Recent Transposition of an LTR-Retrotransposon in the Gene Coding for S Receptor Kinase is Responsible for a Novel Self-Compatible Phenotype of Radish (Raphanus Sativus L.)

Self-incompatibility (SI) responses of radish (Raphanus sativus L.) are determined by two tightly linked genes encoding an S receptor kinase (SRK) and an S-locus cysteine-rich protein/S locus protein 11 (SCR/SP11), respectively. A radish showing an almost self-compatible (SC) phenotype was identied in this study. Inheritance patterns showed that this SC phenotype was dominant over an SI phenotype. In addition, this SC phenotype co-segregated with an S haplotype in an F 2 population. This SC radish contained an RsS-26 haplotype in which duplicate SRK-like genes were previously identied. Full-length sequences of two SRK-like genes of 18,133-bp and 6,200-bp in length were obtained from radish with the RsS-26 haplotype (designated as RsSRK-26-1 and RsSRK-26-2, respectively). Duplicate SCR/SP11-like genes were also identied in the radish with the RsS-26 haplotype. Phylogenetic analyses indicated that both duplicate SRK-like and SCR/SP11-like genes were closely related to other known SRK and SCR/SP11 genes, respectively. No critical mutation was found in the coding region of SRK-like or SCR/SP11-like gene. However, a 4,146-bp intact LTR-retrotransposon was identied in the third intron of RsSRK-26-1 of the SC radish. Interestingly, this LTR-retrotransposon was not detected in three other breeding lines containing the same RsS-26 haplotype. Except for this LTR-retrotransposon, only two single nucleotide polymorphisms (SNPs) were identied in intronic regions between normal and mutant RsSRK-26-1 alleles. While normal transcription was observed for radish showing RsSRK-26-1 and SI phenotypes in these three breeding lines, no transcript of RsSRK-26-1 was detected in the SC radish, suggesting that recent transposition of an LTR-retrotransposon in the RsSRK-26-1 gene might be responsible for the SC phenotype of radish.


Introduction
Radish (Raphanus sativus L.) is one of the most important root vegetables in East Asian countries such as China, Japan, and Korea. As a cross-pollinated crop, radish shows hybrid vigor. Its F 1 hybrid varieties that harness such hybrid vigor have been developed (Curtis 2011). Self-incompatibility (SI) is de ned as inhibition of pollen tube growth in pistils during self-pollination. It has been commercially used for the production of F 1 hybrid seeds in radish (Singh et al. 2001). SI systems are widespread in plants. They can promote outcrossing and increase genetic diversity. More than 100 families and almost 40% of angiosperms have adopted SI systems (Igic et al. 2008).
Generally, a single locus initially named as S (Sterility) is involved in SI responses. Two tightly linked genes consisting of male and female determinants of SI are positioned in the S locus. Nucleotide sequences of these two SI determinants are highly polymorphic and multi-allelic. Since at least two genes are present in the S locus, combinations of these genes can lead to S haplotypes (Stone and Goring 2001; Watanabe et al. 2012). Types of SI systems are largely categorized into heteromorphic and homomorphic SI. In the case of heteromorphic SI, cross-pollination is possible between different oral morphs. Candidate genes responsible for distyly have been reported in Primula vulgaris (Li et al. 2016) and Fagopyrum esculentum (Yasui et al. 2012).
Homomorphic SI systems are generally classi ed into gametophytic SI (GSI) and sporophytic SI (SSI) depending on inheritance patterns of SI phenotypes. Such inheritance patterns are determined by genotypes of haploid pollen in GSI and diploid pollen parents in SSI (Silva and Goring 2001;Muñoz-Sanz et al. 2020). GSI systems based on RNase and S-locus F-box proteins have been extensively studied in Solanaceae, Plantaginaceae, and Rosaceae families (Franceschi et al. 2012). The Papaver system based on programmed cell death has been mainly studied in poppy (Wheeler et al. 2009). Meanwhile, SSI has been extensively studied in Brassicaceae including radish (Sobotka et  Two tightly linked genes encoding S receptor kinase (SRK) and S-locus cysteine-rich protein (SCR)/S locus protein 11(SP11) or SP11/SCR have been revealed as female and male determinants of SSI, respectively (Schopfer et al. 1999;Takasaki et al. 2000;Takayama et al. 2000). Another gene encoding S locus glycoprotein (SLG) has been identi ed in the S locus, although its precise role remains uncertain (Nasrallah et al. 1985;Watanabe et al. 2012). Sequences of SLG and S domain of SRK genes are highly homologous within the same S haplotypes (Sato et al. 2002;Lim et al. 2002;Okamoto et al. 2004). Based on sequence diversity of SLG and SRK genes, S haplotypes are classi ed into class I and class II Sato et al. 2002). Generally, S haplotypes of class I are dominant over those of class II and small noncoding RNAs and DNA methylation are involved in such dominance relationships (Tarutani et al. 2010;Yasuda et al. 2016).
SI was rst used in production of F 1 hybrid seeds of cabbage in 1940s by Japanese seed companies.
Since then, SI has been exploited for economical production of F 1 hybrid seeds of radish and Brassica crops (Muñoz-Sanz et al. 2020). To implement SI systems in radish F 1 hybrid breeding programs, speci c S haplotypes of parental lines should be identi ed to avoid cross-incompatibility between parental lines. At least 35 S haplotypes have been identi ed in radish (Haseyama et al. 2018). More than 50 and 100 S haplotypes have been identi ed in Chinese cabbage (Nou et al. 1993) and cabbage (Ockendon 2000), respectively.
For e cient identi cation of S haplotypes, polymerase chain reaction-restriction fragment length polymorphism (PCR-RFLP) methods have been developed for radish and Brassica crops (Brace et al. 1993(Brace et al. , 1994Nishio et al. 1994Nishio et al. , 1996Nishio et al. , 1997Sakamoto et al. 2000;Lim et al. 2002). Although PCR-RFLP is more e cient than laborious pollination tests, it has pitfalls such as complicated band patterns of digested PCR products and di culty in speci c PCR ampli cation of SLG or SRK genes due to the presence of multiple homologous SRK-like genes (Boyes et al. 1991;Kim and Kim 2018). To overcome these problems, a new S haplotyping system has been developed based on SLL2 and SP6 genes positioned at borders of S core regions and 31 S haplotypes have been identi ed from diverse breeding lines in our previous studies (Kim et  showing a SC phenotype was identi ed and the critical mutation responsible for the SC phenotype was revealed in this study. In addition, e cient application of this SC phenotype in radish breeding programs is discussed.

Plant materials
A radish breeding line (JNUR1537) showing an SC phenotype was introduced from a seed company (Farm Hannong Co., Ltd., Republic of Korea). Detail pedigree of this breeding line is unknown. JNUR1537 was crossed with JNUR1123 to produce F 1 hybrids by hand-pollination. An F 2 population was produced by self-pollination of F 1 hybrids. Three breeding lines (DBRL294-2, DBR2085 and DBR2086) containing the RsS3 haplotype were used to compare SI phenotypes with JNUR1537. S haplotypes of these three breeding lines have been identi ed in a previous study (Kim and Kim 2019). Evaluation of SI phenotypes were performed in greenhouses. Single plants were covered with mesh cages to prevent cross-pollination.
No insect pollinator was introduced to mesh cages. PCR ampli cation, genome walking, and sequencing of PCR products Total genomic DNAs were extracted from leaf tissues using a cetyl trimethylammonium bromide (CTAB) method (Doyle and Doyle 1987). PCR ampli cation and sequencing of PCR products of SLL2, SP6, and SRK genes of parental lines of the F 2 population were carried out following methods described in a previous study (Kim and Kim 2019). The detail protocol for high-resolution melting (HRM) analysis has been described in a previous study (Seo et al. 2020). SYTO®9 green uorescent nucleic acid stain (Thermo Fisher Scienti c, Waltham, MA, USA) was used as a dye. HRM peaks were obtained using a LightCycler® 96 system (Roche Molecular Systems, Pleasanton, CA, USA). Primer sequences of HRM markers are shown in Supplementary Table 1.
Genome walking was performed using a Universal GenomeWalker kit (Takara Bio) according to the manufacturer's instructions. Total genomic DNAs used to construct genome walking libraries were isolated from young seedlings using a DNeasy Plant Mini Kit (QIAGEN, Valencia, CA, USA). All PCR products were visualized on 1.5% agarose gels after ethidium bromide staining. Subsequently, PCR products were puri ed using a QIAquick PCR Puri cation Kit (QIAGEN). Sequencing was performed by a specialized company (Macrogen, Seoul, Republic of Korea). RNA extraction and reverse transcription polymerase chain reaction (RT-PCR) Total RNAs were extracted from oral buds using a RNeasy Plant Mini Kit (QIAGEN). cDNAs were synthesized using a cDNA synthesis kit (SuperScript™ III rst-strand synthesis system for RT-PCR, Invitrogen, Carlsbad, CA, USA). RT-PCR ampli cation was performed with an initial denaturation step at 94°C for 3 min, followed by 35 cycles of 94°C for 30 s, 68°C for 30 s, and 72°C for 2 min. It was nished with a nal 10-min extension step at 72°C. Primer sequences used in RT-PCR are listed in Supplementary  Table 1. Radish tubulin sequence (Rs395780) identi ed from a draft genome sequence (Jeong et al. 2016) was used as a control.

Construction of phylogenetic trees
Genomic DNA, cDNA, or deduced amino acid sequences of SLL2, SP6, SRK, and SCR/SP11 genes were aligned using a BioEdit software (Hall 1999). Large gaps in alignments were removed using Gblocks program (Castresana 2000). Phylogenetic trees were produced using MEGA version X (Kumar et al. 2018) with a neighbor-joining method. Node support of the phylogenetic tree was assessed using 1,000 bootstrap replicates.

Results
Discovery of an SC radish breeding line and inheritance patterns of the SC phenotype Radishes derived from JNUR1537 showed an almost SC phenotype. Compared with an SI radish accession (Fig. 1A), a large number of pods were formed in the SC radish without any insect pollinators (Fig. 1B). This SC plant was cross-pollinated with a SI breeding line (JNUR1123). F 1 hybrids showed the SC phenotype, indicating that the SC phenotype was dominant over the SI phenotype. Subsequently, an F 2 population was produced from self-pollination of F 1 hybrids. Ten individuals of the F 2 population were grown in a greenhouse. Each individual plant was separately covered with mech cages. SI phenotypes clearly segregated among F 2 plants (Table 1).  Fig. 4).
After genotyping F 2 seedlings, ve individuals of each genotype were separately grown in cages. While SC phenotypes were observed in both homozygous RsS3 and heterozygous individuals, few pods and seeds were formed in homozygous RsS32 individuals (Table 2). In addition, SI phenotypes of 10 F 2 individuals previously grown in isolated greenhouse were matched with genotypes of S haplotypes (Table  1). These results implied that the SC phenotype was probably related to S core genes such as SRK and SCR/SP11. Assembly of full-length genomic DNA sequences of S core genes in the SC radish To identify any defects in SRK and SCR/SP11 genes in the SC radish, full-length sequences of SRK and SCR/SP11 genes were analyzed. To identify the authentic SRK among duplicate SRK-like genes, full-length genomic DNA sequences of closely related SRK-like genes were obtained in this study. Using primers designed based on conserved regions among radish and three Brassica class II SRK genes ( Supplementary Fig. 5), 5' S domain sequences of SRK were isolated. After obtaining sequences of the last exon7 using genome walking, a full-length SRK sequence was assembled by connecting 5' S and 3' kinase domains through long PCR ampli cations (Fig. 2). In addition, approximately 2-kb 5' sequences containing a putative promoter was obtained by genome walking, resulting in a 18,133-bp full-length sequence. Meanwhile, a 6,200-bp full-length sequence of another putative SRK gene was obtained using genome walking PCRs with primers designed based on dissimilar sequences between two putative SRK genes (Fig. 2).
Since partial S domain sequences of the larger putative SRK gene containing a large-sized 11,865-bp intron 3 were identical to those of the SRK-26 deposited in the GenBank (LC341218), the S haplotype of the SC radish might be identical to the RsS-26 haplotype designated by Haseyama et al. (2018). Indeed, nucleotide sequences of SLG of the SC radish were identical to those of SLG-26 (LC341241) of the RsS-26 haplotype. In addition, partial sequences identical to those of RsSCR-26 (LC325812) were ampli ed in the SC radish. However, additional 552-bp full-length SCR/SP11 gene sharing 91.0% sequence identities with RsSCR-26 was obtained by genome walking in the SC radish, implying that both SRK and SCR/SP11 genes might be duplicated in the RsS-26 haplotype. The novel SCR/SP11 gene was closely related to other known SCR/SP11 genes ( Supplementary Fig. 6). This novel gene was designated as RsSCR-26-2. Its sequence was deposited into GenBank under the accession number of MZ383800. The full-length RsSCR-26-2 contained intact exons. Following the uni ed nomenclature suggested by Haseyama et al. (2018), the RsS3 haplotype was renamed as RsS-26 to avoid confusion.
Coding sequences of two putative SRK genes were also intact without any premature stop codons. They shared 91.7% nucleotide sequence identities with each other. Since S core gene sequences were identical to those of the RsS-26 haplotype, the large and small putative SRK genes were designated as RsSRK-26-1 and RsSRK-26-2, respectively. Full-length sequences of both SRKs were deposited into GenBank under accession numbers of MZ383801 and MT241389, respectively. The phylogenetic tree of radish SRK genes showed that RsSRK-26-1 and RsSRK-26-2 were closely related to each other (Fig. 3). Similar to other class II SRK genes, the RsSRK-26-1 gene contained a large-sized intron 3. However, a relatively small intron 3 was identi ed in the RsSRK-26-2 gene ( Supplementary Fig. 7), suggesting that the RsSRK-26-1 gene might be the genuine SRK in the RsS-26 haplotype.
Identi cation of a critical mutation responsible for the radish SC phenotype Interestingly, a transposable element-like sequence was found in the large-sized intron 3 of RsSRK-26-1 (Fig. 2). Target site duplication (TSD) of 4-bp ('GGAC') was found at anking regions of this element. LTR sequences of 284-bp in length positioned at both ends were perfectly identical to each other. This novel LTR-retrotransposon was designated as RsCopia1.
Since the ORF encoding a polyprotein was intact and both LTR sequences were identical to each other, RsCopia1 was assumed to be recently transposed into the RsSRK-26-1 gene. To investigate whether all RsS-26 haplotypes contained RsCopia1 in the SRK gene, three breeding lines (DBRL294-2, DBR2085 and DBR2086) found to contain RsS-26 haplotypes in the previous study (Kim and Kim 2019) were analyzed. Results of PCR ampli cation and sequencing showed no RsCopia1 insertion in any of these three breeding lines (Fig. 4A). Except for the RsCopia1 insertion, only two single nucleotide polymorphisms (SNPs) were identi ed in the intron 3 between normal and mutant RsSRK-26-1 alleles. Nucleotide sequence of the normal allele was deposited into GenBank with accession number of MT241388.
Unlike SC JNUR1537 containing the mutant SRK allele, three breeding lines harboring normal SRK alleles showed SI phenotypes (Fig. 4C). Signi cantly reduced numbers of pods and seeds were produced in these three SI breeding lines (Fig. 4C, Supplementary Fig. 8). Transcripts of the mutant RsSKR-26-1 were not detected in the SC radish compared with the normal RsSRK-26-1 in three SI breeding lines (Fig. 4B).
There was only a single SNP in putative promoter regions between normal and mutant RsSRK-26-1 alleles. These results suggest that transposition of RsCopia1 in the RsSRK-26-1 gene might be responsible for the inactivation of SRK and the resulting SC phenotype.

Discussion
Identi cation of duplicate SRK-like genes in the radish RsS-26 haplotype Full-length genomic DNA sequences of duplicated putative SRK and SCR/SP11 genes were obtained from the radish RsS-26 haplotype in this study (Fig. 2). Duplication of S core genes including SCR/SP11 and SRK and its effects on SI responses have been reported in Brassica rapa (Takada et al. 2005(Takada et al. , 2017 and Leavenworthia alabamica (Chantha et al. 2013), a member of Brassicaceae family. Takada et al. (2017) have demonstrated that duplicated SUI1 and PUI1 genes corresponding to SRK and SCR/SP11, respectively, control intraspeci c unilateral incompatibility in B. rapa. In the case of Leavenworthia, a novel SI system might have evolved from paralogs (LaLal2 and LaSCRL) of SRK and SCR/SP11 genes after loss of the original S locus, which is common in Arabidopsis, Brassica, and Leavenworthia (Chantha et al. 2013). The phylogenetic tree indicated that duplication of SRK homologs in the radish RsS-26 haplotype occurred more recently than that in SUI1 and LaLal2 genes (Fig. 3).
However, it was unlikely that duplication was very recent since signi cant sequence and length polymorphisms existed between duplicated SRKs (Fig. 2). Since duplicate SCR/SP11 genes were also identi ed, the entire S core region might be duplicated in the RsS-26 haplotype. Alternatively, two separate S core regions might have been merged by homologous recombination-mediated translocation. Further studies are needed to elucidate the exact duplication event and effects of S core region duplication. Isolation of full-length S core regions of the RsS-26 haplotype might provide a clue to resolve these issues.
As shown in B. rapa (Takada et al. 2005(Takada et al. , 2017 and Leavenworthia (Chantha et al. 2013), SI responses in the Brassicaceae family are complex processes. However, they are exible enough to adopt duplicate paralogous SRK and SCR/SP11 pairs and restore SI systems after losing the original S locus under su cient selection pressure. Multigene family of SRK homologs in Brassicaceae (Cock et al. 1995;Suzuki et al. 1997;Pastuglia et al. 1997;Kai et al. 2001) represents a potential source of such plastic evolution of SI systems. In radish, a total of 61 SRK homologs have been identi ed from two draft genome sequences (Kim and Kim 2018). Further studies such as functional characterization of duplicate SRK and SCR/SP11 genes and analysis of their effect on the strength of SI responses are needed to determine implications of S core region duplication in the evolution of SI systems in Brassicaceae family. The radish RsS-26 haplotype harboring duplicate S core genes represent a valuable material for such studies in the future.

Identi cation of a radish breeding line showing a SC phenotype and its application in radish breeding
A radish showing a SC phenotype was identi ed in this study. To the best of our knowledge, this is the rst study to report an SC radish. Among Brassica species, several SC mutants have been previously reported. In Brassica rapa, SC mutants have been identi ed from two cultivars, Yellow Sarson (Fujimoto et al. 2006a) and Dahuangyoucai (Zhang et al. 2013), containing similar mutant S haplotypes. Another SC phenotype induced by gene conversion from SLG to SRK has been reported (Fujimoto et al. 2006b). In addition, an SC B. rapa has been arti cially developed by silencing of SCR/SP11 using RNAi (Jung et al. 2012). In the case of B. oleracea, deletion of exon1 and 2 of SRK is responsible for an SC phenotype (Nasrallah et al. 1994). Recently, eight quantitative trait loci (QTLs) controlling a SC phenotype have been identi ed in an inbred line of B. oleracea (Xiao et al. 2019).
Duplicate SRK-like genes were identi ed from the SC radish in this study. Some evidences indicated that the large-sized RsSRK-26-1 might be a genuine SRK. First of all, transcription of RsSRK-26-1 was inactivated in the SC radish in contrast to three other SI breeding lines containing normal RsSRK-26-1. These results showed a direct relationship between SI phenotypes and RsSRK-26-1. In addition, the RsSRK-26-1 gene contained a large-sized intron 3 as shown in other Brassica class II SRK genes ( Supplementary Fig. 7). Further functional studies are needed to clarify exact roles of both duplicate SRK genes. Since the RsSRK-26-2 gene contained intact exons and its transcripts were more abundant than those of RsSRK-26-1 (Fig. 4B), further functional characterization of RsSRK-26-2 might be an intriguing topic.
An intact LTR-retrotransposon was identi ed in the large-sized intron 3 of RsSRK-26-1 of the SC radish in this study. Since this element was transposed into an intronic region, this insertion might not have any effect on transcription of SRK. Transcripts of RsSRK-26-1 were not detected in the mutant allele. Because there was only one single SNP in approximately 2.0 kb putative promoter regions between mutant and normal RsSRK-26-1 alleles, insertion of RsCopia1 might be responsible for blockage of transcription. DNA methylation of promoter regions of RsSRK-26-1 is assumed to be induced by transposition of RsCopia1, although further functional analyses are required. DNA methylation is known to be involved in silencing of transposable elements in plants ( Bartels et al. 2018). Similarly, transcripts of the SRK gene are not detected in SC Yellow Sarson probably due to insertion of an LTR-retrotransposon in the intron 1, although there is no critical mutation in their promoter regions (Fujimoto et al. 2006a). In another case, reduced expression of the FLC gene is caused by insertion of an LTR-retrotransposon in the rst intron (Michaels et al. 2003 Fig. 9).
The SC radish identi ed in this study would be a valuable material for radish F 1 hybrid breeding. Due to unstable SI phenotypes of some inbred lines, inadvertent self-pollination of maternal lines of F 1 hybrids frequently can result in a low genetic purity of F 1 cultivars. For this reason, male-sterility has replaced SI systems as a more stable genetic emasculation tool. Although male-sterility is used for F 1 hybrid breeding, propagation of inbred parental lines should be performed by self-pollination of parental lines using high concentrations of CO 2 . However, if the SC phenotype is introgressed to parental lines, such expensive treatment with CO 2 might be unnecessary. Regarding xed varieties, the SC phenotype might greatly improve seed yields. Taken together, the SC radish identi ed in this study will become an important material for radish breeding programs. Figure 1 Phenotypes of self-incompatible and self-compatible radish breeding lines at the stage of seed harvest.