Gill Transcriptome Sequencing and De Novo Annotation of Acanthogobius ommaturus in Response to Salinity Stress

Acanthogobius ommaturus is a euryhaline fish widely distributed in coastal, bay and estuarine areas, showing a strong tolerance to salinity. In order to understand the mechanism of adaptation to salinity stress, RNA-seq was used to compare the transcriptome responses of Acanthogobius ommaturus to the changes of salinity. Four salinity gradients, 0 psu, 15 psu (control), 30 psu and 45 psu were set to conduct the experiment. In total, 131,225 unigenes were obtained from the gill tissue of A. ommaturus using the Illumina HiSeq 2000 platform (San Diego, USA). Compared with the gene expression profile of the control group, 572 differentially expressed genes (DEGs) were screened, with 150 at 0 psu, 170 at 30 psu, and 252 at 45 psu. Additionally, among these DEGs, Gene Ontology (GO) analysis indicated that binding, metabolic processes and cellular processes were significantly enriched. Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways analysis detected 3, 5 and 8 pathways related to signal transduction, metabolism, digestive and endocrine systems at 0 psu, 30 psu and 45 psu, respectively. Based on GO enrichment analysis and manual literature searches, the results of the present study indicated that A. ommaturus mainly responded to energy metabolism, ion transport and signal transduction to resist the damage caused by salinity stress. Eight DEGs were randomly selected for further validation by quantitative real-time PCR (qRT-PCR) and the results were consistent with the RNA-seq data.


Introduction
The salinity is one of the most significant factors for regulating distribution, abundance and diversity of aquatic animals [1], and it is also one of the main environmental factors that exerts selective pressure on aquatic organisms. The sea surface salinity would rise with the increase of temperature under global warming [2] Changes in environmental salinity will directly affect osmotic pressure of aquatic organisms [3], which has been predicted to have a profound impact on fish at the molecular and cellular level. Fish need to make adaptive changes to maintain physiological function, and conduct compensatory adjustment through changing distribution range [4].
Acanthogobius ommaturus is a large, demersal fish of the Gobiidae, which is economically important across its distribution area [5]. It is a euryhaline fish that could adapt rapidly and maintain homeostasis in a wide range of salinities (from freshwater to salinity 30) [6]. Studies on A. ommaturus mainly focused on the comparative analysis of individual fecundity and morphology [7], fishery biology [8], name of species [9], and genetic diversity and structure by molecular markers [10]. However, the potential mechanism of salinity adaptation in A. ommaturus, especially at the molecular level, such as comprehensive molecular pathway response, has not been well studied. It is worth noting that the Acanthogobius flavimanus, which is closely related to the A. ommaturus and distributed in East Asia, has invaded California, USA [11,12] and southern Australia [13]. As a euryhaline fish with strong viability and rapid growth, A. ommaturus may be also capable to invade other waters leading to the imbalance or collapse of the local marine ecosystem. Therefore, it is very necessary to conduct the potential mechanism of salinity adaptation of A. ommaturus.
With the development of high-throughput sequencing technology, RNA-seq provides an opportunity to study transcriptome variation in organisms under different environmental conditions. Transcriptome sequencing can not only detect almost all effective genes expressed in specific cells or organs, but also provide a comprehensive understanding of the regulatory mechanisms involved in specific biological processes based on the structure and function of differential genes [14]. In addition, transcriptome sequencing allows simultaneous analysis of all physiological process, including metabolism [15], protein homeostasis [16] and other regulatory cellular processes [17]. Thus, transcriptome technology addresses key gene expression patterns of non-model organisms in specific environments and enables detection of unknown genes and discovery of new transcripts. In recent years, transcriptome technology has been widely used in transcriptome assembly and annotation of many non-model fish, such as Megalobrama amblycephala [18], Pelteobagrus fulvidraco [19] and Atlantic salmon [20].
In this study, to further know about the osmoregulatory mechanisms of A. ommaturus in response to salinity stress, RNA-Seq was used to sequence transcripts of gill tissues of A. ommaturus under salinity stress. This is the first transcriptome study of A. ommaturus, which will provide important information for enriching its genetic resources. Moreover, the main purpose of this study is to explore and identify the genes and pathways of gill organs that play a potential role in salt tolerance by using RNA-Seq.

Ethics Statement
All animal experiments were conducted in accordance with the guidelines and approval of the respective Animal Research and Ethics Committees of Ocean University of China. The field studies did not involve endangered or protected species. In addition, frost anesthesia was performed to minimize the suffering of all A. ommaturus specimens.

Experimental Design and Sampling
Acanthogobius ommaturus specimens (initial weight 41.88 ± 5.52 g, initial length 19.23 ± 1.91 cm) were collected from the coastal water of Qingdao, China. All specimens were dispatched into four aquariums (80 × 60 × 40 cm) with recirculating aerated natural seawater (25 psu) at a density of 20 fishes per aquarium for one week of acclimation. During the acclimation and experimental periods, A. ommaturus was fed to satiation with small shrimp twice daily (08:30 and 16:00), and the remaining feed and feces were siphoned out and 1/3 of the water in each aquarium was replaced with pre-aerated water each day. After 7 days of acclimation, healthy fish with strong individual vitality were used for salinity stress experiments. In the manner of acute stress, each box was a salinity gradient, and the four salinity gradients were 0 psu, 15 psu, 30 psu and 45 psu, respectively. In this experiment, since the optimum salinity range of A. ommaturus is 5-15 psu [21], salinity 15 was selected as the control salinity. Three groups of experiments with salinity of 0 vs. 15, 15 vs. 30 and 15 vs. 45 were analyzed. After 48 h salinity stress, three experimental individuals with similar size, vigor and health were immediately euthanized and sampled in each experimental group. Gills tissues were collected and stored at −80 • C for RNA extraction.

RNA Isolation and Illumina Sequencing
Total RNA was extracted from the gill tissues using the TRIzol reagent (Invitrogen, Carlsbad, USA) according to the instructions, and genomic DNA was removed using DNase I (TaKaRa, Otsu, Japan). RNA purity was checked using the NanoPhotometer ® spectrophotometer (IMPLEN, Calabasas, CA, USA). RNA concentration was measured using Qubit ® RNA Assay Kit in Qubit ® 2.0 Flurometer (Life Technologies, Carlsbad, CA, USA). RNA quality was determined by an Agilent 2100 Bioanalyzer.
Only high-quality RNA samples (OD260/280 = 1.9-2.1, OD260/230 ≥ 2.1, RIN ≥ 9.5, 28/18S ≥ 1.0, > 30 µg) were used to construct these quenching libraries. The RNAs from tissues of salinity gradient were pooled in equal amounts and mRNA was extracted from the total RNA using magnetic beads with Oligo (dT) probes. Fragmentation buffer was applied to lyse the mRNA into fragments with a suitable size and the fragmented mRNA was used to construct a cDNA library using TruSeq Stranded mRNA LTSample Prep Kit (Illumina, San Diego, CA, USA). Then, the sequencing was carried out by Novogene Company (Beijing, China) on Illumina Hiseq 2000 platform.

De Novo Assembly and Functional Annotation
All raw reads in FASTQ format quality control relied on NGS QC ToolKit software (v2.3.3, National Institute of Plant Genome Research, New Delhi, India) [22] and clean reads were obtained by removing raw reads containing adapter, ploy-N (N ratio > 10%) and low quality reads (quality scores ≤ 5). The Q20, Q30, GC (Guanine and Cytosine)-content and sequence duplication level of the clean data were calculated. All the downstream analyses were based on clean data with high quality. Transcriptome de novo assembly was carried out with Trinity software (version 2.0.6 Broad Institute, Cambridge, MA USA) with a min_kmer_cov set to 2 by default and all other parameters set to their defaults [23]. After assembly, the transcripts were clustered and the redundant ones were removed and the remaining sequences were defined as unigenes. All of the assembled unigenes were searched against the Nr (NCBI non-redundant protein sequences) database, the eukaryotic orthologous group (KOG) database, COG (Clusters of Orthologous Groups of Proteins) database, Pfam (Protein family) database, SwissProt databases, the Kyoto Encyclopedia of Genes and Genomes (KEGG) database and Gene Ontology (GO) database using BLASTx to identify the proteins that had the highest sequence similarity with the given transcripts to retrieve their function annotations, and a typical E-value cut-off was set at < 1.0 × 10 −5 [24].

Differential Expression Analysis
To investigate the different gill transcriptome responses of A. ommaturus to salinity fluctuation, we analyzed the number and biological functions of differentially expressed genes (DEGs) in three treatment groups. Firstly, we used BWA-mem [25] to map all unigenes to a multicast file. Then, the expression levels of all unigenes were normalized using RSEM [26] and Bowtie2 [27] to determine the number of fragments per kilobase of exon model per million mapped fragments (FPKM). DEGs of three treatment salinity were identified using the edgeR package [28], and |log 2 FC| ≥ 1 (which corresponds to FC = 2) and FDR ≤ 0.05 were used as the filtering thresholds. In addition, in order to further understand the function of DEGs under salinity stress, DEGs were used to determine the Gene Ontology (GO) term and the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway. Blast2GO (https://www.blast2go.com/) software was used to obtain the GO annotations of unigenes for describing biological processes, molecular functions and cellular components [29]. The KEGG Automatic Annotation Server (KAAS) (http: //www.genome.jp/kaas-bin/kaas_main) system was used for pathway reconstruction, respectively [30].

Quantitative Real-Time PCR
In order to verify the reliability of the transcriptome sequencing results, eight differentially expressed genes (DEGs) with high expression levels were randomly selected, and primers were designed with Primer 6.0 (Table S1). Firstly, reverse transcription was carried out from each total RNA sample using the PrimeScript RT reagent Kit with gDNA Eraser (TaKaRa, JPN) with RT Primer Mix of Random 6 mers and Oligo dT Primer. Furthermore, standard curves were constructed to identify the ideal dilution times of cDNA samples and were used as calibrators. A total of 12 cDNA samples were diluted in 10-fold with nuclease-free water and were used as templates for PCR. Furthermore, the qRT-PCR analysis was designed following instructions for SYBR ® Premix Ex TaqTM (Tli RNaseH Plus) RR420A with a StepOnePlus (TaKaRa, JPN). A reaction system of 20 µL was amplified, including 2.0 µL of diluted cDNA template, 10µL of SYBR Premix Ex Taq (2×), 0.4 µL of each of the forward and reverse primers and ROX Reference Dye (50×) and 6.8 µL of nuclease-free water. The amplification processes consisted of a holding stage of 30 s at 95 • C, followed by 40 cycles of 5 s at 95 • C and 35 s at 52 • C. All reactions were performed in triplicates. Melting curve analysis was performed to determine the target specificity, and melting curve temperature was 61.9-86.7 • C. The relative expression levels of all target unigenes were calculated by the 2 −∆∆CT analysis method (∆CT = CT target unigene − CT reference gene , ∆∆CT = ∆CT treatment − ∆CT control ), and β-actin was used as the reference gene for qRT-PCR normalization.

Statistical Analysis
Statistical analyses were performed using SPSS 18.0 software. Data expressed as the mean ± standard deviation (SD) from three independent replicates. Significant differences between samples were analyzed by Duncan's tests at a significance level of 0.05.

Illumina Sequencing and Assembly
In brief, RNA-Seq was carried out on libraries constructed from gill of A. ommaturus. After quality trimming and adapter clipping, results from RNA-Seq produced 12 libraries that consisted of 638,902,551 clean reads with an N50 value equal to 1111. The error rate was less than 0.03, Q30% (the rate of bases for which quality is greater than 20) was over 93%, and GC percentage was around 47-49%, indicating high-quality sequencing and the feasibility of the subsequent analyses. After de novo assembly using trinity software (version 2.0.6 Broad Institute, Cambridge, MA, USA), 131,225 unigenes and 88,228,207 assembled bases were assembled. The overall sequencing results are shown in Table  S2. All raw sequence data were deposited in NCBI Sequence Read Archive (SRA) under accession number SRP258637.

Unigenes Functional Annotation and Classification
Unigenes were subjected to annotation analysis by comparing with Nr, KOG, COG, SwissProt, Pfam, KEGG, and GO databases. Results show that a total of 46,443 unigenes (35.38%) were annotated in at least one database, with 39,920 annotated unigenes (30.42%) had a significant BLAST hit against Nr database. The detailed annotation results are listed in Table 1. For top-hit species matched against Nr database, 51.4% of the matched unigenes indicated similarity with Boleophthalmus pectinirostris (51.4%), followed by Larimichthys crocea (3.8%) and others (31.9%) ( Figure S1).

Analysis of Differentially Expressed Genes (DEGs)
To elucidate the gene expression pattern under salinity fluctuation, we firstly compared the numbers of DEGs of three treatment groups. A total of 150 (59 up-and 91 down-regulated), 170 (39 upand 131 down-regulated), and 252 (159 up-and 93 down-regulated) genes were differently expressed with |log 2 FC|≥1 (which corresponds to FC = 2) and FDR ≤ 0.05 in 0 vs. 15 psu, 15 vs. 30 psu and 15 vs. 45 psu, respectively. As shown in Figure 1, three DEGs were present in all three comparisons.

Analysis of Differentially Expressed Genes (DEGs)
To elucidate the gene expression pattern under salinity fluctuation, we firstly compared the numbers of DEGs of three treatment groups. A total of 150 (59 up-and 91 down-regulated), 170 (39 up-and 131 down-regulated), and 252 (159 up-and 93 down-regulated) genes were differently expressed with |log2FC|≥1 (which corresponds to FC = 2) and FDR ≤ 0.05 in 0 vs. 15 psu, 15 vs. 30 psu and 15 vs. 45 psu, respectively. As shown in Figure 1, three DEGs were present in all three comparisons.

GO and KEGG Pathway Enrichment Analysis of DEGs
To identify the functional changes potentially associated with the salinity changes of A. ommaturus, GO and KEGG pathway enrichment analyses were conducted. Gene ontology analysis of the three comparisons was performed to ascribe GO categories to 150, 170 and 252 DEGs, respectively. Compared to the control group (15 psu), 101 of 150, 109 of 170 and 167 of 252 significant DEGs were classified into 37 GO terms, 34 GO terms and 39 GO terms at 0 psu, 30 psu and 45 psu, respectively. The distribution of GO terms indicated that cellular processes (GO: 0009987) and metabolic processes (GO: 0008152) in biological processes, cells (GO: 0005623) and cell parts (GO: 0033643) in cellular components, and binding (GO: 0005488) and catalytic activity (GO: 0003824) in molecular functions were most significantly enriched under salinity stress. The detailed information regarding significant GO terms is shown in Figure 2.
KEGG pathways provide valuable information for studying the specific biological, metabolic processes and molecular mechanisms under salinity stress in A. ommaturus. With the increase of salinity, DEGs were enriched in 170, 168 and 234 different pathways, among which 7, 9 and 11 pathways (q value < 0.05) were significantly enriched in 0 vs. 15, 15 vs. 30 and 15 vs. 45 groups, respectively. Compared to the control group, most DEGs were enriched in signal transduction pathways, metabolic pathways, digestive and endocrine systems at 0 psu, 30 psu and 45 psu, respectively. For the KEGG pathway analysis, the dominant pathways were as follows: cytokinecytokine receptor interaction (ko04060), metabolism of xenobiotics by cytochrome P450 (ko00980) and proximal tubule bicarbonate reclamation" (ko04964) in 0 vs. 15, 15 vs. 30 and 15 vs. 45, respectively. The significantly enriched pathways in the top 20 were shown in Figure 3.

GO and KEGG Pathway Enrichment Analysis of DEGs
To identify the functional changes potentially associated with the salinity changes of A. ommaturus, GO and KEGG pathway enrichment analyses were conducted. Gene ontology analysis of the three comparisons was performed to ascribe GO categories to 150, 170 and 252 DEGs, respectively.

Identification of DEGs Related to Salinity Changes
As a non-model species with only limited gene function annotation resources, we based Nr annotation and manual literature searches on the combination of enrichment analysis to further discuss candidate DEGs potentially associated with salinity adaptation and osmoregulation. These DEGs were categorized into three functional categories including energy metabolism, ion transporters and signal transduction ( Table 2) by GO enrichment analysis. Imputed putative functions of these genes are covered in the discussion.

Identification of DEGs Related to Salinity Changes
As a non-model species with only limited gene function annotation resources, we based Nr annotation and manual literature searches on the combination of enrichment analysis to further discuss candidate DEGs potentially associated with salinity adaptation and osmoregulation. These DEGs were categorized into three functional categories including energy metabolism, ion transporters and signal transduction ( Table 2) by GO enrichment analysis. Imputed putative functions of these genes are covered in the discussion.

Validation of Transcriptomic Data by qRT-PCR
Eight differentially expressed genes (DEGs) were randomly selected to validate the reliability of the transcriptome in this study. Melting-curve analysis revealed a single product for all of the tested genes. For these candidate genes, the variation trend in expression was concordant between qRT-PCR data and transcriptomic data, excluding two genes (86284_c2_g1 in 0 vs. 15 group and 91346_c0_g1 in 15 vs. 30 group) did not match perfectly. Therefore, the results indicated the transcriptomic data were credible (Figure 4).

Validation of Transcriptomic Data by qRT-PCR
Eight differentially expressed genes (DEGs) were randomly selected to validate the reliability of the transcriptome in this study. Melting-curve analysis revealed a single product for all of the tested genes. For these candidate genes, the variation trend in expression was concordant between qRT-PCR data and transcriptomic data, excluding two genes (86284_c2_g1 in 0 vs. 15 group and 91346_c0_g1 in 15 vs. 30 group) did not match perfectly. Therefore, the results indicated the transcriptomic data were credible (Figure 4).

Discussion
Salinity is one of the decisive environmental factors affecting fish survival and growth [31], food intake [32], disease [33] and distribution [34]. However, limited evidence is available to explain how salinity stress affects the cellular physiological effects of A. ommaturus. RNA-Seq analysis is considered to be a robust method for evaluating transcriptional responses to different experimental conditions, especially in non-model organisms without reported genomes [35]. In this study, 156, 165, 167 and 151 million clean reads were obtained for four salinity gradients of 0 psu, 15 psu, 30 psu and 45 psu, respectively. For most eukaryotic transcriptomes, sequencing up to 100 million reads can quantify precisely genes and transcripts that have low expression levels [36]. According to the analysis of RNA sequencing data, Q30 is more than 93%, and the N50 length of unigenes is 1111 bp, which proves the reliability of raw data. From the distribution of species annotated by nr database, 20,505 of 39,920 unigenes (51.4%) have been annotated to the species of Boleophthalmus pectinirostris. This is probably due to the closer phylogenetic relationship between two species since both are in the order of Gobioidei, which is also supported the accuracy of raw data. In addition, to further confirm gene expression profiles, qRT-PCR analysis of eight selected DEGs was conducted. Although there were some differences in gene expression between qRT-PCR and RNA-Seq, the overall gene expression trends were consistent. In conclusion, the sequencing results of this study are accurate and reliable, which is suitable for further analysis.
To further explore the research results, we performed GO and KEGG enrichment analysis of differential expressed genes (DEGs). The results of GO enrichment analysis indicated that DEGs were significantly enriched in binding (GO: 0005488), metabolic processes (GO: 0008152), and catalytic activities (GO: 0003824). Previous studies have confirmed these three functions play an important role in osmoregulatory mechanisms. Firstly, the 14-3-3 protein inhibit Ca 2+ activated Cl − channels by binding (GO: 0005488) to calmodulin to maintain osmotic pressure homeostasis in the euryhaline teleost Fundulus heteroclitus [37]. Acipenser brevirostrum appeared to have successfully adapted to higher salinity, probably due to the osmotic pressure balance maintained by the metabolism (GO: 0008152) of non-esterified fatty acids [38]. In addition, previous studies had shown that the catalytic

Discussion
Salinity is one of the decisive environmental factors affecting fish survival and growth [31], food intake [32], disease [33] and distribution [34]. However, limited evidence is available to explain how salinity stress affects the cellular physiological effects of A. ommaturus. RNA-Seq analysis is considered to be a robust method for evaluating transcriptional responses to different experimental conditions, especially in non-model organisms without reported genomes [35]. In this study, 156, 165, 167 and 151 million clean reads were obtained for four salinity gradients of 0 psu, 15 psu, 30 psu and 45 psu, respectively. For most eukaryotic transcriptomes, sequencing up to 100 million reads can quantify precisely genes and transcripts that have low expression levels [36]. According to the analysis of RNA sequencing data, Q30 is more than 93%, and the N50 length of unigenes is 1111 bp, which proves the reliability of raw data. From the distribution of species annotated by nr database, 20,505 of 39,920 unigenes (51.4%) have been annotated to the species of Boleophthalmus pectinirostris. This is probably due to the closer phylogenetic relationship between two species since both are in the order of Gobioidei, which is also supported the accuracy of raw data. In addition, to further confirm gene expression profiles, qRT-PCR analysis of eight selected DEGs was conducted. Although there were some differences in gene expression between qRT-PCR and RNA-Seq, the overall gene expression trends were consistent. In conclusion, the sequencing results of this study are accurate and reliable, which is suitable for further analysis.
To further explore the research results, we performed GO and KEGG enrichment analysis of differential expressed genes (DEGs). The results of GO enrichment analysis indicated that DEGs were significantly enriched in binding (GO: 0005488), metabolic processes (GO: 0008152), and catalytic activities (GO: 0003824). Previous studies have confirmed these three functions play an important role in osmoregulatory mechanisms. Firstly, the 14-3-3 protein inhibit Ca 2+ activated Cl − channels by binding (GO: 0005488) to calmodulin to maintain osmotic pressure homeostasis in the euryhaline teleost Fundulus heteroclitus [37]. Acipenser brevirostrum appeared to have successfully adapted to higher salinity, probably due to the osmotic pressure balance maintained by the metabolism (GO: 0008152) of non-esterified fatty acids [38]. In addition, previous studies had shown that the catalytic function (GO: 0003824) of enzymes regulated ion changes and osmotic pressure when crustaceans were exposed to salinity fluctuations [39]. KEGG pathway analysis can enhance our understanding of how salinity fluctuations affect the osmoregulation of A. ommaturus. For the 0 vs. 15 group, cytokine receptor interaction and JAK STAT signaling pathway related to signal transduction was upstream pathways regulating immune and osmoregulation [40]. Steroid hormone biosynthesis and amino acid metabolism associated with metabolic pathways involved in osmoregulation were significantly enriched in the 15 vs. 30 group. Steroid hormone biosynthesis in metabolic pathways played pivotal roles in response to salinity stress and was inextricably linked to other pathways in aquatic animals [41]. The variation of amino acids will directly lead to the change of protein content [38], which facilitates more rapid acclimation to environmental changes than relying solely on enzymatic mechanisms [42]. Regarding the annotated path map ( Figure S2) in the 15 vs. 45 group, pancreatic secretion, mineral absorption, proximal tubule bicarbonate reclamation and bile secretion have a common feature of increasing the ATPase or ATP content to provide energy for the exchange of sodium and potassium ions. In these pathway maps, we found high expression of genes involved in osmotic pressure regulation, such as cystic fibrosis transmembrane conductance regulator (CFTR), Na + -K + -2Clcotransporter 1 (NKCC1), and phosphoenolpyruvate carboxykinase (PEPCK). In addition, the energy metabolism, ion transporters and signal transduction related DEGs and their potential functions of salinity response were discussed below.

DEGs Related to Energy Metabolism
In order to maintain osmotic balance, the osmoregulatory processes are regulated by several kinds of enzymes, and the synthesis and operation of these proteins requires large amounts of energy [4]. Stress induced by salinity changes has been associated with enhanced reactive oxygen species (ROS) generation, causing oxidative damage [43]. Glutathione S-transferase (GST) is a key enzyme of cellular detoxification systems that defends cells against ROS [44]. However, compared with the control salinity, GST was down-regulated at 0 psu and 30 psu in this study. These differences in GST response may highlight differences in detoxification capacity between species and different tissues [45]. We hypothesized that the down-regulated expression of GST at 0 psu and 30 psu was related to the salinity tolerance of A. ommaturus. Glutamine synthetase (GS) activity may facilitate the production of glutamine, proline, and other organic solutes characteristic of osmotic and pH adjustments under stress conditions [46]. GS expression was significantly up-regulated under high salinity stress (30 psu and 45 psu). We speculated that the high expression of GS promoted the production of proline and other amino acids, which was widely used in hyperosmotic fermentation systems as an osmotic pressure protectant [47]. It is worth noting that phosphoenolpyruvate carboxykinase (PEPCK) was up-regulated at 30 psu. PEPCK catalyzes the conversion of oxaloacetate into phosphoenolpyruvate, which is the first rate-limiting step in gluconeogenesis [48]. The increased expression of PEPCK in the gills of A. ommaturus indicated that the endogenous glucose increased by enhancing the gluconeogenesis to provide the extra energy to maintain the ion balance under the change of salinity [48]. In brief, we hypothesized that genes associated with energy metabolism provided energy for cell survival and ultimately increased the ability of A. ommaturus to survive in salinity changes.

DEGs Related to Ion Transporters
Osmotic balance is maintained by the coordinated transport between water and ions in the intestines, gills and kidneys in teleosts [49]. The gill is an osmoregulartory organ; several classical ion transporters and channels were found in this study, such as Aquaporin (AQP1 and AQP3), transient receptor potential cation channel subfamily V member 4 (TRPV4), Na + -K + -2Clcotransporter 1(NKCC1), ATP-sensitive inward rectifier potassium channel 15(KCNJ15) and sodium potassium ATPase beta subunit (NKA) [50][51][52][53]. Tilapia is one of the typical experimental models for euryhaline fish. It would be more informative if we discuss the similarity and differences in the patterns of gene expression of gill ion transporters and channels between tilapia and A. ommaturus. To acclimate in a high salinity environment, NKA transports 3Na + outward in exchange for 2K + , creating low intracellular Na + and a highly negative charge within the cell; the Na + gradient is used to transport Na + , K + and 2Cl − into the cell through a basolateral NKCC1; Cl − then leaves the cells down an electrical gradient through an apical CFTR [54]. In this osmoregulatory mechanism, the up-regulation of NKA, NKCC1a and CFTR in the gills of A. ommaturus does make sense; the similarity in the patterns of gene expression in Oreochromis mossambicus [55] adapted to seawater stress. The currently accepted model for active NaCl secretion by ionocytes in seawater-acclimated teleosts consists primarily of the cooperative action of three major ion-transport proteins: NKA, NKCC1, and CFTR [54]. Under low salinity environment, the basolateral NKA activity is probably involved in driving uptake of NaCl, possibly in conjunction with an apical V-type H + -ATPase, via apical Na + channels (90406_c5_g1, log 2 FC = 0.231) and Cl -/HCO 3 exchangers [56]. The up-regulation of NKA plays an important role in freshwater adaptation of A. ommaturus and O. mossambicus [55]. In the same way, the up-regulation of NKA has been found in Salvelinus namaycush and Salvelinus fontinalis [57], and in Stenogobius hawaiiensis [50] in both high and low salinity environments. NKA provides the major driving force for both ion secretion and absorption [58], this may be closely related to the up-regulated expression of NKA in both high and low salinity environments. Teleost fishes are able to acclimatize to seawater by secreting excess NaCl by means of specialized "ionocytes" in the gill epithelium [59]. The main function of NKCC is ion secretion [59,60], which explains that the up-regulation of NKCC under high salinity environment and down-regulated under low salinity environment in A. ommaturus. Unlike A. ommaturus, the up-regulation of NKCC in O. mossambicus occurred during freshwater adaptation, which was surprising [55]. Moreover, ion absorption function Na + /Clcotransporter (NCC) and Na + /H + exchanger 3 (NHE3), were up-regulated in freshwater adaptation of O. mossambicus [58,60], but this significant phenomenon was not found in A. ommaturus. Although the expression of NHE3 in A. ommaturus was up-regulated in the low salinity environment, the log 2 FC value was only 0.167. The similarity and differences in the patterns of gene expression in salinity adaptation between A. ommaturus and O. mossambicus suggests the likelihood that there is more than one type of ionocyte in most teleost fishes adaptable to freshwater, while ion-secretory ionocytes in seawater seem to be of only one type in all teleosts [55]. Prolactin (PRL) is a freshwater adapting hormone in teleosts that acts by reducing water permeability and promoting ion uptake by osmoregulatory organs, such as gills [61]. PRL release in a hypoosmotic environment is induced by an increase in PRL cell volume brought about by the influx of water via a membrane-bound AQP3 [62] and Ca 2+ influx through the stretch-activated channel TRPV4 [63]. PRL may play an important role in the control of water and electrolyte balance through PRLR expressed in the osmoregulatory organs in the marine teleost [64]. Despite, the significantly up-regulated expression of PRLR instead of PRL, we suspected that the up-regulated expression of AQP3 and TRPV4 at 0 psu were the direct reason that promoted the secretion of PRL and led to the up-regulated expression of PRLR to adapt to freshwater stress. We found that the expression patterns of AQP3 and TRPV4 genes in A. ommaturus and O. mossambicus were similar, both of which were up-regulated under freshwater stress [62,63]. In addition, AQP1 in gills of Acanthopagrus schlegeli [65] and Dicentrarchus labrax [66] decreased with the increase in salinity, which was consistent with the results of this experiment. In the same way, FXYD also plays an important role in osmoregulation [67]. In general, the patterns of gene expression of osmoregulatory mechanisms between A. ommaturus and O. mossambicus were largely similar. The synergistic effect of ion transportations and channels elicits complex regulatory strategies and ultimately increases tolerance to salinity stresses.

DEGs Related to Signal Transduction
The adaptability and stimulatory response of fish to salt stress depends on the effective mechanism of osmosensing and osmotic stress signaling [68]. Furthermore, osmosensory signals are not only rapidly transduced within cells, but may also be amplified and distributed to many types of downstream osmotic effects [68,69]. The cAMP as a signal factor participates in the regulation of osmotic pressure by mediating the increase of Na + uptake by cells in the internal environment [70]. At the same time, the cAMP signaling pathway stimulates the production of arachidonic acid metabolites, which regulates the production of cortisol and glucagon, osmotic regulation and cellular fatty acid signaling in fish [71]. The "cAMP-responsive element modulator (CREM)" and "cAMP-specific 3 , 5 -cyclic phosphodiesterase 4B (PDE4B)" were both down-regulated at 0 psu and up-regulated at 45 psu. We suspected that the down-regulated expression of CREM and PDE4B at 0 psu inhibits Na + uptake and increases water transport, which maintains the balance of osmotic pressure. On the contrary, the up-regulation expression of CEMR and PDE4B at 45 psu enhances metabolic and osmolarity-regulated signals, providing energy and signaling mechanisms for maintaining osmolarity balance. Strangely enough, high affinity cAMP-specific and IBMX-insensitive 3 , 5 -cyclic phosphodiesterase 8A did decrease at 30 psu, whether the specific mechanism of action related to the salinity tolerance of 30 psu is yet to be further explored. Furthermore, prolactin receptor (PRLR) was up-regulated at 0 psu for A. ommaturus. For freshwater adaptation, prolactin (PRL) is known to reduce Na + efflux and water permeability, and is up-regulated under freshwater stress (0 psu) [72], and PRL initiates its role by binding to a specific cell surface PRL receptor (PRLR) [73]. In this way, PRLR participates in osmotic pressure regulation by affecting the binding of PRL hormones, and it indirectly regulates sodium ions to maintain the stability of osmotic pressure [64]. It is predicted that all teleost species may have the PRL/PRLR system to adapt to sudden osmotic changes in the environment, and that PRLs are widely expressed in osmotically regulated organs [64]. These signals have important functions in the signal cascade of osmotic regulation. It can be used to amplify or transmit osmotic sensory signals and then regulate the functional characteristics of various downstream osmotic effects.

Conclusions
In this study, RNA-seq analysis was performed on A. ommaturus exposed to four salinity gradients (0 psu, 15 psu, 30 psu, and 45 psu), and the obtained high quality data could be the best guarantee of the reliability of experimental results. After the verification of this experiment, these changes are related to metabolic processes, ion transport, signal transduction and other characteristics. We also found that the similarity in the patterns of gene expression of gill ion transporters and channels in salinity adaptation between A. ommaturus and O. mossambicus. Compared with high salinity stress, A. ommaturus is better able to adapt to low salinity stress. This information will improve our understanding of salt tolerance of A. ommaturus at the molecular level and provide relevant transcriptome information for further study.

Conflicts of Interest:
The authors declare no conflict of interest.