Rapid genomic changes in allopolyploids of Carassius auratus red var. (♀) × Megalobrama amblycephala (♂)

To better understand genomic changes in the early generations after polyploidisation, we examined the chromosomal consequences of genomic merger in allotetraploid hybrids (4 nF1) (AABB, 4n = 148) of Carassius auratus red var. (RCC) (AA, 2n = 100) (♀) × Megalobrama amblycephala (BSB) (BB, 2n = 48) (♂). Complete loss of the paternal 5S rDNA sequence and the expected number of maternal chromosomal loci were found in 4 nF1, suggesting directional genomic changes occurred in the first generations after polyploidisation. Recent studies have reported instability of newly established allotetraploid genomes. To assess this in the newly formed 4 nF1 genome, we performed fluorescence in situ hybridisation on an allotetraploid gynogenetic hybrid (4 nG) (AABB, 4n = 148) and an allopentaploid hybrid (5 nH) (AABBB, 5n = 172) from 4 nF1 (♀) × BSB (♂) with 5S rDNA gene and centromere probes from RCC, the original diploid parent. The expected numbers of maternal chromosomal loci were found in 4 nG, while chromosomal locus deletions and chromosome recombinations were detected in 5 nH. These observations suggest that abnormal meiosis did not lead to obvious genomic changes in the newly established allotetraploid genomes, but hybridisation with the original diploid parent resulted in obvious genomic changes in the newly established allotetraploid genomes, as was found for the maternal genome.

Polyploidisation is a significant evolutionary process that results in rapid speciation [1][2][3][4][5][6] . Many diploids species are actually ancient polyploids that have undergone diploidisation. Recent studies, mostly in plants, suggested that allopolyploid formation could induce various types of genomic changes, including directional sequence elimination, random structural changes, and chromosome structure [7][8][9][10][11][12][13] . Importantly, a variety of genomic changes has been shown to result in diploid-like chromosome pairing [14][15][16][17] . Although the potential contribution of genomic changes to the evolutionary success of polyploidy has been widely recognised, virtually no information is available on how newly established genomes have evolved after polyploidisation.
Cytogenetics studies using fluorescence in situ hybridisation (FISH) have reported chromosomal changes in many polyploid species [25][26][27][28][29] . To further understand allopolyploid genome evolution in a broad context, we used the first generation 4 nF 1 hybrids and their backcross progenies to explore potential genomic changes on polyploidisation. We determined the response of two dispersed chromosomal loci (5S rDNA and centromere) that may be particularly important for de novo allopolyploidy, because they are likely to be most vulnerable to genetic changes after polyploidisation and (or) hybridisation [29][30][31] . Our results revealed rapid genomic changes occurred in the first generations after polyploidisation, and indicated instability of the newly established allotetraploid genome. The findings of this study provide new insights into chromosomal evolution in vertebrates.
A total of 340 clones were sequenced to examine the different patterns of the 5S rDNA sequences; 60 clones were from RCC, 40 clones were from BSB, and 80 clones were from each of the 4 nF 1 , 4 nG, and 5 nH hybrids (Table 1). Based on the BLASTn analysis, all the sequences from all five hybrids were confirmed as 5S rDNA repeat units. The 5S rDNA sequences from RCC fell into three distinct families (designated class I: 203 bp; class II: 340 bp; and class III: 477 bp), while the 5S rDNA sequences from BSB formed one family (designated class IV: 188 bp). All three RCC-derived families (class I, class II and class III) were detected in 4 nF 1 and 4 nG, while the BSB-derived family (class IV) was not found in 4 nF 1 and 4 nG. All four classes were detected in 5 nH (Table 1).

Southern blot hybridisation.
Genomic DNA from RCC, BSB and 4 nF 1 was digested with HindIII and ScaI. Southern hybridisation was performed using the 5S rDNA sequence from BSB as a probe. This probe hybridised with the genomic DNA from BSB, but not with the genomic DNA from RCC and 4 nF 1 (Fig. 3). This result implies that the paternal 5S rDNA cluster is completely deleted in 4 nF 1 .
Fluorescence in situ hybridisation. FISH hybridisation of the RCC-derived class I (203 bp, GenBank: GQ485555) 5S rDNA gene probe to the RCC and BSB metaphase chromosomes yielded eight 5S rDNA gene loci in RCC ( Fig. 4A; Table 2), but none in BSB (Table 2). It was expected that the eight 5S rDNA loci will also    be present in the 4 nF 1 , 4 nG, and 5 nH hybrids because they were derived from RCC. However, we found that, while all eight 5S rDNA gene loci detected in the metaphase chromosomes of 4 nF 1 and 4 nG and were similar to the eight RCC loci (Fig. 4B,C; Table 2), only five of the 5S rDNA gene loci were detected in the 5 nH metaphase chromosomes ( Fig. 4D; Table 2).   FISH hybridisation of the class II (340 bp, GenBank: GQ485556) 5S rDNA gene probe to the RCC and BSB metaphase chromosomes yield four 5S rDNA gene loci in RCC (Fig. 5A; Table 2), but none in BSB ( Table 2). The chromosomal locus map for RCC revealed two large 5S rDNA gene loci on homologous submetacentric chromosomes, and two small 5S rDNA gene loci on homologous subtelocentric chromosomes (Fig. 5A). Similarly, as expected, two large and two small 5S rDNA gene loci were found on homologous submetacentric chromosomes and homologous subtelocentric chromosomes, respectively, in 4 nF 1 (Fig. 5B) and 4 nG (Fig. 5C). Unexpectedly, in 5 nH, one large 5S rDNA gene locus was located on a submetacentric chromosome and another was located on a metacentric chromosome (Fig. 5D), suggesting that these two large 5S rDNA gene loci were not located on homologous chromosomes.
FISH hybridisation of the class III (477 bp, GenBank: GQ485557) 5S rDNA gene probe to the RCC and BSB metaphase chromosomes yield eight 5S gene loci in RCC (Fig. 6A; Table 2), but none in BSB ( Table 2). As expected, the eight RCC-derived 5S rDNA gene loci were detected in the metaphase chromosomes of 4 nF 1 , 4 nG and 5 nH and were similar to the loci in RCC (Fig. 6B-D; Table 2).

Discussion
Polyploidisation may increase genomic variation rates and is important for the formation in new polyploid species 4 . Evidence for genomic variations, including fragment loss, chromosomal rearrangement, and rDNA loci changes have been reported in both synthesised polyploid 7,27,32 and natural polyploid species 25,26 . Genomic variations usually occur in the early generations after polyploidisation, possibly reflecting instability in newly established polyploid genomes 26,27 . The results of the present study support previous observations that genomic changes occur in newly established polyploid genomes, and reveal that these changes can begin as early as the first generation after polyploidisation.
Because of incompatibility between homeologous chromosomes, hybridization can boost genomic change 33,34 . The frequency of genomic change has been associated with divergence of the diploid parental genomes 7 . The 4 nF 1 hybrid was formed by combining the two diploid genomes from RCC and BSB, two fish species in the family Cyprinidae, that belonged to different subfamilies (Cyprininae and Cultrinae) 21 , implying RCC and BSB are genetically distinct. In 4 nF 1 , not only was the paternal 5S rDNA unit deleted entirely, but so were the paternal sox 21,35 and hox (unpublished data) gene families, suggesting that a large number of genomic changes had occurred in the newly established allotetraploid genome. A variety of genomic changes can result in diploid-like chromosome pairing, which has been reported to prevent meiotic irregularities and improve the efficiency of gamete production in polyploid species 16,36 . However, there is still no direct evidence that large numbers of genomic variations or unstable individuals are selected for during the establishment of polyploid species 27,37-39 . In previous study, we found that diploid-like chromosome pairing was not restored in 4 nF 1 22,23 . We speculated that mass deletion of paternal genetic material gave rise to excessive genomic modification in 4 nF 1 , which prevented diploid-like chromosome pairing, and resulted in weak fertility and the generation of gametes with a different genetic composition. To avoid extinction, the unstable 4 nF 1 individuals may have entered a novel evolutionary trajectory by abnormal meiosis, and produced diploid gamete with two sets of RCC-derived chromosomes. Thus, unexpectedly, we obtained better fertile autotetraploids among the progenies of 4 nF 1 , and successfully established an autotetraploid fish line (F 2 -F 9 ) 24 .
In some cases, it has been shown that hybridization had more effect on the change in genomic and gene expression than polyploidization 40,41 . In our study, 4nG result from genome doubling of germ cell, 5 nH was obtained by hybridization of 4nF 1 (♀ ) × BSB. Thus, the 5S rDNA units and chromosomal loci (5S rDNA and centromere) remained intact in the 4nG genome, but obvious variations were found in 5 nH. In addition, our data also revealed the elimination of the entire paternal 5S rDNA unit, and stabilisation of the maternal 5S rDNA units and chromosomal loci in the allotetraploid hybrids, implying that the paternal genome underwent greater polyploidisation-associated modifications than the maternal genome. Similar findings have been reported in polyploid plants 30,42,43 . The nucleo-cytoplasmic hypothesis might be an explanation for the apparent paternal genome lability. This hypothesis predicts that the paternal genome of a newly formed allopolyploid evolves most rapidly because the maternal cytoplasmic background leads to paternal genome instabilities 7 . However in 5 nH, the newly established maternal allotetraploid genome showed obvious variations in chromosomal loci, while the parental 5S rDNA units remained intact, suggesting that the maternal genome was more unstable than the parental genome. These results are opposite to those predicted by the nucleo-cytoplasmic hypothesis. We speculate that the genetic variations in the maternal chromosomal loci may be attributed to instability of the newly established PCR amplification and sequencing. One pair of primers (5SP1: 5′ -GCTATGCCCGATCTCGT CTGA-3′ and 5SP2R: 5′ -CAGGTTGGTATGGCCGTAAGC-3′ ) was designed and synthesised to amplify the 5S rDNA repeats directly from genomic DNA by PCR. The PCR reactions and sequencing were performed as described by Qin et al. 44 . Sequences were analysed using ClustalW software (http://www.ebi.ac.uk/Tools/msa/clustalw2/).

Southern blot hybridisation.
Genomic DNA (10 mg) from all the samples from RCC, BSB and 4nF 1 was completely digested with the restriction endonucleases HindIII and ScaI, submitted to 0.8% agarose gel electrophoresis, and transferred onto Hybond-N1 membrane 45 . The 5S rDNA sequences were labelled with Dig-11-dUTP (Roche), which was used as a probe, and hybridised with the filter-immobilised DNA. Hybrid signal detection was performed with a DIG detection kit II (Innogent, China).
Fluorescence in situ hybridisation. Chromosome preparation was carried out on the kidney tissues of all samples, according to the procedures reported by Liu et al. 21 The FISH probes for the 5S gene and species-specific centromere were amplified by PCR using 5SP1 and 5SP2R primer, and the primer 5′ -TTCGAAAAGAGAGAATAATCTA-3′ and 5′ -AACTCGTCTAAACCCGAACTA-3′ , respectively. The FISH probes were produced by Dig-11-dUTP labelling (using a nick translation kit; Roche, Germany) of the purified PCR products. FISH was performed according to the method described by He et al. 46 For each type of fish hybrid, 200 metaphase spreads (20 metaphase spreads in each sample) of the chromosomes were analysed.