Functional feeds marginally alter immune expression and microbiota of Atlantic salmon (Salmo salar) gut, gill, and skin mucosa though evidence of tissue-specific signatures and host-microbe coadaptation remain

Abstract

Background: Mucosal surfaces of fish provide cardinal defense against environmental pathogens and toxins, yet these external mucosae are also responsible for maintaining and regulating beneficial microbiota. To better our understanding of interactions between host, diet, and microbiota in finfish and how those interactions may vary across mucosal tissue, we used an integrative approach to characterize and compare immune biomarkers and microbiota across three mucosal tissues (skin, gill, and gut) in Atlantic salmon receiving a control diet or diets supplemented with mannan-oligosaccharides, coconut oil, or both. Dietary impacts on mucosal immunity were further evaluated by experimental ectoparasitic sea lice (Lepeophtheirus salmonis) challenge.

Results: Fish grew to a final size of 646.5 g ± 35.8 during the 12-week trial, with no dietary effects on growth or sea lice resistance. Bacterial richness differed among the three tissues with the highest richness detected in the gill, followed by skin, then gut, although dietary effects on richness were only detected within skin and gill. Shannon diversity was reduced in the gut compared to skin and gill but was not influenced by diet. Microbiota communities clustered separately by tissue, with dietary impacts on phylogenetic composition only detected in the skin, although skin and gill communities showed greater overlap compared to the gut according to overall composition, differential abundance, and covariance networks. Inferred metagenomic functions revealed preliminary evidence for tissue-specific host-microbiota coadaptation, as putative microbiota functions showed ties to the physiology of each tissue. Immune gene expression profiles displayed tissue-specific signatures, yet dietary effects were also detected within each tissue and peripheral blood leukocytes. Procrustes analysis comparing sample-matched multivariate variation in microbiota composition to that of immune expression profiles indicated a highly significant correlation between datasets.

Conclusions: Diets supplemented with functional ingredients, namely mannan-oligosaccharide, coconut oil, or a both, resulted in no difference in Atlantic salmon growth or resistance to sea lice infection. However, at the molecular level, functional ingredients caused physiologically relevant changes to mucosal microbiota and host immune expression. Putative tissue-specific metagenomic functions and the high correlation between expression profiles and microbiota composition suggest host and microbiota are interdependent and coadapted in a tissue-specific manner.

Keywords: Atlantic salmon; Coconut oil; Fish microbiome; Functional feeds; Gene expression; Host–microbiota interactions; Immune regulation; Mannan-oligosaccharides; Sea lice.

Fig. 1

Microbiota composition detected across mucosal tissues of Atlantic salmon fed diets supplemented with functional ingredients. Fish received a control diet (Control), a 1% mannan oligosaccharide supplementation (MOS), a 96% lipid replacement using coconut oil (CoconutOil) or a combination of the two treatments (CocoMOS). Alpha diversity was tested by two-way ANOVA with tissue showing global effects on richness (A) and Shannon diversity (B). Due to tissue-diet interactions, dietary effects on richness and diversity within each tissue were also tested by one-way ANOVA and Dunnett’s post-hoc (* p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001). Beta diversity is displayed by tissue using principal co-ordinates analysis (PCoA) of weighted (C) and unweighted (D) UniFrac distances with statistical values taken from PERMANOVA. Within tissue dietary effects on microbiota composition were only detected in the skin according to both weighted (p = 0.039) and unweighted UniFrac (E). Pairwise PERMANOVA showed the MOS and Coconut oil diets to significantly altering communities relative to the control in unweighted UniFrac (E)

Fig. 2

Keystone microbiota of Atlantic salmon associated with mucosal tissues and dietary treatments according to differential abundance testing and network analysis. A log2-fold-change plot (A) shows the results of pairwise differential abundance (DA) conducted between tissues, while controlling for diet using DESeq2 (FDR corrected q ≤ 0.05 and log2-fold change|≥ 1). Within tissue dietary effects were also tested, with only one dietary DA ASV identified (Gill: Control v. MOS) (A). Bacterial genera are listed on the y-axis, points are colored by phylum, and shape identifies the pairwise treatment comparison for which the ASV showed DA. Positive fold-changes indicate an increased abundance in the first group in the comparison, and vice versa. Microbiota networks (B–D) depict the top 50 most connected ASV (nodes) according to sparse inverse co-variance networks reconstructed from the gut (B), gill (C), and skin (D) microbiota datasets. Network nodes are colored by phylum, while node size is positively correlated with Laplacian centrality, and edges are colored by positive (green) and negative (red) covariance relationships

Fig. 3

Pairwise tissue-specific differences in inferred metagenomic pathways for Atlantic salmon mucosal microbiota. Metagenomic functions were inferred using PICRUSt2. Pathway abundance was compared across tissues using a Kruskal Wallis test followed by a Tukey’s post-hoc with BH-FDR corrections. Significant differences in pathways abundance were considered at p < 0.01 and effect size > 0.5. Out of 404 inferred MetaCyc pathways, 60 showed significant differences across tissue, with only a subset of those shown here. A full list of all differentially abundant inferred metagenomic functions and pathways can be found in Additional file 2: Supplemental Table S4-S6. REDCITCYC—reductive TCA cycle; NADSYN—NAD synthesis; ASPASN-PWY—superpathway of L-aspartate and L- asparagine; FERMENTATION-PWY—mixed acid fermentation; PWY3781—aerobic respiration I; DENITRIFICATION-PWY—nitrate reduction I; MET-SAM-PWY—superpathway of S-adenosyl-L-methionine biosynthesis; GLUCOSE1METAB-PWY—glucose and glucose-1-phosphate degradation

Fig. 4

Immune gene expression across mucosal tissues of Atlantic salmon fed diets supplemented with functional ingredients. A set of systemic-adaptive-immunity genes (A, C) were assayed in the gut, gill, skin and peripheral blood lymphocytes (PBL), while a set of mucosal-innate immunity markers (B, D) were assayed in the three mucosal tissues. Transcript abundances, shown on a log scale, were inferred from a global Bayesian model using efficiency corrected qPCR data (A, B). Dashed lines show a significant pairwise difference between tissues among fish receiving the control diet (FDR corrected p ≤ 0.05), while significant pairwise differences in expression between two or more dietary treatments within a specific tissue is denoted by *. Principal components analysis plots (C, D) show the multivariate sample ordinations based on the systemic-adaptive-immunity genes (C) and mucosal-innate-immunity genes, with overlaid eigenvector loadings indicating the contribution of each gene

Fig. 5

Multivariate Procrustes analysis comparing microbiota composition to host immune gene expression profiles across mucosal tissues of Atlantic salmon. Principal co-ordinates analysis (PCoA) sample ordinations of abundance-weighted (A, C) and unweighted (B, D) microbiota phylogenetic composition (UniFrac) (arrows) were mapped to sample ordinations based on host mucosal-innate (A, B) or systemic-adaptive immune (C, D) gene expression profiles (points) taken from the same set of samples. Longer lines between a sample gene expression eigenvalue (points) and its concurrent microbiota eigenvalue (arrows) indicates greater discordance between datasets for that sample. Inset violin plots display the distribution of Procrustes residuals by tissue for each plot. Significant correlations (p < 0.001, 999 permutations) were detected in all comparisons