In dioecious diploid persimmon species , the Y chromosome-encoded small-RNA gene OGI is responsible for expression of maleness, through the repression of its autosomal counterpart, a gene called MeGI. Importantly, hexaploid cultivated Oriental persimmon has evolved a more plastic sex determination system from the dioecious system, possibly through epigenetic regulation of OGI and MeGI. In hexaploid Oriental persimmon, genetically male individuals, carrying at least one copy of Y-chromosome , exhibit monoecy, where both male and female flowers are produced by the same tree, although they often produce only female flowers. This monoecious system is based on the semi-inactivation of OGI by the presence of a highly methylated retrotransposon insertion, named Kali, within the OGI promoter region, and on the variable DNA methylation of the MeGI sequence . With the exception of a few genetically female cultivars, which do not carry the OGI sequence but occasionally bear male flowers, this male production is consistent with the change of epigenetic regulation on MeGI. Therefore, the OGI/ MeGI system is thought to be conserved as the regulator of sex determination in hexaploid persimmon as in diploid persimmon, large plastic pots despite the fact that expression of OGI in hexaploid persimmon has never been formally observed, possibly because it is extremely restricted both temporally and spatially.

In hexaploidy persimmon trees, the morphological difference between male and female developing buds is first visible in June. Typically, all floral buds produced on a branch develop similarly, either into male or female flowers. The first flowering year, all branches bear female flowers. In subsequent years, female branches tend to produce female branches but there is a low frequency of transition from female to male. Once a branch is male, it typically produces male branches in subsequent years, but there is a low frequency of reversal to femaleness as well. These observations are consistent with the hypothesis that occasional activation of OGI expression results in the accumulation of small-RNA and -DNA methylation on MeGI, leading to the production of male flowers. Subsequently, maintenance of maleness is dependent on the maintenance of the DNA methylation status of MeGI. According to this hypothesis, the frequency of OGI activation should be directly associated with male conversion. Here, we used segregating sibling plants that carry the OGI gene to investigate the genetic factors that regulate male conversion in hexaploid Oriental persimmon. Investigating the genetic architecture of traits with multiple alleles and mixed inheritance patterns in polyploid species is challenging. Genotyping in polyploids is complicated by the possibility of more than two alleles at each locus, and the existence of different heterozygous states. For instance, in the case of autohexaploid persimmon with hexasomic inheritance, we can define five heterozygous states; AAAAAa , AAAAaa , and AAAaaa AAaaaa to Aaaaaa .

Quantitative genotyping of such loci using qPCR is a possible solution, but it is time-consuming. Recent advances in sequencing and genotyping technologies now allow calling high density single nucleotide polymorphism , and accurate determination of allele ratio and allele dosage for polyploid genomes. Genome-wide association analysis, using quantitative genotypes, coupled with realistic genetic models could shed light on the genetic basis of complex traits. In this study, to examine the genetic factors underlying female-tomale conversion in hexaploid Oriental persimmon, we collected high density genotype information in a segregating population exhibiting biases in male flower ratio. Considering the nature of polysomic inheritance, we developed genome-wide correlation/association analyses for polyploid persimmon using two different models : an additive model using quantitative genotypes in the form of allelic ratio, and a diploidized additive model using diploidized genotypes. The results led to the identification of genetic regions and candidate genes potentially involved in the regulation of female-to-male conversion, providing novel insights into the genetic basis of flexible sexuality after adaptation to polyploidization or domestication.ddRAD-sequencing libraries were prepared according to previous reports using 107 individuals of the YTF1 population and their two parents ‘Yamatogosho’ and ‘Taishu’. The libraries were sequenced on an Illumina HiSeq4000 platform, and generating PE100 reads at the Vincent J. Coates Genomics Sequencing Laboratory, University of California Berkeley. The reads were aligned to the reference genome of Diospyros lotus, a wild relative close to Oriental persimmon 23 using the Burrows–Wheeler Aligner with default parameters.

Based on these parameters, it is possible that some allelic and potential paralogous polymorphisms were occasionally derived from non-specific mapping. The resulting sam files were converted to bam files and subsequently vcf format using bcf/vcftools and Varscan. For genome-wide association/correlation analyses, only the 83 individuals carrying the OGI allele were used . Individual genotypes were only considered if coverage > 20. A total of 95,639 heterozygous markers were selected by using bcftools with the following options: minor allele frequency within the population > 0.05 and data were available for at least 50% of the individuals. For detection of transmission ratio distortion, only the 91 individuals of the YTF1 population with total coverage > 200 Mb were used, to decrease the possibility of detecting the false transmission distortion when using low coverage data . A total of 39,344 loci were selected, with average coverage between 60 and 200 for each of the 91 individuals, minimum coverage in the parents > 60, and by using bcftools with the following options: minor allele frequency > 0.01 and max-missing ¼ 1. Transmission distortion ratios of alternative alleles were calculated as the log value of [allele ratio in the YTF1/allele ratio in the parents]. Here, for the ‘allele ratio in the parents’ values, we calculated the alternative allelic read coverages in the two parents independently, and averaged them. Significant transmission distortion was detected by standardizing the transmission distortion ratios using z-transformation with the threshold of P-value < 1E-10.Genome-wide correlation analyses based on quantitative genotypes using an additive model resulted in several peaks, the strongest of which was located on the sex-chromosome .23 This wasAfter compensating for the effect of OGI allele dosage , the peaks on the sex chromosome, both with quantitative and diploidized genotypes, was significantly reduced . Although the trends were almost identical to those in the original analysis , some peaks, especially those associated with the male conversion rate , became sharper than in the original model. Some of the original putative peaks were also reduced after compensation for the OGI dosage effect, presumably because their genotypes were accidentally similar to those of OGI. Some of the major peaks, especially on Chr. 5, 8, and 9, were common to the two models . On the other hand, peaks on Chr. 1, 2, and 4 were specific to additive model . Here, we focussed on the common candidate loci on Chr. 5, 8, and 9. To understand the combination effect of OGI and these three loci, multiplex regression tests were performed between the male conversion rate and the quantitative/diploidized genotypes , resulting in r 2 values of 0.611 and 0.548 for the quantitative and diploidized genotypes, respectively. Importantly, the accuracy of the regression was significantly increased in comparison to the test without compensation for the OGI allele dosage . These results highlight the importance of quantitative genotyping and compensation for allele dosage in GWASs in polyploids. Both the additive and diploidized additive models showed consistent associations between Chr. 5 and male conversion rate . The haploblock including the peak summit was maintained over a 3 Mb region , and contained 184 genes . For Chr. 8, the peak spanned the region between 3.8 and 8.8 Mb, included 433 genes , and was associated with male conversion rate in both the additive and diploidized additive models . The quantitative genotypes for the peak summit of the Chr. 9 peak observed in the additive model ranged from nulliplex to duplex . The peak covered 4 Mb of the sub-telomeric region and contained 262 genes . For all three of these peaks identified from the diploidized additive model analysis, nulliplex individuals showed significantly higher male conversion rate than heterozygous individuals , suggesting that loss of that particular allele is associated with positive regulation of male conversion. This is consistent with the results from the additive model on Chr. 8, raspberry container which indicated that fewer copies of the alternative alleles were positively correlated with male conversion rate as well . The genotypes of the male parent ‘Taishu’ and the YTF1 individuals with high male conversion were consistent in the highest peak on Chr. 5 . On the other hand, the peaks of Chr. 8 and Chr. 9 showed inconsistent genotypes between the male parent ‘Taishu’ and the YTF1 individuals with high male conversion rateGenome-wide analyses in polyploid genomes are often challenging in terms of genotype calling, and haplotype phasing. 

For genotyping, we adopted the allele ratio per locus with high coverage for quantitative genotypes, as previously reported in potato and in blueberry. This approach, combined with the Pearson correlation was successful at detecting the effect of Y-chromosome dosage for OGI activation and other potential candidates. Furthermore, the compensation with OGI allele dosage with two approaches, partial correlation , and rrBLUP with the covariate of OGI dosage , improved our ability to detect candidate loci. On the other hand, issues involving haplotype phasing still remain to be solved, since polyploids have more than two haplotypes per reference region, which are generally difficult to define, not only for conventional binary genotypes but also for quantitative genotypes. For instance, in our hexaploidy samples, the presence of many homologous haploblocks caused LD inconsistent with the order of theSNPs markers, despite of the F1 segregation line in which LD decay completely depends on recombinations , as observed in autotetraploid blueberry as well. In autohexaploid sweet potato with 90 chromosomes , 96 linkage groups were generated, using only double-simplex SNPs showing a Mendelian segregation ratio in an F2 progeny. Such construction of well-separated LGs would allow QTL analysis with conventional tools, but the targets would be limited to simplex loci and, unless the genomes were significantly different, much of the genomic space would not be included. Here, we were able to include multiplex alleles in our genome-wide association analysis, and our results suggest that a simple GWAS approach can be used to identify polysomic candidates, if sufficient markers are available. Our results indicated that, in the additive model based on quantitative genotypes, individuals with higher dosages of OGI have a higher probability of male conversion. If we hypothesize that male conversion is associated with OGI activation, we can propose the following two hypothesis for cis-regulation of OGI to maintain OGI silencing : multiple trans-acting factors can access the OGI promoter, each with slightly different sequence specificity, or epigenetic cis-factors of OGI, such as DNA/histone methylation, are modified independently, and thus, more copies of OGI alleles result in higher probability of OGI expression release. On the other hand, in both the additive and the diploidized additive models, all the loci on Chr. 5, 8, and 9 significantly affects male conversion rate . This suggests that, although recessive alleles at these loci lack the function to maintain OGI suppression, the other alleles also vary in their ability to suppress OGI and/or have additive effects for silencing of OGI . Such a complex situation is not uncommon when dealing with polysomic genetic factors, and exemplifies how they can contribute to the acquisition or fine tuning of traits in a way that is not possible in a diploid situation.SNPs markers, despite of the F1 segregation line in which LD decay completely depends on recombinations , as observed in autotetraploid blueberry as well. In autohexaploid sweet potato with 90 chromosomes , 96 linkage groups were generated, using only double-simplex SNPs showing a Mendelian segregation ratio in an F2 progeny. Such construction of well-separated LGs would allow QTL analysis with conventional tools, but the targets would be limited to simplex loci and, unless the genomes were significantly different, much of the genomic space would not be included. Here, we were able to include multiplex alleles in our genome-wide association analysis, and our results suggest that a simple GWAS approach can be used to identify polysomic candidates, if sufficient markers are available. Our results indicated that, in the additive model based on quantitative genotypes, individuals with higher dosages of OGI have a higher probability of male conversion. If we hypothesize that male conversion is associated with OGI activation, we can propose the following two hypothesis for cis-regulation of OGI to maintain OGI silencing : multiple trans-acting factors can access the OGI promoter, each with slightly different sequence specificity, or epigenetic cis-factors of OGI, such as DNA/histone methylation, are modified independently, and thus, more copies of OGI alleles result in higher probability of OGI expression release. On the other hand, in both the additive and the diploidized additive models, all the loci on Chr. 5, 8, and 9 significantly affects male conversion rate . This suggests that, although recessive alleles at these loci lack the function to maintain OGI suppression, the other alleles also vary in their ability to suppress OGI and/or have additive effects for silencing of OGI . Such a complex situation is not uncommon when dealing with polysomic genetic factors, and exemplifies how they can contribute to the acquisition or fine tuning of traits in a way that is not possible in a diploid situation.