Introduction

Cytoplasmic male sterility (CMS), a maternally inherited trait that prevents plants from producing functional pollen, is a widespread phenomenon in higher plants. The previous studies have led to the belief that unusual and often chimeric open reading frames (ORFs) arising from mitochondrial genome rearrangement are responsible for CMS (Hanson and Bentolila 2004; Chase 2007; Fujii and Toriyam 2008; Chen and Liu 2014). In many cases, specific dominant nuclear genes termed “restorer of fertility” (Rf) have been found, which can suppress the male sterile phenotype at different molecular levels, such as genomic, post-transcriptional, translational or post-translational, and metabolic levels, and restore the fertility in plants carrying the CMS mitochondrial genome (Chen and Liu 2014). CMS/Rf systems have not only facilitated hybrid seed production, but are also an ideal model for studying cytoplasmic inheritance and the crosstalk between the mitochondrial and nuclear genomes.

The restoration of one CMS-associated gene by several different Rf genes is a common phenomenon in many plants. In maize, Rf2, in combination with one of three other Rf genes (Rf1, Rf8, and Rf*) can fully or partially restore the fertility of T-CMS maize (Wise et al. 1999). In radish, Ogura CMS can be restored by two Rf genes, Rfo and RsRf3 (Uyttewaal et al. 2008; Wang et al. 2013). Similarly, A1-CMS of sorghum and BT-CMS, HL-CMS, and WA-CMS of rice also have two Rf genes (Liu et al. 2004; Klein et al. 2005; Wang et al. 2006; Jordan et al. 2010; Luo et al. 2013). To date, more than ten Rf genes have been isolated from different plant species. These include Rf2 for maize T-CMS (Cui et al. 1996), PPR592 for petunia pcf-CMS (Bentolila et al. 2002), Rfo and Rfk for radish Ogura-CMS and Kosena-CMS (Brown et al. 2003; Desloire et al. 2003; Koizuka et al. 2003), PPR13 for sorghum A1-CMS (Klein et al. 2005), Rf1a and Rf1b for rice BT-CMS (Wang et al. 2006), Rf17 for rice CW-CMS (Fujii and Toriyam 2009), Rf2 for rice LD-CMS (Itabashi et al. 2011), Rf5 for rice HL-CMS (Hu et al. 2012), Rf4 for rice WA-CMS (Tang et al. 2014), bvORF20 for sugar beet Owen CMS (Kitazaki et al. 2015), and Rfp for rapeseed pol CMS (Liu et al. 2012, 2016). All the cloned Rf genes encode a pentatricopeptide repeat-containing (PPR) protein, except Rf2 of maize, bvORF20 of sugar beet, and Rf17 and Rf2 of rice.

Polima CMS (pol CMS), discovered by Fu (1981), was the first practically valuable CMS used in hybrid breeding of rapeseed. Since the mapping of Rfp-to-B. napus linkage group 18 by Jean et al. (1997), several researchers have identified tightly linked molecular markers at the Rfp locus. Recently, Liu et al. (2012) mapped the Rfp locus to a 29.2-kb region on chromosome A09 of B. napus and predicted that a PPR gene was the most likely candidate for Rfp. Using transgenic rapeseed plants, they confirmed that the PPR gene, indeed, functions as Rfp (Liu et al. 2016).

Pol cytoplasm has been transferred into Chinese cabbage (Brassica rapa L. AA) by interspecific cross and has become an important CMS type for hybrid breeding in Chinese cabbage (Ke and Song 1989; Ke et al. 1992). By extensive test crossing, several restorer lines that can fully or partially restore pol CMS have been screened in Chinese cabbage (Ke and Zhang 1993; Zhang and Ke 1994). However, it remains unclear whether the Rf genes of different restorer lines were the same. In the restorer line 01s325, Xu et al. (2014) mapped BrRfp locus on chromosome A09 and identified two markers, SSR03 and SSR2528, which co-segregated with the BrRfp locus. However, the restorer genes have not been cloned and the molecular mechanism of fertility restoration by the pol CMS/Rf system needs to be investigated.

In this study, two Rf loci were confirmed from four restorer lines of pol CMS in Chinese cabbage by allelism analysis. Moreover, using simple sequence repeat (SSR) markers developed from B. rapa reference genome and InDel markers derived from whole-genome resequencing of two parental lines, BrRfp1 of restorer line 92s105 was mapped to a 12.6-kb region that contained a potential candidate gene encoding PPR. Based on the candidate gene sequences of the restoring and nonrestoring alleles, a molecular marker co-segregating with BrRfp1 locus was developed, which would be useful for breeding via marker-assisted selection (MAS) and for the screening of new restorer lines.

Materials and methods

Plant materials

Four restorer lines, 92s105, 01s325, 00s109, and 88s148, with very large phenotypic differences were screened from Chinese cabbage germplasms (Ke and Zhang 1993; Zhang and Ke 1994). The lines 92s105, 01s325, and 00s109 could fully restore the fertility of pol CMS (Supplementary Fig. S1c), whereas 88s148 could only partially restore it (Supplementary Fig. S1b). The line 94c9 is a pol CMS line. For allelism analysis of the restorer loci in the four restorer lines, five crosses, 92s105 × 01s325, 92s105 × 00s109, 00s109 × 01s325, 01s325 × 88s148, and 92s105 × 88s148, were performed to produce the F1 generation. These five F1 plants were subsequently testcrossed with 94c9. For fine mapping of the restorer gene, the BC1F1 population was produced from the backcross between 94c9 (the acceptor parent) and 92s105 (the donor parent), and one male fertile (heterozygous in BrRfp1 locus) individual from the BC1F1 population was self-pollinated to produce BC1F2 population.

All the seeds used in this study were obtained from the Chinese cabbage research team at the College of Horticulture, Northwest A&F University, Yangling, China. All the plants were grown from seedlings in a greenhouse in the early January and were transplanted to the experimental field at the Northwest A&F University, in late February of 2015 and 2016. The plants were vernalized naturally. The male fertility of each individual was assessed by observing at least three flowers during the flowering period twice, at 7-day interval.

Development of SSR and InDel markers

To develop SSR markers, sequences containing five SSR markers identified from the restorer line 01s325 by Xu et al. (2014) were downloaded from the B. rapa reference genome (http://brassicadb.org/brad/). The SSR sites were searched using the software SSRHunter (Li and Wan 2005). To develop InDel markers, genomes of the two parental lines, 94c9 and 92s105, were sequenced with HiSeq X Ten (Macrogen, Shenzhen, China) at 99- and 91-fold sequencing depths. After removing duplication and finding the variant with Isaac Variant Caller (IVC), the information for each variant was classified by chromosomes or scaffolds based on the B. rapa reference genome available at NCBI Website. Only the insertions or deletions >4 bp were chosen to design the InDel primers. The designing of these primers was based on the B. rapa reference genome region, and 150-bp upstream and downstream of the InDel sites, respectively. The primers for the SSR and InDel markers were designed by Primer Premier 5.0 (http://www.premierbiosoft.com/primerdesign/). All the primers used in this study were synthesized by Sangon Biotech Co., Ltd (Shanghai, China).

Genomic DNA extraction, PCR, and gel electrophoresis

Genomic DNA was extracted from young leaves by the cetyl trimethylammonium bromide (CTAB) method as described by Porebski et al. (1997). PCR was performed in 10-μL reaction mixtures containing 1-μL DNA template (50 ng/μL), 0.3-μL forward primer (10 μM), 0.3-μL reverse primer (10 μM), and 5-μL 2 × Taq Master Mix for PAGE (Vazyme, Nanjing, China). The PCR protocol was as follows: 94 °C for 4 min followed by 28–30 cycles of 94 °C for 30 s, 55 °C for 30 s, and 72 °C for 30 s, and a final extension at 72 °C for 5 min. The PCR products were separated on 9% nondenaturing polyacrylamide gel.

Construction of genetic and physical maps and identification of candidate gene

The segregation data of the polymorphic molecular markers and the phenotypic data for each individual in the population were used to construct the genetic linkage map by the JoinMap 4.0 software with a logarithm of the odds (LOD) threshold score of 6.0. The physical map of BrRfp1 locus was constructed using the physical position of each marker on the chromosome in the NCBI B. rapa reference genome. Gene annotation of the delimited region between the two closest markers was performed according to the annotation results of B. rapa reference genome in NCBI (http://www.ncbi.nlm.nih.gov/genome/?term=Brassica+rapa) and BRAD (Brassica database, http://brassicadb.org/brad).

Cloning and sequence analysis of the candidate gene

The specific primers for the candidate gene were designed according to the reference genome of B. rapa. The PCR products were purified using a DNA Gel Extraction Kit (Tiangen, Beijing, China) and used for TA cloning. The purified PCR products were introduced into the pMD18-T Simple Vector (Takara, Dalian, China) and transformed into E. coli strain DH5α. The recombinant plasmids were sequenced by Sangon Biotech Co., Ltd (Shanghai, China). The complete coding sequences of the eight Chinese cabbage lines were submitted to GenBank under the following accession numbers: 92s105 (KX671967), 01s325 (KX671968), 00s109 (KX671969), 12CF29 (KX671970), xiakang2 (KX671971), 94c9 (KX671972), 12CF9 (KX671973), and 88s148 (KX671974). Sequence alignment was performed with MegAlign of the DNA Star Lasergene package (http://www.dnastar.com). The PPR motif prediction of the candidate gene was carried out with PPRFinder Online (http://www.plantppr.com).

Results

Allelism analysis of the restorer locus in the four restorer lines

To analyze the allelism of the restorer locus, five populations were generated (Table 1). All the testcross progenies generated from 92s105 × 01s325, 92s105 × 00s109, and 00s109 × 01s325 were fully fertile. However, the testcross progenies generated from 01s325 × 88s148 and 92s105 × 88s148 were fertility segregates and the segregation ratio of fully fertile to partially fertile and male sterile plants best fitted a 2:1:1 ratio (Table 1). These results indicated that the restorer genes of 92s105, 01s325, and 00s109 were at the same locus (designated as BrRfp1), but the restorer gene of 88s148 was at a different locus (designated as BrRfp2). The results of allelism analysis were consistent with the phenotypic observation that 92s105, 01s325, and 00s109 restore the fertility of pol CMS completely, whereas 88s148 restores the fertility of pol CMS partially. In addition, BrRfp1 is epistatic to BrRfp2 in the restoration of fertility for pol CMS.

Table 1 Results of BrRfp allelism analysis of four restorer lines
Full size table

Genetic analysis of BrRfp1 and preliminary mapping

A BC1F1 population containing 487 individuals was generated from a backcross between the male sterile line 94c9 and the restorer line 92s105. Phenotypic investigation showed the segregation of 247 fertile individuals and 240 sterile individuals. The χ 2 test revealed that the segregation ratio was consistent with the expected ratio of 1:1 (χ 2 = 0.101, P > 0.05), indicating that the male fertility was controlled by a single dominant gene, the same as the restorer line 01s325 (Xu et al. 2014).

For preliminary mapping of BrRfp1, five SSR markers on chromosome A09, identified by Xu et al. (2014) from the restorer line 01s325, were first screened in the two parental lines of BC1F1 population; however, no polymorphism was observed except for SSR2331. Based on the B. rapa reference genome sequence containing these five SSR markers, 121 more SSR markers were developed. Six markers (SSR14, SSR16, SSR18, SSR1301, SSR1424, and SSR1447) (Supplementary Table S1) exhibited polymorphism in the two parental lines. Thereafter, these seven SSR markers were used to assay 487 individuals of the BC1F1 population and the BrRfp1 locus was flanked by SSR1301 and SSR1424 with a genetic distance of 0.4 and 0.8 cM, respectively (Fig. 1a). To further narrow the delimited region of BrRfp1 locus, 100 primer pairs that amplified about 1.0-kb products were designed. Three primer pairs (A620, A801, and A802) only amplified products from the restorer line 92s105; these three primer pairs were used as dominant markers. When the primer pairs could amplify product from both the parental lines, the product was sequenced to locate the InDel site. Two such sites were found after sequencing the amplified product and then developed into InDel markers (InDel817 and InDel820). Using these five markers (Supplementary Table S1), the previously selected recombinants were tested. The BrRfp1 locus, co-segregating with A801 and A802, was localized to a region between A620 and InDel 817 (Fig. 1a). The genetic distances of A620 and InDel817 were both 0.2 cM, and the physical distance between A620 and InDel 817 was about 132.5 kb.

Fig. 1
figure 1

Genetic and physical maps of the BrRfp1 locus. a Genetic map of BrRfp1 locus based on 487 BC1F1 population. b High-resolution linkage map of BrRfp1 locus based on 2485 BC1F2 population. c Physical map of molecular markers in b on the chromosome A09 (the physical position on the right side is based on the B. rapa reference genome in NCBI)

Full size image

Fine mapping of BrRfp1 locus using whole-genome resequencing data

To fine map the BrRfp1 locus, a BC1F2 population with 2485 individuals generated from one heterozygous male fertile individual in the BC1F1 population was produced. To develop more InDel markers in the delimited region, the genome sequences of the two parental lines, 94c9 and 92s105, were obtained by whole-genome resequencing. Based on the whole-genome resequencing data, 24 InDel markers were developed and 9 of them showed polymorphism in the parental lines (Supplementary Table S1). A620, a dominant marker, which is not suitable for detecting recombination events in BC1F2 population, was replaced by a co-dominant marker, Indel351, distanced about 1.5 kb from A620. InDel351 and InDel817 were used to screen the recombination events among BC1F2 populations. A total of 41 recombinants were detected, including 24 recombination events for InDel351 and 17 recombination events for InDel817. These recombinants were further detected by eight other InDel markers. The segregation data of these ten markers and the male fertility survey data of BC1F2 population were applied to construct a high-resolution linkage map (Fig. 1b). All these markers were mapped at a 1.0-cM region around the BrRfp1 locus. Three of these markers were located on one side of the BrRfp1 gene, six were on the other side, and one (InDel441) co-segregated with the BrRfp1 gene. The order of these markers on the linkage map was consistent with their physical position on chromosome A09 of B. rapa reference genome (Fig. 1c). Finally, the BrRfp1 gene of Chinese cabbage was mapped to a region, about 12.6 kb, between InDel431 and InDel451, with a genetic distance of 0.12 cM and 0.04 cM, respectively (Fig. 2).

Fig. 2
figure 2

Identification of BrRfp1 as a candidate gene. BrRfp1 is restricted to the region between the markers InDel 431 and InDel 451. This region is about 12.6 kb on the chromosome A09 based on the reference genome of B. rapa and there are three annotated genes in this region

Full size image

BrRfp1 candidate gene identification and sequence analysis

According to the gene annotations of this 12.6-kb region in BRAD and NCBI, there are three genes in this region (Table 2), but only one (LOC103842088) encodes for the PPR protein. Therefore, LOC103842088 was considered as the candidate gene of BrRfp1 in Chinese cabbage.

Table 2 Gene annotation in mapped region of B. rapa reference genome
Full size table

Using the primers Rfp088 designed against B. rapa genome sequences present in NCBI, the full-length sequences of LOC103842088 from the two parental lines, 92s105 and 94c9, were cloned and sequenced. The sequences analysis revealed that LOC103842088 from 92s105 and 94c9, both had no introns and encoded a PPR protein, belonging to P-class, with 650 amino acids comprising 16 PPR motifs. There were 47 single-nucleotide polymorphisms (SNPs) in the coding region between 92s105 and 94c9 (Supplementary Fig. S2). These SNPs had no effect on the deduced protein length but resulted in mutations of 32 amino-acid residues (Supplementary Fig. S3). Liu et al. (2016) verified that Rfp, the restorer gene for pol CMS in rapeseed, was a PPR gene on chromosome A09. The deduced protein sequences of the BrRfp1 candidate gene and Rfp were compared. There was a difference in five amino-acid residues; however, none of them were located on the eight key code positions indicated by Liu et al. (2016) (Supplementary Fig. S3).

To identify the functional variations between the restoring BrRfp1 and nonrestoring Brrfp1 alleles in Chinese cabbage, the genes from other three restorer lines (01s325, 00s109, and 88s148) and three male sterile lines (xiakang2, 12CF9, and 12CF29) were sequenced. The deduced protein sequences from 01s325 and 00s109 were identical to each other, and had differences in three amino-acid residues compared to 92s105. In 88s148, the sequence of the first six PPR motifs was the same as that of 92s105, but the sequence after ten PPR motifs was the same as that of 94c9 (Supplementary Fig. S3). These results are consistent with the allelism analysis and phenotypic observation in the four restorer lines. In addition to the two rapeseed sequences, Rfp and rfp (Liu et al. 2016), ten deduced protein sequences (including four restoring genes and six nonrestoring genes) were aligned. Among these deduced protein sequences, differences in three amino-acid residues located in the 11th PPR motif of all the six nonrestoring genes were found (Fig. 3).

Fig. 3
figure 3

Structure and homology analyses of the BrRfp1 sequence. a Deduced protein sequence structure of BrRfp1. b Sequence alignment of the eleventh PPR motif in four restoring and six nonrestoring alleles. The mutant amino acids are indicated in black. The red rectangle indicates the mutations at all the nonrestoring alleles (color figure online)

Full size image

Development of a co-segregation molecular marker of BrRfp1

Based on the nucleotide polymorphisms of the candidate gene, BrRfp1, between the restorer lines and sterile lines, a dominant marker SC718 (Supplementary Table S1) was developed (Supplementary Fig. S2). Individuals (2485 in number) of BC1F2 population were detected using SC718 and the results were entirely consistent with the phenotype (Fig. 4). To detect the transferability of SC718 in the different populations, the F2 population, 14F2Q21, derived from the cross 94c6 × tkxf was tested using SC718 and another co-segregating marker, InDel441. SC718 also co-segregated with the phenotype, but InDel441 showed no polymorphism (data not shown). Therefore, SC718 could be used as a reliable molecular marker for MAS breeding and screening of new restorer germplasms containing the BrRfp1 locus in Chinese cabbage.

Fig. 4
figure 4

Co-segregation analysis in the partial BC1F2 population using SC718 marker. M DL2000 marker, F male fertile plant, S male sterile plant

Full size image

Discussion

Among the four restorer lines of Chinese cabbage used in this study, 92s105, 01s325, and 00s109 could fully restore pol CMS, but 88s148 could only partially restore pol CMS. Both allelism and candidate gene sequence analysis indicated that the restorer locus of 88s148 was a new locus different from other three restorer lines. The pol cytoplasm of Chinese cabbage was introduced from rapeseed by an interspecific cross (Ke and Song 1989, Ke et al. 1992). In both the Chinese cabbage and rapeseed, restorer lines having different restorer capabilities were determined (Fu et al. 1989; Ke and Zhang 1993). The phenomenon of the partial restoration of infertility was also found in maize. In maize, except for the full restorer gene Rf1, two partial restorer genes, Rf8 and Rf*, for T-CMS were found, which had a distinct way of processing the T-urf13 mitochondrial transcripts (Dill et al. 1997; Wise et al. 1999). In Chinese cabbage and rapeseed, it had been found that some pol CMS lines were sensitive to temperature (Fu et al. 1989; Zhang and Hao 2001). These germplasms could be applied to “two lines” hybrid breeding. The previous studies were all focused on mapping and cloning of the restorer genes with full restorer capability, and rarely considered the partial restorer and temperature sensitive restorer. The cloning of partial restorer and temperature sensitive restorer genes would provide more insights into the molecular mechanisms of nuclear–mitochondrial interactions in the pol CMS/Rf system.

For the development of both the SSR and InDel markers, a great deal of genomic sequence information is required. The release of Chinese cabbage whole-genome sequence (Wang et al. 2011) was a great convenience for the development of the SSR markers; however, this strategy is inefficient and time consuming. In this study, only six (5%) SSR markers showed polymorphism between 94c9 and 92s105 among the 121 SSR markers derived from B. rapa reference genome sequences. Furthermore, 100 primer pairs, which were used to develop the InDel markers after sequencing and comparison of the amplification product, were designed according to the sequences of the delimited region downloaded from the B. rapa reference genome. Using this method, only two (2%) InDel markers were developed. For developing more InDel markers, the whole genomes of 94c9 and 92s105 were sequenced at about 100-fold sequencing depths. By comparing the sequenced genomes to the B. rapa reference genome, a large number of SNPs and InDels were found. Only in the preliminary mapped region (about 132.5-kb), there were 1520 SNPs and 280 InDels. Twenty-four InDel primers were designed and synthesized; nine (37.5%) of them showed amplification polymorphism. These InDels could greatly accelerate the process of fine mapping the target gene. Finally, BrRfp1 from the restorer line 92s105 was successfully mapped to a 12.6-kb region. Therefore, the whole-genome resequencing is a very efficient method to develop InDel markers.

When BrRfp1 locus was mapped to a 132.5-kb region using the BC1F1 population, there were two genes, Bra026927 (BRAD) and LOC103843088 (NCBI), distanced about 50.3 kb, which encoded the PPR proteins. The phenomenon of the distribution of PPR genes in clusters is also found in other mapped restorer loci. For example, in the petunia Rf locus, there are two PPR genes adjacent to each other, sharing 93% similarity in the predicted protein sequences (Bentolila et al. 2002). The Rfo locus of radish was also found to contain three clustered genes encoding similar PPR proteins (Desloire et al. 2003; Brown et al. 2003). In rice, the Rf loci of BT-CMS, HL-CMS, and WA-CMS were all mapped to a region about 310 kb on the chromosome 10 containing nine genes encoding the PPR proteins (Komori et al. 2004; Wang et al. 2006; Hu et al. 2012; Tang et al. 2014). In Arabidopsis thaliana, most Rf-PPR-like (RFL) genes are mainly clustered in two regions on chromosome 1 (Fujii et al. 2011). Rf and RFL genes are clustered in the chromosomal regions possibly because of their rapid and dynamic evolution through gene duplication and functional divergence (Fujii et al. 2011).

Because the restorer lines for pol CMS were found from B. napus, B. rapa, and B. juncea (AABB), the restorer gene for pol CMS was believed to be present on the A genome of the Brassica genus (Fu et al. 1989; Yang and Fu 1991). Liu et al. (2016) verified a PPR gene on the chromosome A09 of rapeseed, which functioned as an Rfp in transgenic plants. In rapeseed, the changes in eight amino-acid residues located at positions 2, 5, and 35 of the PPR motifs were considered as the cause of functional divergence. In this study, BrRfp1 of the restorer line 92s105 was also mapped to a PPR gene on the chromosome A09, which shared high similarity to Rfp in B. napus, with only five different amino-acid residues present in the predicted protein sequences. In the restorer lines 01s325 and 00s109, the deduced protein sequences of BrRfp1 were highly similar to Rfp, having only two different amino-acid residues (Supplementary Fig. S3). All the different amino-acid residues were not in the key position pointed out by Liu et al. (2016). Therefore, the restorer gene, BrRfp1, for pol CMS in Chinese cabbage is probably identical to the restorer gene, Rfp, in B. napus. Chinese cabbage and rapeseed, which originated independently in China and Europe, are two species of the Brassica genus. Although Chinese cabbage and rapeseed have independently evolved more than 1000 years (Nagaharu 1935) ago, they may have the same restorer gene for pol CMS. The probable reason for this could be that the cytoplasm and restorer gene of pol CMS originated from ancient B. rapa and then integrated into B. napus genome with the hybridization between ancient B. rapa and ancient B. oleracea. Whether the restorer gene for pol CMS in B. juncea is identical to that in B. rapa and B. napus needs further investigation.

In angiosperms, the PPR gene family is one of the largest gene families in nearly all the genomes sequenced so far (Cheng et al. 2016). In B. rapa, there are 592 PPR members; 347 of them are P-class PPR proteins (http://www.plantppr.com). Almost all the PPR proteins are targeted to plastids and mitochondria, where they play important roles in post-transcriptional processes, including RNA stabilization, RNA cleavage, intron splicing, RNA editing, and in the initiation of translation (Barkan and Small 2014). One PPR motif has the ability to recognize a single nucleotide, and a specific RNA sequence is recognized by an array of PPR domains (Yin et al. 2013; Barkan and Small 2014). A few studies reported that the amino-acid residues at the positions 2, 5, and 35 of a PPR motif were the molecular determinants for RNA-binding specificity (Fujii et al. 2011; Yin et al. 2013). For example, there are two different amino-acid substitutions at the position 5 in two different PPR motifs of a nonrestoring allele of Rfk in radish CMS (Koizuka et al. 2003). An amino-acid mutation at the position 35 of the third PPR motif results in RFL9 losing its function of processing the rps3 and orf240 transcript in the mitochondria of Arabidopsis thaliana (Arnal et al. 2014). The comparison of the four restoring and six nonrestoring alleles for pol CMS in Chinese cabbage and rapeseed revealed that the three amino-acid residue variations at the positions 2, 4, and 6 of the 11th PPR motif were present in all the nonrestoring alleles (Fig. 3). The alteration of Trp406 to Val, caused by the change of TG1216 to GT at the position 2, could probably account for the loss of the restoration function (Supplementary Fig. S3).

Author contribution statement

HZ constructed the mapping populations, performed marker development and mapping analysis, and wrote the paper. JW assisted with phenotypic survey and DNA extraction, ZD, MQ, and LH assisted to extract DNA. YR and QL provided valuable research ideas. LZ designed and supervised the study.