Analyzing homoeolog expression provides insights into the rediploidization event in gynogenetic hybrids of Carassius auratus red var. × Cyprinus carpio

Rediploidization is considered to be a part of the evolutionary history of allotetraploids, and resulted in the emergence of novel epigenetic regulatory activities. To study the changing patterns of gene expression following the reduction of a genome by 50%, we used RNA-seq and quantitative real-time PCR (qPCR) to investigate total gene expression and homoeolog expression in three hybrids of a C. auratus red var. (2n = 100, ♀) (R) and C. carpio (2n = 100, ♂) (C) (i.e., F1, F18, and G4) and their original parents. A comparison of homoeolog expression between G4 and F18 identified 7 genes (0.22%) that exhibited novel R/C homoeolog expression patterns in G4, while 4 genes (0.12%) were affected by R/C homoeolog silencing. We determined the direction and extent of the homoeolog expression bias (HEB). The C-HEB genes (i.e., nrp1a and igf1) and R-HEB genes (i.e., fgf23 and esm1) provided insights into the effects of the dominance of one parental homoeolog expression on growth regulation. This dominance may contribute to the rapid growth of G4 fish. Our findings may be relevant for clarifying the relationship between growth heterosis and differences in homoeolog expression patterns.


Results
Transcriptome sequencing. To clarify the effects of rediploidization on transcript abundance, three hybrid lineages of the C. auratus red var. × C. carpio with different ploidy levels (i.e., F 1 , F 18 , and G 4 ) and their original parents were analyzed 15,16 (Table 1 and Fig. 1). Transcriptome sequencing produced 86.5 Gb of raw data for 15 libraries of the three hybrids and their original parents (Table S1). All short-read data were deposited in the Short Read Archive with the following accession numbers: SRX668436, SRX175397, SRX668453, SRX177691, SRX671568, SRX671569, SRX668467, SRX1610992, and SRX2347299.
Statistical mapping of RNA-seq data. After eliminating the read adapters and low quality reads, clean reads (77.7%) from nine libraries for three hybrid lineages with different ploidy levels were mapped to reference transcriptomes of the maternal (R) and paternal (C) parents to obtain the total gene expression profiles. Clean reads of six C. auratus red var. and C. carpio libraries (84.4%) were mapped to their respective reference transcriptomes (Tables S2 and S3). The 11,998 genes commonly expressed among all samples were used for total gene expression analyses. Meanwhile, the homoeolog expression levels of the three hybrid lineages were detected using another method (described in method "Specific mapping of the R/C homoeologs"). The clean reads for the three hybrid lineages were mapped to reference transcriptomes of the parents based on the threshold values of the specific SNPs. The analysis of R and C homoeolog expression involved only 3,540 genes. Differential total gene expression in gynogenetic diploid fish. A comparison of total gene expression among the three hybrid groups (i.e., F 1 , F 18 , and G 4 ) based on hierarchical clustering revealed that the hybrid groups could clearly be separated from their parents ( Fig. 2A). Additionally, the G 4 allodiploid was highly correlated with its parental F 18 allotetraploid ( Fig. 2A). The MA-plots of the comparison of G 4 with F 1 and F 18 are presented in Fig. 2B,C, respectively. Additionally, details regarding the differential total gene expression levels are provided in Fig. 2D. We identified 507 (4.75%) and 1,846 (21.95%) genes that were differentially expressed between G 4 and the original maternal C. auratus red var. and paternal C. carpio, respectively. The analysis of total gene expression levels indicated that G 4 was heavily biased toward the original maternal C. auratus red var.. Based on the comparison between G 4 and F 18 , we identified 691 genes (5.71%) that were differentially expressed (Table 2), which was a consequence of rediploidization. In contrast, the comparison between G 4 and F 1 revealed more than double the number of differentially expressed genes (1,358, 12.93%) (Fig. 2D). Meanwhile, a comparison among G 4 , F 1 , and F 18 , and the original maternal C. auratus red var. revealed that the gene expression levels   Pearson correlation coefficients were calculated for all pairwise comparisons, and are presented in a heatmap following unsupervised clustering. (B) Comparison of the expression levels between G 4 and F 1 allodiploid hybrids. Black dots between two blue lines represent the genes with similar expression levels and those that are outside the blue lines represent the genes with significantly different expression levels (log 2 FC > 2 and FDR < 0.05). (C) Comparison of the expression levels between G 4 allodiploid and F 18 allotetraploid hybrids.
(D) Bold text indicates the total number and fraction of genes differentially expressed in each comparison. Proportions of the total number of differentially expressed genes that are up-regulated were also given in boxes. For example, 507 genes were differentially expressed between R and G 4 groups. Of these, 329 were up-regulated in C. auratus red var. and 178 were up-regulated in G 4 .
in G 4 tended to be low (Fig. 2D). These results suggest that the global gene expression in G 4 may have gradually decreased to a novel level among diploid lineages. We next investigated the expression level changes resulting from the rediploidization events and compared it with the hybridization and tetraploidization events (They had been described in latest reports 24 ). Differences in total gene expression levels were identified by comparing each hybrid offspring with the original maternal C. auratus red var. and paternal C. carpio. Compared with the C. carpio expression levels, we detected 25 and 77 upand down-regulated genes common to the three hybrid offspring (Fig. S1). However, only two down-regulated genes were identified in all three hybrid offspring lineages in a comparison with C. auratus red var. expression levels (Fig. S1).
Novel expression pattern, silencing, and homoeolog expression bias. A gene ontology analysis (level 2) indicated that some genes were not simultaneously expressed in the liver of three hybrid individuals. These genes were mainly associated with metabolic or catalytic processes (Fig. S2). The novel expression pattern had been detected in 116 genes. Among of these, 83 genes (0.69% in G 4 ) exhibited expression silencing in F 18 and re-expressed in G 4 (Table S4). While the remaining 33 genes (0.28%) expressed in F 18 , but silenced in G 4 (Table S4). To investigate changes in homoeolog expression levels accompanying a rediploidization after an allotetraploidization, the R/C homoeolog expression patterns in F 1 , F 18 , and G 4 were analyzed. During a comparison between G 4 and F 18 , three (0.08%) and five (0.14%) genes were detected as R and C homoeologs with novel expression patterns in G 4 , respectively. We also identified two genes (0.06%) as silenced R homoeologs, and another two genes (0.06%) as silenced C homoeologs. These results suggest that a rediploidization accelerates the emergence of novel R/C homoeologous gene expression and silencing.
Principal component analysis (PCA) confirmed a strong correlation between R and C homoeolog expression levels in all hybrids (Fig. 3A). Yet, the expression levels of R and C homoeologs in hybrids showed variation based on their genotype. The distribution of the R and C homoeolog expression levels for the three analyzed hybrids, which was consistent with the PCA results, is presented in Fig. 3B. The differences between the R and C homoeolog expression levels in G 4 were smaller than those in F 18 . Additionally, the R and C homoeolog expression levels were more closely correlated in F 18 than in F 1 . These results imply that the R and C homoeolog expression levels become increasingly related with the development of new generations with differing ploidy levels. After clarifying the R and C homoeolog expression levels, we obtained the distribution of the log 2 fold change (FC) values between R and C in three hybrid lineages (Fig. 4A). After hybridizations, the phenomena of R-and C-HEB were observed for increasing numbers of genes as the number of generations increased. Meanwhile, the number of differentially expressed R and C homoeologs gradually decreased from F 1 to G 4 (Fig. 4A). For a more thorough   comparison of R and C homoeolog expression levels, we divided the genes into those exhibiting HEB and those that did not (no-HEB) based on a specific differential gene expression threshold. After identifying the HEB genes in the three hybrids, we focused on R-and C-HEB genes to identify which genes maintained a strong HEB. According to the HEB analysis, 73 and 47 genes were determined to be R-and C-HEB genes in the three analyzed hybrid lineages, respectively ( Fig. 4B,C).
To further investigate the effects of rediploidization on R and C homoeolog expression levels, we compared and classified only G 4 and F 18 genes. The HEB remained unchanged for most genes between F 18 and G 4 . These genes were then divided into the following three categories: no-HEB (58.47%), R-HEB (13.84%), and C-HEB (4.21%) ( Table 3). These observations suggest that the parental condition remained during a rediploidization. Meanwhile, the HEB of some genes changed because of a rediploidization. A total of 420 genes (11.86%) that exhibited HEB in F 18 , including 283 R-HEB genes (7.99%) and 137 C-HEB genes (3.87%), exhibited no-HEB in G 4 . Additionally, 260 (7.34%) and 143 (4.04%) genes with no-HEB in F 18 exhibited R-and C-HEB in G 4 , respectively (Table 3). A few genes underwent considerable changes. For example, six genes (0.17%) changed from C-HEB to R-HEB, while the opposite change occurred in three genes (0.08%) ( Table 3).

Homoeolog expression bias of growth-regulating genes identified by RNA-seq and qPCR.
A previous study concluded that the G 4 fish grow 30% faster than the parental F 18 fish 14 . Using RNA-seq technology, we analyzed the transcript levels of 92 growth-regulating genes (Table S5). The HEB status was obtained for only 34 of these genes (Table S6) because HEB was known for only 3,540 genes of the 11,998 genes included in the total gene expression analyses ( Table 2). Most of the growth-regulating genes (142 genes, 95.30%) were not differentially expressed between G 4 and F 18 . Only the dual specificity phosphatase 22 (dusp22) gene, growth hormone secretagogue receptor (ghsra) gene, fibroblast growth factor 19 (fgf19) gene, and glypican-4 (gpc4) gene were up-regulated in G 4 . In contrast, the bone morphogenetic protein 10 (bmp10) gene and two endothelial cell-specific molecule-1 (esm1) genes were down-regulated in G 4 (Tables 2 and Table S5). Under the condition of no-HEB in paternal F 18 , the neuropilin 1 (nrp1a) gene exhibited C-HEB in G 4 , while the fibroblast growth factor 23 (fgf23) and esm1 genes exhibited R-HEB (Tables 3 and Table S6).
To study the relationship between HEB and rapid growth, we determined the HEB of six growth regulated genes (i.e., igf1, igf2, ghr, tab1, bmp4, and mstn) in three tissues (i.e., liver, muscle, and ovaries) of G 4 and F 18 fish using homoeolog-specific qPCR (Fig. 5). Interestingly, we observed that the C homoeolog of the mstn gene was silenced in the muscle of G 4 and F 18 fish (Fig. 6). Novel expression patterns of the C homoeolog of the mstn gene were observed in the liver of G 4 fish, while this homoeolog was silenced in F 18 (Fig. 6). We also observed differences in the extent of R/C-HEB in the three analyzed tissues (Fig. 6). Specifically, igf1 exhibited an overall R-HEB in the muscle and liver of G 4 fish, while bmp4 and mstn exhibited an overall C-HEB in the ovaries of F 18 fish. In summary, R-HEB was the predominant expression bias in the liver, muscle, and ovaries.

Discussion
Continuous changes to homoeolog expression accompany to new generations with different ploidy levels. In this study, the initial hybridization involved the merging of C. carpio and C. auratus red var. genomes. Following a genome duplication in the F 2 diploid hybrid gametes, a bisexual fertile allotetraploid population was obtained 15 . Gynogenesis was then exploited to breed the G 1 diploid hybrid, in which the genome consisted of only half of the allotetraploid genome 16 . The three hybrids and the two associated genome-level changes enabled us to elucidate the relationship between genomic variation and growth differences in hybrids.
Characterizing the structural and functional changes in hybrids with different ploidy levels is a major challenge because of the complexities of genome structures and regulatory pathways. In particular, it is still unclear whether the two original parental genomes contribute equally to the gene expression levels of hybrids. We explored this uncertainty in detail using a systematic approach for dissecting the relative contributions of two homoeologs (i.e., R and C) to total gene expression levels. We also investigated the changes to homoeolog expression accompanying a genome duplication and rediploidization events. Only a few studies have focused on the differential expression of homoeologs in vertebrates because of a lack of fertilized hybrids. However, studying homoeologs originating from different species has been considered useful for clarifying the genetic structure and specific phenotypic changes occurring in hybrids with differing ploidy levels 25,26 . Subsequent genome-wide assessments have relied primarily on microarrays, which may not be appropriate for distinguishing the expression of closely related genes 10,27 . Therefore, these studies have been unable to separate the contributions of different homoeologs, particularly the R and C homoeologs originating from two diploid parental species. In this study, we used two novel analyses based on RNA-seq and qPCR to study the HEB under the condition that the gene transcripts from the parental species could be distinguished. A subsequent analysis of growth-regulating genes provided insights into growth heterosis resulting from a rediploidization.
The genome reduction following a hybridization altered the parental allotetraploid gene expression levels. Our total gene expression analysis revealed that 5.97% of the genes were differentially expressed between G 4 and F 18 , while 21.5% of the genes were differentially expressed between F 1 and F 18 (Fig. 2). Additionally, analyses of homoeolog expression indicated that the number of genes exhibiting HEB decreased in new generations with ploidy-level changes (Fig. 2). These observations imply that a rediploidization leads to novel expression patterns, which are not simply the result of a reversal of the tetraploidization process. Compared with their parents, a new population was observed to exhibit differences related to growth 14 , random amplification of polymorphic DNA and microsatellite characteristics 14 , and reproduction 17 . These changes were closely related to gene expression, especially homoeolog expression levels. Additionally, the maternal R-HEB phenomenon was detected in three hybrids (Table 3) 24 . Maternal expression bias is frequently observed in hybrid fish, including the M. amblycephala × Culter alburnus 8 , Oncorhynchus mykiss 28 , and S. salar 28 hybrids. Our PCA results indicated that the R and C homoeolog expression levels were increasingly closely linked during the reconstruction of the hybrid genotype and the development of new generations (Fig. 3). The hybrid genome gradually started to change, with the gene expression levels inherited from the original parents slowly reaching a similar level. These gene expression level changes suggest the hybrid lineages tended to reach a steady state. However, tetraploidizations and rediploidizations were associated with increases in the number of R-HEB genes. This change also influenced hybrid phenotypes, and induced various genetic changes conducive for adaptations. Recent studies have deduced the possible mechanisms underlying changes to total gene and homoeolog expression levels. Allelic interactions and gene redundancies were considered the major causes of alterations to non-coding RNA, DNA, and methylation, which resulted in additional changes to the hybrid transcriptomes 11,29 .
An analysis of homoeolog expression related to hybrid characteristics has produced relatively precise information regarding the associated regulatory mechanisms 30 . To investigate the transcriptome shock caused by a rediploidization and the resulting increase in the growth rate of G 4 fish, we analyzed the expression of growth-regulating genes. The four up-regulated and three down-regulated genes enabled additional investigations  Table 3. Differences in homoeolog expression bias between the allotetraploid (F 18 ) and gynogenetic allodiploid (G 4 ) hybrids. R = C denotes equal expression; R > C and R < C denote R-biased and C-biased expression, respectively. a Based on the significance differential homoeolog expression comparison of R and C homoeologues (P < 0.05 in comparisons; Fisher's exact test). b the ratio of R and C homoeologs greater than 1 was considered as potential R-biased in hybrids. Conversely, it represent as potential C-biased.
into the relationship between gene expression levels and growth rates. However, we did not conduct any related phenotypic analyses. Thus, we were only able to obtain the annotation details of these genes (i.e., dusp22, ghsra, fgf19, esm1, and gpc4), which are important for controlling cell division and regulating growth [31][32][33][34] . Furthermore, bmp10 encodes a potent inhibitor of endothelial cell migration and growth 35 . The down-regulated expression of this gene likely promotes growth (Tables 2 and Table S5). We used the analysis of HEB among growth-regulating  genes to investigate the relationship between HEB and phenotypes, and to characterize the mechanism underlying heterosis in polyploid hybrids (Tables 3 and Table S6). Our RNA-seq results suggested that nrp1a was associated with C-HEB, while fgf23 and esm1 were related to R-HEB (Tables 3 and Table S6). An assessment of the localization of the expression of six key growth-regulating genes revealed that igf1 exhibits C-HEB in the muscle and liver. The original paternal parent C. carpio grows faster than the maternal parent C. auratus red var. 24 . The C-HEB of igf1 likely promotes the dominance of the paternal growth characteristic in G 4 , possibly inducing a rapid growth rate. Interestingly, the diploid G 4 retained the inherited characteristics of the parental F 18 . The silencing of the C homoeolog was also observed for the mstn gene (Figs 5 and 6). This may have been due to genomic imprinting, implying that the regulation of gene expression is mediated by one parental genome, while the genetic material inherited from the other parent is silenced in the hybrid 36 . Some genes in hybrids always exhibit single-genomemediated expression 36 . An earlier study concluded that mutations in mstn always result in increased muscle mass and strength in vertebrates, making these individuals considerably stronger than their peers 37 . However, we determined that the expression pattern for a C homoeolog of mstn in the liver of G 4 fish differed from that of F 18 fish. The re-expression of mstn reflected the new regulatory mechanism influencing growth that may contribute to an increased growth rate in G 4 individuals. This mstn expression pattern was similar to that observed in F 1 24 . However, additional studies are required to verify that a rediploidization accelerates the emergence of new phenotypes via changes to the HEB and homoeolog silencing.

Conclusion
Gynogenetic diploid offspring provide unique opportunities to study the evolutionary effects of rediploidization. The associated transcriptome shock was clarified by a novel analysis of genes exhibiting HEB. In this study, we examined this phenomenon in detail using qPCR and homoeolog-specific sequences in the transcriptome. Our data revealed the direction and extent of HEBs and homoeolog silencing changes accompanying to hybridization, tetraploidization, and rediploidization. Additionally, our findings may be useful for more comprehensively characterizing of the relationship between novel growth phenotypes and homoeolog expression differences. Additional studies on this topic might contribute to elucidation of the mechanism regulating growth heterosis.

Materials and Methods
Animal materials. Fertile allodiploid hybrids (F 1 and F 2 , 2n = 100) were obtained from the hybridization between C. auratus red var. (2n = 100, ♀) (R) and C. carpio (2n = 100, ♂). Then, the unreduced gametes from F 2 hybrids lead to emergence of bisexual fertile allotetraploid hybrids (4n = 200). Until now, the allotetraploid lineage continued into F 25 by successive self-crossing. A fertile diploid gynogenetic fish lineage was obtained from the induction of gynogenesis in the female fertile allotetraploid hybrids (2n = 100). The fertilization of gynogenetic fish was performed with the eggs with UV-irradiated sperm from common carp. This technology had ensured the continuous of diploid gynogenetic fish lineage (G 1 -G 10 ), which could be considered as the model of rediploidization event regarding the allotetraploid lineage. The results of FISH showed that the genotype of three hybrid lineages were half of the R and C (1:1) genomes (Table 1).
All experiments (2012-2015) were approved by the Animal Care Committee of Hunan Normal University. We followed the animal experimentation guidelines of the Science and Technology Bureau of China. The experimental fish were kept in an indoor freshwater tank which maintained to carry suitable environmental conditions in terms of photo-period, water temperatures, forage etc. in the Engineering Center of Polyploidy Fish Breeding of the National Education Ministry located at Hunan Normal University, China. Fish were deeply anesthetized with 100 mg/L MS-222 (Sigma-Aldrich, St. Louis, Missouri, USA) before being dissected. Three mature females from each ploidy group, including diploid C. auratus red var., diploid C. carpio, F 1 allodiploid hybrids of C. auratus red var. × C. carpio, F 18 allotetraploid hybrids of C. auratus red var. × C. carpio, and G 4 allodiploid hybrids of C. auratus red var. × C. carpio (2-year-old individuals) were collected.
Their ploidy levels were confirmed by measuring the DNA contents of their erythrocytes via flow cytometry and by direct counting of chromosome numbers from metaphase spreads. To prepare metaphase spreads, we cultured the red blood in DMEM solution for 68-72 h at 25.5 °C and 5% CO 2 . Cells were harvested by centrifugation, and then treated in a hypotonic solution (0.075 M KCl) at 26 °C for 25-30 min. Samples were fixed in a methanol-acetic acid (3:1, v/v) solution with three changes. Then, the samples were placed on cold slides, air-dried, and stained in 4% Giemsa solution for 30 min. The chromosome numbers were observed in metaphase spreads of 15 individuals to determine the ploidy levels.
cDNA generation, library construction, and RNA sequencing. After anesthetizing the fish with 2-phenoxyethanol, the liver, muscle, and ovary tissues were excised and immediately placed in RNAlater for storage based on the manufacturer's instructions (Ambion Life Technologies, Carlsbad, CA, USA). Total RNA extracts were treated from the harvested tissues according to a standard Trizol protocol (Invitrogen) after the RNAlater was removed. Total RNA was treated with a DNA-free ™ DNA Removal Kit (Ambion) to remove any contaminating genomic DNA. The purified RNA was quantified using a 2100 Bioanalyzer system (Agilent, Santa Clara, CA, USA).
We fragmented 2 μg isolated mRNA with fragmentation buffer. The resulting short fragments were reverse transcribed and amplified to produce cDNA. An Illumina RNA-seq library was prepared according to a standard high-throughput method 38 . The cDNA library concentration and quality were assessed by Qbit (Invitrogen) and the Agilent 2100 Bioanalyzer, after which the library was sequenced using the Illumina HiSeq. 2000 platform. The RNA-seq experiment was conducted with three biological replicates. The transcriptome data were generated using a 101-bp paired-end setting. After removing the read adapters and low quality reads, clean reads from each Specific mapping of the R/C homoeologs. For detection of changes in homoeolog expression due to a rediploidization event, we obtained the genome of the maternal parent C. auratus red var. (http://rd.biocloud. org.cn/) (39,069 transcripts) and the genome of the paternal parent C. carpio (http://www.carpbase.org/) (52,610 transcripts). A database of full-length transcripts was used as a resource for gene expression details (Table S2). Additionally, analyses of putative orthologs between R and C may help to identify true orthologs. Thus, the R and C sequences were aligned by using the reciprocal BLAST (BLASTN) hit method, with an e-value cut off of 1e −20 41 . Two sequences were defined as orthologs if each one was the best hit for the other, and the aligned sequences contained at least 300 bp. We used putative R and C orthologs as the reference sequences for detecting the homoeolog expression of hybrids. By using custom perl scripts, we calculated the homoeolog expression levels over the SNPs between the R and C orthologs were applied for calculating homoeolog expression levels. The mapping of reads to the reference transcriptome for three hybrids and their original parents was completed by means of TopHat2 program 42 .
Analyses of homoeolog expression bias. Prior to analyses, the expression level data were normalized by using Cufflink program (version 2.1.0) 43 . We also restricted the data analysis with the number of read counts (≥ 1) of genes to remove the negative effects of background expression noise in all biological replicates. The abundance or the coverage of each transcript was normalized by using the number of reads per kilobase of exon per million mapped reads (RPKM) 44 . Gene expression levels were estimated based on the RPKM values of the reads. The false discovery rate (FDR) was used for determining the threshold P value in multiple tests and analyses. Meanwhile, unigenes with an FDR ≤ 0.05 and a fold change > 2 were considered differentially expressed genes. The mapping results were analyzed with the DEGseq package of the R program (version 2.13) (R Foundation for Statistical Computing, Vienna, Austria) 45 .
Comparisons of the R/C homoeolog expression levels among three hybrids were used to identify the genes with potentially novel expression patterns (e.g., new expression of a gene in the liver) and silenced genes in the hybrids according to the standards described by Yoo et al. 9 . To screen the newly expressed or silenced genes, we analyzed the homoeolog expression categories associated with these genes, which were simultaneously expressed among the three hybrids and their original parents. For homoeolog expression bias (HEB) analyses, we compared R and C homoeolog expression levels in the three hybrids. The log 2 FC of homoeologue expression value (R vs C) > 2 had been considered as the threshold values of R-HEB. Contrarily, the log 2 FC (C vs R) > 2 had been considered as the threshold values of C-HEB. The extent and direction of differential homoeolog expression were assessed according to Ren et al. 24 .
Homoeolog-specific qPCR. To elucidate the relationship between HEB and rapid growth, we examined the HEB of six important genes (i.e., igf1, igf2, ghr, tab1, bmp4, and mstn) previously observed to regulate growth rates 13,46,47 . However, HEB information was lacking for some of these genes. So the homoeolog-specific qPCR had been used to detect the homoeolog expression level. Total RNA was extracted from the liver, muscle, and ovaries. First-strand cDNA was synthesized using AMV reverse transcriptase (Fermentas, Canada) with an oligo (dT) 12-18 primer at 42 °C for 60 min and 70 °C for 5 min. To further detect the changes in homoeolog expression pattern of growth-regulating genes, we obtained sequence information for the R and C transcripts of a housekeeping gene (i.e., β-actin) and six key growth-regulating genes (i.e., igf1, igf2, ghr, tab1, bmp4, and mstn) from the C. auratus red var. and C. carpio transcripts. A comparison of the R and C homoeolog sequences by using Bioedit program (version 7.0.9) 48 revealed the presence of homoeolog-specific loci in all seven genes. The regions containing homoeolog-specific loci were used to design R/C homoeolog-specific primers for qPCR analysis 24 . To obtain highly sensitive specific primers, we completed cross amplifications by using ABI Prism 7500 Sequence Detection System (Applied Biosystems, USA) 24 . The PCR amplification conditions were as follows: 50 °C for 5 min, 95 °C for 10 min, and 40 cycles at 95 °C for 15 s and 60 °C for 45 s. Each test was conducted three times to improve accuracy of the results. Finally, relative quantities were calculated, and a melting curve analysis was used to verify the generation of a single product. Each sample was used in triplicate for the assays and to generate standard curves. The expression data for each homoeologous gene was normalized against that of β-actin according to 2 −ΔΔCt method 49 . The β-actin expression level in the hybrids was estimated by using the ratio of transcript abundance to gene copy number using PCR and qPCR conducted with co-extracted DNA and RNA samples. The β-actin expression levels in somatic organs and gonads were similar among the three hybrids and their original parents 50,51 . Data deposition. The transcriptome data were submitted to NCBI (accession number: accession numbers: SRX668436, SRX175397, SRX668453, SRX177691, SRX671568, SRX671569, SRX668467, SRX1610992, and SRX2347299). The remaining data are available within the article and its Supplementary Information files or available from the authors upon request.