Abstract Simple Summary Smoltification is a developmental process that preadapts Atlantic salmon for a life in seawater. Suboptimal smoltification and poor timing of transfer to seawater is associated with increased mortality. MicroRNAs (miRNAs) are small non-coding genes. They regulate gene expression post-transcriptionally as part of the miRNA induce silencing complex (miRISC) where they guide miRISC to particular mRNAs (target genes). The aim of this study was to identify Atlantic salmon miRNAs expressed in liver that are associated with smoltification and adaptation to seawater as well as to predict their target genes. In total, 62 guide miRNAs were identified, and by their expression patterns they were clustered into three groups. Target gene predictions followed by gene enrichment analysis of the predicted targets indicated that the guide miRNAs were involved in post-transcriptional regulation of important smoltification associated biological processes. Some of these were energy metabolism, protein metabolism and transport, circadian rhythm, stress and immune response. Together, the results indicate that certain miRNAs are involved in the regulation of many of the important changes occurring in the liver during this developmental transition. Abstract Optimal smoltification is crucial for normal development, growth, and health of farmed Atlantic salmon in seawater. Here, we characterize miRNA expression in liver to reveal whether miRNAs regulate gene expression during this developmental transition. Expression changes of miRNAs and mRNAs was studied by small-RNA sequencing and microarray analysis, respectively. This revealed 62 differentially expressed guide miRNAs (gDE-miRNAs) that could be divided into three groups with characteristic dynamic expression patterns. Three of miRNA families are known as highly expressed in liver. A rare arm shift was observed during smoltification in the Atlantic salmon-specific novel-ssa-miR-16. The gDE-miRNAs were predicted to target 2804 of the genes revealing expression changes in the microarray analysis. Enrichment analysis revealed that targets were significantly enriched in smoltification-associated biological process groups. These included lipid and cholesterol synthesis, carbohydrate metabolism, protein metabolism and protein transport, immune system genes, circadian rhythm and stress response. The results indicate that gDE-miRNAs may regulate many of the changes associated with this developmental transition in liver. The results pave the way for validation of the predicted target genes and further study of gDE-miRNA and their targets by functional assays. Keywords: smoltification, seawater adaptation, microRNAs, small-RNA sequencing, liver, Atlantic salmon, microarray transcriptome 1. Introduction Atlantic salmon is one of the most successful aquaculture species. The total worldwide production of farmed Atlantic salmon was around 2638 kilotons in 2020 and 1821 kilotons of this was produced in the North Atlantic area [[34]1]. The majority of farmed salmon in the North Atlantic area is produced by Norway and UK (Scotland) (77% and 11%, respectively) [[35]1]. The rapid growth in the salmon aquaculture industry is due to continuous improvement in production practices [[36]2]. However, salmon aquaculture is still facing several challenges including high mortality after sea transfer associated with deficient smoltification and higher susceptibility to infection diseases [[37]3,[38]4,[39]5]. Studies of salmonids smoltification have also revealed changes in expression of genes involved in immune responses [[40]6,[41]7,[42]8,[43]9]. Smoltification (also known as Parr–Smolt transformation) is a complex preparatory developmental process that transforms parr to smolt for a successful life in the marine environment [[44]10,[45]11,[46]12]. Smoltification involves changes in physiology (e.g., increased metabolic rate, increased seawater tolerance and alterations in lipid metabolism) [[47]13], morphology (e.g., smolts acquire silver skin pigmentation and more streamlined body shape) and behavior (e.g., downstream movement and the loss of territorial behavior) [[48]12,[49]14,[50]15]. Smoltification is a highly energy-demanding process and it is associated with decrease in liver glycogen, whole-body lipid content and muscle content in smolt [[51]16,[52]17]. An increase in mitochondrial enzymes activity [[53]18], enzymes of glycolysis, fatty acid and lactate metabolism (i.e., phosphofructokinase (PFK), β-hydroxyacyl-coenzyme A dehydrogenase (HOAD) and lactate dehydrogenase (LDH)) in the liver of Atlantic salmon have also been reported. Elevated levels of thyroid hormones could be responsible for this increase [[54]18,[55]19]. It has also been reported that β-oxidation capacity in liver increased significantly prior to seawater transfer which gives liver an important role in energy production during this period [[56]16]. Citrate synthase activity in liver, gill and kidney was also enhanced during smoltification [[57]18,[58]20] and did not change after seawater transfer [[59]21]. In nature, smolting is stimulated by the increasing day length (photoperiod) in spring and seasonal temperature fluctuations [[60]10]. The aquaculture industry takes advantage of this photoperiod-dependence in the production of seawater-tolerant juvenile salmon [[61]22]. Smolting in salmon aquaculture is artificially achieved by exposing parr reaching a desirable size to a short photoperiod for several weeks and then returning them to continuous light [[62]23]. Optimal smoltification and correct timing of seawater transfer (SWT) are crucial for normal development, growth and health of farmed salmon [[63]24,[64]25]. Various tests for assessing smolt readiness and quality are employed, including salinity tolerance test, seawater challenge hypo-osmoregulatory test (SCHT), mRNA expression and enzymatic activity of gill Na^+/K^+-ATPase (NKA) and measurement of hormones involved in smolting [[65]24,[66]26,[67]27,[68]28]. Salinity tolerance test and SCHT are inexpensive and relative quick to perform. The SCHT is used by many commercial hatcheries in Norway and Canada [[69]24]. Studies showed that despite satisfactory reactions to salinity tolerance test and SCHT, some hatchery-reared smolts perform poorly at sea in terms of survival and growth rate compared to their naturally produced smolts [[70]29]. Additionally, gill NKA activity in freshwater at the peak of smolting does not always predict long-term growth in seawater [[71]30]. In recent years, incomplete smoltification is reported as one of the most important causes of mortality and contributor to disease development in the period following sea transfer in Norwegian salmon farms [[72]5,[73]25]. According to the annual summary of fish health in Norway [[74]25] from the Norwegian Veterinary Institute, 52.1 million of 290 million smolts transferred to sea died prior to harvest in 2020. The high mortality rate of farmed salmon in the period following sea transfer represents a major fish welfare problem as well as a major economic loss for the aquaculture industry. Better measures for optimal smoltification and transfer to sea at the correct time and good follow up during the first period in seawater are required to reduce mortalities and improve welfare of farmed salmon [[75]25]. Research on smoltification in the past 30 years was particularly directed to gain a better understanding of optimal timing for transfer of smoltified juveniles into ocean net pens [[76]3,[77]26]. In recent years, more studies of gene expression associated with smoltification have been carried out to better understand this developmental transition as well as to search for new biomarkers [[78]7,[79]22,[80]31]. MicroRNAs (miRNAs) are single-stranded, non-coding RNAs (typically ~22 nucleotides in length) that regulate a large variety of biological processes at the post-transcriptional level [[81]32,[82]33]. Mature miRNAs are processed by a cascade of several nuclear and cytoplasmic enzymatic processing steps [[83]34]. Most miRNAs are transcribed into long primary miRNAs (pri-miRNAs) that are processed further into hairpin-structured precursor miRNAs (pre-miRNAs) and finally short miRNA duplexes [[84]35]. At the final step, one of the strands of the miRNA duplex (termed as “mature miRNA” or “guide miRNA”) is loaded into and retained in Argonaute (AGO) proteins, forming the miRNA-induced silencing complex (miRISC). The other strand of the miRNA duplex (termed as “passenger miRNA”) is released from AGO proteins and degraded. The guide miRNAs direct the miRISCs to their target messenger RNAs (mRNAs), usually by binding partially to the 3′-untranslated region (3′-UTR) of the target transcripts. This leads to translational repression or degradation of target mRNAs [[85]36]. The guide miRNA may originate from the 5′ end of the pre-miRNA (referred to as “5p”) or the 3′end of the pre-miRNA (referred to as “3p”) [[86]34,[87]37]. Atlantic salmon miRNAs are among the best characterized in teleost [[88]38]. Due to the salmonid specific genome duplication [[89]39,[90]40], the number of miRNA gene families is larger in salmonids than in any other teleost [[91]38,[92]41]. Recently, the transcriptomes from liver, head kidney and gills from different stages of the smoltification process were characterized by full-length error corrected mRNA sequencing. This provided a catalog of genes and splice variants that are expressed in liver during this developmental process [[93]42]. Furthermore, the 3′UTRs from the full-length sequenced mRNAs were analyzed regarding their potential as miRNA targets. This resulted in a comprehensive resource of predicted mRNA targets for any of the mature miRNAs in Atlantic salmon [[94]43]. The miRNA expression analysis often shows that one of the mature miRNAs appears as the highly expressed one compared to the other mature miRNA processed from same precursor. This is a consistent pattern shown in several studies [[95]38,[96]41]. The explanation to this difference in abundance of the two mature miRNAs from one precursor is that the more abundant one is the biologically important guide miRNA incorporated into miRISC while the other one with the much lower abundancy, is the passenger miRNAs that are degraded [[97]37]. Here, in cases where there is a large difference in abundance between miRNAs from same precursor, we refer to the abundant mature miRNAs (10 times higher read counts than its corresponding mature from same precursor) as the biologically important guide miRNAs. Previous mammalian studies have characterized some miRNAs that were highly expressed in liver [[98]44,[99]45], suggesting that they are important regulators in liver development [[100]46], liver homeostasis, ion metabolism and lipid, and cholesterol biosynthesis [[101]47,[102]48,[103]49]. Studies in teleost fish have also reported that miRNAs are involved in immune response [[104]50,[105]51,[106]52] and response to environmental stimuli [[107]53]. A study of hepatic miRNAs in rainbow trout and farmed carp indicated that miRNAs are involved in hepatic energy metabolism [[108]54,[109]55] while a study in Atlantic salmon revealed several miRNAs associated with lipid metabolism in liver [[110]56]. Furthermore, miR-122-5p, miR-8163-3p, miR-148-5p and miR-101-3p have been reported as highly expressed in the liver of marine teleost, indicating that they serve a common important function in this organ [[111]38,[112]57]. Our recent study conducted in the head kidney of Atlantic salmon indicated that miRNAs are involved in post-transcriptional regulation of gene expression during smoltification and adaptation to seawater in this organ [[113]58]. Gene ontology analysis of the predicted target genes showed that they were enriched in head kidney specific biological processes associated with smoltification and seawater adaptation [[114]58]. So far, similar studies have not been carried out in liver, an organ that plays an important role in storage (lipids, carbohydrates, vitamin A and iron) and detoxification in teleost [[115]59]. In addition, the liver has a central position in amino acid and carbohydrate metabolism and in the synthesis and export of many proteins [[116]60]. The aim of the present study was to characterize miRNA expression changes associated with smoltification and adaptation to seawater in the first month following seawater transfer. Investigating the post-transcriptional interaction between miRNAs and their predicted target mRNAs during this critical period of Atlantic salmon life may provide a better understanding of how the fine-tuning of gene expression in liver may help in facilitating this developmental transition. 2. Materials and Methods 2.1. Experimental Fish Trial and Samplings The experiment was carried out at the Nofima’s Research Station for Sustainable Aquaculture (Sunndalsøre, Norway) in accordance with the Guidelines of the EU-legislation (2010/63/EU), as well as with the Norwegian legislation on animal experimentation. The experimental fish were not exposed to any pain or distress. They were solely killed for the use of their tissues in this project and, thus, approval from the Norwegian Food Safety Authority was not required. The experimental fish were from SalmonBreed commercial strain SalmonBreed-Model SB-Optimal. The most important characteristic of this strain is good growth combined with a balanced weighting for health properties through family selection. A total of 70 fish were selected for this experiment. The fish were kept in one tank supplied with running water throughout the experimental period and they were fed commercial dry feed (Skretting, Norway). The fish used in this study were the same experimental fish as described in Shwe et al. [[117]58]. From the start of feeding, the fish were kept in freshwater with an average water temperature of 13 °C and 24 h continuous light. The average water temperature was then dropped to 8 °C 2 weeks before the start of smoltification process. The initial smoltification process started by decreasing daylight from 24 h to 12 h and increasing water temperature from 8 °C to 13 °C for 5 days, followed by 12 °C for 41 days. Subsequently, the daylight was increased to 24 h and the water temperature lowered to 8 °C for the final stage of smoltification ([118]Table 1). The seawater challenge test was performed once a week in the last 3 weeks before SWT using a salinity of 35‰. Seawater challenge test, blood plasma ions (Cl^−, Na^+ and Mg^2+) level and the change to silvery skin color indicated that the experimental fish were smoltified 81 days after onset of the experiment. The smoltified smolts with average weight 72.4 ± 8.7 g were then transferred to seawater. The average weight at 1-week post-SWT and one-month post-SWT was 63.2 ± 8.5 g and 98.4 ± 14.9, respectively. No mortality of smolts was observed during smoltification or after SWT. Table 1. Photoperiod and water temperature during the experimental trial. Experimental Days Hours of Light per Day (h) Water Temperature (°C) Water Type Day 0 24 8 Fresh water Day 1–5 12 13 Fresh water Day 6–47 12 12 Fresh water Day 48–60 24 12 Fresh water Day 61–81 24 8 Fresh water Day 82–111 24 8 seawater [119]Open in a new tab The liver samples were collected at six time points ([120]Table 2), T1: parr, 1 day prior to light treatment, T2: halfway through light treatment at the change from 24 h to 12 h light and temperature to 12 °C (47 days post-onset of light treatment (POL)), T3: three quarters into the light treatment period (67 days POL), T4: smolt, 1 day prior to SWT (81 days POL), T5: 1 week after SWT (88 days POL), and T6: 1 month after SWT (111 days POL). The experimental conditions such as day light, water temperature, average weight of experimental fish and water type at each sampling points are provided in [121]Table 2. Ten fish were euthanized with an overdose of anesthetic metacain (MS-222; 0.1 g/L) and killed by a blow to the head prior to weighing and sampling at each time point. The collected liver samples were frozen immediately in liquid hydrogen and stored at −80 °C. Seven fish from each time points were used for RNA extraction and small RNA sequencing, while RNA from five of these fish was also used for microarray analysis. Table 2. Time-points and conditions where liver samples were collected. Group Sample Collection Time Points Light ^1 Temp. ^2 Weight ^3 Water Type Sampling ^4 T1 Parr, 1 day prior to light treatment 24 8 29.4 ± 5.6 Fresh water Day 0 T2 Halfway into light treatment 12 12 52.6 ± 5.9 Fresh water Day 47 T3 Three quarters into light treatment 24 8 63.9 ± 10.1 Fresh water Day 67 T4 Smolt, 1 day prior to SWT 24 8 72.4 ± 8.7 Fresh water Day 81 T5 One week after SWT 24 8 63.2 ± 8.5 Seawater Day 88 T6 One month after SWT 24 8 98.4 ± 14.9 Seawater Day 111 [122]Open in a new tab ^1 Hours with day light. ^2 Water temperature in degree Celsius (°C). ^3 Average weight in gram of the experimental fish collected at each time points. ^4 Sampling day within the experimental period. 2.2. Total RNA Extraction for Sequencing and Microarray Analysis Total RNA was isolated using the mirVanaTM miRNA Isolation Kit (Ambion, Life Technologies, Carlsbad, CA, USA) according to the manufacturer’s protocol. The RNA quality and quantity were determined using NanoDropTM1000 Spectrophotometer (Nanodrop ND-1000, Thermo Fisher Scientific, Wilmington, DE, USA). The integrity of total RNA (RIN value) was measured using the Agilent 2100 Bioanalyzer in combination with an Agilent 6000 Nano Chip (Agilent Technologies, Santa Clara, CA, USA). Extracted total RNA was stored at −80 °C. 2.3. Small-RNA Library Preparation and Sequencing Library construction and sequencing of 42 liver samples were carried out at the Norwegian High-Throughput Sequencing Centre (NSC; Oslo, Norway). Seven liver samples from each of the time points T1-T6 were selected for sequencing. The NEBnext^® multiplex small RNA Library Prep Set (New England Biolabs, Inc., Ipswich, MA, USA) was used to construct libraries for 42 liver samples in accordance with manufacturers protocol. One µg total RNA from each sample were used as input for preparation of the libraries followed by 5′ and 3′ adapter ligation, reverse transcription, PCR amplification and size selection of 140–150 bp fragments using 6% polyacrylamide gel. Sequencing was performed on a NextSeq 500 from Illumina (Illumina, Inc, San Diego, CA, USA), producing 75 bp single end reads. All sequenced samples have been submitted to the NCBI Sequence Read Archieve Centre (SRA) ([123]https://www.ncbi.nlm.nih.gov/sra accessed 20 April 2022) with accession bioproject number PRJNA665200 and will automatically be released by NCBI at the publication of this study. 2.4. Processing of Small-RNA Reads and DESeq2 Expression Analysis Data processing and quality control of small-RNA reads were performed according to the procedure described in Shwe et al. [[124]58]. The quality of raw reads was checked using FASTQC software (v.0.11.8). The raw reads that passed quality control were trimmed using Cutadapt (v.2.3) Python package (v.3.7.3) [[125]61]. Trimmed reads were processed further by size filtering to discard all reads that were shorter than 18 nucleotides (nts) or longer than 25 nts. An additional FASTQC analysis was performed to make sure that there were no adapter sequences or poor-quality reads in our final data set of clean reads. Two samples from smoltified fish (T4) and saltwater-adapted fish (T6) were used in mirDeep2 analysis (v.0.0.7) [[126]62] for miRNA discovery as described in Woldemariam et al. [[127]38], but no Atlantic salmon miRNAs other than those already described in the Atlantic salmon miRNAome [[128]38] were discovered. Subsequently, clean reads were aligned to the reference index of all known Salmo salar mature miRNAs [[129]38] using STAR aligner software with default parameters except modified with parameter—alignIntronMax 1 (v.2.5.2b) [[130]63]. The output files of STAR alignment (BAM format) were processed further in R-studio using the feature Counts function from Rsubread package (v.1.34.2) to produce matrices [[131]64]. These count tables were used as input in the DESeq2 R package (v.1.24.0) for miRNA differential expression analysis by comparing each of the time points T2, T3, T4, T5 and T6 with T1. DESeq2 performs an internal normalization by estimating the size factor for each sample. The size factor is estimated by first calculating geometric mean for each gene across all samples. The counts for a gene in each sample is then divided by this geometric mean. The median of these ratios in a sample corresponds to the size factor for that sample [[132]65]. All miRNAs with log2 fold-change ≤−1.0 or ≥1.0, Benjamini-Hochberg adjusted p-value ≤ 0.05 and with average normalized read counts > 30 in at least at one comparison were defined as differentially expressed miRNAs (DE-miRNAs). Subsequently, read counts of DE-miRNAs originating from the same precursor were compared. If one showed >10 times more reads than the other mature from the same precursor, this one was assumed to be the biologically active guide miRNA and subsequently used in the in silico target analysis and gene enrichment analysis. Additional unsupervised hierarchical clustering with complete linkage and spearman correlation was performed with DE-miRNA log2-fold changes as input using hclust function from the stats package (v.3.6.1) in R. Heatmap2 from R-package gplots (v.3.0.1.1) was used to plot heatmaps of DE-miRNAs grouped by the hierarchical clustering analysis. 2.5. Microarray Analysis The expression profiling of differentially expressed mRNAs (DE-mRNAs) in the liver of Atlantic salmon was performed at NOFIMA (Ås, Norway) using 44 k DNA oligonucleotide microarray containing 60-mer probes to protein coding genes (Salgeno-2, [133]GPL28080). The oligonucleotide microarray for Atlantic salmon were designed at Nofima and annotated with bioinformatics package STARS [[134]66]. Microarrays were manufactured by Agilent Technologies (Inc., Cedar Creek, TX, USA), and the reagents and equipment were from the same source. One-color hybridization was used, and each sample was analyzed with separate array. Microarray was performed on 28 of the 42 liver samples selected for small-RNA sequencing, representing six time points (T1–T6), with five fish per time point except T5 and T6 where four fish were analyzed. Total RNA (220 ng) from each sample was used as input for cDNA synthesis, amplification and Cy3 labeling of cRNA using a LowInput QuickAmp Labeling Kit according to the manufacturer’s protocol. The labeled/amplified cRNA were purified using Qiagen’s RNeasy Mini Kit (QIAGEN group, Hilden, Germany). The quantity and quality of the purified cRNA was assessed by NanoDropTM1000 Spectrophometer (Nanodrop ND-1000, Thermo Fisher Scientific, Wilmington, DE, USA). Cy3-labeled cRNA (1650 ng) was used as input to prepare the hybridization mix for each sample using Gene Expression Hybridization Kit. Slides were hybridized in oven (17 h, 65 C, rotation speed 0.01 g). The hybridized slides were washed and scanned with SureScan Microarray Scanner (Agilent Technologies, Santa Clara, CA, USA). Nofima’s bioinformatic package STARS [[135]66] was used for subsequent data processing of mRNA array data. Differential expression analyses were carried out by comparing each of the time points (T2–T6) with T1. The transcripts/mRNAs with log2 fold-changes ≤−0.80 or ≥−0.80 and p < 0.05 (t-test) were defined as significantly changed and termed differentially expressed mRNAs (DE-mRNAs). Enrichment analysis of DE-mRNAs was performed as described in Krasnov et al. [[136]67] by comparing the numbers of DE-mRNAs per functional category using GO and STARS annotation data sets. Additional pathway analysis was carried out using KEGG annotation data set. Functional categories or pathways with ratio ≥2 and Yates’ corrected chi-square (p ≤ 0.05) were defined as overrepresented in DE-mRNA enrichment analysis. 2.6. In Silico Target Gene Predictions and Enrichment Analysis of Predicted miRNA Targets Prediction of guide DE-miRNA targets was carried out against the DE-mRNAs. Firstly, to provide 3′UTR sequences for in silico prediction, probe sequences of the 5708 DE-mRNAs on the microarray were utilized to identify matching transcripts in the Atlantic salmon full-length (FL) mRNA transcriptome [[137]42]. Analysis was performed with the BLASTN tool in the BLAST+ package (v.2.9.0+) [[138]68]. Match criteria of 90% sequence identity and 91% query coverage were used to classify an alignment of a probe sequence as identifying its matching transcript in the FL-transcriptome. The set of FL-transcripts that were classified as matches to at least one DE-mRNA was used as input in the MicroSalmon GitHub repository ([139]http://github.com/AndreassenLab/MicroSalmon/ accessed 20 April 2022) [[140]43]. This produced a list of transcripts that were identified both as being DE-mRNAs and predicted as targets of the gDE-miRNAs revealed in this study (FL-targets). The gene symbols of FL-targets were retrieved from both MicroSalmon [[141]43] and the Universal Protein Resource (UniProt) ([142]https://beta.uniprot.org/ accessed 20 April 2022) using the FL-target transcripts annotation information as input. Gene ontology enrichment analysis was performed using PANTHER Overrepresentation Test (version 16.0) ([143]http://pantherdb.org/ accessed 20 April 2022) [[144]69]. Homo sapiens was chosen as reference gene list for the enrichment analysis as this is the most complete functional annotated database while teleost is rather incompletely annotated. The annotation data sets GO biological process complete and Reactome pathways were used to identify enriched biological processes (BP) and gene pathways associated with DE-miRNA targets, respectively. Fold enrichment (FE) ≥2 and Fisher’s Exact test with False Discovery Rate (FDR) less than 0.05 as calculated by the Benjamini–Hochberg procedure were used as thresholds in the enrichment analysis. Subclasses that were related to the same functional category or pathway category were grouped together in the PANTHER analysis and sorted by most specific subclasses. The grouped outputs also show the related biological processes ranked from general to specific biological processes (see [145]Supplementary Files in [146]Section 3.6). 3. Results 3.1. RNA Library Preparation and Small RNA Sequencing Total RNA from 42 liver samples collected before, during smoltification and post-SWT was successfully extracted, and all samples were subsequently small-RNA sequenced. The number of raw reads obtained from small-RNA sequencing of liver samples (n = 42) ranged from 5.7 to 12.3 million. The quality-filtered (Phred score >32), adapter-trimmed and size-filtered reads for each sample ranged from 3.3 million to 8.4 million. The clean reads uniquely mapped as mature Salmo salar miRNAs in each sample ranged from 49.9 to 79.6%. An overview of all samples including their RNA concentration, RIN value, read numbers, reads uniquely mapped as mature miRNAs and the SRA accession numbers are given in [147]Table S1. 3.2. miRNAs with Differential Expression Changes in Liver during Smoltification and Post SWT To identify DE-miRNAs, we compared the miRNA expression before smoltification (parr) (T1) against the ongoing smoltification period (T2, T3), smolt (T4) and samples from post-SWT (T5 and T6). An overview of conditions at different sampling points is given in [148]Table 1. A total of 88 miRNAs (Benjamini–Hochberg adjusted p-value ≤ 0.05, log2 fold-change ≤−1.0 or ≥1.0 and average normalized read counts >30) belonging to 47 miRNA families were differentially expressed relative to T1 on at least one of the five time points. The relative expression changes of the 88 DE-miRNAs at each timepoint and the mature miRNA sequences of each of the DE-miRNAs is given in [149]Table S2. A heatmap illustrating changes in expression pattern of all 88 DE-miRNAs at all time points is given in [150]Figure S1. There were 62 among these 88 that were identified as guide DE-miRNAs (gDE-miRNAs), meaning that they were the highly abundant mature miRNA of the two miRNAs originated from the same precursor or both mature miRNAs if present in similar amounts (see definition in introduction and methods). Read counts of the gDE-miRNAs are given in [151]Table S3. The annotated gDE-miRNAs in this study agreed with the previously characterized differences in abundance between miRNAs from same precursor [[152]38,[153]41]. Hierarchical clustering analysis of these 62 gDE-miRNAs revealed 3 major clusters. The results are illustrated in the heatmap in [154]Figure 1. Figure 1. [155]Figure 1 [156]Open in a new tab Heatmap and hierarchical clustering of the 62 guide DE-miRNAs. Each row represents a miRNA, and each column represents the expression changes at each time points relative to T1 (pre-smolt, one day before smoltification). T2–T4 and T5–T6 are relative expression changes during smoltification period and post-SWT period, respectively. The dendrogram and the row side colors on the left show the three major clusters of DE-miRNAs (Cluster 1-orange, Cluster 2-red, Cluster 3-green). The direction of expression changes in terms of log2 fold-change is illustrated by the color key above the heatmap. The annotation (1) indicates liver-specific miRNAs and (2) indicates some miRNAs with large changes from T1 to T4 (smoltified fish). Cluster 1 consisted of 18 gDE-miRNAs belonging to 15 miRNA families ([157]Table 3). This cluster was characterized by larger increases in miRNA expression occurring during smoltification including some gDE-miRNAs with modest changes 1 week post-SWT ([158]Figure 1). The second cluster consisted of 17 gDE-miRNAs from 11 miRNA families including ssa-miR-novel-16 which has so far only been discovered in Atlantic salmon ([159]Table 3). The expression pattern common to cluster 2 gDE-miRNAs was a small upregulation of their expression during smoltification (T2–T4) that continued in a larger increase and peaked post-SWT (T5–T6). The gDE-miRNAs included in cluster 3 were those characterized by a decrease in their expression relative to T1. The expression of some miRNAs in this cluster decreased gradually from T1 to T6 while others decreased at later time points ([160]Figure 1). Common to all these, except for ssa-miR-206-3p, were that the downregulation peaked after SWT. There were 27 miRNAs in this cluster belonging to 19 miRNA families. Notably, six of these miRNAs have only been discovered in Atlantic salmon (ssa-miR-novel-1, ssa-miR-novel-2-5p, ssa-miR-novel-10-3p, ssa-miR-novel-12-5p, ssa-miR-novel-13-5p and ssa-miR-novel-16-3p) ([161]Table 3). Table 3. Overview of differentially expressed guide miRNAs in cluster 1, 2 and 3. Cluster 1. Cluster 2 Cluster 3 ssa-miR-1-3p ssa-let-7e-5p ssa-let-7b-3p ssa-miR-15d-5p ssa-miR-19a-2-3p ssa-let-7j-3p ssa-miR-15e-5p ssa-miR-29b-1-5p ssa-miR-15a-3p ssa-miR-17-5p ssa-miR-93a-5p ssa-miR-15bf-5p ssa-miR-18bc-5p ssa-miR-93b-5p ssa-miR-15e-3p ssa-miR-20a-5p ^2 ssa-miR-101b-3p ^2 ssa-miR-30a-1-2-5p ssa-miR-101a-3p ^1 ssa-miR-146a-5p ssa-miR-30a-3-4-5p ssa-miR-106a-5p ssa-miR-146b-5p ssa-miR-30e-5p ssa-miR-106b-5p ssa-miR-200b-3p ssa-miR-122-1-3p ssa-miR-125b-1-3p ssa-miR-200c-3p ssa-miR-122-2-3p ^1,2 ssa-miR-130a-2-3p ssa-miR-200d-3p ssa-miR-203b-3p ssa-miR-142b-5p ssa-miR-218a-5p ssa-miR-205b-5p ssa-miR-146d-5p ssa-miR-218b-5p ssa-miR-206-3p ssa-miR-153a-3p ^2 ssa-miR-221-3p ssa-miR-210-1-5p ssa-miR-216c-3p ssa-miR-222a-5p ssa-miR-214-1-2-3p ssa-miR-462b-5p ssa-miR-222b-3p ssa-miR-214-3-3p ssa-miR-551a-3p ssa-miR-novel-16-5p ssa-miR-301a-5p ssa-miR-551b-3p ssa-miR-455-3p ssa-miR-457ab-5p ^2 ssa-miR-8163-3p ssa-miR-8163-5p ^1,2 ssa-miR-novel-1-3p ssa-miR-novel-2-5p ssa-miR-novel-10-3p ssa-miR-novel-12-5p ssa-miR-novel-13-5p ssa-miR-novel-16-3p ^2 [162]Open in a new tab The annotation (^1) indicates liver-specific miRNAs and (^2) indicates miRNAs with large changes from T1 to T4. 3.3. Liver Specific DE-miRNAs, ARM-SHIFT and Potential Biomarker miRNAs Three of the DE-miRNAs (ssa-miR-101a-3p, ssa-miR-122-2-3p and ssa-miR-8163-5p, [163]Table 3) are among those highly expressed in the liver of teleost fish and are assumed to have liver-specific functions [[164]38,[165]57]. The expression of the liver-specific ssa-miR-101a-3p (cluster 1, [166]Figure 1) showed an increase during smoltification and 1 week post-SWT followed by a slight decrease in expression 1-month post-SWT. The liver-specific ssa-miR-122-2-3p (cluster 3, [167]Figure 1), on the other hand, showed a decrease in expression during smoltification followed by an even larger decrease in expression after SWT. The liver-enriched miRNA, ssa-miR-8163-5p, also belonged to cluster 3. This miRNA was characterized by a decrease in expression but differed from ssa-miR-122-2-3p by peaking 1 week post-SWT (T5, [168]Figure 1). Ssa-miR-novel-16-5p increased in expression during smoltification and post-SWT while ssa-miR-novel-16-3p decreased in its expression across all time points. Interestingly, the mature 5p and 3p showed a significantly inverse correlation relationship from T1 to T6 with Spearman’s rho coefficient of −0.83 and p = 0.04. Taken together, this revealed that there was a change of arm dominance of the mature miRNAs processed from the miR-novel-16 precursor. The change was from 3p being the major expressed mature miRNA at pre-smolt stage (433 normalized read counts at T1, 69 normalized read counts at T6) to 5p being the major expressed miRNA post-SWT (85 normalized read counts at T1, 197 normalized read counts at T6). Such changes are referred to as arm shifts. In this case, the arm shift was occurring gradually over the experimental timepoints measured. The expression changes of some gDE-miRNAs (miR-20a-5p, miR-153a-3p, miR-101b-3p, miR-457ab-5p, miR-novel-16-3p, miR-122-2-3p and miR-8163-5p) were rather large from T1 to T4 (smoltified fish, 1 day prior to SWT) ([169]Figure 1), indicating potential biomarkers for identifying smoltified fish. On the other hand, eight miRNAs (miR-140-5p, miR-107-3p, miR-103-3p, miR-130a-3p, miR-148a-3p, miR-199a-3p, miR-22a-3p and miR-21a-5p) showed stable expression across all time points. These eight miRNAs would be suitable reference miRNAs in RT-qPCR analysis of miRNAs changing expression over this developmental transition. Three of these miRNAs miR-183, miR-140 and miR-107 have previously been validated as stable miRNAs suitable as references in RT-qPCR analysis [[170]57].