was performed on all samples as described in Knobloch et al. ( 2021 ) using the universal prokaryotic primer pair S-D-Bact-0341b-S-17 (5 -CCTA CGGGNGGCWGCA G-3 ) and S-D-Bact-0785-a-A-21 (5 -GA CTA CHV GGGTATCTAATCC-3 ) (Klindworth et al. 2013 ) and high-fidelity Q5 pol ymer ase (Ne w England Biolabs).All samples, including the negative PCR products of the extraction controls, w ere bar coded with Nextera XT v2 indices (Illumina), normalized using Sequel-Prep Normalisation Plates (Thermo Fisher Scientific), and sequenced on a MiSeq desktop sequencer (Illumina) with v3 chemistry to generate 300 bp long paired-end reads.
DNA of three selected samples with a high r elativ e abundance of the three dominant gut symbionts were subjected to shotgun metagenomic sequencing.In short, bacterial DN A w as enriched using the NEBNext Microbiome DNA Enrichment Kit (New England Biolabs) follo w ed b y libr ary pr epar ation using the Nexter a Flex kit (Illumina) according to the manufacturer's instructions.Libr aries wer e pooled and sequenced on a MiSeq sequencer as mentioned abo ve .Sequencing gener ated 2.88 Gbp r aw data acr oss 9 781 473 paired-end reads.

Inference of 16S rRNA ASVs and microbial community analysis
Raw 16S rRNA r eads wer e filter ed, trimmed, and processed into amplicon sequence variants (ASVs) with D AD A2 v. 1.12.1 (Callahan et al. 2016 ) implemented in R (R Core Team 2020 ).In short, primer sequences were removed and the forw ar d and r e v erse reads trimmed after 260 bp and 240 bp, r espectiv el y.The settings maxEE and truncQ were both set to 2. After learning and filtering errors with default settings, forw ar d and r e v erse r eads wer e mer ged, sequences outside of the target amplicon r ange r emov ed and c himer as detected with the function removeBimeraDenovo .Taxonomic assignment was performed with the function as-signTaxonom y a gainst a tr aining set of the Silv a SSU database v ersion 138 (Quast et al. 2013 ).ASVs assigned to the kingdom Eukaryota , order Chloroplast or family Mitochondria , as well as 31 ASVs detected pr edominantl y in the negativ e contr ol samples, wer e r emov ed.Micr obial comm unity anal ysis, including statistical anal ysis and plotting community composition, alpha diversity and beta diversity, was performed with R packages phyloseq version 1.42.0 (McMurdie and Holmes 2013 ) and vegan version 2.6-4 (Oksanen et al. 2013 ).P airwise m ultile v el comparison was conducted with the R pac ka ge pairwiseAdonis v ersion 0.4.1 ( https://github.com/pmartinezarbizu/pairwiseAdonis ).A phylogenetic tree of all ASVs for calculating weighted UniFrac distances was created with DE-CIPHER (Wright 2016 ) and FastTree 2 (Price et al. 2010 ).Differential abundance analysis was performed with DeSeq2 version 1.38.1 (Love et al. 2014 ).The Venn dia gr am was produced with the R pac ka ge ampvis2 version 2.7.35 (Andersen et al. 2018 ) with an abundance cutoff of 0.11% and a frequency cutoff of 1%.

MAG binning, functional genome analysis, and phylogenetics
To construct the draft genomes of the three dominant fish gut symbionts, meta genomic r aw r eads wer e quality filter ed with Trimmomatic (Bolger et al. 2014 ) with settings LEADING:3 TRAIL-ING:3 SLIDINGWINDOW:4:15 MINLEN:100, leading to the r emov al of 17.59% of the raw data, and coassembled using Megahit version 1.1.3(Li et al. 2015 ).Quality filtered reads of each sample were then mapped back to the coassembled contigs using Bowtie 2 (Langmead and Salzberg 2012 ).Binning of metagenomeassembled genomes (MAGs) was performed in Anvi'o version 7 (Eren et al. 2015 ) following the "Anvi'o User Tutorial for Meta genomic Workflow" ( https://mer enlab.org/2016/06/22/anviotutorial-v2/).In brief, k-mer frequencies were calculated and open r eading fr ames (ORFs) identified in the contigs using the anvi-gencontigs-database command.HMM (hidden Markov model) profiles wer e gener ated using the command anvi-run-hmms and genes annotated with the command an vi-run-ncbi-cogs .Taxonomic annotation was performed with centrifuge (Kim et al. 2016 ).Anvi'o profiles were merged and imported into the anvi'o inter activ e interface.MAGs were then manually binned based on tetr anucleotide fr equency, taxonomic assignment, and cov er a ge.Genome completeness and contamination of the three resulting MAGs were calculated with Chec kM v ersion 1.1.0(P arks et al. 2015 ) using a lineage-specific w orkflo w.Av er a ge genome cover a ge for each MAG w as calculated, b y mapping the quality filter ed r eads to the MAG using Bowtie 2 and calculating cov er a ge with the samtools depth function (Li et al. 2009 ).Closest cultiv ated r elativ es based on near full-length 16S rRNA genes of each MAG were determined with EzBioCLoud (Yoon et al. 2017 ).Aver a ge amino acid identity (AAI) was calculated using the tool Genome Matrix (Rodriguez-R and Konstantinidis 2016 ) on the web serv er http://env e-omics.ce.gatec h.edu/g-matrix .Closel y r elated genomes for comparison were chosen based on 16S rRNA gene similarity, as mentioned abo ve .ORFs for each MAG were called with prodigal version 2.6.3 (Hyatt et al. 2010 ).Genome statistics were determined with QUAST (Gur e vic h et al. 2013 ) and RAST (Aziz et al. 2008 ).Clusters of orthologous groups of proteins (COGs) wer e pr edicted using RPS-BLAST + a gainst the 2014 r elease of the COG database (Galperin et al. 2015 ).COG category and functional descriptions wer e inferr ed using the cdd2cog.plscript (Leimbac h 2016 ).Assigning KEGG Orthology (KOs) was performed using the BlastKOALA web service (Kanehisa andGoto 2000 , Kanehisa et al. 2016 ).For those genes not found in specific KEGG pathwa ys , a tBLASTn search of the corresponding candidate KO proteins against the MAGs was performed to verify their absence.Secondary metabolite gene clusters were searched with the web application of antiSMASH version 7 (Blin et al. 2017 ) with default settings .Genes in volv ed in short c hain fatty acid (SCFA) pr oduction w ere sear ched through the gutSMASH web server (Pascal Andreu et al. 2021 ).Carbohydr ate-activ e enzymes wer e pr edicted with the dbCAN3 web server (Zheng et al. 2023 ) with all detection tools selected.
Phylogenetic trees based on the 16S rRNA gene sequence of each MAG was constructed in ARB (Ludwig et al. 2004 ).In brief, sequences were aligned against the global SILVA SSU alignment with the SIN A w eb-tool (Pruesse et al. 2012 ) and merged with the SILVA SSU database version 138.1 (Quast et al. 2013 ) in ARB.Maxim um-likelihood tr ees wer e calculated with PhyML (Guindon and Gascuel 2003 ) for each MAG including closely related sequence.

Histology and FISH
Fixed gut sections were dehydrated in successive baths of ethanol and xylene, and then embedded in paraffin.Sections of 5 μm thic kness wer e cut on a CM1800 microtome (Leica) with MX35 Ultr a micr otome blades (Thermo Scientific), follo w ed b y deparaffinizing in xylene and a washing step in 100% ethanol.FISH of probes to the bacterial 16S rRNA subunit was conducted as pr e viousl y described in Knobloch et al. ( 2019 ) with Cy3-labelled pr obes tar geting the Mycoplasma sp.(equal mixtur e of pr obes MYC542 and MYC629), the unclassified Ruminococcaceae (probe RUM1447) and the Brevinema sp.(probe BRV1455), as well as the pr e viousl y described Alexa488-labelled universal bacterial probe EUB338 (Amann et al. 1990 ) ( Table S3 , Supporting Information ).The hybridization buffer contained 40% formamide.Hybridized sections were stained with Fluoroshield antifade containing DAPI (Sigma) and visualized with a model BX51 epifluorescence micr oscope (Ol ympus).Epifluor escence ima ges wer e pr ocessed with daime v. 2.2 (Daims et al. 2006 ).

Temporal succession of the gut microbiota and dominance of three bacterial taxa
To describe the gut microbial community composition of juvenile farmed Arctic char, 16S rRNA gene sequence amplicons were analysed at three different time points over the course of 8 weeks from 52 farmed fish with an av er a ge weight of 2.2 ± 0.2 g at time T0, 4.5 ± 1.1 g at time T1, and 9.8 ± 0.7 g at time T2.The gut microbiome was dominated by the genus Mycoplasma , comprising 62.8% of the av er a ge r elativ e abundance acr oss all a ge gr oups (Fig. 1 A).Within Mycoplasma , a single ASV made up 97.1% of the r elativ e abundance, with most other ASVs having only a single nucleotide difference (data not shown).Though not detected in all samples, the second and third most abundant taxa were Brevinema and an unclassified Ruminococcaceae , both also dominated by a single ASV, and accounting for 12.3% and 6.3% of the av er a ge r elativ e abundance, r espectiv el y.All other gener a made up less than 20% of the av er a ge r elativ e abundance.
There was considerable variability in the microbial community composition both between a ge gr oups and between individuals, with the Mycoplasma -associated ASV being the only ASV shared between all samples (Fig. 1 B).The presence of the Brevinema sp. and unclassified Ruminococcaceae a ppear ed to incr ease ov er time, with a higher percentage of fish containing either one of the species at T2 (87%) than at T0 (41%) (Fig. 1 B).Other taxa, such as an unknown Bacillaceae and Bacillus sp.decreased markedly o ver time .A significant difference in the microbial communities between a ge gr oups was detected based on Bray-Curtis dissimilarities (PERMANOVA, F(2, 49) = [3.4148],P = .001)(Fig. 1 C) and w eighted UniF rac distances (PERMANOVA, F(2, 49) = [2.1006],P = .047)( Fig. S1 , Supporting Information ).The number of observed ASVs ranged from 5 to 129 with an average of 42 ASVs across all a ge gr oups .T her e wer e significant differ ences between the number of observed ASVs (ANOVA, F(2, 49) = [4.189],P = .0209)and Shannon diversity (ANOVA, F(2, 49) = [7.534],P = .0014)between time points, with the Shannon diversity significantly decreasing from T0 to T1 and to T2 (Fig. 1 D).
Inter estingl y, Mycoplasma , Brevinema , and the unclassified Ruminococcaceae were also present on the skin of the fish, together accounting for 41.8% of the r elativ e abundance, wher eas they wer e only detected at a low r elativ e abundance of 0.7% in the tank water samples (Fig. 1 A).

Absence of gut content strongly influences the gut microbial composition
Due to the large observed interindividual variability in gut microbial composition, the microbial community was examined in relationship to gut content filling on 24 additional fish from time point T2 of which eight fish were characterized as having empty mid-and hind-guts, nine as having partially filled guts and se v en as having full mid-and hind-guts (Fig. 2 A).The gut microbial community of fish with empty guts was dominated by the unclassified Ruminococcaceae , which was significantly more abundant in these samples than in samples of fish with partially full or full guts (DE-Seq2, p adj < 0.001) (Fig. 2 B).In ad dition, the n umber of observed ASVs (ANOVA, F(2, 21) = [19.11],P < .0001)and Shannon diversity (ANOVA, F(2, 21) = [14.84],P < .0001)was significantl y differ ent between the state of gut filling, with fish with empty guts having sig-nificantly lo w er v alues compar ed to fish with partiall y full or full guts (Fig. 2 D).This discrepancy was also highlighted by a significant difference in the microbial community composition within the gut content and feed groups based on Bray-Curtis dissimilarities (PERMANOVA, F(3, 31) = [26.736],P = .001),with pairwise comparisons showing significant differences between empty and both full and partially filled guts ( p adj = 0.006 each) and between the feed samples and all three gut groups ( p adj = 0.006 each), but not between the full and partially filled guts ( p adj = 1) (Fig. 2 E).
To e v aluate the contribution of micr oor ganisms in the fish feed to the gut micr obial comm unity, empty guts, full guts, and feed samples wer e compar ed to eac h other (Fig. 2 C).This sho w ed that only one ASV, a member of the Clostridiaceae , was detected in all sample types above a relative abundance of 0.1%.The three dominant ASVs in the gut, Mycoplasma , Brevinema , and the unclassified Ruminococcaceae , as well as a member of the genus Rhodococcus wer e shar ed between the empty and full guts and contributed to over half of the relative abundance of the combined communities, but were not detected in the feed above the selected thresh- old.The feed and full guts shared 103 ASVs which contributed to 27.6% of the ov er all r elativ e abundance between samples.Separ atel y, the full gut and feed harboured 242 and 248 ASVs not shared with the other sample types, contributing to 4.6% and 14.3% of the r elativ e abundance, r espectiv el y.

Spatial distribution of gut symbionts
Micr oscopic observ ation of the gut micr obiota at all time points using 16S rRNA FISH showed colonization of the gut epithelia by Mycoplasma sp. and the unclassified Ruminococcaceae .The Brevinema sp. was not detected using the selected pr obes, possibl y due to site inaccessibility of the targeted 16S rRNA region or a lack of adhesion of the bacteria to the gut epithelium.The Mycoplasma sp. a ppear ed slightl y elongated with a length of ∼0.6 μm and was distributed as single cells or in clusters of up to 8 μm thickness on the outer layer of the intestinal mucosa (Fig. 3 A).Howe v er, the epithelium was not cov er ed uniforml y with Mycoplasma with man y areas being void of bacteria.The unclassified Ruminococcaceae was spherical-shaped and ∼1 μm in diameter (Fig. 3 B).It formed dense clusters up to 10 μm thickness and was found pr edominantl y in fish with empty guts.16S rRNA FISH confirmed that Mycoplasma and the unclassified Ruminococcaceae were the dominant bacteria in the gut of the sampled Arctic c har, with fe w other bacterial morphotypes detected with the universal bacterial probes.

Phylogeny of the dominant gut symbionts and comparison to wild Arctic char
To perform phylogenetics and understand the putative function of the three dominant gut symbionts, DNA samples enriched in Mycoplasma sp., Brevinema sp., and Ruminococcaceae were selected for meta genomic anal ysis.Assembl y and MAG binning of the meta genomic datasets led to the r ecov ery of three medium-to highquality MAGs [defined according to Bo w ers et al. ( 2017)] for the Mycoplasma sp., Brevinema sp., and unclassified Ruminococcaceae , designated A C_MYC01, A C_BRV01, and A C_RUM01, r espectiv el y (Table 1 ).
The closest cultivated relatives to the A C_MYC01, A C_BRV01, and AC_RUM01 based on near full-length 16S rRNA gene sequence similarity, were Mycoplasma moatsii (93.45% sequence identity), Brevinema andersonii (90.97%), and Paludicola psychrotolerans (89.81%), r espectiv el y.Comparison of av er a ge AAI to members of the r espectiv e or closel y r elated gener a sho w ed less than 56% to each of the MAGs ( Table S1 , Supporting Information ), below the suggested threshold of 65% for species of shared genera  (Konstantinidis and Tiedje 2007 ), indicating that they likely fall within novel genera.Ho w ever, for this stud y, the y will continue to be named according to the taxonomic classification of their r espectiv e ASVs.Alignment against the nonredundant Silva 138 SSU database and phylogenetic analysis placed the AC_MYC01 and AC_BRV01 separ atel y into clades with three other uncul-tiv ated bacteria pr e viousl y detected in the gut of fish species and within the order Mycoplasmatales and Brevinematales , respectiv el y (Fig. 4 A and B).AC_RUM01 occupied a separate branch to se v er al uncultur ed bacteria in the order Oscillospirales pr e viously detected in human, animal and environmental samples (Fig. 4 C).To determine if the dominant bacteria detected in farmed Arctic char might be obligate symbionts essential for animal health or gut function, a comparison was made to the gut microbiota of 35 wild juvenile Arctic char.This sho w ed that only the ASV corresponding to the Ruminococcaceae bacterium was shared (100% sequence similarity) between the wild and farmed fish ( Fig. S2 , Supporting Information ).In total, 29 of the wild fish harboured this specific Ruminococcaceae which contributed to an average relative abundance of 49.4% of the gut microbiota.The second and third most abundant ASVs belonged to the genus Deefgea and Propionibacterium with 7.6% and 3.8% of the av er a ge r elativ e abundance, r espectiv el y.Neither Mycoplasma nor Brevinema were detected in the wild fish gut microbiota.

Functional attributes of the dominant gut symbionts
An ov ervie w of the genome c har acteristics of AC_MYC01, A C_BRV01, and A C_RUM01 are shown in Table 1 .AC_MYC01 had the smallest genome size with 0.85 Mbp, follo w ed b y AC_RUM01 with 1.30 Mbp and AC_BRV01 with 1.50 Mbp.This corresponded with the number of detected ORFs, being 874, 1262, and 1397 for A C_MYC01, A C_RUM01, and A C_BRV01, r espectiv el y.The GC content ranged from 26.2% for AC_MYC01 to 29.2% for AC_BRV01.
The genomes of the three dominant bacteria differed in metabolic potential and cellular function with AC_MYC01 having, for instance, compar ativ el y fe wer COGs in categories "Ener gy pr oduction and conversion" (COG category C), "Amino acid transport and metabolism" (E) and "Cell wall/membr ane/env elop biogenesis" (M), AC_BRV01 having more COGs in category "Cell motility" (N) and AC_RUM01 having fewer COGs in category "Carbohydrate tr ansport and metabolism" (G) compar ed to their genome sizes (Fig. 5 B).
Hypothesized metabolic pathways of the three bacteria are presented in Fig. 5 (A) with a complete list of KOs (KEGG Orthology groups) in Table S2 ( Supporting Information ).In AC_MYC01 the centr al ener gy pr oduction and carbohydr ate metabolism was gl ycolysis via the Embden-Meyerhof-Parnas (EMP) pathway.It further contained phosphotr ansfer ase systems (PTS) for glucose, fructose, and l -ascorbate transport, the former two likely being degraded by the EMP pathway to pyruv ate.Pyruv ate could be further metabolized to lactate by lactate dehydrogenase, but genes wer e lac king to metabolize pyruvate to acetyl-CoA.l -ascorbate could be degraded to d -xylulose-5P, an intermediate in the pentose phosphate pathway, with the UlaG enzyme substituted with a lactonase Ms0025 as described for Mycoplasma synoviae (Korczynska et al. 2014 ).Similar to other Mycoplasma spp .,A C_MYC01 did not have complete pathways for amino acid biosynthesis (Himmelr eic h et al. 1996, Arr aes et al. 2007, Santos et al. 2011 ).Howe v er, it contained a complete pyrimidine ribonucleotide biosynthesis pathway and both nucleotide sugar biosynthesis pathways for glucose to UDP-glucose and galactose to UDP-galactose were present, as well as the interconversion of UDP-glucose to UDPgalactose through UDP-glucose 4-epimerase .T he genome further contained all genes involved in F-type ATPase which could be involved in gliding motility on host cells as pr e viousl y described for other Mycoplasma (Tulum et al. 2020 ).AC_MYC01 only contained genes coding for fiv e carbohydr ate activ e enzyme (C AZy) families , namel y Carbohydr ate Ester ase Famil y 9 (CE9), Gl ycoside Hydr olase Family 170 (GH170), and Glycosyl Transferase Family 2, 4, and 58 (GT2, GT4, and GT58) (Fig. 5 D).
In AC_BRV01, gl ycol ysis was also the centr al ener gy and carbohydrate metabolism.It contained various sugar transporters, in-cluding PTS for glucose , maltose , fructose , mannose , cellobiose or diacetylchitobiose , alpha-glucoside , N -acetylgalactosamine , and N -acetylglucosamine .T he latter two substrates being present in large amounts in mucin (Liu et al. 2021 ) and hence could point to w ar ds m ucin utilization and degr adation.This is further highlighted by the presence of several genes coding for gl ycoside hydr olase, including GH2 β-galactosidases that can cleav e linka ges of Gal-β1,3-GalNAc and GH29 fucosidases that can cleave fucose linked to mucin O -glycans (Raba and Luis 2023 ).AC_BRV01 also contained genes for converting mannose and N -acetylglucosamine to d -fructose-6-phosphate via mannose-6-phosphate isomerase and N -acetylglucosamine-6-phosphate deacetylase, r espectiv el y, as well as for maltose and cellobiose to d -glucose-6-phosphate via maltose-6 -phosphate and 6-phosphobeta-glucosidase, r espectiv el y.Hence, AC_BRV01 could use se v er al substrates for energy production.AC_BRV01 could further conv ert pyruv ate to acetyl-CoA and to acetate via phosphate acetyltr ansfer ase and acetate kinase.Similar to the Mycoplasma sp., AC_BRV01 did not have any complete known pathways for amino acid biosynthesis, instead containing an ABC transporter for arginine , lysine , and histidine .It also contained most genes in the initiation pathway of fatty acid biosynthesis and ABC transporters for phospholipids and lipopol ysacc harides.AC_BRV01 had se v er al genes r equir ed for fla gella assembl y and c hemotaxis, whic h could confer it mobility.
The unclassified Ruminococcaceae AC_RUM01 contained full pathw ays for p yruvate oxidation, phosphoribosyl diphosphate (PRPP) biosynthesis, nucleotide sugar biosynthesis, and UDP-Nacetyl-d -glucosamine biosynthesis.Acetyl-CoA fr om pyruv ate oxidation could be further converted to acetate, via the phosphate acetyltr ansfer ase-acetate kinase pathwa y.AC_R UM01 contained all genes involved in fatty acid biosynthesis, as well as multiple genes in the porA pathway and pyruvate to acetate-formate gene clusters ( Fig. S3 , Supporting Information ) involved in the production of short-chain fatty acids (Amador-Noguez et al. 2010, Guo et al. 2019 ).The genome further contained all genes for lysine biosynthesis and the Shikimate pathway, producing chorismite, a precursor for aromatic compounds .AC_R UM01 had genes coding for se v er al ABC tr ansporters and PTS including those for the amino acids ar ginine, l ysine, and histidine.It also contained a PTS for N -acetylglucosamine.and a gene coding for Glycoside Hydr olase Famil y 84 (GH84) β-N -acetylglucosaminidases potentiall y clea ving GlcNAc , indicating the utilization of by-pr oducts fr om pr e viousl y degr aded m ucin (Raba and Luis 2023 ).Compar ed to the other two genomes , AC_R UM01 contained secondary metabolite biosynthetic gene clusters .T hese were predicted to produce a cyclic-lactone-autoinducer, a ranthipeptide, and an unspecified ribosomally synthesized and post-tr anslationall y modified peptide product (RIPP), involved in quorum sensing and crossinhibition (Sturme et al. 2002, Chen et al. 2020 ) ( Fig. S4 , Supporting Information ).

Discussion
Using a longitudinal a ppr oac h, it is shown that the gut microbiota of farmed Arctic char changes within the first weeks post hatch, similar to pr e vious r eports on the succession of the gut microbiota in fish (Bledsoe et al. 2016, Keating et al. 2021 ).Due to the r elativ el y short period of investigation, it was not possible to conclude when or if the gut microbial community had reached a stable state.Ho w e v er, Mycoplasma r emained the dominant taxon throughout the study period, follo w ed b y Brevinema and Ruminococcaceae , the latter two becoming more abundant as age progressed.This com- m unity pr ofile differ ed, but also sho w ed similarities, to those of pr e viousl y described wild and farmed Arctic c har, whic h can vary widely depending on geographic location, habitat and diet (Nyman et al. 2017, Hamilton et al. 2019, Element et al. 2020, 2021 ).Nyman et al. ( 2017 ) analysed the gut microbiota of on-growing Arctic char fed experimental and control diets, showing that the dominant taxa were Photobacterium and Leuconostocaceae , with a mean r elativ e abundance of 14.2% and 13.6%, r espectiv el y.Mycoplasma , Brevinema , and Ruminococcaceae were absent from the gut comm unity.In contr ast, Element et al. ( 2020) also detected Mycoplasma and Brevinema as two dominant taxa in the gut of wild Arctic c har, with av er a ge r elativ e abundances v arying, depending on the habitat of the fish.Phylogenetic analysis and a comparison to wild fish from Icelandic waters , pro vide further evidence that the three dominant gut taxa in this study are fish-associated bacteria, while highlighting likely species or str ain-le v el differences between hosts .T he Ruminococcaceae strain, in particular, could hav e coe volv ed with wild Arctic c har in Iceland and confer certain benefits to its host, as pr e viousl y suggested for other fish symbionts (Kim et al. 2021, Rasmussen et al. 2023 ).Detection of the three dominant strains on the fish skin in high relative abundances could be explained by a transfer of the str ains fr om faecal matter in the tank water to the skin and their affinity to adhered to a surface as biofilms.Ho w e v er, a low abundance of the taxa in the tank water and higher abundance on the skin could also point to w ar ds the bacteria having mor e than one nic he or modes of transmission.
The large inter-individual variability of the gut microbiota detected in the present study has also been described for gut microbiota of other fish species (Gatesoupe et al. 2016, Knobloch et al. 2021 ) and likely reflects the presence of only few stable taxa and the detection of otherwise alloc hthonous, or tr ansient, bacteria.A comparison to the feed samples underpins this hypothesis, as a part fr om the thr ee dominant taxa, onl y one bacterium, a Rhodococcus , was shared between full and empty guts while not being detected in the feed samples.A study by Karlsen et al. ( 2022 ) sho w ed that the autochthonous gut microbiota of Atlantic salmon was m uc h less div erse if excluding feed-associated bacteria.In their study, both Mycoplasma and Ruminococcaceae were the dominant digesta-specific taxa in the gut.Ho w e v er, it a ppear ed that a dominance of Mycoplasma precluded a high r elativ e abundance of Ruminococcaceae and vice versa.This supports our findings that Mycoplasma and Ruminococcaceae alternate in high r elativ e abundance due to environmental changes in the gut environment.In the present study, the dominance of Ruminococcaceae was clearly linked to an empty mid-and hind-gut and points to w ar d a shift in comm unity structur e de pending on n utrient availability.In their study on wild Arctic char, Element et al. ( 2020 ) found that Mycoplasma and Brevinema both had a lo w er r elativ e abundance in ov erwintering fish compar ed to those in the other seasons .T his suggests that other taxa, possibly with similar functions to the unclassified Ruminococcaceae , take over during such period of possible prolonged fasting in the wild.
16S rRNA FISH analysis showed that both Mycoplasma and Ruminococcaceae build thick clusters of cells in the mucus layer on the intestinal epithelium, which could facilitate nutrient exchange, but also enable r a pid r emov al once the mucus is shed (Ringø et al. 2007 ).A similar spatial occurrence of bacterial cells on the epithelial surface was observed in the distal gut of juvenile rainbow trout in which Mycoplasma was also the dominant taxon (Rasmussen et al. 2021 ).In addition, a study by Cheaib et al. ( 2021 ), employing 16S rRN A FISH, sho w ed that Mycoplasma also inhabited areas further up the digestive tract in Atlantic salmon, including the stomach lining and pyloric caecum.Whereas not studied in the present work, these sites could be potential reservoirs leading to the r a pid recolonization of Mycoplasma in the mid-and hind-gut when digesta passes through the intestine.
Genome analysis of the three dominant gut symbionts sho w ed that the gut microbiome of farmed Arctic char could provide several benefits to its host.Similar to other Mycoplasma , the genome of AC_MYC01 was small with few specialized metabolic pathwa ys .This reduction of genes has pr e viousl y been associated with a high adaptation to a host-associated lifestyle (Cheaib et al. 2021 ), thus not necessitating endogenous biosynthesis in an environment c har acterized by high nutrient av ailability.The lac k of pathwa ys in volved in the biosynthesis of amino acids and vitamins, both pr e viousl y detected in salmonid-r elated Mycoplasma (Rasmussen et al. 2021(Rasmussen et al. , 2023 ) ), further suggests that these are not the main reasons for the symbiotic relationship between Mycoplasma and salmonids.Instead, its small genome size, adapted to the host gut envir onment, pair ed with the ability to r a pidl y r ecolonize the gut after less fa vorable conditions , could be the main feature enabling this symbiosis.
The Brevinema sp. had similar gene functions to Mycoplasma , albeit a higher number of sugar and other nutrient transporters.
This along with potential mobility could confer it a spatial niche a part fr om Mycoplasma in the m ucus lining, while enabling cr ossfeeding and efficient scavenging for available nutrients.Brevinema is often detected as a salmonid-associated gut symbiont (Brown et al. 2019, Gupta et al. 2019, Li et al. 2021 ), but its functional attributes have so far remained elusive .T he draft genome of AC_BRV01 highlights the potential role of Brevinema as a welladapted nutrient scavenger and cohabitant of the salmonid gut along with Mycoplasma .
Members of the class Clostridia , including Ruminococcaceae , are often involved in SCFA production in animal intestines (Barcenilla et al. 2000, Zhang et al. 2018, Lan et al. 2023 ).The genome of AC_RUM01 contained genes involved in SCFA production, which could be beneficial to the host (Li et al. 2022 ).Although the guts wer e c har acterized as empty, r emnants of the digesta needed for SCFA pr oduction wer e likel y still pr esent due to continuous feeding, and hence r a pid turnov er of nutrients in the gut.In addition, its genome contained genes involved in mucin degradation and utilization of by-products from previously degraded mucin.This could be an additional energy source and confer it a growth adv anta ge when other nutrients become limited once the digesta has left the mid-and hind-gut.Interspecies cross-feeding of gut micr oor ganisms has pr e viousl y been described in other animals (Solden et al. 2018, Luo et al. 2021 ) and could aid in creating a stable micr obial comm unity composition, making it less susceptible to invasion or perturbation from external sources (Culp and Goodman 2023 ).Furthermor e, degr adation and feeding on m ucin is a factor for maintaining a pr otectiv e barrier and contributing to intestinal homeostasis, thereby contributing to host health (Paone and Cani 2020 ).Further, AC_RUM01 contained full pathways for the biosynthesis of amino acids, possibly enabling its r a pid gr owth e v en in nutrient limited conditions .T he takeo ver of the gut microbiome by Ruminococcaceae when the gut is temporarily empty could be coordinated be the production of secondary metabolites enabling r a pid pr olifer ation and suppr ession of other species when conditions are suitable (Uhlig and Hyland 2022 ).Such clear alternation in r elativ e abundance between dominant gut symbionts has not yet been shown for fish and points to w ar d a strategy that could pr e v ent the colonization of harmful bacteria in a fluctuating and dynamic environment.

Conclusion
T his study pro vides an in-depth o v ervie w of the gut microbiota in farmed Arctic char and the functional attributes of its dominant resident symbionts .T hese insights are also relevant for other salmonid species due to the frequent presence of these bacterial taxa among salmonid gut microbiota.We show that there is a clear alternation in the r elativ e abundance of the main symbionts depending on passage of the digesta with different functional c har acteristics possibl y ada pted to either a state of high or low nutrient a vailability.T his presents a previously undiscover ed str ategy to maintain a high bacterial abundance in the gut during fluctuating environmental conditions and thereby prevent colonization of micr oor ganisms that could be harmful to the host.The presence of the same sequence type of an unclassified Ruminococcaceae in both wild and farmed Arctic char further raises the questions of how essential this bacterium is for host health or disease resistance.Further research is also needed to better understand the interactions between the cohabitating Mycoplasma and Brevinema , as well as their impact on host health and productivity.Tar geted gr owth studies and pathogen challenge experiments between fish with and without each symbiont will provide further insights into these questions, as well as the into the potential of targeted gut microbiome modulation for improved aquaculture performance of salmonid species.

Figure 2 .
Figure 2. Micr obial comm unity composition of Arctic c har guts with differ ent amounts of gut content.(A) Example of gut sample c har acterized as empty, partially filled, and full.(B) Mean r elativ e abundance of gener a detected in gut samples c har acterized as empty ( n = 8), partiall y filled ( n = 9), and full ( n = 7) at time T2; unclassified and low abundant taxa are summarized under "Unknown/Others".(C) Venn dia gr am of shared taxa between empty guts, full guts, and feed (partially filled gut group excluded); values indicate the number of shared and not shar ed ASVs; percenta ges in br ac kets indicate percentage of relative abundance.(D) Alpha diversity metrices of gut microbial communities between empty, partially filled, and full guts; significance based on one-way ANOVA and Tuk e y's post hoc test ( * * : p adj < 0.01; * * * : p adj < 0.001; NS: nonsignificant).(E) NMDS plot of Bray-Curtis dissimilarities between microbial communities in empty, partially filled, and full guts, as well as feed samples (stress = 0.08).

Figure 3 .
Figure 3. FISH images of hind-gut sections with Mycoplasma-specific probes (A) and Ruminococcaceae-specific probes (B).Left: DAPI stain; middle: universal bacterial probes labelled with Alexa488 (green); right: taxon-specific probes labelled with Cy3 (red).Arrow pointing toward bacterial cells lining the epithelium.The sections shown were taken from fish at time point T2.Bar: 10 μm.

Figure 4 .
Figure 4. Maximum-likelihood phylogenetic trees for AC_MYC01 (A), AC_BRV01 (B), and AC_RUM01 (C) with related bacteria based on near full-length 16S rRNA gene sequences.Bootstr a p v alues ( > 50%) ar e giv en as percenta ges at the br anc hing points and ar e based on 100 r esamplings.Related sequences of bacteria r etrie v ed fr om fish ar e mark ed in bold.Bars show 0.10 substitutions per n ucleotide position.

Figure 5 .
Figure 5. Ov ervie w of the functional properties for the MAGs AC_MYC01, AC_BRV01, and AC_RUM01.(A) Hypothesized metabolic pathways reconstructed for the MAGs; (B) number of COGs per COG category for each MAG; COG functional categories: A, RNA processing and modification; B, c hr omatin structur e and dynamics; C, ener gy pr oduction and conv ersion; D, cell cycle contr ol, cell division, and c hr omosome partitioning; E, amino acid transport and metabolism; F, nucleotide transport and metabolism; G, carbohydrate transport and metabolism; H, coenzyme transport and metabolism; I, lipid transport and metabolism; J, translation, ribosomal structure, and biogenesis; K, transcription; L, r eplication, r ecombination and repair; M, cell wall/membrane/envelop biogenesis; N, cell motility; O, post-translational modification, protein turnover, chaperones; P, inorganic ion transport and metabolism; Q, secondary metabolite biosynthesis, transport, and catabolism; R, general function prediction; S, function unknown; T, signal transduction mechanism; U, intracellular trafficking, secretion, and vesicular transport; V, defence mechanism; Y, nuclear structure; and Z, cytoskeleton.(C) Ov ervie w of the putative functions and interaction of the gut microbiome in empty and full states; (D) dbCAN analysis comparing the functional CAZy family classifications in the MAGs.

Table 1 .
Genome c har acteristics of MA Gs A C_MYC01, A C_BRV01, and A C_RUM01.