Integrative microRNAome analysis of skeletal muscle of Colossoma macropomum (tambaqui), Piaractus mesopotamicus (pacu), and the hybrid tambacu, based on next-generation sequencing data

Colossoma macropomum (tambaqui) and Piaractus mesopotamicus (pacu) are good fish species for aquaculture. The tambacu, individuals originating from the induced hybridization of the female tambaqui with the male pacu, present rapid growth and robustness, characteristics which have made the tambacu a good choice for Brazilian fish farms. Here, we used small RNA sequencing to examine global miRNA expression in the genotypes pacu (PC), tambaqui (TQ), and hybrid tambacu (TC), (Juveniles, n = 5 per genotype), to better understand the relationship between tambacu and its parental species, and also to clarify the mechanisms involved in tambacu muscle growth and maintenance based on miRNAs expression. Regarding differentially expressed (DE) miRNAs between the three genotypes, we observed 8 upregulated and 7 downregulated miRNAs considering TC vs. PC; 14 miRNAs were upregulated and 10 were downregulated considering TC vs. TQ, and 15 miRNAs upregulated and 9 were downregulated considering PC vs. TQ. The majority of the miRNAs showed specific regulation for each genotype pair, and no miRNA were shared between the 3 genotype pairs, in both up- and down-regulated miRNAs. Considering only the miRNAs with validated target genes, we observed the miRNAs miR-144-3p, miR-138-5p, miR-206-3p, and miR-499-5p. GO enrichment analysis showed that the main target genes for these miRNAs were grouped in pathways related to oxygen homeostasis, blood vessel modulation, and oxidative metabolism. Our global miRNA analysis provided interesting DE miRNAs in the skeletal muscle of pacu, tambaqui, and the hybrid tambacu. In addition, in the hybrid tambacu, we identified some miRNAs controlling important molecular muscle markers that could be relevant for the farming maximization.


Background
Colossoma macropomum (tambaqui) and Piaractus mesopotamicus (pacu) are fish species that have widely accepted in the consumer market. In addition, they present desirable characteristics for an intensive breeding environment as rapid growth and optimal adaptation to artificial feeding [1].
Noncoding RNAs have become a very important tool for carrying out different types of evolutionary and expression profile experiments. In such way, many databases and protocols for annotation have been developed [2][3][4].
It is known that Hybrid individuals possess desirable characteristics for production, such as high growth rate, higher resistance to disease, and higher quality of meat [5]. For this reason, several producers have chosen to cultivate the hybrid between the induced crossing of the tambaqui female and the pacu male, the tambacu. Tambacu, in spite of the few genetic information, has the capacity to be more resistant to parasites and stress [6]. Furthermore, tambacu is considered a fish with great potential for Brazilian aquaculture, since it presents a high and fast growth rate, including the skeletal muscle and higher resistance to low temperatures, which contribute to increasing their rusticity [7], thus representing an important model to study. The skeletal muscle is directly involved in the growth of the fish, corresponding to about 35-60% of the body weight of the animal [8]. This abundant muscle mass enables the survival of the animals in the aquatic environment and is commercially important for animal production, since it is one of the most important food sources for human diet [9].
With the focus of massive data analysis, non-coding RNAs have been highly explored in recent years considering their strong impact on controlling several biological processes [10,11]. Among non-coding RNAs, miRNAs (miRNAs), a well reported class of small noncoding RNAs, are known to perform a fine regulation of gene activity in a post-transcriptional mode upon their association with RNA-induced silencing complex (RISC) and binding via base-complementarity to target mRNAs, resulting in their translational repression and/or in their degradation [12,13].
Some miRNAs are known to be specifically or highly expressed in the cardiac and/or skeletal muscles and have been dubbed "myo-miRs" [14]. In fish, recent studies have brought information regarding the evolution and genomic organization of myo-miRNAs [15].
Considering the important role of miRNAs in cell physiology, the aim of our study was to evaluate the global miRNA profile in skeletal muscle of pacu, tambaqui, and the hybrid tambacu, to investigate differentially expressed miRNAs and the ontology analysis considering the comparison between the hybrid with and its parental genotypes, to better understand the molecular signaling pathways involved in tambacu muscle development, growth, and maintenance. In this way, the present work can be an important source of information to supports studies that address the advantages of adopting hybrids for cultivation, which are still scarce [7].

Results
Early juvenile fish Colossoma macropomum, Piaractus mesopotamicus, and hybrid tambacu were studied. All the extracted RNA samples were analyzed by NanoVue (GE Healthcare), and Bioanalyzer (Agilent Technologies) and only samples with a RIN ≥ 8 were selected for the sequencing procedures.
After preprocessing steps, the number of reads decreased to approximately 50% due to the removal of low-quality reads and the adapters, and the removal of sequences not matching the minimum/maximum size corresponding to miRNAs (Fig. 1, Table 1, and Additional file 1). We obtained a total alignment average of 51,78%, considering all samples analyzed. As a result of featureCounts processing (see Additional files 2 and 3), a count matrix was generated for each miRNA present in the reference (see Additional file 4). As part of the pipeline, Principal Component Analysis (PCA) (Fig. 2a) and Dispersion Analysis (Fig. 2b) were run to cluster the samples based on the expression values and to observe the quality of the data. We observed the samples separated into clusters among samples, indicating that the different groups really correspond to different genotypes. Also, the Dispersion Analysis showed that below a mean read count of 10, the dispersion of the data increased dramatically. Thus, miRNAs presenting a mean read count < 10 were filtered out to keep only miRNAs not presenting high dispersion levels.
Differentially expressed (DE) miRNAs groups were observed in a row-clustered heatmap containing all the comparison genotype pairs (TC vs. PC, TC vs. TQ, and PC vs. TQ) and all the miRNAs detected as DE (Fig. 3). Based on color histogram, it was possible to observe the variations in terms of expression levels between up-regulated and down-regulated miRNAs throughout the three genotypes analyzed. A relative high number of DE miRNAs are observed in the data. It Is possible to detect such a balance of DE miRNAs, where there is no comparison only with up-or down-regulated miRNAs (Figs. 4 and 5).
DE analysis showed 8 upregulated and 7 downregulated miRNAs for TC vs. PC genotype pair ( Fig. 4a and d, Table 2 and Additional file 5), 14 upregulated and 10 downregulated miRNAs for TC vs. TQ genotype pair ( Fig. 4b and e, Table 3 and Additional file 6), and 15 upregulated and 9 downregulated miRNAs for the PC vs. TQ genotype pair ( Fig. 4c and f, Table 4 and Additional file 7). A full list exclusive for DE miRNAs can be observed on Additional file 8.
According to the Venn diagram data (Figs. 6 and 7), we analyzed the miRNA distribution pattern between the parental species and the hybrid, and it was possible to observe some DE miRNA-sharing between the different genotype's pairs. For example, considering the upregulated set of miRNAs, miR-144-3p and miR-10a-5p appeared to be upregulated in TC compared to both PC and TQ (Figs. 6 and 8). This leads us to think that the upregulation of these miRNAs could be a result of a multifactorial characteristic involving both profiles given that such miRNAs are upregulated only when the contrast involves the hybrid and such miRNAs are not DE between parental species. A similar characteristic was observed in the down-regulated miRNAs, where miR-199-3-3p appears downregulated in TC when compared with both PC and TQ. These miRNAs are considered as TC-exclusive (in terms of expression levels), as they appear to be up-or down-regulated only in the hybrid compared to the parental species, but not between parental species (Fig. 8). Also, we detected some up-or down-regulated miRNAs only in the hybrid, representing a specific characteristic of this genotype.
We then assigned different miRNAs in 6 different categories regarding inheritance characteristics related to expression levels, i.e., (i) TC-exclusive (for those miR-NAs differentially expressed only in TC compared with both parental species), (ii) PC-inherited (for those miR-NAs with expression patterns similar to PC expression levels), (iii) TQ-inherited (for those miRNAs with expression patterns similar to TQ expression levels), (iv) TC vs. PC-exclusive (for miRNAs that are differentially expressed only when TC and PC are involved in comparisons), (v) PC vs. TQ-exclusive (for miRNAs that are differentially expressed only when PC and TQ are involved in comparisons), and (vi) TC vs. TQ-exclusive (for miRNAs that are differentially expressed only when TC and TQ are involved in comparisons). miRNAs are also discriminated in up-(▲) and downregulated (▼) (Fig. 8).

Discussion
Next-generation sequencing has been widely applied for global analysis in several studies involving miRNAs and skeletal muscle. Nachtigall et al. employed nextgeneration sequencing in Nile Tilapia (Oreochromis niloticus), to obtain information regarding evolutionary pathways with emphasis on muscle miRNAs [15]. The authors identified that there are large syntenic blocks in the genome, possibly being linked to a common function of such miRNAs, and as well specific role of miR-499 expression [18]. Gomes et al. also used next generation sequencing to analyze samples of liver and skin from tambaqui (Colossoma macropomum) to characterize and identify the expression levels of miRNAs and the interaction with target genes having D. rerio as reference [19]. The analysis performed showed that although there are tissue-specific miRNAs expression profile, some miRNAs can be shared between different tissues, as liver and skin. The most expressed miRNAs in both tissues enriched signaling pathways that control several biological processes with a large gene network.
Previous studies in our group have also analyzed the expression of some muscle specific miRNAs in the skeletal muscle of fish involving fasting and re-feeding treatments and muscle development [20,21]. The authors observed that some miRNAs (miR-1, miR-133, miR-155, miR-206, and miR-499) presented a possible role in the regulation of factors related to muscle cell proliferation and differentiation and with muscle performance and metabolism, modulating the rate of protein synthesis and degradation.
In the present study, global analysis of miRNAs in the three genotypes identified a small number of differentially expressed miRNAs considering the comparison between the genotypes pairs. The comparison between tambacu and P. mesopotamicus (TC vs. PC) showed 15 DE miRNAs, 8 upregulated and 7 downregulated. Considering the comparison between tambacu and C.macropomum (TC vs. TQ), we observed 24 DE miRNAs, 14 upregulated and 10 downregulated, and considering the comparison between P. mesopotamicus and C. macropomum (PC vs. TQ), we observed 24 DE miRNAs, 15 upregulated and 9 downregulated. Interestingly, the majority of the miRNAs were specific for each genotype pair. No miRNA was shared between the 3 genotype pairs, both up and downregulated. Between the upregulated DE miRNAs, 9 overlapped between PC vs. TQ and TC vs. TQ and 2 overlap between TC vs. PC and TC vs. TQ. Considering the downregulated DE miRNAs, 3   overlap between PC vs. TQ and TC vs. TQ, and only one overlaps between TC vs. PC and TC vs. TQ. No miRNA was differentially expressed between all of the three comparison pairs at the same time. Some miRNAs show characteristics that make it possible to track their inheritance pattern by looking into the different groups. For example, some miRNAs appear to be exclusive for the comparison between TC vs. TQ (miR-187, miR-18a, miR-190a, miR-206-3p, miR-221-3p, miR-27b-3p, miR-29a, miR-363-3p, and miR-727-5p), some of which were exclusive in the comparison between PC vs. TQ (miR-146b, miR-212-5p, miR-2188-3p, miR-26a-2-3p, and miR-489). Interestingly, some miRNAs presented a characteristic of being inherited specifically of one genotype, and also, there were some miRNAs that appeared to be up-or down-regulated only in the hybrid (miR-10a-5p, miR-144-3p, and miR-199-3-3p). These miRNAs presented (up or downregulated) behavior, exclusively in the hybrid, not differentially expressed between the parental genotypes.
Ontology analysis by Enrichr, considering the Danio rerio validated interaction data, showed that the experimentally validated target genes of the DE miRNAs were grouped into several Biological Processes, Cellular Components, and Molecular Functions categories. The most enriched processes were oxygen homeostasis, nuclear  euchromatin and transcriptional repressor activity, and RNA polymerase II activating transcription factor binding, respectively. The more enriched genes in the signaling pathway involved in these processes were alas2, jun, and lmo2. miR-206-3p, identified in the present study, was differentially expressed in the comparison pair TC vs. TQ, targets jun and vegfa genes. Jun is a member of the Activator Protein 1 (AP-1) transcription factor family that regulates cell proliferation and differentiation, apoptosis, cellular migration, inflammation, and cell-cell interaction [22]. On the other hand, vegfa is involved in vascular development and new blood vessel formation [23,24] and stimulates endothelial cell migration by activating AP-1 transcription factor jun [25]. Skeletal muscle is the most abundant source of VEGFA [26,27], and skeletal-muscle-specific VEGFA knockout changed angiogenesis in muscle fibers [28].
As jun and vegfa genes are targets of miR-206-3p and, as this miRNA was observed to be differentially expressed in the comparison pair TC vs. TQ, this is an indication that this miRNA can represent a remarkable characteristic for the hybrid, since these genes are involved in the maintenance of blood irrigation that, in turn, controls the oxygen rates in the tissues.
The miRNA involved in the control of the alas2 and lmo2 genes was miR-144-3p. This miRNA appeared upregulated in the comparison pairs TC vs. PC and TC vs. TQ and did not appear in the comparison between the parental genotypes PC vs. TQ.
ALAS2 is one of the isozymes of ALAS (5-aminolevulinate synthase) involved in vertebrate heme biosynthesis. This gene is expressed preferentially in erythroid cell-specific mitochondrial enzymes [29,30] and catalyzes the biosynthesis of bulk heme for hemoglobin production [31][32][33][34]. ALAS2 is also regulated by hypoxia-inducible factor HIF. Khenchaduri and colaborators observed the overexpression of ALAS2 in cardiac myoblasts submitted to chronic hypoxia, with a corresponding increase in cellular heme levels. They concluded that, similar to erythroid cells, ALAS2 is positively regulated by hypoxia in cardiac myoblasts with an increase in heme levels [35].
According to Zhang et al., the upregulation of alas2 during hypoxia is directly mediated by a transcription factor, hypoxia-inducible factor 1 (Hif1) [36], and an increased Hif-1 that occurs under low oxygen concentration conditions also promotes the activation of genes involved in hypoxia, such as vegf [37]. Moreover, it has been shown that the transcription factor LIM domain only 2 (Lmo2) is required for angiogenesis and is important for angiogenic remodeling of the existing capillary network into mature blood vessels [38][39][40]. Besides, Lmo2 forms common transcription complexes with GATA-binding protein 2 (GATA2) and cooperatively regulate VEGF-induced angiogenesis [41].
Therefore, it is possible that the interaction between miR-144-3p and alas2 and lmo2 genes can regulate important pathways related to muscle vascularization, oxygen homeostasis, and oxidative metabolism. We believe that this is an important characteristic that probably could provide, for the hybrid tambacu, more resistance to environments with lower oxygen content, once this miRNA was upregulated only in the hybrid, in the comparison pairs TC vs. PC and TC vs. TQ.
In our study, miR-144-3p was identified as DE in TC vs. PC and TC vs. TQ, and the miR-206-3p described as DE only in the comparison pair TC vs. TQ demonstrated that these miRNAs may represent an important species-specific feature. The hybrid tambacu is a very popular fish in Brazil, especially in the south and southeast regions and combines important characteristics for commercial fish farms, such as resistance to low temperatures during the tropical winter, similar to pacu, and faster growth, similar to tambaqui [1,6]. Also, juvenile tambacu exhibits a total capacity for compensatory growth when submitted to short periods of food deprivation, followed by refeeding, a fact that could contribute to reducing production costs [42].
Although there are few studies showing the real potential of the hybrid for aquaculture production [42,43], some of them have been done to better understand the genetics and molecular characteristics of this genotype. Recently, Gomes et al. used Next Generation Sequencing technologies for the analysis of muscle and skin transcriptome of C. macropomum and the hybrid tambacu, observed differences in the specificity and gene expression levels in both the muscle and skin of the hybrid tambacu and the parental tambaqui [44]. The authors demonstrated that in the muscle of tambaqui, upregulated genes involved in muscle contraction, such as the myosin family, actin and myomesin, and catalytic protein genes. The hybrid tambacu muscle presented upregulated genes controlling oxidative stress, amino acid metabolism, the ubiquitin family, defense, and heat shock protein (HSP) family genes, involved in stress. In our study, we can emphasize that the interaction of jun and vegfa genes with miR-206-3p (TC vs. TQ) and lmo2 and alas2 genes with miR-144-3p (TC vs. PC and TC vs. TQ) could provide an evolutionary advantage to the hybrid tambacu once these genes are controlling pathways related to blood vessel modulation, oxygen homeostasis, and oxidative metabolism. Taking this into account, the differential regulation of these miRNAs may represent an explanation of why the hybrid has a better adaptation from exposure to hypoxic environments [38,45]. Besides, it could indicate an important function of miRNAs miR-206-3p and miR-144-3p in the control of the robustness and faster growth of the hybrid tambacu [46].
In addition, the most significant molecular function enriched by the targets of the up-regulated miRNAs in the tambacu muscle was transcriptional repressor activity. The inhibition of this function reinforces our hypothesis that these miRNAs could be related to the robustness and faster growth of the hybrid, since their activity may promote enhanced transcriptional activity, which combined with translation, is an essential process for muscle hypertrophy and growth [47].

Conclusions
Using comparative global miRNA analysis in the hybrid tambacu and their parental genotypes Piaractus mesopotamicus and Colossoma macropomum, we demonstrated, for the first time, miRNAs differentially expressed in the skeletal muscle of the genotype comparison pairs (TC vs. PC, Fig. 6 Venn Diagram showing the relationship between up-regulated miRNAs throughout the different groups. The number marked in the overlapping areas shows the common DE miRNAs TC vs. TQ, and PC vs. TQ). Furthermore, we showed that the hybrid expressed miRNAs that control important molecular muscle markers with potential applications in aquaculture management systems. Our study also provides insight for further investigations involving the validation of these miRNAs and their signaling pathway components in muscle phenotype in these genotypes.

Methods
Experimental design, sample collection, Total RNA extraction, and small RNAs sequencing Pure early juvenile fish Colossoma macropomum, Piaractus mesopotamicus, and hybrid tambacu (Juveniles, n = 5 per genotype) were obtained at the Aquaculture Center of UNESP -CAUNESP, and were maintained in aerated tanks in the Agribusiness Technology Agency (APTA), Presidente Prudente, Sao Paulo, until fish attained~10 g. Fish were kept in a natural photoperiod (12 h light/ dark), and water temperature around 28°C. Dissolved oxygen and pH were monitored daily, and ammonia, nitrite, and nitrate were monitored weekly. Fish were fed with Guabi-Pirá 28 fish feed. Excess benzocaine (concentration over 250 mg/L) was applied to euthanize the animals. Afterward, body weight and total length were determined, and fast muscle samples were collected from the epaxial region near the dorsal fin. Muscle  samples were then frozen in liquid nitrogen for molecular analysis. Total RNA was isolated using TRIZOL (Invitrogen), following the manufacturer's specifications. The total extracted RNA was treated with DNase (Life Technologies, USA), following the manufacturer's specifications. All extracted RNA samples were analyzed by NanoVue Spectrophotometer (GE Healthcare) and Bioanalyzer (Agilent Technologies) for checking the total concentration and the RNA Integrity Number (RIN). The RNA extraction resulted in high quality RNA, and the RIN obtained was ≥8. Samples were submitted to the sequencing, comprehending 15 sequenced samples in total. The samples were shipped to LC Sciences (Houston, TX, USA) in dry ice, and the entire process of library preparation and sequencing by Illumina platform was carried out. Libraries preparation were carried out using the TruSeq Small RNA Library Preparation Kit, and samples were run in 50 cycle single-end Illumina HiSeq 2500 Fast Mode. After sequencing, the data were downloaded through a link provided by the company.

MiRNA analysis
The processing pipeline consisted initially of analyzing the data regarding the read quality (Phred Score). For this purpose, the software FastQC (http://www. bioinformatics.babraham.ac.uk/projects/fastqc) version 0.10.1 was used. After obtaining the sequencing quality information, adapters were removed using the software Fastx-toolkit (http://hannonlab.cshl.edu/fastx_toolkit/ index.html). Reads not containing adapters were kept in libraries. Reads that, after removing the adapters, presented less than 17 nucleotides in length were permanently excluded. A filter based on read-quality parameters using the Fastx-toolkit software was also carried out. In such way, 90% of the read content has a phred score of at least 30. Reads not satisfying these quality prerequisites were also permanently excluded. The libraries were then filtered by max length using a custom Perl script (Additional file 9), where reads longer than 26 nucleotides were discarded. The remaining reads were used to obtain an expression profile of known miRNAs, based on known miRNAs lists of Danio rerio, using data from mature and precursor miRNAs obtained from the miRBase database version 22 [4]. For return more alignment than other species, we selected Danio rerio as a reference for this study, which is considered a good biological model. For the differential expression analysis of miRNAs, mature sequences were aligned against the precursor sequences using GMAP [48], using standard parameters, to create an annotation file in GFF format, making it possible to locate the mature sequences in the precursor sequences and differentiate alignments between the 5′ and 3′ miRNAs.
After building the annotation file, the filtered reads were aligned against the precursor references using Hisat2 [49] with standard parameters. The software fea-tureCounts [50] was applied over the alignments to collect the read-count for each miRNA aligned against the precursor reference and matching the GFF annotation file. The counts were then analyzed to detect differentially expressed (DE) miRNAs using the Bioconductor/R package DESeq2 [51].
A priori, we performed contrasts between the hybrid and each parental species, always having the hybrid as a base of comparison (TC vs. PC and TC vs. TQ). A posteriori, for further understanding of the traits inherited by each parental, in terms of miRNA expression profiles, we also have run a contrast analysis between the two parental genotypes. Such a second contrast has pacu as a base of comparison (PC vs. TQ). In this way, three contrasts were obtained: TC vs. PC, TC vs. TQ, and PC vs. TQ (see Additional files 5, 6 and 7).
The list of DE miRNAs detected by DESeq2 analysis was applied to the drawing of Venn Diagrams to visualize the DE miRNA sharing characteristics of each  The R package gplots was applied with the DE miR-NAs list for obtain heatmaps, which gives us a graphical idea of the expression variations between all the contrasts.
Target genes were detected using MirTarBase [16] database and the results were applied to ontology analysis by using Enrichr [17] to retrieve information regarding pathways in which the target genes are involved. The MirTarBase search comprises only experimentally validated interactions classified as "strong evidence".
Interaction networks were assembled using Cytoscape 3.7.1 [52] based on MirTarBase recovered data for a better observation of the interaction between DE miRNAs and their validated targets. Two sets of networks have been assembled: (i) a network containing validated interactions for Homo sapiens, Mus musculus, Rattus norvegicus, and Danio rerio, and (ii) a network containing only Danio rerio validated interactions.
All bioinformatics steps can be observed in a flowchart presented (Additional file 10).