The following outline of the presentation of the results has been chosen; firstly, to give a picture of the overall results from all treatments, before more detailed comparisons are made between the two pairs of treatment for which the cause of differences can be interpreted and discussed to achieve the goals of the study, i.e. fish fed the FOS and the FOS-BC diets and those fed the FOS-BC and GOS-BC diets. This approach will help us to understand the effects of supplementation of probiotic, BC, to prebiotic, FOS, and the influence of alteration of prebiotic combined with BC.
Overall performance
The fish grew well throughout the experiment showing thermal growth coefficients (TGCs) averaging about 3.1 (Fig. 2). Fish in the FOS-BC group grew significantly faster than those in the control group, showing TGCs of 3.23 and 2.96, respectively, during the 10 weeks of feeding. Fish in the other treatments showed intermediate TGCs. Feed intake and feed conversion ratios, which averaged 847g ± 8 (SEM) and 1.12 ± 0.02, respectively, showed no significant differences among the four treatments.
Overall results of gut histology
The distal intestine and pyloric caeca of the fish from the four treatments showed largely normal morphological characteristics, but some individuals from all diet groups showed abnormal morphology that ranged from mild to severe. Figures 3a and 3b illustrate the observations made regarding signs of inflammation in the distal intestine, i.e. regarding cell infiltration and loss of distal intestine enterocyte vacuoles, respectively. The results showed no significant differences between treatments. The same was observed regarding infiltration of inflammatory cells in mucosa and lipid accumulation (steatosis) in pyloric caeca, (i.e. inflammation and steatosis, Figs. 3c and 3d, respectively).
Overall results of microbiota profiling
The absolute bacterial DNA levels measured by qPCR analysis did not show significant differences between any of the three experimental diets. However, the variation between samples within treatment was large (Additional File 1. Figure S1). Generally bacterial DNA levels in digesta was higher compared to the levels in mucosa.
Alpha diversity, i.e. number of different species within a sample, measured as observed species and Shannon indices, are presented in Additional File 1: Figures S2a and S2b for digesta and S2c and S2d for mucosa. In the digesta samples, significant differences between the treatments were observed (observed species: p = 0.05 and Shannon index: p = 0.05). Fish fed the control diet showed the highest value for observed species, those fed the GOS-BC diet showed the lowest with the difference being significant (observed species: p = 0.04). The mucosa samples did not show significant diet effects (observed species: p = 0.22 and Shannon index: p = 0.09) among the fish fed different diets.
Beta diversity, i.e. differences in bacterial species between samples taking into account species differences as well as the abundance of the species, was evaluated by PERMANOVA analysis based on Bray-Curtis dissimilarity matrix. For the digesta samples (Fig. 4a) significant differences among treatments were observed (F = 1.99, R2 = 0.176 and p = 0.04). On the other hand, the mucosa samples (Figs. 4b) did not show significant differences in beta diversity (F = 1.24, R2 = 0.117 and p = 0.23) among different treatments.
In the digesta, at the phylum level, Firmicutes and Proteobacteria dominated the bacterial communities, representing more than 80% of the average relative abundance in all treatments (Additional File 1: Figure S3a). At the genus level, the lactic acid bacteria group, represented mainly by Lactobacillus and Leuconostoc comprised around 50% of the average relative abundance in all treatments (Fig. 5a). In the mucosa, the most abundant phyla were Firmicutes and Proteobacteria followed by Tenericutes. Together they accounted for approximately 80% of averaged relative abundance in all the treatments (Additional File 1: Figure S3b). The dominant genera in mucosa were Lactobacillus, Aliivibrio and Leuconostoc which comprised around 50% averaged relative abundance per feeding group (Fig. 5b). The complete list of genera in digesta which showed significant changes in their abundance among treatments are presented in Additional File 2: Tables S1.
The Random Forest model performed well for predicting the results of the four treatments regarding the digesta samples, but not the mucosa samples, as indicated by 0.25 and 0.812 OOB (out of bag) error obtained, respectively (Additional File 2: Tables S2 and S3). Therefore, in the following, we mainly focus on digesta-associated microbiota. The model classified the treatments, FOS-BC and GOS-BC, quite precisely with 87.5% predicting accuracy for the digesta samples. In the digesta samples, the most important taxon that allowed discrimination of fish fed diets supplemented with BC from the other fish, was P. acidilactici (Fig. 6a). In the mucosa, it was the fifth most important discriminatory taxon (Fig. 6b). Both digesta (Fig. 6c) and mucosa (Fig. 6d) samples from the FOS-BC and GOS-BC diet fed fish had higher relative abundance of P. acidilactici compared to the FOS and control diet fed fish.
Overall results of transcriptome profiling: The RNA-seq data showed raw read counts ranging from 20.4–42.8 million reads with an average count of 30.1 million per sample. Uniquely mapped reads ranged between 15–32 million among the samples having an 71% of average unique mapping efficiency. Compared to the other treatments the GOS-BC treatment showed the highest number differently expressed genes (DEGs, Benjamini-Hochberg adjusted p < 0.1, Table 1) and annotated DEGs among treatments were presented in Additional File 3. Gene ontology (GO) enrichment analysis indicated enrichments of biological process within the statistical criteria only for upregulated genes in GOS-BC vs FOS-BC and GOS-BC vs Control. The complete list of summarized GO terms generated from upregulated genes from respective comparison is available in Additional File 2: Table S4. There were no enriched biological processes for the comparisons among the other treatments due to the low number of DEGs.
Comparisons
|
Differentially expressed genes (DEGs)
(q < 0.1, FC > 1.5)
|
Table 1
Number of differentially expressed genes (DEGs) resulted from pairwise comparisons of treatments
|
Total
|
Upregulated
|
Downregulated
|
FOS-BC vs FOS
|
34
|
27
|
6
|
GOS-BC vs FOS-BC
|
220
|
174
|
46
|
FOS vs Control
|
07
|
04
|
03
|
FOS-BC vs Control
|
07
|
02
|
05
|
GOS-BC vs Control
|
537
|
269
|
268
|
Overall results of global metabolome profiling: In total, 747 and 655 metabolites were detected respectively in distal intestine digesta, and blood plasma samples collected from the various treatments. The number of significantly altered metabolites among fish fed different diets are presented in the Table 2. All the detected metabolites highlighting the significantly altered metabolites in each of the comparisons between treatments are presented in the Additional File 3: Files S4 and S5 for digesta and plasma, respectively. The highest number of significantly different metabolite levels in the digesta samples were observed for the comparison between the GOS-BC and the control treatments, and the lowest between the FOS-BC and the FOS treatments i.e. 227 and 27, respectively. The metabolites showing significant differences in the GOS-BC vs control comparison indicate effects on several steps in the metabolism of most nutrients, such as amino acid and peptides, carbohydrate, fatty acid, phospholipid and sterol metabolism, nucleotide, cofactors and vitamins, as well as xenobiotics. Regarding the FOS-BC vs FOS comparison, the few significant effects were seen generally scattered over the metabolic map, not showing clear effects on any metabolic pathway. Both FOS and FOS-BC showed quite similar number of significantly altered metabolites compared to the control. Fish fed the three supplemented diets showed high levels of metabolites linked to methylation of lysine and/or carnitine biosynthesis (such as N6-methyllysine, N6, N6, N6-trimethyllysine and deoxycarnitine) from the control fed fish, which did not clearly differ between fish fed the experimental diets.
Comparisons
|
Significantly altered metabolites in digesta (p ≤ 0.05)
|
Significantly altered metabolites in plasma (p ≤ 0.05)
|
Table 2
Number of significantly altered metabolites obtained from pairwise comparisons of treatments
|
Increased
|
Decreased
|
Increased
|
Decreased
|
FOS-BC vs FOS
|
14
|
13
|
03
|
19
|
GOS-BC vs FOS-BC
|
86
|
23
|
18
|
34
|
FOS vs Control
|
60
|
56
|
104
|
48
|
FOS-BC vs Control
|
63
|
63
|
65
|
60
|
GOS-BC vs Control
|
165
|
62
|
103
|
101
|
Random Forest modeling of results from digesta and plasma indicated 0.31 and 0.25 OOB error, respectively, regarding prediction of the four treatments (Additional File 2: Tables S5 and S6, respectively). In both the digesta and the plasma, the model performed very well at classifying the control and GOS-BC treatment but encountered more difficulty at correctly distinguishing FOS group from FOS-BC group. Although some differences were observed, many of the changes in plasma and digesta metabolites mirrored each other by dietary treatment, for example some of the metabolites important to group separation in digesta were similar to those in plasma (Additional File 1: Figure S4). Among those were metabolites important for methylation of protein lysine and/or carnitine biosynthesis (such as N6-methyllysine, N6, N6, N6-trimethyllysine and deoxycarnitine) and microbiota-linked metabolism (N-methylhydantoin).
Overall results of SCFA levels: The metabolome analyses of plasma samples did not show significant treatment effects, neither regarding the major SCFAs (acetic acid, butyric acid, and propionic acid) nor the minor (valeric acid and hexanoic acid, and branched chain fatty acids, 2-methylbutyric acid, isovaleric acid) (Additional File 2: Table S7). Regarding the digesta samples on the other hand, butyric and valeric acid showed significant treatment effects (Table 3). For both acids, significantly lower values were observed for the GOS-BC treatment compared to the control, whereas the FOS and FOS-BC showed intermediate and quite similar results.
|
SCFA concentrations in digesta of distal intestine (ng/ml) ¤
|
Table 3
SCFA concentrations in distal intestinal digesta of the fish from four treatments
|
Control
|
FOS
|
FOS-BC
|
GOS-BC
|
Acetic acid
|
8.4E + 04
± 3.9E + 04
|
8.3E + 04
± 2.6E + 04
|
6.5E + 04 ± 1. 8E + 04
|
1. 6E + 05 ± 7. 5E + 04
|
Butyric acid
|
83 ± 11a
|
68 ± 7 ab
|
60 ± 4 ab
|
54 ± 3 b
|
Propionic acid
|
121 ± 17
|
101 ± 9
|
87 ± 7
|
88 ± 6
|
Valeric acid
|
41 ± 4 a
|
34 ± 3 ab
|
32 ± 7 ab
|
28 ± 1 b
|
Hexanoic acid
|
254 ± 22
|
225 ± 13
|
213 ± 1
|
203 ± 7
|
2-Methylbutyric acid
|
21 ± 4
|
16 ± 2
|
16 ± 2
|
12 ± 1
|
Isobutyric acid
|
25 ± 3
|
20 ± 1.5
|
21 ± 2
|
19 ± 1
|
Isovaleric acid
|
13 ± 2
|
14 ± 1
|
14 ± 1
|
12 ± 1
|
¤ Mean value ± SEM are presented for n = 8 samples |
Different letters indicate significant changes between treatments for SCFAs (p ≤ 0.05). |
Table 4. Composition of experimental diets for post-smolt Atlantic salmon
|
Trial feeds (5mm pellet size)
|
Diet composition (g/ 100g)
|
Control
|
FOS
|
FOS-BC
|
GOS-BC
|
Fish Meal
|
15.0
|
15.0
|
15.0
|
15.0
|
Soya SPC
|
11.0
|
11.0
|
11.0
|
11.0
|
Wheat Gluten
|
7.2
|
8.0
|
8.0
|
8.0
|
Maize gluten
|
5.0
|
5.0
|
5.0
|
5.0
|
Pea protein
|
15.0
|
15.0
|
15.0
|
15.0
|
Guar Meal
|
8.0
|
7.0
|
7.0
|
7.0
|
Wheat
|
11.0
|
10.8
|
10.9
|
10.0
|
Fish Oil
|
13.2
|
11.5
|
11.5
|
11.5
|
Rapeseed Oil
|
10.4
|
11.1
|
11.1
|
11.1
|
Vit + min + AA
|
4.3
|
4.9
|
4.9
|
4.9
|
Yttrium
|
0.1
|
0.1
|
0.1
|
0.1
|
FOS
|
-
|
0.1
|
0.1
|
-
|
GOS
|
-
|
-
|
-
|
1.0
|
Bactocell
|
|
|
0.03
|
0.03
|
Water change
|
-0.1
|
0.5
|
0.5
|
0.5
|
Analyzed moisture (%)
|
5.8
|
5.4
|
5.7
|
6
|
Energy (bomb calorimetry, MJ/kg)
|
24.2
|
24.2
|
23.8
|
24.1
|
Crude FAT (%)
|
28.5
|
27.9
|
27.6
|
28.1
|
Crude protein (%)
|
43.2
|
43.5
|
43.9
|
43
|
Beta glucan, nucleotides and krill were added only to the experimental diets in equal amounts.
Overall results of association analysis of gut microbiota and metabolites
The Spearman correlation analysis showed significant differences in specific microbe–metabolite correlations between the treatments. In the correlation analyses 436 digesta metabolites with the human metabolome database (HMDB) IDs were included. The circos plot and the heat map for microbe-metabolite correlations in digesta samples from comparisons between FOS-BC and FOS, and GOS-BC and FOS-BC treatments are presented in Figs. 7 and 8, respectively. The circos plots indicated that carbohydrates, cofactors and vitamins, amino acids and lipids were closely correlated with genera belonging to Firmicutes, Actinobacteria and Proteobacteria phyla for both the comparisons presented. Heatmaps show expansion of the results shown in the circos plots. The comparison between the GOS-BC and FOS-BC treatments showed the highest number of associations between microbiota and metabolites, and most of them were significant. In the heatmaps, statistically significant results (p < 0.05) are indicated with asterisks. Correlations values (R) and p-values for the specific microbe–metabolite correlations are presented in Additional File 4.
Supervised multivariate analysis on the combined data matrix of microbiota (at genus level) and metabolome and in the digesta with the OPLS-DA method pointed out clearer separation between GOS-BC and FOS-BC treatments than between FOS-BC and FOS treatments as indicated by the first component (Additional File 1: Figure S5). Variable importance plot (not shown) based on the OPLS-DA model was used to identify differential microbes and metabolites contributing to the separation of one group compared to the other (Variable Importance on Projection, VIP values > 1 and correlation coefficients p < 0.05). The list of microbiota and metabolites fulfilling the above statistical criteria in each comparison between dietary treatments are presented in the Additional File 5. Genus Pediococcus was identified as an important variable in both FOS-BC and GOS-BC treatments from the control and the FOS treatment.
Comparison of the FOS and the FOS-BC treatment
As mentioned above, the fish in the FOS and the FOS-BC treatments showed no significant difference in growth rate, FCR or histological appearance of the pyloric caeca and distal intestine. The results of metabolome analyses of digesta and plasma indicated very few significant alterations among the more than 750 identified metabolites, indicating no important metabolic differences between fish fed these two diets.
Regarding the microbiota results the FOS and FOS-BC treatments did not show significantly different alpha diversity, neither in digesta (Additional File 1: Figures S2a - S2b) nor mucosa samples (Additional File 1: Figures S2c - S2d). On the other hand, beta-diversity from the FOS-BC treatment showed clear separation from those in the FOS treatment for the digesta samples (p = 0.009, Figure 4a), but not for the mucosa samples (Figure 4b). The number of differentially abundant genera was 19, and 15 of them showed highest abundance in the FOS-BC fed fish. Pediococcus and Staphylococcus were among the genera showing increase. The complete list of genera in digesta which showed significant changes in their abundance in FOS-BC group compared to the FOS group is shown in Additional File 2: Table S1. Further, fish fed the FOS-BC diet had higher relative abundance of P. acidilactici compared to the fish fed the FOS diet in both digesta and mucosa samples (Figure 6c and 6d).
The global transcriptomic changes in the distal intestine of fish fed the FOS-BC diet compared to those fed the FOS diet showed a low number of DEGs (27 up- and 6 down-regulated, Table 1, list of differentially expressed annotated genes in Additional File 3: File S1). Results of GO enrichment analysis did not indicate enrichments of biological processes within the statistical criteria for particular comparison due to the low number of DEGs.
Regarding the results of association analysis of gut microbiota and metabolite, Circos plots showed that 4 different classes of metabolites, carbohydrates, cofactors and vitamins, amino acids and lipids were closely correlated with genera belonging to Firmicutes, Actinobacteria, Proteobacteria and Epsilonbacteoeota phyla (Figure 7a). Similar observations were made for the metabolites in plasma (data not shown). As shown in the heatmap (Figure 8a), the 15 genera showing increase in the FOS-BC treatment compared to FOS treatment showed positive correlation with the 10 - 11 significantly changed metabolites in the respective comparison. Some of these correlations were statistically significant (p < 0.05). Genus Pediococcus showed positive and significant associations with 10 metabolites including lactose, ergosterol, chiro-inositol and ribose (Additional File 4: File S1).
Supervised multivariate analysis on the combined data matrix of microbiota and metabolome in the digesta pointed out less clear separation between FOS-BC and FOS treatments (Additional File 1: Figure S5). The FOS-BC treatment showed 41 separating factors from the FOS treatment (Additional File 5: File S1). The genus Pediococcus was identified as an important variable in FOS-BC group distinguishing it from the FOS group.
Comparison of the FOS-BC and the GOS-BC treatment
As mentioned above, the fish in the FOS-BC and the GOS-BC treatments showed no significant difference in TGC, FCR or histological appearance of the pyloric caeca and distal intestine.
Regarding microbiota profiling of the digesta samples the results showed that replacement of GOS for FOS in the FOS-BC diet caused a significant decrease in alpha diversity (observed species, p = 0.046 and Shannon index, p-value = 0.0046) (Additional File 1: Figures S2a - S2b), although the mucosa samples did not show significant diet effects on alpha diversity (Additional File 1: Figures S2c - S2d).
Beta diversity in the digesta samples, revealed that the microbiota in fish from the GOS-BC treatment clustered close to, but distinct from that of the FOS-BC treatment (p = 0.02, Figure 4a). The mucosa samples (Figures 4b) did not show significant differences in beta diversity among the samples.
Fish fed the GOS-BC diet, compared to those fed the FOS-BC diet, showed reduction in 24 genera including Kurthia, Savagea, Staphylococcus, Vagococcus and Peptostreptococcus (Additional File 2: Table S1). Moreover, both digesta and mucosa samples from the fish fed the GOS-BC diet had higher relative abundance of P. acidilactici compared to the fish fed the FOS-BC diet (Figure 6c and 6d).
Global transcriptome analyses showed major differences in the distal intestinal tissue between fish fed the GOS-BC diet and FOS-BC diet. In the fish fed GOS-BC diet 220 genes were differentially expressed, 174 up- and 46 down-regulated compared to the fish fed FOS-BC diet (Table 1, Additional File 3: File S2). Among the upregulated genes in fish fed with GOS-BC diet were cysteine knot cytokine members, interleukin 17 and receptors, Il17a, il17a/f1 and i17ra; TNF superfamily members and receptors tnfrsf1b, tnfrsf1, tnfrsf9a and tnfsf18; beta trefoil cytokine family member il-1rl; and a number of chemokines (Additional File 3: File S2). The fish in the GOS-BC treatment also showed an increase in expression of transcripts of NADPH oxidases family of enzymes, dual oxidases (duox and duox2) and NADPH oxidase activator 1 (noxa1a and noxo1b) and key antioxidant enzyme, glutathione peroxidase 1b (gpx1b).
The GO enrichment analysis was performed for the genes upregulated in the GOS-BC treatment compared to the FOS-BC treatment. The summarized GO terms generated from enriched nonredundant biological function GO terms are presented in Figure 9 for the upregulated genes. The complete list of non-redundant GO terms generated from upregulated genes is available in Additional File 2: Table S4. Among the enriched GO biological process terms were immune response, apoptotic process, inflammatory response, response to stress and reactive oxygen species metabolic process.
The global metabolome profiling showed that replacement of FOS in the FOS-BC diet with GOS significantly altered a high number of metabolites in both digesta and plasma (Table 2, Additional File 3: Files S4 and S5 respectively for digesta and plasma). Unique for the GOS-BC treatment were high levels of long chain saturated, long chain monounsaturated, long chain polyunsaturated as well as branched fatty acids, most pronounced for digesta (Additional File 3: File S4). Among those metabolites were n-3 (EPA, 20:5n-3 and docosapentaenoic acid, DPA, 22:5n-3) and n-6 fatty acids (linoleic acid; 18:2n-6, eicosadienoic acid, 20:2n-6, arachidonic acid, 20:4n-6, adrenic acid, 22:4n-6 and dihomo-gamma-linolenic acid; 20:3n-6, DPA, 22:5n-6 and tetracosahexaenoic acid, 24:6n-3). The GOS-BC fed fish also showed increased levels in the digesta of acetylcarnitine, propionylcarnitine, butyrylcarnitine compared to FOS-BC fed fish, as well as compared to the other treatments. Levels of several sphingomyelins, ceramides and hexosylceramides were also increased distinctively in the GOS-BC fed fish compared to the FOS-BC fed fish and fish from the other treatments.
No significant changes in SCFA levels in plasma or digesta were observed in the comparison between GOS-BC and FOS-BC treatments (Additional File 2: Table S7, and Table 3 for plasma and digesta samples respectively).
The association analysis of gut microbiota and metabolite presented in the Circos plot showed that seven different classes of metabolites including nucleotides, carbohydrates, peptides, cofactors and vitamins, xenobiotics, amino acids and lipids were closely correlated with genera mainly belonging to Firmicutes, Actinobacteria and Proteobacteria phyla (Figure 7b). Similar observations were made for the metabolites in plasma (data not shown). All the 24 genera showing decrease in GOS-BC treatment compared to FOS-BC treatment displayed positive correlation with 54 – 56 metabolites (Figure 8b). Some of these correlations were statistically significant (p < 0.05). Negative correlations were observed between significantly increased metabolites including n-3 and n-6 polyunsaturated fatty acids and the altered bacterial genera in GOS-BC treatment compared to the FOS-BC treatment (Additional File 4: File S2).
Supervised multivariate analysis on the combined data matrix of microbiota and metabolome in the digesta with the OPLS-DA method showed a clear separation between GOS-BC and FOS-BC treatments as indicated by the first component (Additional File 1: Figure S5). Variable importance plot identified showed 115 separating factors in the GOS-BC group compared to the FOS-BC treatment (Additional File 5: File S2).