The need for high-resolution gut microbiome characterization to design efficient strategies for sustainable aquaculture production

Abstract

Microbiome-directed dietary interventions such as microbiota-directed fibers (MDFs) have a proven track record in eliciting responses in beneficial gut microbes and are increasingly being promoted as an effective strategy to improve animal production systems. Here we used initial metataxonomic data on fish gut microbiomes as well as a wealth of a priori mammalian microbiome knowledge on α-mannooligosaccharides (MOS) and β-mannan-derived MDFs to study effects of such feed supplements in Atlantic salmon (Salmo salar) and their impact on its gut microbiome composition and functionalities. Our multi-omic analysis revealed that the investigated MDFs (two α-mannans and an acetylated β-galactoglucomannan), at a dose of 0.2% in the diet, had negligible effects on both host gene expression, and gut microbiome structure and function under the studied conditions. While a subsequent trial using a higher (4%) dietary inclusion of β-mannan significantly shifted the gut microbiome composition, there were still no biologically relevant effects on salmon metabolism and physiology. Only a single Burkholderia-Caballeronia-Paraburkholderia (BCP) population demonstrated consistent and significant abundance shifts across both feeding trials, although with no evidence of β-mannan utilization capabilities or changes in gene transcripts for producing metabolites beneficial to the host. In light of these findings, we revisited our omics data to predict and outline previously unreported and potentially beneficial endogenous lactic acid bacteria that should be targeted with future, conceivably more suitable, MDF strategies for salmon.

Fig. 1. Effect of mannan (0.2% inclusion rate) on host and gut microbial community structure and function.

A Sampling strategy for studying the effect of mannan on the temporal dynamics of the Atlantic salmon gut microbiota. T0 (parr), T1 (pre-smolts), T2 (smolts), and T3 (post-smolts) represent the different sampling time points. The experimental groups are labeled as CTR (Control), MC1 (Diet 1), MC2 (Diet 2), and MN3 (Diet 3), indicating the different diets administered to fish. B The mean body weight in all the experimental groups, stratified by sampling time (n = 390 samples). Boxplots show medians and Interquartile Range (IQR). P-values were determined by the Kruskal–Wallis test with False Discovery Rate (FDR) control for multiple testing. C Alpha diversity (Shannon diversity index), stratified by sampling time (n = 156 samples). Boxplots show medians and IQR. P-values were determined by the Kruskal–Wallis test for comparisons involving more than two groups, and the Wilcoxon test for two-group comparisons, with FDR control for multiple testing. D Beta diversity was assessed using Bray–Curtis dissimilarity for 16S rRNA gene data obtained from the hindgut samples (n = 156 samples). The effect of sampling time was tested with PERMANOVA. Each dot represents individual samples colored by sampling time (T0 (parr), T1 (pre-smolts), T2 (smolts), and T3 (post-smolts)), as indicated in the legend. E Beta diversity was assessed through Bray–Curtis dissimilarity for 16S rRNA gene data obtained from the hindgut samples (n = 156 samples). The effect of different diets was tested with PERMANOVA. Each dot represents individual samples colored by different MDFs (CTR (Control), MC1 (Diet 1), MC2 (Diet 2), and MN3 (Diet 3)), as indicated in the legend. F PCA plot showing the differences in host gene expression between MDFs (MC1 (Diet 1), MC2 (Diet 2), and MN3 (Diet 3)) and control samples (CTR) from the hindgut (n = 142 samples). G Top 20 most abundant genera in all the groups based on 16S rRNA gene data. Other genera with relative abundance less than 1% were shown as “Other genus”. Statistically significant differences between MDFs (MC1 (Diet 1), MC2 (Diet 2), and MN3 (Diet 3)) and control samples (CTR) were calculated at each sampling point using the Wilcoxon test, with significance levels indicated by stars: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Colors represent row Z-scores of each microbial genus (brown: high; turquoise: low). A total of 144 samples were analyzed out of 156, excluding samples from T0 (parr). H PCA plot showing the differences in metatranscriptome expression between MDFs (MC1 (Diet 1), MC2 (Diet 2), and MN3 (Diet 3)) and control samples (CTR) from the hindgut (n = 139 samples). In D, E, F and H, the numbers on the axes represent the variance explained by the principal components.

Fig. 2. Effect of β-mannan (4% inclusion rate) on host and gut microbial community structure and function.

A Sampling strategy for studying the effect of β-mannan on the temporal dynamics of the Atlantic salmon gut microbiota. The experimental groups are labeled as CTR (Control) and 4%MN3 diet (4% β-mannan diet). B Boxplot showing the mean body weight in all the experimental groups (n = 36). Boxplots display medians and IQR. P-values were determined by the Wilcoxon test with FDR control for multiple testing. C Alpha diversity (Shannon diversity index), stratified by sample type (hindgut, n = 35 samples; and pyloric caeca, n = 34 samples). Boxplots display medians and IQR. P-values were determined by the Wilcoxon test with FDR control for multiple testing. D Beta diversity was assessed using Bray–Curtis dissimilarity for 16S rRNA gene data obtained from the hindgut samples (n = 35 samples). The effects of the diet were tested with PERMANOVA. Each dot represents individual samples colored by diet, as detailed in the legend. E Beta diversity was assessed using Bray–Curtis dissimilarity for 16S rRNA gene data obtained from the pyloric caeca samples (n = 34 samples). The effects of the diet were tested with PERMANOVA. Each dot represents individual samples colored by diet, as detailed in the legend. F Top 20 most abundant genera in all the groups based on 16S rRNA gene data (n = 69 samples). Other genera with relative abundance less than 1% were shown as “Other genus”. Statistically significant differences between the 4% β-mannan (4% MN3) and control (CTR) samples were calculated using the Wilcoxon test, with significance levels indicated by stars: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Colors represent row Z-scores of each microbial genus (brown: high; turquoise: low). G PCA plot showing the differences in gene expression between the 4% β-mannan (4%MN3) and control samples from the hindgut (n = 37 samples). H PCA plot showing differences in metatranscriptome expression between 4% β-mannan (4%MN3) and control (CTR) samples from the hindgut (n = 37 samples). In D, E, G and H, the numbers on the axes indicate the variance explained by the principal components.

Fig. 3. Selected metabolic features of the salmon gut microbiome as inferred from genome and metatranscriptome comparisons.

The different metabolic pathways, including host and dietary carbohydrate depolymerization, glycolysis, tricarboxylic acid (TCA) cycle and SCFA production, are displayed for each population MAG. The graphical representation includes different carbohydrates, CAZymes, and cellular features based on functional annotations depicted as numbered boxes or abbreviated gene names, which are additionally listed in Supplementary Data 4d. Features are included if a gene was expressed at either the smolt (T2) or post-smolt (T3) stage from either the control or MDF (MC1, MC2, MN3) diets. The main carbohydrates predicted to be utilized (beta-glucans, xylans, galactans, and chitin), SCFAs (e.g., acetate), and organic acids (e.g., lactate and succinate) are represented by large colored arrows. GTBD-Tk inferred taxonomy is included. Gene names and abbreviations are also provided in Supplementary Data 4d.