Determination of dosage compensation and comparison of gene expression in a triploid hybrid fish

Transcriptome assembly

To examine the changes in the global transcriptomic profile in triploid of Ctenopharyngodon idellus and Megalobrama amblycephala (GB), we obtained nine liver transcriptomes from maternal Ctenopharyngodon idellus (GC), paternal Megalobrama amblycephalae (BSB), and triploid offspring GB (Fig. 1).

Functional analyses

Using BLASTX (e-value ≤ 1e⁻⁶) against NCBI-NR, Swiss-Prot, Kyoto Encyclopedia of Genes and Genomes (KEGG), Clusters of Orthologous Groups (COG) and Gene Ontology (GO) databases (alignment length ≥100 bp), 28,950 sequences from paternal BSB, 29,110 sequences from maternal GC, and 29,255 sequences from triploid GB were identified as annotated sequences. The sequence distribution of annotated sequences in the above five public databases and the e-value distribution of annotated genes are shown in Additional file 1. After BLASTX alignment, we performed GO analysis (level 2). The distribution of gene annotations showed the function differences between the parents and their hybrids (Additional file 2). To obtain more accurate information about the gene expression in the three species, our next analysis was focused on the 13,893 shared genes (Additional file 3).

Differential expression between diploid and triploid species

To investigate expression level in the two diploid parents and their triploid offsprings, a total of 157,878 contigs from nine individuals were clustered by CD-HIT, and the 95,702 reference transcript contigs were obtained from clustering (Additional file 4). Then, the total reads from the nine samples were mapped to the 95,702 reference transcripts using BLAST-like alignment tool (Blat) (Additional file 5) [33]. According to the mapping results, we detected the silenced genes (GB = 0, GC > 10, and BSB > 10) and novel genes based on the read counts (GB > 10, GC = 0, and BSB = 0) in triploid offspring, the 27 genes appeared to be silenced, and two genes exhibited a novel expression pattern (Additional file 6).

To detect significant differentially expression, false discovery rate (FDR) < 0.001 and the absolute value of log₂ ratio > 1 were used as thresholds in comparison of the two parents and their triploid offsprings. In all comparisons, the percentage of genes showing differential expression between the F₁ triploids and the two parents was asymmetric (P < 0.05; Fisher’s exact test). Comparison of the expression level in the two parents revealed that 2,446 genes were up-regulated in paternal BSB, and 2,376 genes were up-regulated in maternal GC (Fig. 2a and d). We compared the gene expression in paternal BSB and triploid GB, and we determined that 2,138 genes were up-regulated in BSB, and 1,257 genes were up-regulated in GB (Fig. 2b and d). Then, we compared the expression of maternal GC and triploid GB; 2,483 genes were up-regulated in GC, and 1,516 genes were up-regulated in GB (Fig. 2c and d).

To detect whether the phenomenon of dosage effect occurred in triploidstriploid, the comparison of the value of predicted triploid expression level (PT-ELV, also known as in silico mid-parents C₂ + B) and the value of actual triploid expression level (AT-ELV) of GB was performed (see methods). The 4,048 genes (29.1%) had exhibited up-regulated expression in PT-ELV of GB and only 81 genes (0.6%) had shown up-regulated expression in AT-ELV of GB (Fig. 3a and c). The above results were obviously showing that the negative dosage effect of maternal GC-homoeologous chromosomes had occurred in triploid offspring. Based on the existence of dosage effect, we had hypothesized the value of predicted diploid expression level (PD-ELV, also known as in silico mid-parents C + B) and compared it with the AT-ELV. The 2,441 genes that were significantly differentially expressed in triploids included 2,232 (16.1%) up-regulated genes in PD-ELV of GB and 209 (1.5%) up-regulated genes in AT-ELV of GB (Fig. 3b and c). Our results shed insight into that both the mechanism of negative dosage effects and another unknown mechanism result in triploid expression level decreasing to the diploid state.

Expression patterns under dosage effect

As a prerequisite of the dosage effect found in triploid, it shed us insight into the expression level raised from one paternal set of chromosomes and one maternal set of chromosomes in triploid. For better understanding ELD and HEB under dosage effects, we had established 12 categories including mid-parents (XI and XII), up/down expression (I, II, III, IV, V, and VI), and ELD (VII, VIII, IX and X) to assess differential gene expression (see Methods). Among of 13,893 shared genes, 2,749 genes (19.8%) were detected as ELD category (Fig. 4a). Maternal GC-ELD including 1,645 genes (11.8% of all genes, categories IX and X) had exhibited more influence than paternal BSB-ELD (1,104 genes, 7.9% of all genes, categories VII and VIII) in triploid (Fig. 4a). Categories VII and X (GC vs BSB = 1.8 vs 1) represented the up-regulated ELD, while down-regulated ELD (GC vs BSB = 1.3 vs 1) was detected in categories VIII and IX in triploid (Fig. 4a). The results showed that the number of HEB genes was unbalanced in triploid with respect to the original parent was inclined to maternal GC genome (paternal BSB bias vs maternal GC bias = 1,104 vs 1,645) (Fig. 4a). To compare triploid GB with paternal BSB, we examined the 1,536 up-regulated genes (IV, V, VI, X, and XII) and 2,170 down-regulated genes (I, II, III, IX, and XI). Compared with maternal GC, the 1,144 up-regulated genes (IV, V, VI, VII, and XI) and 2,021 down-regulated genes (I, II, III, VIII, and XII) was examined in triploid (Fig. 4a). The gene number related to down- or up-regulation had a global mRNA preference toward down-regulation (up-regulation vs down-regulation = 70 vs 586). In addition, 65.4% (9,083 genes, categories of no changes) showed similar expression levels in the parents.

The expression level of growth genes in the hybrid

After detecting the ELD of growth-regulated genes in triploid, eleven genes exhibited a paternal BSB-ELD, and 13 genes were showed a maternal GC-ELD (Fig. 4a). The percent of maternal GC-ELD (22.8%) of growth genes was higher than that of the total genes (11.8%). The percent of paternal BSB-ELD (19.3%) of growth genes was higher than that of the total genes (8.9%). The percent of parent ELD in growth-related genes was more than other genes in triploid. Eleven genes were considered to be mid-parent genes, and the remaining 12 growth-related genes showed no change in expression levels (Fig. 4a). The 21.1% of growth-related genes in the “No Change” category was lower than the 65.4% of total genes in that category (Additional file 7). These results suggest that there are more changes in growth-related gene expression in triploid than in other gene functions.

Real-time quantitative PCR (qPCR) validation

To validate the quality of RNA sequencing (RNA-Seq) data and the reliability of triploid expression level compared to both parents, we chose 10 representative differentially expressed genes (igfbp2b, igfbp5a, smad7, gdf6a, igf1, ctnnb1, igf2b, ppm1bb, gdf2, and insra) and performed qPCR on biological replicates in triplicate. The same trends in expression levels of these genes were detected by qPCR as were obtained from the RNA-Seq data analysis (Fig. 5). These results indicate that RNA-Seq data and associated analysis methods can be used to accurately detect differentially expressed genes.

Dosage effect in triploid fish

To investigate whether a regulation mechanism was operating on gene dosage in a triploid genome context, 13,893 shared genes were used in our analysis because other genes may be errors in the assembly or the result of differential transcript expression in the nine individual livers. We compared the gene number of AT-ELV and PT-ELV among the total genes, and most of the total differentially expressed genes (4,048, 29.1% in total shared genes) were down-regulated; these results (number of up-regulated genes vs down-regulated genes = 1 vs 49.9), suggesting that dosage effects would occur in 4,048 genes of triploid (Fig. 3). A recent report suggested the silencing of one of three sets of alleles would result in transcript levels in triploid fish being decreased to the diploid state [10]. In contrast to both X chromosomes being partially repressed in Caenorhabditis elegans hermaphrodites (XX), the dosage effect resulted in the silencing of one X chromosome in vertebrates [34]. The existence of two identical sets of chromosomes in the nucleus would induce dosage compensation, which could result in the silencing of one set of maternal chromosomes (GC) in triploid (GB). The similar result was detected in salmon [17, 35]. However, the comparison of AT-ELV and PD-ELV was used to assess the percent of up- and down-regulated genes (see Methods). The results showed that the expression level (AT-ELV) of 2,232 genes was lower than ones’ in the diploid state (PD-ELV), while only 209 genes (AT-ELV) showed higher than ones’ in the diploid state (PD-ELV). That give insight into other mechanisms occurred in triploid offsprings, such as the level of methylation variation accompanied by triploidization, might act on the down-regulation of gene expression of some alleles, results in the expression level in triploid decreasing to lower than that in the diploid state [36]. The above results helped us understanding maternal GC-dosage effect on the parts of genes in triploid offspring.

Homoeolog expression bias and expression level dominance

After charting the dosage effect in triploid fish, we used a novel method to analyze the state of expression levels between triploids and their diploid parents. This is the first report of this phenomenon in triploid fish (previously referred to as genomic dominance in plants [22, 32]). The 12 categories of expression level patterns were described above. Our results showed that the percentage of up-/down-regulated genes was 19.8% (2,749 genes) (Fig. 4a). Our method to analyze the negative dosage effect was feasible. Although we used the AT-ELV as the normalization state, the down- vs up-regulated ratio was 10.7:1 in triploid compared to the AT-ELV with PT-ELV (Fig. 4a). This suggested that the expression level of homoeologous genes was regulated after the genomes merged, which is the potential force behind the differential epigenetic regulation of the hybrid [22, 36]. Therefore, compared with the pattern of no change in initial predictions of triploid expression levels, the number of genes in the “no change” category was reduced because of the more feasible and detailed method that we used for classification based on expression levels.

To detect the global gene expression, the negative dosage effect of silencing of one set of maternal GC homoeologs was used in our analysis. Further analysis of triploid expression level compared with either of the diploid parents demonstrated the preferential transcription of maternal GC homoeologs in triploid (Fig. 4a). This phenomenon was commonly described in polyploidy [37] and refers to the pattern of redundant genes being silenced [38]. Approximately 25% of genes showed evidence of ELD in four allopolyploid cottons based on RNA-Seq data [22]. In triploid Squalius alburnoides, the vasa gene illustrated genome ELD in the gonad, and the β-actin gene exhibited the same phenomenon in the gonad and liver [10]. In addition to ELD, a second phenomenon was also described: middle expression levels were found in the polyploid based on the relative expression levels of the two parents. In our study, the 10,488 genes (75.5%, XI, XII, and No Change categories) showed expression levels that were regulated by homoeologs from both parents (Fig. 4a). These phenomena were always described in hybrids and polyploids based on the total gene analysis [22, 39, 40]. However, the phenomenon of middle expression levels exhibited organ-specific expression. For example, the rpl8 and gapdh genes only show a mid-parents expression level in the liver of triploid individuals [10].

Expression patterns of growth-related genes

The liver plays a major role in metabolism and has a number of functions, including the regulation of growth and development in fish. The study of the expression level of growth-related genes in triploid individuals is central to understanding the mechanism of the hybrid system. Here, we applied next-generation sequencing technology to study the relationship between growth rate and gene expression in a triploid. The dosage effect was evident in the global gene expression in the liver as we showed above. Genome-wide ELD shows maternal GC-HEB in the growth genes of triploid individuals.

In our study, 57 growth-related genes were screened from the categories of global gene expression (Fig. 4a). Four genes (7%) were up-regulated, which was higher than the percentage of total genes (0.5%) (P < 0.05; Fisher’s exact test). For example, igfbp2b and igfbp5a serve as a carrier protein for igf-1, which binds to igf-1 inside the liver, allowing growth hormone to continuously act upon the liver to produce more igf-1 [41]. Up-regulated expression of these genes will help organisms to accumulate and prolong the half-life of the insulin-like growth factors (Table 2; Fig. 5f and g). Another up-regulated gene, igf-1, and a paternal BSB-ELD gene, igf2b, were shown to play roles in the promotion of cell proliferation (Table 2; Fig. 5a and b). The up-regulated expression of igf in triploid was considered to play a crucial role in its faster growth rate relative to diploids [12]. The last up-regulated growth-related gene, insra, is a transmembrane receptor that is activated by insulin and igf, and it belongs to the class of tyrosine kinase receptors (Table 2; Fig. 5h) [42]. The up-regulation of insra resulted in an enhancement of the regulation of glucose homeostasis. Additionally, down-regulated expression of ppm1bb is known to be a negative regulator of cell stress response pathways, and overexpression of this phosphatase is reported to cause cell-growth arrest (Table 2; Fig. 5i) [43].

The other expression patterns were dominance and mid-parents expression level patterns that included 35 growth-related genes (Fig. 4a); these patterns provided insight into new expression level patterns in triploids. For example, paternal BSB-ELD was evident for the gdf6 and gdf2 genes (Table 2; Fig. 5c and e), which are members of the BMP family and the TGF-β superfamily that regulator cell growth and differentiation in both embryonic and adult tissues. These genes also promote bone and joint formation [44]. The hybrid individual had higher expression levels than maternal GC. These small changes in expression level contributed to changes in growth regulation. In addition, compared to the diploid GC, triploid had a significantly higher growth rate [6]. Therefore, these mechanisms might play important roles in the regulation of growth by changing some growth-related gene expression levels. Maternal GC-ELD gene smad7 enhances muscle differentiation and plays a role in the negative feedback of TGF-β signaling (Table 2; Fig. 5j) [45]. These observations agreed with observations from some previous reports of polyploid fish [46, 47]. The middle-parents gene ctnnb1 indicated that its expression was positively regulated by paternal BSB and resulted in up-regulated expression in triploid.

The differences in expression levels in triploid and the inbred diploid parents gave us a platform to investigate the rapid growth in triploid individuals. However, we should also investigate the gene expression changes that indirectly result in a change in growth traits. More research on these subjects will help us understand how the growth-related function was regulated in triploids. However, the observed results suggested that the rapid growth in triploids could be regulated by genes with a negative dosage effect.

Mechanism of various expression patterns

Recent evidence showed that dosage compensation resulted in novel epigenetic regulation in triploids [17]. The current challenge is determining which changes in regulatory mechanisms explain the observed differences in gene expression levels and the evolution of complex phenotypes [35, 48]. Epigenetic instability in polyploids was described recently [49, 50]. Increased gene copy numbers from different species usually lead to changes in gene expression. This change usually destroys the steady state of the regulatory adaptations that were selected in the parents [50]. However, these abundant expression level patterns in polyploids provide important materials for adapting to various situations. The hybrids are likely to display regulatory alterations. These changes involved the silencing or activation of genes and DNA transposition of the Spm/CACTA family; these changes were described in allopolyploids of Arabidopsis thaliana [51, 52]. Our study also showed the activation of two genes in the liver transcriptome (Additional file 6). Possible mechanisms include small inhibitory RNA and epigenetic pathways that mediate the expression levels together with dosage compensation in triploids [35, 53].

Animal material

For this study, all experiments were approved by Animal Care Committee of Hunan Normal University and followed guidelines statement of the Administration of Affairs Concerning Animal Experimentation of China. Experimental individuals were fed in a pool with suitable illumination, water temperature, dissolved oxygen content, and adequate forage for 19 months in the Engineering Center of Polyploidy Fish Breeding of the National Education Ministry located at Hunan Normal University, China. Triploid hybrids of female grass carp (Ctenopharyngodon idellus, GC, Cyprininae, 2n = 48) × male blunt snout bream (Megalobrama amblycephala, BSB, Cultrinae, 2n = 48) were successfully obtained by distant hybridization as a result of human selection (Fig. 1b and c) [6]. The 5S rDNA locus has been used to identify triploid hybrids that possessed 72 chromosomes with two sets from maternal GC and one set from paternal BSB [6]. Triploid hybrid of GC (♀) × BSB (♂) was abbreviated as GB hybrids. Nine individuals (three hybrids and six parents) were collected for our studies. The information about fish samples including body traits (body length, body height, and weight) and DNA content were obtained at the time of the experiment (Additional file 8).

The ploidy levels of the nine individuals were distinguished by a metaphase chromosome assay of cultured blood cells (Fig. 1a). After anesthetizing the fish with 2-phenoxyethanol, liver tissue was excised carefully to avoid gut contamination. The fish were treated humanely. All of the experiments were approved by the Animal Care Committee of Hunan Normal University and the Administration of Affairs Concerning Animal Experimentation guidelines stated approval from the Science and Technology Bureau of China. Samples were cut into small pieces and immediately pulled into RNALater (Ambion, AM7021, USA) at −80 °C following the manufacturer’s instructions. Total RNA was extracted from liver tissue of the BSB, GC, and GB samples. After RNALater was removed, the samples were homogenized using a pestle and mortar. RNA was isolated according to the standard trizol protocol, and agarose gel electrophoresis and the optical density at 260 nm (OD260)/OD280 ratio was used to assess RNA quality. A TURBO DNA-free kit was used to remove DNA contamination.

Illumina sequencing and assembly of the Illumina contigs

Poly (A) mRNA isolation was performed using oligo (dT) beads after total RNA collection. Fragmentation buffer was added to generate short fragments of mRNA. Using these short fragments as templates, first-strand cDNA was synthesized by a random hexamer primer. Second-strand cDNA was then synthesized using buffer, dNTPs, RNaseH, and DNA polymerase I. Short fragments were purified with the QiaQuick PCR extraction kit (Qiagen) and resolved with elution buffer. These fragments were separated by agarose gel electrophoresis after adding sequencing adapters. PCR amplification templates of the suitable fragments were selected. During the quality control steps, the Agilent 2100 Bioanalyzer and ABI StepOnePlus Real-Time PCR System were used to qualify and quantify the sample library. Finally, the nine libraries from the nine individuals (six parents and three triploids) were sequenced using an Illumina HiSeq™ 2000/2500.

After raw reads were produced by sequencing, the read adaptors and low quality reads were removed. Transcriptome de novo assembly was carried out with a short-reads assembly program (Trinity) [54], using three independent software modules called Inchworm, Chrysalis, and Butterfly. Principal component analysis (PCA) of nine liver transcriptomes was applied to examine the contribution of each transcript to the separation of the classes [55, 56] (Additional file 9).

Gene annotation

Contig annotation was performed using the five public databases. BLASTX alignment (e-value ≤ 1e⁻⁶) between contigs and protein databases was performed, and the best-aligned results were used to decide the sequence direction of contigs (Additional file 1). After screening the sequences (alignment length ≤ 100 bp), accession numbers of the genes were obtained from the BLASTX results. Then, GO terms of annotation sequences were obtained through Ensembl BioMart [57]. WEGO software was used to analyze the GO annotation (Additional file 2) [58]. For pathway enrichment analysis, we mapped all differentially expressed genes to terms in the KEGG database and looked for significantly enriched KEGG terms (Additional file 10).

Mapping and differential expression

To obtain the shared transcripts in the three species, the reference transcripts were merged from the BSB, GC, and GB contigs using CD-HIT with 95% as the threshold [59]. Then, we utilized the merged sequences as the reference transcript because this database was built using transcripts from both parents and the hybrid offspring. The total clean reads were aligned against the merged sequences using Blat [33]. Then, information about the expression level in the three species was reflected by the number of aligned reads.

Mapped, filtered, and sorted reads were analyzed with the DEGseq package in R software version 2.13 (R Foundation for Statistical Computing, Vienna, Austria) [60]. Differential expression was assessed in triploids and their diploid parents using Fisher’s exact tests [61]. The abundance or the coverage of each transcript was determined by read counts and normalized using the number of reads per kilobase exon per million mapped reads (RPKM) [62]. The RPKM value of the read density reflected the molar concentration of a transcript in the starting sample after normalizing for the RNA length and total read number in the measurements. This facilitated a transparent comparison of transcript levels within and between samples. Herein, we defined gene expression as the average sequence expression of a gene, and a species comparison was shown (Fig. 3a, b, and c).

Dosage compensation in triploid fish

The absolute values of the log₂Ratio ≤ 1 were used as the threshold to judge the significance of the gene expression difference. Expression values above the threshold were described as upregulated and those below the threshold were described as downregulated.

To effectively analyze the dosage effects in triploid, we first set the PT-ELV (χ_triploid) according to the composition of the genome: two sets of genomes from maternal GC and one set from paternal BSB [6]. The value was constructed from two parts in which one is half the BSB value of gene expression (χ_BSB) and the other is the GC values of gene expression (χ_GC) (χ_triploid = 1/2χ_BSB + χ_GC). If no dosage effect happens in triploids, the gene expression level of triploids will float along with χ_triploid. However, comparing the AT-ELV with PT-ELV in triploids revealed that most genes were down-regulated. In this situation, we assumed that the dosage effect occurred in maternal GC homoeolog of triploids similar to other triploid individuals [10] and set up the PD-ELV (χ_diploid = 1/2χ_BSB + 1/2χ_GC). Comparing the AT-ELV and PD-ELV, the number of differentially expressed genes showed trends of up- and down-regulation in triploid fish.

Analyses of expression level dominance and homoeolog expression bias

We explored the data to identify candidate novel expression (new expression of a gene in liver) and homoeolog silencing patterns (no expression of one homoeolog) in the hybrids. Novel expression was inferred when both parental species had no reads for a gene, yet hybrids displayed more than 10 RPKM. If both parental species had more than 10 RPKM, but hybrids had zero counts for the same gene, this was considered silencing. These two cases were eliminated from further analysis, and we focused on genes that were expressed among both the diploid parents and triploid offspring.

In triploid offspring, the total liver genes were affected by a negative dosage effect. Genes that were identified as differentially expressed in the hybrid relative to the diploid parents were binned into 12 possible expression classes of differential expression (Fig. 4a), ELD, mid-parents, and up/down expression (outside the range of either parent), according to Rappet et al. (2009) [32]. Briefly, genes were parsed into these 12 categories (using Roman numerals; see Fig. 4a), depending on the relative expression levels between triploid and the diploid parents. Examined in this manner, genes may display mid-parents (XI and XII), paternal BSB-ELD (VII and VIII), maternal GC-ELD (IX and X), expression lower than both parents (I, II, and III), or expression higher than both parents (IV, V, and VI). For each of the 12 categories above (which are based on joint expression levels for both homoeologs), we calculated the RPKM value of reads to examine the gene expression for each homoeolog pair. The FDR was used to determine the threshold P value in multiple tests and analyses. FDR < 0.001 and the absolute value of log₂ ratio ≤ 1 were used as thresholds to judge the significance of gene expression differences between two species. For each gene, the expression level of the two diploid parents was estimated and classified into three situations; then, the expression level of triploid hybrid for the same gene was exhibited in the three situations (Fig. 4b, c, and d).

qPCR analysis

According to the expression level of transcriptome data, we had detected the expression of β-actin among of BSB, GC and GB. The expression level of β-actin in liver of triploid was also decreased to one’s in diploid state. So β-actin could be considered as the references gene in qPCR. The total RNA that was extracted from the liver tissue was used for qPCR analysis. qPCR analysis was performed using the Prism 7500 Sequence Detection System (Applied Biosystems) with a miScript SYBR Green PCR kit (Qiagen). qPCR was performed on biological replicates in triplicate (and triplicate technical qPCR replicates). The amplification conditions were as follows: 50 °C for 5 min and 95 °C for 10 min, followed by 40 cycles at 95 °C for 15 s and 60 °C for 45 s. The average threshold cycle (Ct) was calculated for each sample using the 2^-ΔΔCt method and normalized to β-actin. Lastly, a melting curve analysis was completed to validate the specific generation of the expected product.