Modulation of hepatic miRNA expression in Atlantic salmon (Salmo salar) by family background and dietary fatty acid composition

Abstract This study finds significant differences in hepatic fatty acid composition between four groups of Atlantic salmon (Salmo salar) consisting of offspring from families selected for high and low capacities to express the delta 6 desaturase isomer b and fed diets with 10% or 75% fish oil. The results demonstrated that hepatic lipid metabolism was affected by experimental conditions (diet/family). The fatty acid composition in the four groups mirrored the differences in dietary composition, but it was also associated with the family groups. Small RNA sequencing followed by RT‐qPCR identified 12 differentially expressed microRNAs (DE miRNAs), with expression associated with family groups (miR‐146 family members, miR‐200b, miR‐214, miR‐221, miR‐125, miR‐135, miR‐137, miR_nov_1), diets (miR‐203, miR‐462) or both conditions. All the conserved DE miRNAs have been reported as associated with lipid metabolism in other vertebrates. In silico predictions revealed 37 lipid metabolism pathway genes, including desaturases, transcription factors and key enzymes in the synthesis pathways as putative targets (e.g., srebp‐1 and 2, Δ6fad_b and c, hmdh, elovl4 and 5b, cdc42). RT‐qPCR analysis of selected target genes showed expression changes that were associated with diet and with family groups (d5fad, d6fad_a, srebp‐1). There was a reciprocal difference in the abundance of ssa‐miR‐203a‐3p and srebp‐1 in one group comparison, whereas other predicted targets did not reveal any evidence of being negatively regulated by degradation. More experimental studies are needed to validate and fully understand the predicted interactions and how the DE miRNAs may participate in the regulation of hepatic lipid metabolism.

in the degradation of the target gene transcript or inhibition of its translation. This leads to a post-transcriptional downregulation of the target gene protein (Chekulaeva & Filipowicz, 2009;Hausser & Zavolan, 2014;Krol et al., 2010). Research over the past decade has revealed several hundreds of miRNA genes, each of which has the potential to target several target transcripts. Thus, miRNAs control the expression of a large number of genes and seem to be the major post-transcriptional regulators of cellular gene networks in vertebrates (Friedman et al., 2009). Studies in teleost fish have shown that miRNAs participate in the regulation of early development, apoptosis, the maintenance of tissue-specific functions, reproduction and immune response Bizuayehu & Babiak, 2014;Chen et al., 2019). Studies on miRNAs in commercially important fish species have indicated that they also participate in the regulation of economically interesting traits like growth or food conversion (Andreassen et al., 2016;Mennigen, 2016). and humans. These fatty acids exert a range of health benefits through their molecular, cellular and physiological actions (Bou et al., 2017a;Bou et al., 2017b;Calder, 2014). EPA and DHA are synthesized from the essential fatty acid 18:3n-3 in a cascade of reactions consisting of elongation (catalysed by elongase2 and elongase5), desaturation (by delta 5 desaturase and delta 6 desaturase) and a final peroxisomal beta-oxidation step (by acetyl co-A oxidase) (Sprecher, 2000).
Factors such as diets, life stage, genotype and growth are known to influence the capacity for EPA and DHA syntheses in Atlantic salmon (Tocher et al., 2000;Rosenlund et al., 2001;Bell et al., 2002;Torstensen et al., 2005;Leaver et al., 2011;Thomassen et al., 2012).  (Kjaer et al., 2008;Morais et al., 2009;Moya-Falon et al., 2005;Ruyter et al., 2003). The genetic background is also important for the EPA and DHA composition in salmon, and the DHA content of salmon fillet was recently identified as a highly heritable trait (h = 0.46) (Horn et al., 2018). Nonetheless, the correlation between the liver and muscle content of EPA and DHA seems to be low (Horn et al., 2019), indicating that the capacity for EPA and DHA syntheses in liver is less important for the levels of these fatty acids in the muscle. Various genes are involved in hepatic lipid metabolism (Morais et al., 2011;Torstensen et al., 2009). Many of these lipogenic genes are regulated by the transcription factors, sterol regulatory element binding proteins (SREBP) srebp1 and srepb2 (Minghetti et al., 2011). Srepb1 plays a crucial role in the regulation of fatty acid biosynthesis, whereas cholesterol biosynthesis is regulated by Srebp2. Several other genes and gene networks involved in lipid metabolism have been identified in transcriptome studies of salmon, and a connection between fatty acid accumulation, dietary lipid content and immune response has been revealed (Martinez-Rubio et al., 2012;Martinez-Rubio et al., 2013;Skugor et al., 2010).
Characterizations of miRNAs associated with lipid metabolism in teleost fish have been carried out by manipulation of dietary lipids in tilapia, rabbit fish and rainbow trout (Mennigen et al., 2014;Tao et al., 2017;Zhang et al., 2014). These studies have revealed smaller groups of differentially expressed miRNAs (DE miRNAs) likely to be involved in the regulation of lipid metabolism. Genes that might be the target transcripts were identified in a few cases and shown to be among the key genes in lipid metabolism gene networks. The miRNAome is well characterized in Atlantic salmon (Andreassen et al., 2013;Woldemariam et al., 2019), and several miRNAs that respond to viral infection and that are likely to regulate inflammatory response have been identified Woldemariam et al., 2020). Nonetheless, to authors' knowledge, there are no studies of miRNAs and their putative regulatory roles in the salmon lipid metabolism. The molecular mechanism leading to the reported dietary effect from EPA and DHA on expression of genes in the omega-3 synthesis pathway (Kjaer et al., 2008;Morais et al., 2009;Moya-Falon et al., 2005;Ruyter et al., 2003) could, e.g., involve posttranscriptional regulation by miRNAs.
This study investigates how a diet that was either rich in fish oil and low in rapeseed oil (75FO) or low in fish oil and rich in rapeseed oil

| Fatty acid composition of the diets 10FO and 75FO
The diets were formulated to contain two different levels of fish oil, 10% fish oil (10FO) or 75% fish oil (75FO) of the oil fraction. Table 1 provides the chemical composition, whereas Table 2 provides the fatty acid composition of the two diets. The resulting fatty acid composition of the 75FO diets thereby contained a higher percentage of EPA (C20:5n-3), docosapentaenoic acid (DPA, C22:5n-3) and DHA (C22:6n-3) than the 10FO.
The sum of EPA and DHA in the 75FO was 22.1% (of total fatty acids) compared to 6.9% in the 10FO. The 18:3n-3 was, nonetheless, higher in 10FO (7.5%) than in 75FO (2.8%). The 75FO had a higher percentage of saturated fatty acids (SFAs) than 10FO, constituted mainly by 14:0, 16:0 and 18:0. The percentage of monounsaturated fatty acids (MUFAs) was also higher in 10FO than in 75FO, resulting mostly from a 2.3 times higher level of 18:1n-9 in the 10FO. The sum of n-6 fatty acids was higher in 10FO compared to 75FO and was dominated by 18:2n-6 (linoleic acid) that was 2.4 times higher in 10FO than in 75FO.

| Measurements of fatty acid composition of the liver
The fatty acid composition of the 36 liver samples (9 fish per group) was analysed by trans-methylating the lipids re-dissolved in FISH chloroform using 2,2-dimethoxypropane, methanolic HCl and benzene at room temperature, as described by Mason (Mason & Waller, 1964).
The methyl esters of fatty acids were then separated in a gas chromatograph (Hewlett Packard 6890) with a split injector, SGE BPX70 capillary column (length 60 m, internal diameter 0.25 mm and thickness of the film 0.25 μm), flame ionization detector and HP Chem Station software. The carrier gas was helium. The injector and detector temperatures were 300 C. The oven temperature was raised from 50 to 170 C at a rate of 4 C min -1 and thereafter raised to 200 C at a rate of 0.5 C min -1 and finally to 300 C at a rate of 10 C min -1 . The relative quantity of each fatty acid was determined by measuring the area under the peak in the gas chromatograph spectrum corresponding to the specific fatty acids. 2.5 | cDNA synthesis and miRNA expression measurements using RT-qPCR

| Isolation of RNA
The miScript assays were used for cDNA synthesis and qPCR as The normalized read counts from the small-RNA-sequenced samples were utilized to compare the stability of ssa-miR-25-3p, ssa-miR-92a-3p, ssa-miR-181a-3p, ssa-miR-455-5p, ssa-miR-107-3p and ssa-miR-17-5p across all groups, an approach similar to the one applied in Johansen and Andreassen (2014) to select candidate reference miRNAs. They all showed high stability across all groups, and three of these miRNAs (ssa-miR-25-3p, ssa-miR-92a-3p and ssa-miR-181a-3p) were analysed in all samples using RT-qPCR. The following normfinder analysis carried out as described in Johansen and Andreassen (2014) showed that ssa-miR-25-3p and ssa-miR-92a-3p were the best reference gene combination with a combined stability value of 0.002.
These two miRNAs were used as reference genes in the miRNA expression analysis.
2.6 | Small-RNA sequencing and expression analysis (DESeq2) The library construction was performed at the Norwegian Genomics Next, the reads from each of the 12 samples were aligned to a reference miRNAome that consisted of all known S. salar mature miRNAs (Andreassen et al., 2013;Woldemariam et al., 2019). The reads mapping with edit distance 1 or less to the mature reference sequences were counted. DE miRNAs were identified using the DESeq2 package (Anders & Huber, 2010). Rows with fewer than two reads for each condition were discarded from the analysis. The four groups were compared, and a threshold of P-adjusted value <0.1 (adjusted according to Benjamini-Hochberg procedure) was applied to report the putative DE miRNAs.

| In silico predictions of target transcripts
Target gene predictions were carried out using RNAhybrid. The analysis was performed with conditions of helix constraint 2-8 and no G:U in seed, allowing only target genes that had perfect "seed" matches to be detected (Rehmsmeier et al., 2004). The minimum free energy threshold for RNA hybrids was set to −18 kcal mol -1 to retrieve results (target site matches) from RNA hybrids that had a high stability (Peterson et al., 2014). The 3 0 UTR sequences from all S. salar transcripts in Genbank (Non-redundant mRNA, NM entries in RefSeq, NCBI) were used as an input in the in silico analysis.
Gene ontology (GO) annotations for the predicted target genes (biological process and molecular function) were received for each of the target genes using UniProt database (http://www.uniprot.org/ uploadlists/). These GO annotations were used to identify the sub-set of predicted targets that were associated with lipid metabolism.  measurements of fatty acid composition along with results from twoway ANOVA tests for significant differences are provided in Table 3.

|
The relative liver composition of DHA, the sum of the polyunsaturated n-3 fatty acids (PUFA) and the sum of EPA and DHA, was significantly affected by diets, family groups and their interaction. The relative EPA levels (20:5n-3) were 1.2-1.3 times higher in the HIGH groups than in the LOW groups showing EPA levels of 1.7% (10FO/ HIGH) and 8.6% (75FO/HIGH). The relative DHA levels were 1.5-1.8 times higher in the HIGH groups than in the LOW groups with DHA levels of 7.9% (10FO/HIGH) and 27.8% (75FO/HIGH). Nevertheless, the increased content of fish oil in the diet of the 75FO groups resulted in a much higher percentage of PUFA (including EPA and DHA) in the 75FO group than in the 10FO group (Table 3).
The diet, family background and their interaction also affected the relative level of the sum of SFAs ( P SFA, Within the 75FO group, the HIGH group had a lower liver level of several fatty acids such as 16:1n-9, 18:1n-9, 18:1n-7 and 20:1n-9 than the LOW group.
Six miRNAs showed relative differences associated with only family groups (ssa-miR-200b-5p, ssa-miR-214-5p, ssa-miR-221-5p, ssa-miR-125a-3p, ssa-miR-135c-5p and ssa-miR-137a-3p). Whereas ssa-miR-135c-5p and ssa-miR-137a-3p showed a relative decrease in the HIGH groups, the other four miRNAs showed relative increases in the HIGH groups. The novel Atlantic salmon miRNA (ssa-miR-nov-1-5p) showed an expression associated with both diet and family background. This was also the only miRNA that revealed interaction between the two conditions. This miRNA showed a relative increase in HIGH vs. LOW family groups although it showed a relative decrease in the 75FO vs. 10FO diet groups.

| In silico analysis predicted 37 lipid metabolism genes as putative target genes
The putative target genes of the DE miRNAs were predicted using the 3 0 UTR sequences of all S. salar transcript sequences in GenBank as input (see Section 2). This analysis showed 1069 predicted target genes for the 12 DE miRNAs, ranging from 20 matches by ssa-miR-125a-3p to 167 matches by ssa-miR-203a-3p and ssa-miR-nov1-5p (data not shown).
The UniProt database (GO annotations) was used to identify a sub-set of target genes with functions in the lipid metabolism gene pathways. One of the DE miRNAs, ssa-miR-200b-5p, did not reveal any matches to such lipid metabolism genes. The remaining 11 DE miRNAs showed one or more matches to 37 lipid metabolism genes. ØSTBYE ET AL. 7

FISH
The target genes along with the DE miRNAs predicted to target their 3 0 UTRs are provided in Table 5. Supporting File S6 also shows GO annotations and GenBank accession numbers.
Acyl-CoA oxidase 1 (acox1) was predicted as target for four DE miRNAs, whereas acyl-CoA synthetase long-chain family member 1 (acsl1), lysosomal acid lipase/cholesteryl ester hydrolase (lich) and acyl-protein thioesterase (lypa1) were predicted as target for three DE miRNAs. Each of nine other genes (including acot5, srebp2 and s27a2) showed target site matches for two DE miRNAs. The remaining genes, including three desaturases (d6fad-c, d6fad-b and d5fad), were targeted by a single DE miRNA. Among other genes that were predicted as targets were the two key lipid metabolism enzymes: elongation of very long-chain fatty acids-like 4 (elovl4) targeted by ssa-miR-221-5p and polyunsaturated fatty acid elongase (elovl5_b) targeted by ssa-miR-135c-5p. Both of these enzymes are essential for the synthesis of omega-3 fatty acids.
On the contrary, each of the DE miRNAs was in most cases predicted to target several genes. Ssa-miR-135c-5p was, e.g., predicted to target nine genes, whereas ssa-miR-137a-3p, ssa-miR-203a-3p and ssa-miR-221-5p were predicted to target six genes each.

| Gene expression changes in predicted target genes and miRNA/target comparisons
The seven predicted target genes acox1, cpt1, srebp-1, srebp-2, Δ6fad_b, Δ6fad_c and Δ5fad and the Δ6 desaturase gene Δ6fad_a were further analysed by RT-qPCR in the complete materials. These are key genes involved in omega-3 fatty acid synthesis (Δ5fad, Δ6fad_a, Δ6fad_b, Δ6fad_c), regulation of lipid synthesis as transcription factors (srebp1, srebp2) or lipid degradation (oxidation) (cpt1, acox). The two genes acox and cpt1 did not reveal any significant changes when groups were compared, whereas the effect of diet on Δ6fad_c was on the borderline of significance (P = 0.07). In the remaining five genes, there was at least one group comparison that showed a significant difference. The results from the two-way ANOVA analysis of these genes are summarized in Table 6.
The ANOVA analysis showed a highly significant effect of the diet on the expression of the three genes Δ5fad, Δ6fad_a and Δ6fad_b, with decreased expression in 75FO compared to 10FO. A similar trend towards a dietary effect on gene expression was observed in Δ6fad_c (towards a decrease in 75FO groups). A significant effect of family background (HIGH vs. LOW groups) was also detected on the gene expression of Δ5fad and Δ6fad_a. When the two family groups were fed 75FO, the gene expression of Δ5fad and Δ6fad_a was downregulated in HIGH compared to LOW.
There was a significant effect of family background on the gene expression of sterol regulatory element binding transcription factor 1 (srebp1) with a relative increase in HIGH compared to LOW and a strong trend towards a dietary effect. The expression of srebp2 was affected only by the diet with a downregulation in fish fed 75FO vs.
10FO in the HIGH group.
The revealed differences in the expression of some of the target genes allowed for a comparison of the abundance of the miRNAs and their predicted targets. Sterol regulatory element binding transcription factor 1 (srebp-1) was predicted as target for ssa-miR-203a-3p. The results showed that there was a relative decrease in srebp-1 in the 75FO diet group (significant in 10FO/HIGH vs. 75FO/LOW, Table 6).
In accordance with srebp-1 being negatively regulated (degraded) by an ssa-miR-203a-3p-guided RISC, this miRNA showed a relative increase in expression when same groups were compared (Table 4).
Sterol regulatory element binding transcription factor 2 (srebp2) was predicted as target for ssa-miR-nov-1-5p. Nevertheless, both the miRNA and the predicted target showed expression changes in the same direction (10FO/HIGH vs. 75FO/HIGH). Both Δ6fad_b and Δ6fad_c were also predicted as target for ssa-miR-nov-1-5p. These genes showed a lower expression in the 10FO/LOW vs. 75FO/LOW comparisons, which was significant in Δ6fad_b (Table 6). Nevertheless, also in this case the DE miRNA showed a change in the same direction (decrease, Table 4). The gene Δ5fad was predicted as target for ssa-miR-462b-5p, but also in this case, both the gene and the miRNA expression changed in the same direction in group comparisons (Tables 4 and 6). Peroxisomal acyl-coenzyme A oxidase 1 (acox1) and carnitine-o-palmitoyltransferase 1 (cp1) were also among the predicted targets. Nonetheless, they did not reveal any significant expression changes in their mRNAs (Table 6).
T A B L E 5 Predicted target genes that are part of lipid metabolism gene networks
A higher inclusion level of fish oil, and therefore EPA and DHA, in the 75FO diet induced downregulation of all the desaturases, except for d6fad_c, compared to the 10FO diet high in rapeseed oil (Table 6).
This is in agreement with findings in other studies reporting that there is a reduced synthesis capacity of EPA and DHA and a decreased gene expression of desaturases, elongases and sterol regulatory element binding protein 1 (srepb1) when Atlantic salmon is fed high-dietary levels of fish oils, whereas high-dietary vegetable oil levels resulted in the opposite effect (Kjaer et al., 2008;Morais et al., 2009;Moya-Falon et al., 2005;Ruyter et al., 2003). Even though the 10FO groups had a higher gene expression level of the desaturases than the 75FO groups, the expression differences in srepb1, which regulate desaturase expression, were less affected by diet (P = 0.08) in the present study. Also, despite a selection of parental individuals into HIGH and LOW groups based on d6fad_b expression, the differences in the gene expression that were observed between the two family groups were not significant for this desaturase in the materials. The first-generation offspring did not show any obvious added effect on d6fad_b expression from crossing the parental individuals from families with highaverage d6fad_b expression. This could indicate that there is genetic variation in several genes within the parental generation that affects the d6fad_b expression and not necessarily variation in the same genes that led to the average high expression in the different parental HIGH families. On the contrary, both the EPA/DHA percentage and other key genes like d5fad, d6fad_a and srebp-1 did reveal significant differences associated with family background.
Studies have shown that other fatty acids also influence omega-3 synthesis (and desaturase expression). The synthesis capacity for EPA and DHA is partly regulated by the availability of the precursors 18:3n-3. The increased expression of genes of the omega-3 synthesis pathway was, e.g., observed in Atlantic salmon fed a vegetable oil high in 18:3n-3 compared to salmon fed a fish oil low in 18:3n-3 (Gillard et al., 2018). Another fatty acid, 18:1n-9, which is the most abundant fatty acid in rapeseed oil and is found at high levels in the 10FO diet, has also been shown to stimulate the gene expression of the Δ5and Δ6fads in salmon hepatocytes (Kjaer et al., 2016). Compared to the 75FO diet, the rapeseed oil-based 10FO diet had higher levels of both 18:3n-3 and 18:1n-9. In accordance with other studies (Gillard et al., 2018), the gene expression of the Δ5and Δ6fads was higher in the 10FO compared to the 75FO (Table 6). Nonetheless, as revealed in the measurements of fatty acid composition (Table 3), the increased availability of 18:3n-3 and 18:1n-9 in the 10FO diet did not stimulate omega-3 synthesis to produce EPA and DHA to levels gained from the 75FO diet. Atlantic salmon has all the enzymes for EPA and DHA synthesis, but, as demonstrated here, the capacity is limited. Zheng et al. (2009) showed that the molecular mechanism that leads to EPA suppressing the omega-3 synthesis in salmon cells could be by suppressing the activity of the Δ6fad promoter. Nonetheless, five of the miRNAs identified as differentially expressed in this study could, as discussed in Section 4.2, also be involved in such diet-triggered regulation by targeting key lipid metabolism genes.
The transcription factor Srebp-2 predominantly regulates cholesterol biosynthesis, and its expression is increased by the depletion of cholesterol (Horton et al., 2002). A suppression of the cholesterol synthesis pathway was also observed in Atlantic salmon fed a cholesterol-supplemented diet (Kortner et al., 2014). Here, the diets affected the srebp2 gene expression with a higher expression in the group fed less fish oil (significant in 10FO/HIGH vs. 75FO/HIGH), indicating higher biosynthesis of cholesterol in the low-fish-oil diets.
The last step in the synthesis of DHA involves a chain shortening of 24:6n-3 to DHA by the enzyme acyl-CoaA oxidase. There were no significant differences in the gene expression of acox (Table 6) that could have explained the differences in hepatic DHA content of the four groups. There were also no effects of either diets or family background on cpt1 gene expression level involved in mitochondrial β-oxidation (Table 6). In conclusion, the differences in diet and family background resulted in different hepatic fatty acid compositions and differences in the expression of several lipid key genes. These findings are largely in agreement with previous studies.

| Putative regulatory roles of DE miRNAs in S. salar lipid metabolism
The analysis of relative expression differences revealed 12 mature miRNAs with changed expressions. Some of these miRNAs were differentially expressed when comparing family groups, others were differentially expressed when comparing differences in diets and the expression of four miRNAs was associated with both family groups and diet. As the expression of several miRNAs was affected by family background independent of diet, the initial selection of the parental individuals in two family groups using Δ6fad_b expression levels led to a selection into two family groups that also differed in their expression of certain miRNAs. Nonetheless, as the diet affected the expression of six miRNAs, their expression seems to be modulated by dietary levels of fish oil and vegetable oil.
One species-specific mature miRNA (ssa-miR-nov1-5p) and the teleost-specific miR-462b-5p were among the DE miRNAs, indicating they may have regulatory roles in the salmon lipid metabolism.  (Chen et al., 2014). Also, ssa-miR-146a (identical to miR-146b in Ahn et al., 2013) was reported as a regulator of adipogenesis by suppressing the SIRT1-FOXO cascade.
In silico predictions of target genes are a common approach to further elucidate the role of a certain DE miRNA. One limitation of such predictions when studying non-model species is that the 3 0 UTRs of most genes, including those genes suggested as targets in other vertebrate studies, are poorly characterized. Also, a large proportion of the predicted targets will be false positives . Nevertheless, such predictions may narrow down the genes relevant to study further by experimental approaches. One finding in the present in silico analysis was that ssa-miR-137a-3p was predicted to target cdc42. If there is a conserved miRNA/target interaction in vertebrates, one would expect that the target site sequence in the 3 0 UTR of the target transcript is also conserved across species.
Interestingly, the same miRNA ortholog (miR-137) has been suggested by Shin et al. to control adipogenesis in human adipose cell lines, also by targeting the same orthologous target transcript (cdc42) (Shin et al., 2014).
When there is a lack of evidence from comparative studies of particular miRNA-target interactions, direct measurements of the expression changes in the predicted targets could add evidence that they are true targets. One predicted DE miRNA-target gene pair did reveal reciprocal change in abundance in one group comparison (ssa-miR-203a-3p and srebp-1), whereas the others did not. Nonetheless, the negative regulation mechanism executed by the miRISC when directed to a target gene by the guide miRNA depends on the kind of Argonaute homologue that are part of RISC. There are several Argonaute proteins in fish, and whereas one slices the target transcript (leading to degradation of the target mRNA), the others lead to repression of target gene protein expression by, e.g., translational inhibition. The Argonaute proteins are not well characterized in Atlantic salmon, but investigations in other teleosts indicate that cleavage of the target mRNA is not the dominant regulatory mechanism in fish (Chen et al., 2017). Thus, although one predicted DE miRNA-target gene interaction showed signs of reciprocal change in abundance in the groups compared as expected for RISCmediated degradation of target mRNA, the absence of such relationship in the other predicted DE miRNA-target gene pairs does not rule out that they are true targets. This is also the case for the two genes acox1 and cp1 as the mRNA expression, not protein expression, was investigated here.
Some orthologues of the DE miRNAs identified are known to be important, not only in lipid metabolism and adipogenesis but also in low chronic inflammatory processes associated with pathological accumulation of lipids in vertebrates (Arner & Kulyte, 2015). The 10FO diet group, and particularly the 10FO/LOW group, showed a significantly lower percentage of the anti-inflammatory EPA (Table 3).
Reduced omega-3 levels and increased proinflammatory omega-6 fatty acid levels may be associated with metabolic imbalance in the liver (Scorletti & Byrne, 2013). Interestingly, the miRNAs ssa-miR-125a-3p and ssa-miR-221 and the three miR-146 family members are all associated with inflammation in the adipose tissue (Arner & Kulyte, 2015). For example, Mir-146a is shown to inhibit oxidized low-density lipoprotein-induced lipid accumulation and inflammatory response by targeting toll-like receptor 4c (Chen et al., 2016;Yang et al., 2011). All these miRNAs, as well as the teleost-specific ssa-miR-462b-5p, showed changed expressions in the inflammatory phase of viral disease in Atlantic salmon (Woldemariam et al., 2020). Whether the observed differential expression in these particular miRNAs and in the relative level of EPA could influence the immune response capacity of the groups would be an interesting topic for future work.
In summary, applying the 12 DE miRNAs as an input in target gene predictions, a limited number of key genes in the Atlantic salmon hepatic lipid metabolism were predicted as putative targets. One of these miRNA-target gene interactions was supported by a similar study in mammals. More experimental studies in Atlantic salmon are, nonetheless, needed to validate and fully understand the predicted miRNA-target gene interactions. Such knowledge may further help understand whether the identified DE miRNAs participate in the regulatory networks that control lipid metabolism in Atlantic salmon.