Dietary Immunostimulant CpG Modulates MicroRNA Biomarkers Associated with Immune Responses in Atlantic Salmon (Salmo salar)

MicroRNAs (miRNAs) are key regulators in fish immune responses. However, no study has previously characterized the impact of polyriboinosinic polyribocytidylic acid (pIC) and formalin-killed typical Aeromonas salmonicida (ASAL) on miRNA expression in Atlantic salmon fed a commercial diet with and without immunostimulant CpG. To this end, first, we performed small RNA deep sequencing and qPCR analyses to identify and confirm pIC- and/or ASAL-responsive miRNAs in the head kidney of salmon fed a control diet. DESeq2 analyses identified 12 and 18 miRNAs differentially expressed in pIC and ASAL groups, respectively, compared to the controls. Fifteen of these miRNAs were studied by qPCR; nine remained significant by qPCR. Five miRNAs (miR-27d-1-2-5p, miR-29b-2-5p, miR-146a-5p, miR-146a-1-2-3p, miR-221-5p) were shown by qPCR to be significantly induced by both pIC and ASAL. Second, the effect of CpG-containing functional feed on miRNA expression was investigated by qPCR. In pre-injection samples, 6 of 15 miRNAs (e.g., miR-181a-5-3p, miR-462a-3p, miR-722-3p) had significantly lower expression in fish fed CpG diet than control diet. In contrast, several miRNAs (e.g., miR-146a-1-2-3p, miR-192a-5p, miR-194a-5p) in the PBS- and ASAL-injected groups had significantly higher expression in CpG-fed fish. Multivariate statistical analyses confirmed that the CpG diet had a greater impact on miRNA expression in ASAL-injected compared with pIC-injected fish. This study identified immune-relevant miRNA biomarkers that will be valuable in the development of diets to combat infectious diseases of salmon.


Introduction
Worldwide demand for seafood for human consumption, including a growing contribution from aquaculture (~50% in 2016), continues to climb because of a flat or decreasing global wild fisheries production in the face of rising human population [1][2][3]. Consequently, there is great potential for the aquaculture industry to expand. With a variety of species being farmed, Atlantic salmon (Salmo salar) is one of the most economically important species in aquaculture [4]. Infectious diseases have resulted Figure 1. Overview of experimental design. Following 7 weeks of feeding trial, fish fed both diets were subjected to immune challenge by an intraperitoneal (IP) injection of sterile phosphate-buffered saline (PBS), bacterial antigen Aeromonas salmonicida (ASAL), or viral mimic polyriboinosinic polyribocytidylic acid (pIC). mirVana-prepared head kidney (HK) templates from three of each PBS-, Figure 1. Overview of experimental design. Following 7 weeks of feeding trial, fish fed both diets were subjected to immune challenge by an intraperitoneal (IP) injection of sterile phosphate-buffered saline (PBS), bacterial antigen Aeromonas salmonicida (ASAL), or viral mimic polyriboinosinic polyribocytidylic acid (pIC). mirVana-prepared head kidney (HK) templates from three of each PBS-, pIC-, and ASAL-injected fish (control diet only) were selected for deep sequencing based on the qPCR assessed expression of known pIC (i.e., isg15a, mxb, irf7b) and ASAL (i.e., campb, tlr5a, il1b) immune biomarker transcripts. Selected pIC-and/or ASAL-responsive miRNAs were studied by qPCR using all samples from both dietary groups. *mRNA qPCR analyses were conducted using DNase-treated and column-purified total RNA. At the end of the feeding trial, both dietary groups were subjected to immune challenge by an intraperitoneal (IP) injection (25 gauge needle) of bacterial antigen ASAL or viral mimic pIC. Fish were starved for 24 h, after which 4 fish per tank (8 per treatment) were euthanized with an overdose of MS-222 (400 mg L −1 , Syndel Laboratories, Vancouver, BC, Canada) and dissected for time 0 (i.e., pre-injection) head kidney samples. Formalin-killed typical ASAL was obtained in the form of a vaccine (Furogen Dip, Elanco (formerly Novartis), Charlottetown, PE, Canada). The ASAL solution was prepared as in Hori et al. [17], while the pIC (Catalogue # P0913; Sigma-Aldrich, Oakville, ON, Canada) was diluted in sterile phosphate-buffered saline (PBS; Gibco, Carlsbad, CA, USA) at 2 µg µL −1 for injection. Then, 4-5 salmon per tank (i.e., [8][9] per treatment) were lightly anesthetized (50 mg L −1 of MS-222) and injected with 1 µL of pIC, ASAL or PBS per g of wet mass. Fish were then sampled 24 h post-injection as described above. Body weight, fork length, and liver weight of fish were measured. Head kidney samples (50-100 mg) were collected, flash-frozen in liquid nitrogen, and stored at −80 • C until RNA extraction. This study was carried out in accordance with the animal care protocol 17-77-MR, approved by the Institutional Animal Care Committee of Memorial University of Newfoundland.

RNA Isolation
Total RNAs of all collected head kidney samples were extracted using the mirVana miRNA isolation kit (Ambion/Life Technologies, Carlsbad, CA, USA) according to the manufacturer's instructions. The RNA integrity was verified by 1% agarose gel electrophoresis, and RNA purity was assessed by A260/280 and A260/230 using NanoDrop spectrophotometry (Thermo Fisher, Mississauga, ON, Canada). All RNA samples used in this study showed tight 18S and 28S ribosomal RNA bands and A260/230 ratios greater than 2. Also, A260/280 ratios of most of the samples were higher than 1.9; 3 out of 67 samples had A260/280 ratios between 1.7 and 1.9.

Library Preparation and Deep Sequencing
Prior to the selection of the samples for deep sequencing, aliquots of mirVana-prepared total RNAs from all fish fed control diet were subjected to DNase treatment and column purification, as described in Caballero-Solares et al. [36]. These RNAs were subjected to qPCR analyses of known ASAL-(i.e., cathelicidin antimicrobial peptide b (campb), tlr5a, interleukin-1 beta (il1b)) (Caballero-Solares et al. manuscript in preparation) and pIC-(i.e., interferon stimulated gene 15a (isg15a), interferon-induced GTP-binding protein b (alias myxovirus resistance b, mxb), interferon regulatory factor 7b (irf7b)) [5] responsive immune biomarker transcripts, to ensure the efficacy of the immune challenges and to select representative individuals for deep sequencing. The qPCR analyses of these immune biomarkers were conducted as described in Caballero-Solares et al. [36]. Details on the methods and results for immune biomarker mRNA qPCR are provided in Supplemental Table S1. The mirVana-prepared total RNAs from three of each PBS-, ASAL-, and pIC-injected individuals fed control diet were selected for miRNA sequencing analyses (see Supplemental Table S1 for qPCR-based sample selection). Small RNA library construction and sequencing were performed at the Norwegian Genomics Consortium (NGC)'s Genomics Core Facility. All sequencing libraries were generated using the NEBNext Small RNA Library Prep Set for Illumina (New England Biolabs, Ipswich, MA, USA) with 1 µg of total RNA input, following the manufacturers' instructions. In brief, mirVana-prepared total RNAs were ligated with 3 and 5 RNA adapters, followed by reverse transcription (RT) and PCR enrichment using barcoded RT-primers. The resulting cDNA products were purified using 6% polyacrylamide gels, and size selection of fragments (approximately 145-160 bp) was carried out to enrich small RNAs. The sequencing was performed on a NextSeq 500 instrument (Illumina, Inc, San Diego, CA, USA), producing 75 bp single-end reads.

Analysis of Deep Sequencing Data
The quality of raw sequencing reads (fastq files) was assessed using FastQC toolkit (v.0.11.5; http://www.bioinformatics.babraham.ac.uk/projects/fastqc), to ensure that the quality was satisfactory before adaptor sequences were removed using cutadapt (v.1.13) [37]. The trimmed sequence reads were size-filtered to discard reads that were outside the expected size range of mature miRNAs (18-25 nt). The quality of the trimmed and size-filtered reads was checked by a second FastQC analysis. All deep sequencing reads have been submitted to the NCBI Sequence Read Archive (SRA) database (BioProject PRJNA555179).
The clean sequence reads were aligned to a reference index consisting of all known mature miRNAs in Atlantic salmon [22], using STAR aligner software (v.2.5.2b) [38]. The alignment files (BAM format) were further processed in R using the featureCounts function from the Rsubread package to produce count matrices [39]. These count tables were used as input to test for differential expression of miRNAs using the R package DESeq2 [40]. Differentially expressed miRNAs were identified by comparing the ASAL or pIC groups to the PBS group (control) (n = 3 from each experimental condition). miRNAs were considered to be differentially expressed if they had Benjamini-Hochberg adjusted p-value of ≤ 0.10.

Prediction of Target Genes and Their Functional Annotations
The miRNA target prediction tool RNAhybrid [41] was applied to identify the putative target genes of the pIC-and/or ASAL-responsive miRNAs identified by the DESeq2 analyses. The mature miRNA sequences were tested against 3 UTRs from all Atlantic salmon transcripts in the NCBI Reference Sequence database (Refseq; https://www.ncbi.nlm.nih.gov/refseq/). The following parameters were applied in the RNA hybrid analysis: helix constraint 2-8, no G:U in seed, and a minimum free energy threshold of −18 kcal/mol. Gene ontology (GO) terms of the predicted target genes from Atlantic salmon were obtained from UniProt Knowledgebase (http://www.uniprot.org/). Based on the GO term annotations and published studies, a subset of predicted target genes with functions associated with immune response were identified. Cross-reference links from the UniProt database were further used to retrieve organism-specific pathway annotations from the online resource Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (https://www.genome.jp/kegg/pathway.html).

qPCR Analysis of miRNA Expression
The expression of 15 miRNAs (5 pIC-responsive, 7 ASAL-responsive, 3 commonly responsive to both pIC and ASAL) (see Supplemental Table S2 for qPCR primers), selected from the DESeq2 analyses, was quantified by qPCR using samples from all individuals (i.e., 8-9 samples per treatment). In addition to fish fed control diet, the qPCR experiment also included head kidney samples from fish fed CpG containing diet and subjected to the immune stimulations.
cDNA templates for qPCR were synthesized in 20-µL reactions from 400 ng of mirVana extracted total RNA using miScript II RT Kit (Qiagen, Hilden, Germany) as recommended by the manufacturer's instructions. The cDNAs were diluted by adding 180 µL of RNase-free water (Qiagen) for use in the qPCR assays. PCR amplifications were performed in duplicate using 12.5 µL of 2× QuantiTect SYBR Green PCR Master Mix, 2.5 µL of 10× miScript Universal Primer, 2.5 µL specific forward primer (10 µM), 5 µL RNase-free water (Qiagen), and 2.5 µL of diluted cDNA template representing 5 ng of input total RNA. All qPCR assays were conducted in an AriaMx Real-time PCR System (Agilent Technologies, Santa Clara, CA, USA) using 96-well plates. The real-time analysis program consisted of 1 cycle of 95 • C for 15 min, and 40 cycles of 94 • C for 15 s, 55 • C for 30 s and 70 • C for 30 s, followed by a final melting point analysis.
All forward primers were designed based on the mature sequences of miRNAs of interest (Table 1), while a universal primer, provided by the miScript SYBR Green PCR Kit (Qiagen), was used as a reverse primer in each qPCR assay. Quality testing ensured that a single product was amplified (dissociation curve analysis) and that there was no primer-dimer present in the no-template control except for miR-181a-5-3p. Amplification efficiencies [42] were calculated using cDNA synthesized from head kidney RNA samples (n = 6; 2 of each PBS-, ASAL-, and pIC-injected) that had been pooled post-cDNA synthesis. Standard curves were generated using a 4-5-point 1:3 dilution series. Two miRNAs (miR-25-3p and miR-17-5p), suggested as the most suitable normalizers for miRNA expression in Atlantic salmon [43], were used as normalizers in the current study. These normalizers were expressed stably in our qPCR study (i.e., the geometric mean of normalizers' C T less than 0.3 cycles different for injection-matched groups or diet-matched groups) (see Supplemental Table S3 for normalizer C T values). Agilent AriaMx software v1.5 was applied to obtain C T (or Cq) values. The relative quantity (RQ) of each miRNA was determined using a qBase relative quantification framework [44,45], with normalization to both miR-25-3p and miR-17-5p, and with amplification efficiencies incorporated. For pre-injection samples (i.e., T0 samples), the RQs of each miRNA were calibrated against fish fed the control diet, while for IP-injected groups, the RQs of each miRNA were calibrated against PBS-injected fish fed the control diet. The mirVana-prepared total RNAs from three of each PBS-, ASAL-, and pIC-injected individuals fed the control diet were selected for miRNA sequencing analyses. 2 The total number of reads in raw fastq file for each sample. 3 Total number of reads after removing adapters and filtering reads by size (18-25 nt). 4 Percent of trimmed and filtered reads mapped to the reference miRNAome (i.e., all known mature miRNAs of Atlantic salmon) [22]. 5 The accession number of sequencing results for each sample submitted to the NCBI Sequence Read Archive (SRA) database (http://www.ncbi.nlm.nih.gov/sra). All data were deposited under the BioProject PRJNA555179.

Statistical Analyses
All qPCR data (i.e., RQs) were subjected to Grubbs' test to identify potential outliers and then log 2 -transformation to meet the normality assumption. In total, 19 RQ values were identified as statistical outliers in the entire dataset (i.e., out of 765 RQ values), and excluded from the study. Each miRNA of interest had a minimum of 7 samples per treatment. For pre-injection samples, miRNA expression differences between diet groups were determined using a Student's t-test (p < 0.05). For IP-injected groups, miRNA expression differences between treatments and diets were determined using two-way analysis of variance (ANOVA), followed by a Dunnett's test to assess the effect of PAMPs within each dietary group (i.e., pIC/ASAL vs. PBS), and a Student's t-test to assess the dietary effect within treatment groups (p < 0.05). All of the statistical tests above were performed in Prism v7.0 (GraphPad Software Inc., La Jolla, CA, USA).
Principal coordinates analysis (PCoA), permutational multivariate ANOVA (PERMANOVA), and similarity of percentages analysis (SIMPER) were performed using PRIMER (Version 6.1.15; PRIMER-E Ltd, Ivybridge, UK) to explore the differences in qPCR-analyzed miRNA expression among samples from fish fed different diets (control vs. CpG) and in different treatment groups (pre-injection, PBS-, pIC-and ASAL-injected).

Deep Sequencing and Identification of Differentially Expressed miRNAs
A deep sequencing approach was used to discover pIC-or ASAL-responsive miRNAs in the head kidney of salmon fed the control diet. Table 1 provides an overview of the read numbers obtained from the deep sequencing of the samples used in the present study. The total number of raw reads obtained from sequencing for all samples ranged from 9.6 to 34.5 million. After trimming and filtering, the number of clean reads for all samples ranged from 4.8 to 10.0 million reads. More than 68% of clean reads (i.e., after trimming and size filtering) were mapped to a recent update of Atlantic salmon reference miRNAome (i.e., all known mature miRNAs) [22]. The sequencing results of all samples are available in the SRA database of NCBI under the BioProject PRJNA555179 (Table 1).
DESeq2 analyses (adjusted p-value < 0.10) were applied to identify miRNAs that were pIC-or ASAL-responsive in the head kidney of Atlantic salmon. This revealed 12 mature miRNAs that were significantly upregulated in the pIC group when compared to the PBS-injected control group; the expression of these miRNAs was 1.6-to 14.0-fold higher in the pIC group (Table 2). Only one miRNA (miR-106a-3p) showed decreased expression (−1.9-fold) in the pIC group ( Table 2). The comparison of the ASAL group against the PBS-injected control group revealed 16 significantly up-regulated miRNAs; the expression of these miRNAs was 1.5-to 17.2-fold higher in the ASAL group (Table 3). Two miRNAs (miR-722-3p and miR-727a-3p) had decreased expression (−2.1-fold and −1.8-fold) in the ASAL group (Table 3). In addition, 3 miRNAs (miR-146a-1-2-3p, miR-221-5p, miR-146-5p) were upregulated in both pIC and ASAL groups compared with the PBS controls (Tables 2 and 3). The mature sequences and miRBase identities of all pIC-and/or ASAL-responsive miRNAs are given in Supplemental Table S4. Table 2. pIC-responsive miRNAs in the head kidney of fish fed the control diet identified by DESeq2 (n = 3) and confirmed by qPCR (n = 8-9  Table S4. 2 The mean of normalized read counts for all of the samples included in the comparison. 3 Fold-change (pIC/PBS) was converted from the log 2 fold-change (given by DESeq2 analyses). For the down-regulated miRNA, the fold-change value was inverted (−1/fold-change). 4 Adjusted p-values as determined by DESeq2 analyses. 5 For qPCR fold-change calculation (fish fed the control diet), overall fold up-regulation was calculated as 2 A−B as in Xue et al. [46], where A was the mean of log 2 RQ from the pIC group, and B was the mean of log 2 RQ from PBS group. Underlined qPCR fold-change values indicate statistical significance (p < 0.05). N/A = not applicable. 6 miR-27d-2-5p and miR-27d-1-5p are identical except the length difference (24 vs. 23 nt). Therefore, the qPCR assay was generic to both miRNAs. Table 3. ASAL-responsive miRNAs in the head kidney of fish fed the control diet identified by DESeq2 (n = 3) and confirmed by qPCR (n = 8-9 1 miRNAs with bold font were differentially expressed in both pIC and ASAL groups when compared to the PBS-injected control group. The mature sequences and miRBase identities of each miRNA are given in Supplemental Table S4. 2 The mean of normalized read counts for all of the samples included in the comparison. 3 Fold-change (ASAL/PBS) was converted from the log 2 fold-change (given by DESeq2 analyses). For down-regulated miRNAs, fold-change values were inverted (−1/fold-change). 4 Adjusted p-values as determined by DESeq2 analyses. 5 For qPCR fold-change calculation (fish fed the control diet), overall fold up-regulation was calculated as 2 A−B as in Xue et al. [46], where A was the mean of log 2 RQ from ASAL group, and B was the mean of log 2 RQ from PBS group. Underlined qPCR fold-change values indicate statistical significance (p < 0.05). N/A = not applicable.
The predicted target genes of the pIC-and/or ASAL-responsive miRNAs from the DESeq2 analysis were identified by in silico analysis against the 3 UTRs from the Atlantic salmon transcriptome (i.e., mRNA Refseq database). A total of 1591 genes were identified as putative targets of pICand/or ASAL-responsive miRNAs (Supplemental Table S5). The gene ontology annotations of these genes (retrieved from the UniProt database) revealed 130 of them have immune-relevant functions (Supplemental Tables S6 and S7). Within these 130 immune-relevant predicted target genes, 24 and 54 were unique targets associated with pIC-and ASAL-responsive miRNAs, respectively; 52 target genes were in common (Supplemental Tables S6 and S7). Among the immune-relevant predicted targets of pIC-responsive miRNAs, 27 could be mapped to species-specific KEGG pathways; while 35 could be mapped for predicted targets of ASAL-responsive miRNAs. These KEGG pathways included NOD-like receptor signaling pathway, cytokine-cytokine receptor interaction, necroptosis, Toll-like receptor signaling pathway, apoptosis, C-type lectin receptor signaling pathway, RIG-I-like receptor signaling pathway, and cell adhesion molecules (CAMs) ( Table 4). The two KEGG pathways that had the most target genes assigned in both putative target gene lists were NOD-like receptor signaling pathway and cytokine-cytokine receptor interaction (Table 4). Cytokine-cytokine receptor interaction 7 13 sasa04217 Necroptosis 7 11 sasa04620 Toll-like receptor signaling pathway 6 10 sasa04210 Apoptosis 6 7 sasa04625 C-type lectin receptor signaling pathway 5 5 sasa04622 RIG-I-like receptor signaling pathway 4 7 sasa04514 Cell adhesion molecules (CAMs) 4 2 1 Complete list of immune-relevant predicted target genes of pIC-responsive miRNAs is given in Supplemental Table S6. 2 Complete list of immune-relevant predicted target genes of ASAL-responsive miRNAs is given in Supplemental Table S7.

qPCR Validation of DESeq2-Identified pIC-and/or ASAL-Responsive miRNAs
Fifteen miRNAs (5 pIC-responsive, 7 ASAL-responsive, three commonly responsive to both pIC and ASAL) identified as differentially expressed by DESeq2 were successfully subjected to qPCR analyses to confirm the sequencing results using larger numbers of biological replicates (n = 8-9) than were included in the sequencing study (Tables 2 and 3). qPCR analyses on miRNA expression of fish fed the control diet are discussed below. Fold-change values and significance are summarized in Tables 2 and 3. Among deep sequencing-identified pIC-responsive miRNAs, the results from qPCR analyses agreed well (i.e., statistically significant) with the results from DESeq2 analyses for 7 of the 8 miRNAs tested; while not statistically significant, qPCR for miR-135bd-5p revealed the same direction of change as shown by deep sequencing (Table 2). Among the 10 deep sequencing-identified ASAL-responsive miRNAs subjected to qPCR analyses, 5 of these (50%), namely miR-29b-2-5p, miR-146a-5p, miR-146a-1-2-3p, miR-221-5p and miR-727a-3p, were confirmed (i.e., statistically significant) by qPCR; 4 of the remaining miRNAs showed the same direction of change (i.e., up-or down-regulation) as the sequencing results (Table 3).
pathways (e.g., NOD-like receptor signaling pathway, Toll-like receptor signaling pathway, RIG-Ilike receptor signaling pathway or cytokine-cytokine receptor interaction pathway; see Table 4) that are important to host immune responses to viral and/or bacterial infection. This further confirms that the pIC-and/or ASAL-responsive miRNAs identified herein are relevant to the host-pathogen immune response. The DESeq2 analyses of deep sequencing results were performed with a relatively small number of biological replicates (n = 3); therefore, there was a risk of false positives as well as false negatives. To avoid false positives, fifteen miRNAs identified as differentially expressed by DESeq2 were selected for qPCR confirmation using larger numbers of biological replicates (n = 8-9) than were included in the sequencing study. Among deep sequencing-identified pIC-responsive miRNAs subjected to qPCR analyses, only 1 miRNA (out of 8) did not confirm the DESeq2 results. For deep sequencing-identified ASAL-responsive miRNAs, 5 out of 10 were confirmed (i.e., statistically significant) by qPCR analyses. The library preparations of all samples involved in this study utilized identical chemistry and yielded similar percentages of high-quality reads. Therefore, the relative lower validation level of ASAL-responsive miRNA markers is unlikely to have been caused by the

Deep Sequencing and Identification of Differentially Expressed miRNAs
In the current study, small RNA deep sequencing was used to discover miRNAs potentially involved in the antiviral and antibacterial immune responses in Atlantic salmon. The DESeq2 analyses of the head kidney from fish fed the control diet identified 12 and 18 miRNAs differentially expressed in pIC and ASAL groups, respectively, compared to PBS controls. It is well established that PAMPs can be detected by specific PRRs on or within the host cells, triggering immune responses [12]. In fish, pIC and ASAL have been used as models to study differentially expressed mRNAs associated with antiviral responses [5,[13][14][15] and antibacterial responses [16,17], respectively. Similarly, various PAMPs (e.g., lipopolysaccharide, peptidoglycan, pIC) have been shown to modulate immune-relevant miRNA expression in teleosts [21,[47][48][49]. Here we demonstrated that pIC and ASAL can be used to stimulate the immune responses that lead to changes in expression of many miRNAs known to be involved in host responses to infection.
The predicted target gene analyses of pIC-and/or ASAL-responsive miRNAs revealed that each of the miRNAs could regulate from 2 to 21 genes that have immune-relevant functions (Supplementary  Tables S6 and S7). Some of these predicted target genes were mapped to KEGG pathways (e.g., NOD-like receptor signaling pathway, Toll-like receptor signaling pathway, RIG-I-like receptor signaling pathway or cytokine-cytokine receptor interaction pathway; see Table 4) that are important to host immune responses to viral and/or bacterial infection. This further confirms that the pIC-and/or ASAL-responsive miRNAs identified herein are relevant to the host-pathogen immune response.
The DESeq2 analyses of deep sequencing results were performed with a relatively small number of biological replicates (n = 3); therefore, there was a risk of false positives as well as false negatives. To avoid false positives, fifteen miRNAs identified as differentially expressed by DESeq2 were selected for qPCR confirmation using larger numbers of biological replicates (n = 8-9) than were included in the sequencing study. Among deep sequencing-identified pIC-responsive miRNAs subjected to qPCR analyses, only 1 miRNA (out of 8) did not confirm the DESeq2 results. For deep sequencing-identified ASAL-responsive miRNAs, 5 out of 10 were confirmed (i.e., statistically significant) by qPCR analyses. The library preparations of all samples involved in this study utilized identical chemistry and yielded similar percentages of high-quality reads. Therefore, the relative lower validation level of ASAL-responsive miRNA markers is unlikely to have been caused by the chemistry and/or sequencing-related factors. Rather, it is likely attributed to the difference in the sample sizes between sequencing and qPCR (i.e., 3 biological replicates in sequencing vs. 8-9 biological replicates in qPCR). In order to decrease the likelihood of false negatives, our future miRNA deep sequencing studies could employ higher numbers of biological replicates. For the remainder of the discussion, we will focus on the qPCR-confirmed miRNAs that were associated with antiviral and antibacterial immune responses.

miRNAs Associated with Both Antiviral and Antibacterial Immune Responses in Atlantic Salmon Head Kidney
The qPCR results showed that 5 miRNAs (miR-27d-1-2-5p, miR-29b-2-5p, miR-146a-5p, miR-146a-1-2-3p, and miR-221-5p) were significantly up-regulated by both pIC and ASAL stimulations when compared with the PBS-injection controls. The expression of miR-27d-1-2-5p in Atlantic salmon fry was recently shown to be decreased in fish challenged with IPNV at both 7 and 20 days post-infections [50]. In mice, miR-27 was reported to be down-regulated in multiple mouse cell lines and primary macrophages by cytomegalovirus (i.e., a DNA virus) infection; however, upon overexpression it exerted an antiviral function [51]. Compared with the pIC-induced miR-27d-1-2-5p in the present study, the opposite regulation of miR-27 by cytomegalovirus in mouse cells may be influenced by the virus-host interaction and/or difference in viral or PAMP nucleic acid (DNA virus vs. RNA pIC). In Asian seabass (Lates calcarifer), the up-regulation of miR-27 in the spleen was associated with LPS-induced inflammatory immune response [52]. In agreement with our findings, miR-29b was shown to be up-regulated in zebrafish embryos infected with Salmonella typhimurium, and in adult zebrafish infected with Mycobacterium marinum [53]. In contrast, the expression of miR-29 in Nile tilapia (Oreochromis niloticus) was down-regulated in fish challenged with Streptococcus agalactiae at early infection stages (i.e., 72 h) [54]. Interestingly, in mice, miR-29 was shown to suppress the interferon-gamma (IFN-γ) production by targeting IFN-γ mRNA directly, and miR-29 knockdown mice initiated more potent innate and type 1 helper T cell adaptive responses to intracellular bacterial infection [55]. Recently miR-29b-2-5p was also shown to respond to IPNV challenge in Atlantic salmon [50]. However, the functions of miR-27d-1-2-5p and miR-29b-2-5p in the antiviral and antibacterial responses in teleost fish including Atlantic salmon are yet to be determined.
In mammals, miR-146 was shown to regulate inflammatory responses following TLR-dependent pathogen recognition [56]. A study in human monocytes revealed that miR-146 attenuated TLR and cytokine signaling via a negative feedback loop involving the suppression of IL-1 receptor-associated kinase 1 and TNF receptor-associated factor 6 [57]. In fish, the expression of miR-146 increased in olive flounder (Paralichthys olivaceus) infected with VHSV [58], in Atlantic salmon infected with SAV [19], and in half-smooth tongue sole (Cynoglossus semilaevis) infected with Vibrio anguillarum [59]. In humans, studies have revealed that the expression of miR-221 is up-regulated in several types of cancers and was related to cancer cell proliferation [60][61][62]. Yan et al. [63] evaluated the expression of miR-221 in half-smooth tongue sole after challenge with V. anguillarum, and a head kidney cell line stimulated with different PAMPs. Similar to our study, the expression of miR-221 was increased in the liver and spleen of the infected sole, and the in vitro study indicated that both LPS and pIC up-regulated the expression of miR-221 at 6 h post-stimulation [63]. The expression of miR-221-5p was also evaluated in olive flounder in response to VHSV infection, showing increased transcription 24 h post-injection [58]. In olive flounder, miR-221-5p was predicted to target important immune genes (i.e., cd18 and irf5) [58]. Recently, it was also shown to respond to IPNV infection in Atlantic salmon [50]. Taken together with our current results, it appears that miR-146 and miR-221 are involved in both antiviral and antibacterial immune responses and may play critical immune regulatory roles in Atlantic salmon.

miRNAs Only Associated with Antiviral Immune Response in Atlantic Salmon Head Kidney
Of the 4 miRNAs identified by deep sequencing as being responsive only to pIC, 3 (i.e., miR-30e-1-2-3p, miR-181a-5-3p, and miR-462a-3p) were qPCR confirmed as being significantly up-regulated in fish fed the control diet. Although there have been no studies to date reporting the association of miR-30e with antiviral immune response in Atlantic salmon, a previous study in Atlantic salmon infected with the intracellular bacterium P. salmonis did show increased miR-30e expression in both spleen and head kidney [23]. miR-181 and miR-462 were classified as evolutionarily conserved miRNAs associated with immune response in teleosts following viral challenges [18]. The expression of miR-181a was increased in response to red-spotted grouper nervous necrosis virus in a head kidney cell line of half-smooth tongue sole [49]. Similarly, Andreassen et al. [19] showed that SAV infection in Atlantic salmon caused the up-regulation of miR-181c-5p. The predicted target mRNAs of miR-181c-5p in Atlantic salmon included a number of immune-relevant genes such as neutrophil cytosolic factor 1 (ncf1), irf3, and inhibitor of nuclear factor kappa-B kinase subunit alpha-like (ikbka; alias: chuk) [19]. In mammals, it is suggested that the miR-181 family plays a central role in vascular inflammation by controlling critical signaling pathways (e.g., NF-κB signaling) and targets relevant to immune cell activation and homeostasis [64]. Similar to our findings, miR-462 was shown to be up-regulated in the spleen of flounder challenged with megalocytivirus (a DNA virus) infection [65], RTL-W1 (i.e., rainbow trout liver cell line) and Atlantic cod macrophages stimulated with pIC [21,66], and Atlantic salmon challenged with SAV [19]. Atlantic salmon miR-462a-3p was predicted to target macrosialin/CD68, a SAV-responsive gene [19]. Collectively, these results support the hypothesis that miR-30e-1-2-3p, miR-181a-5-3p, and miR-462a-3p have immune-related functions and possibly play important roles in the antiviral immune response in Atlantic salmon.

miRNAs Only Associated with Antibacterial Immune Response in Atlantic Salmon Head Kidney
Among the 6 miRNAs identified by deep sequencing as being responsive only to ASAL, miR-727a-3p was qPCR confirmed as being significantly down-regulated compared with PBS control in fish fed the control diet. Similarly, the expression of miR-727-3p was reduced in the LPS-stimulated blunt snout bream (Megalobrama amblycephala) [67]. It should be noted that miR-727 is likely a teleost-specific miRNA [68], and the role of miR-727a-3p in the antibacterial immune response in fish is yet to be determined.
Taken together, it seems that the 9 qPCR-confirmed miRNAs identified as pIC-and/or ASAL-responsive in our current study have some conserved immune-related functions and may play important regulatory roles in the antiviral and/or antibacterial immune responses in Atlantic salmon.

CpG Supplement Modulated the Expression of Immune-Relevant miRNAs in ASAL-Treated Atlantic Salmon
Unmethylated DNAs containing CpG motifs are PAMPs that are commonly used as immunostimulants in fish [28][29][30][31][32][33]69]. The immune response induced by CpG is mediated through TLR9, a PRR present on the cell surface or within endosomal compartments of immune cells [29]. Three different classes of CpG ODNs (A-, B-, and C-classes) have been characterized based on the backbone structure and sequence composition [28,69]. In humans, A-class CpG ODNs are known to activate type I IFN response, while B-class CpG ODNs (e.g., CpG ODN 1668, used in the current study) are more potent in B cell stimulation [70]. Similarly, Strandskog et al. [71] showed that A-and C-class CpG-ODNs induced strong IFN α/β activity, while B-and C-class CpG ODNs stimulated proliferation of leukocytes in Atlantic salmon. However, the impact of CpG as a functional feed ingredient on the modulation of immune-relevant miRNA expression in fish including Atlantic salmon had not been explored prior to our study.
To fully investigate the immune-modulating property of the CpG-containing diet in Atlantic salmon, we analyzed 15 DESeq2-identified miRNAs associated with immune response on the pre-and post-PAMP stimulation head kidney samples. The current qPCR analyses of candidate antiviral and antibacterial miRNAs showed that CpG supplementation generally suppressed basal expression of many miRNAs studied (i.e., in pre-injection samples). This may lead to the higher basal expression of genes targeted by these miRNAs. In post-injection groups (i.e., PBS and ASAL), however, many of these miRNAs were up-regulated in fish fed the CpG diet compared with fish fed the control diet. As shown by multivariate statistical analyses, dietary CpG had the most significant impact on the miRNA expression in the ASAL treated fish compared with other injection treatments, while the overall dietary CpG impact on candidate miRNAs expression in the pIC-injected fish was not significant (p (perm) = 0.4461). Seven miRNAs (e.g., miR-146a-1-2-3p, miR-192a-5p, miR-221-5p, and miR-29b-2-5p) were the most significant contributing variables to the dissimilarity between the ASAL-treated fish. As discussed above, many of these miRNAs, as shown in previous studies, are involved in immune responses of teleost fish.
Studies involving CpG administration via IP-injection showed that CpG ODN 1668 enhanced the immune responses of Pacific red snapper against Vibrio parahaemolitycus exposure and rock bream against iridovirus (a DNA virus) infection [31,32]. In contrast, studies in olive flounder revealed that CpG ODN 1668 conferred no protection against VHSV challenge and did not modulate the expression of well-known antiviral genes (i.e., mx and isg15), but elicited strong protection and immune response in fish challenged with a unicellular marine eukaryotic parasite Miamiensis avidus [69]. In our study, dietary CpG seemed to have no impact on the expression of candidate miRNAs in pIC-treated individuals, while inducing several miRNAs in the PBS-and ASAL-treated fish. This suggests that CpG ODN 1668 may modulate ASAL-stimulated antibacterial immune response rather than the pIC stimulated antiviral immune response based on candidate miRNA expression profiles. The selection of a CpG ODN appropriate to the characteristics of a specific pathogen (e.g., bacteria, virus or parasite) may be key to designing diets to improve defense against that pathogen. Finally, although the physiological and health-related consequences of the observed miRNA expression changes caused by the CpG diet remain unclear and require further study, it seems that selections of these miRNAs are suitable as immune-system associated biomarkers.

Conclusions
The present study identified and qPCR confirmed 9 miRNA biomarkers of Atlantic salmon response to pIC and/or ASAL immune stimulations. Many of the miRNAs identified herein are involved in immune responses, as shown in many similar teleost immune/pathogen challenge studies (discussed above). Regarding the immune-modulating properties of CpG diet on Atlantic salmon, we applied candidate miRNA biomarkers associated with immune response (identified in the current study) and evaluated the expression changes in pre-and post-stimulation individuals. CpG ODN 1668-containing diet may be useful in modulating the ASAL-triggered antibacterial immune response but not the pIC-triggered antiviral immune response of Atlantic salmon. Since the current study utilized pIC and ASAL rather than live pathogens, in the future it would be interesting to determine if dietary immunostimulant CpG ODN 1668 could have a protective effect in live bacterial pathogen challenges in Atlantic salmon. Finally, we anticipate that the molecular biomarkers identified herein will also be useful in the future development of functional feeds involving immunostimulants.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4409/8/12/1592/s1. Supplemental Table S1. qPCR analyses of known ASAL and pIC immune biomarker transcripts in the head kidney of fish fed the control diet. Supplemental Table S2. miRNA qPCR primers. Supplemental Table S3. CT values of normalizers (miR-25-3p and miR-17-5p) in all analyzed samples. Supplemental Table S4. Mature sequences and miRBase identities of pIC-and/or ASAL-responsive miRNAs (identified by DESeq2 analyses). Supplemental Table S5. Predicted target genes of pIC-and/or ASAL-responsive miRNAs from the DESeq2 analyses. Supplemental Table S6. Immune-relevant predicted target genes of pIC-responsive miRNAs from the DESeq2 analysis. Supplemental Table S7. Immune-relevant predicted target genes of ASAL-responsive miRNAs from the DESeq2 analysis.