Identification of a New Wide-Compatibility Locus in Inter-Subspecific Hybrids of Rice (Oryza sativa L.)

Abstract

As a special class of rice germplasm, wide-compatibility varieties (WCVs) guarantee the fertility of hybrids when there is cross-fertilization between two subspecies. In this study, Chenghui9348 was identified as a new member of the WCV family that improves pollen fertility in an inter-subspecific hybrid. Cytological analysis showed that the abnormal mitosis of microspores resulted in the sterility of pollens at the early bicellular stage in the inter-subspecific hybrid. Furthermore, the new F12 locus, corresponding to improvements in fertility of the indica-japonica hybrid, was found to co-segregate with the RM1047 marker and associated with a region of approximately 630 kb flanked by the D1101 and D1164 markers on chromosome 12. In this region, two putative genes were predicted as the candidates for wide-compatibility genes (WCGs). Sequence analysis revealed that, compared with indica/japonica alleles, deletion/insertion occurred within exons of both putative genes. Together, the present study identified another new WC locus, F12, and offers more opportunities for further exploitation of inter-subspecific hybrids in rice.

1. Introduction

Hybrids between subspecies often lead to hybrid incompatibilities, such as sterility and inviability. Hybrid incompatibility hinders the gene exchange between subspecies, which is a major obstacle to heterosis utilization [1]. Cultivated rice is divided into two species, African rice (Oryza glaberrima Steud) and Asian rice (Oryza sativa L.). Inter-specific hybrids were made for African rice; the New Rice for Africa (NERICA) varieties exhibited high performance in terms of abiotic and biotic stress resistance, and achieved higher yields compared with its parental varieties [2]. The Asian cultivated rice (Oryza sativa L.) was classified into two main subspecies, indica and japonica [3]. The inter-subspecific hybrids have stronger hybrid vigor than the hybrid within the subspecies; however, the sterility barriers in inter-subspecific hybrids prevent the application of heterosis to hybrid rice breeding programs, and the utilization of rice heterosis was limited to indica-indica hybrids, which increased the grain yield by about 20% [1,3]. Fortunately, the discovery of wide-compatibility varieties (WCVs), which allow for the progeny of indica and japonica crossings to exhibit the fertility of normal hybrids, brought breeders powerful tools for exploiting the heterosis between the two subspecies and for further improving grain yield [4].

Nowadays, several genetic mechanisms, such as the duplicate gametophytic lethal model, allelic interaction at a single locus, and epistatic interaction between loci, have been proposed to explain indica/japonica hybrid sterility [5,6].

In 1962, Kitamura first put forward a one-locus sporo-gametophytic interaction model, which could explain the genetic behavior of most hybrid sterility loci [5], and then Ikehashi and Araki demonstrated that a locus named S5 was consistent with this model [4]. According to this model, there are three alleles at the S5 locus: an indica allele, $S_5^i$, a japonica allele, $S_5^j$, and a neutral allele (wide-compatibility allele), $S_5^n$. The hybrid progeny between indica and japonica would be partially sterile, with a $S_5^i$/$S_5^j$ genotype. However, the progenies bearing an $S_5^i$ or $S_5^j$ allele would be fully fertile when WCVs (bearing at least one $S_5^n$ allele) are crossed to indica or japonica. Ikehashi and Araki also found that the S5 locus was located on chromosome 6 by using morphological markers [4], and this result has been further confirmed by using isozymes and molecular markers [7,8]. Then, Qiu and Ji analyzed the S5 locus by using near-isogenic lines (NILs) to delimitate the S5 locus to a 40 kb- and 50 kb-genomic DNA segment, respectively [9,10]. Three open reading frames (ORF3, ORF4, and ORF5) comprise the triallelic system of the S5 locus as shown by Chen et al., and Yang et al. proposed to refer to them as a “Killer-Protector System” [11,12].

However, the progeny showed low fertility when some WCVs were crossed to different indica or japonica varieties [13]. Thus, hybrid sterility was considered to be caused by allelic interactions at many different loci [14,15,16]. By using different cross combinations involving WCVs, many sterility gene loci other than S5 have been found. Since then, more than twenty loci conferring embryo sac sterility and thirty loci conferring male sterility have been found to influence hybrid sterility [17]. Hopefully, it will be possible to overcome the hybrid sterility in rice breeding when there is an understanding of the mechanism of these sterility gene loci. However, we still cannot solve the sterility of the inter-subspecific hybrids, even though the main wide-compatibility allele, i.e., $S_5^n$, has been cloned, and it is necessary to exploit more WCGs to explain the underlying mechanism. Here, we report a newly identified locus conferring intercrossing compatibility in Chenghui9348, a rice variety carrying an S5-i allele that uses 9311 and Lemont as parental varieties. We found that this locus, named F12-C, has extensive effects on both pollen and spikelet fertility. These results will help clone this gene and aid in its marker-assisted selection in rice-breeding programs.

2. Materials and Methods

2.1. Plant Materials and Mapping Populations

The 9311 is an indica cultivar that has an S5-i allele. Nipponbare and Balilla are typical japonica cultivars carrying the S5-j allele at the S5 locus. Chenghui9348, which was developed by the Sichuan Academy of Agricultural Sciences, also has an S5-i allele, as shown by sequencing the S5 locus. A series of hybrids between parents were made in the winter of 2013 in Hainan and planted in the rice-growing season of 2014 in Ezhou, Hubei province. A population of 372 plants from the three-way cross, 9311/Chenghui9348//Nipponbare, were planted in an experimental field of the Hybrid Rice Hainan Experimental Base of Wuhan University in Lingshui, Hainan Province, in November 2013. Another two three-way cross populations (9311/Chenghui9348//Nipponbare, 9311/Chenghui9348//Balilla) were planted in the summer rice-growing season of 2014 in the experimental field of the Hybrid Rice Ezhou Experimental Base of Wuhan University in Ezhou, Hubei Province. The order of the parents in these two three-way cross populations ensures that the population has an $S_5^i$/$S_5^j$ genotype so that it can counteract the effect of the S5 site. All materials were planted with an interval of 16.5 cm. Plots were spaced 23.5 cm apart. The crop was managed following normal commercial practices.

2.2. Evaluation of Pollen and Spikelet Fertility

Six florets from three panicles per plant were collected 1–2 days before flowering and fixed in 70% ethanol. Anthers were mixed and stained with 1% iodine potassium iodide (I₂-KI) solution, and more than 300 pollen grains from each individual were observed by light microscope to estimate the percentage of fertile grains. The fully stained pollens were fertile, and partially stained ones were abortive. The rice spikelet fertility is the ratio of fertile spikelets to total spikelets obtained by counting three panicles on the upper half of the panicles for each plant, as described by Wan [15].

2.3. Cytological and Histological Analysis

The F1 (9311/Nipponbare, Chenghui9348/Nipponbare) hybrid spikelets of different stages were collected from florets and then fixed in formalin fixative (formalin: ethanol (50%, v/v): glacial acetic acid = 18:1:1, v/v/v). The pollen grains were stained in acetocarmine. Analysis of pollen germination on the stigma was performed and the pollen was stained with aniline blue 1 h post-anthesis; ten florets were accounted for per plant and observed by confocal laser scanning microscopy, as described by Zhou [18].

2.4. DNA Preparation and PCR Analysis

DNA was extracted from fresh leaves of each plant and dissolved in TE buffer (10 mM Tris, 0.1 mM EDTA) before the quality test. The eligible samples were diluted to 20 ng/μL with double distilled water (ddH₂O) and stored at 4 °C. Polymerase chain reaction (PCR) was performed in 10 μL reaction volumes containing 10 mM Tris-HCl (pH 8.3), 1.5 mM MgCl₂, 50 mM KCl, 50 μM dNTP, 0.2 μM SSR (simple sequence repeat) primers, 0.5 U Taq polymerase (TaKaRa, Dalian, China), and 20 ng template. The procedure of amplification consisted of a denaturation step (94 °C, 5 min), followed by 35 cycles of 30 s at 94 °C, 30 s at 55 °C, and 30 s at 72 °C, with a final extension of 10 min at 72 °C. The PCR products were electrophoresed on 6% polyacrylamide denaturing gels in 0.5 × TBE buffer and visualized with silver staining. Amplified DNA fragments showing clear polymorphisms between the parental types were used for the analysis of the three-way cross population and construction of the linkage mapping.

2.5. Molecular Marker Development and Assay

All the SSR markers around the F12-C locus region were obtained from Gramene (http://www.gramene.org/microsat/ (accessed on 1 January 2014)) based on the SSR linkage map constructed by McCouch [19]. To obtain more markers at the F12-C locus region, we re-sequenced one of the parental Chenghui9348 individuals and compared it with the sequence of 9311 to find the insertion/deletion (InDel) markers. Divergent InDel markers were designed based on the sequence of 9311 by using the Primer 5.0 software.

2.6. BSA and Linkage Analysis

The individuals with the ten highest (H) and ten lowest (L) pollen and spikelet fertilities were selected from the three-way cross population in 2013 to perform bulked segregation analysis (BSA) [7]. A total of 151 SSR markers that generated polymorphic bands between 9311 and Chenghui9348 were used to screen for polymorphisms between the H and L bulks. Polymorphic SSR markers that were confirmed in the bulks were used to analyze the individuals with the lowest (<30%) and highest (>70%) pollen and spikelet fertilities. Individuals carrying the ‘9311’ allele were scored as 0 and those carrying the ‘Chenghui9348’ allele were scored as 2. QTLs controlling hybrid fertility were determined by interval mapping using QTL IciMapping 4.0 at an LOD threshold of 3.0.

2.7. Sequence Analysis of the Candidate Genes

The coding sequences of candidate genes were amplified in 38 diverse accessions of rice (O. sativa) (Supplemental Table S4). Then, the PCR products were sequenced in TSINGKE (http://www.tsingke.net/ (accessed on 1 January 2014)). The primers used for amplifying the target fragments were designed based on the japonica cultivar Nipponbare genomic sequence data that were available on Gramene (Supplemental Table S2). The sequenced results were aligned to the parents to find insertion/deletion and SNP sites by using AlignX software.

2.8. RNA Extraction and RT-PCR

Total RNA from the mature root, mature stem, mature leaf, and young panicles was isolated using the TRIzol kit. The extracted RNA was treated with DNase (Thermo Fisher Scientific, Waltham, MA, USA) to eliminate genomic DNA contamination. Reverse transcription was performed using oligo (dT) 18 primers (Thermo Fisher Scientific, Waltham, MA, USA) and reverse transcriptase (Invitrogen Life Technologies, Thermo Fisher Scientific, Waltham, MA, USA) Actin primers were as listed (Supplemental Table S2). Amplification of the actin gene occurred at 95 °C for 5 min, 95 °C for 35 s, 55 °C for 35 s, and 72 °C for 35 s for 28 cycles. The 310RT primer was used to amplify ORF1 cDNA fragments. RT-PCR was performed at 95 °C for 5 min, 95 °C for 35 s, 56 °C for 35 s, and 72 °C for 35 s for 28 cycles.

3. Results

3.1. Chenghui9348 Has Wide-Compatibility Genes (WCGs)

Pollen and spikelet fertility were normal in 9311, Chenghui9348, and Nipponbare (over 88%) (

Table 1. Pollen and spikelet fertilities of parents and F₁ hybrids.

Parental Varieties and Crosses	Pollen Fertility (Mean (%) ± SD)	Spikelet Fertility (Mean (%) ± SD)
9311	92.3 ± 2.2	92.0 ± 1.9
Chenghui9348	93.5 ± 2.4	90.2 ± 2.6
Nipponbare	88.0 ± 1.8	91.2 ± 3.2
Balilla	91.0 ± 2.9	90.6 ± 3.8
Chenghui9348/9311	91.0 ± 1.8	87.4 ± 1.8
9311/Nipponbare	24.5 ± 6.2 †	34.5 ± 3.7 †
Nipponbare/9311	20.5 ± 5.4 †	28.3 ± 4.8 †
Chenghui9348/Nipponbare	73.8 ± 4.1 †	70.5 ± 4.2 †
Nipponbare/Chenghui9348	72.3 ± 2.5 †	71.9 ± 4.6 †

^† No significant difference between reciprocal cross at 1% level (t-test).

). However, the pollen and spikelets of F₁ from the 9311/Nipponbare cross showed low fertility rates, with scores of 24.5 ± 6.2% and 34.5 ± 3.7%, respectively (Table 1). When Chenghui9348 was crossed with Nipponbare, the pollen and spikelet fertilities reached 73.8 ± 4.1% and 70.5 ± 4.2%, respectively (Figure 1, Table 1). Further study showed that pollen and spikelet fertilities were 149 91.0 ± 1.8% and 87.4 ± 1.8% in the Chenghui9348/9311 hybrids (Table 1). These results suggested that Chenghui9348 is a wide-compatibility variety (WCV). Moreover, we found that the crosses of Chenghui9348/japonica produced a higher level of fertile hybrids compared to those of the control cross, ‘9311/japonica’, indicating that Chenghui9348 has a wide spectrum of compatibility (Supplemental Table S1). To confirm whether it was the $S_5^n$ wide-compatibility gene (WCG) that caused this phenomenon, we sequenced this gene in Chenghui9348 and the results showed that it had a $S_5^i$ genotype (data not shown). So, we speculated that there were new WCGs in the Chenghui9348 cultivar. Further reciprocal cross results indicated that the pollen and spikelet fertilities of the F₁ plants from the 9311/Nipponbare and Chenghui9348/Nipponbare crosses were close to that of their reciprocal F₁, suggesting that the wide spectrum of compatibility was controlled by the nuclear genome (Table 1).

3.2. WCGs in Chenghui9348 Affect Pollen Development and Pollen Adherence to Stigma

We investigated the developmental process of the pollen from the indica/japonica hybrid by using the carmine acetate dyeing method. The development of the microspores from the Chenghui9348 × Nipponbare F₁ hybrid was normal, as was that of the 9311 × Nipponbare F₁ hybrid, from the microspore mother cell (MMC) formation stage to the microspore stage (Figure 2a–f). We found that the development of the Chenghui9348/Nipponbare F₁ microspores remained normal; however, most of the 9311 × Nipponbare F₁ hybrid microspores only retained a vegetative nucleus and rarely contained a reproductive one at the early bicellular pollen stage, indicating that there were developmental disorders during the first mitosis (Figure 2g,h). Moreover, with the development of the microspores, most microspores of the 9311 × Nipponbare F₁ hybrid had become empty and shapeless shells at the late bicellular pollen stage. Additionally, compared with the F₁ of the Chenghui9348/Nipponbare hybrids, the contents of the microspores, such as starch, almost disappeared in the microspores of the 9311 × Nipponbare F₁ hybrid, and spherical abortion pollen grains can be clearly observed at the mature pollen stage (Figure 2i,j). The staining results indicated that the pollen grains of the Chenghui9348 × Nipponbare F₁ hybrid developed normally, and most pollen grains of the 9311 × Nipponbare F₁ hybrid were aborted at the early bicellular pollen stage.

We further monitored the germination of the pollen grains on the stigmas of the inter-subspecific hybrids by aniline blue staining and observed the grains with a fluorescence microscope. The results showed that more pollen grains adhered to the stigmas and were able to germinate in the F₁ of Chenghui9348 × Nipponbare hybrid compared with the F₁ of 9311 × Nipponbare hybrid (Figure 2k,l). Thus, the reduced spikelet fertility of F₁ hybrids from the cross of 9311 and Nipponbare resulted from pollen abortion and poor pollen adherence to the stigma.

3.3. Identification of a New Wide-Compatibility Locus, F12-C, in Chenghui9348

The abnormally low temperature in the winter of 2013 led to the low pollen and spikelet fertilities, and their frequency distributions showed a continuous distribution from 10 to 50%, with peaks around 20% (Figure 3a,b). In 2014, the distribution of pollen fertility in the 9311/Chenghui9348//Nipponbare population was bimodal, ranging from 10% to 100% with a valley at approximately 60% fertility (Figure 3c). Another three-way cross population (9311/Chenghui9348//Balilla) showed almost the same result (Figure 3d). The segregation of highly fertile and partly sterile individuals deviated from the expected 1:1 ratio, with more individuals in the high-fertility group (from 70–100%) than in the low-fertility group (from 10–50%) (

Table 2. Chi-square test results of low and high pollen fertilities in different populations.

Populations (2014)	Pollen Fertility (10–50%)	Pollen Fertility (70–100%)	χ2	p
9311/Chenghui9348//Nipponbare	247	303	5.70	0.017 *
9311/Chenghui9348//Balilla	112	146	4.48	0.034 *

* Denotes difference at 5% level (t-test).

). This result implied that the deviation is likely due to partial abortion of male gametes of 9311-type in the F₁ plant, which increased the ratio of the Chenghui9348/Nipponbare genotypes against the 9311/Nipponbare genotypes.

To identify the wide-compatibility loci in Chenghui9348, a total of 151 SSR markers distributed on 12 chromosomes that generated polymorphic bands between 9311 and Chenghui9348 were applied to identify markers linked to the loci by the BSA method. Fourteen SSR markers showed polymorphisms between the bulks. Then, we genotyped the individuals with the 50 lowest (<30%) and 20 highest (>70%) pollen and spikelet fertilities by using these polymorphism markers to detect linkage between markers of genetic compatibility. Two polymorphic SSR markers, RM511 and RM313, located on the same chromosome, chromosome 12, were identified to be associated with the F12-C wide-compatibility locus (Fertility 12-Chenghui9348). To confirm the reliability of the linkage between these two markers and F12-C, we screened the whole population and found it was significantly associated with F12-C by a t-test at α = 0.01 in the population. Then, with the average distributions of additional SSR markers on chromosome 12, we established a regional linkage map and conducted composite interval mapping with QTL IciMapping 4.0. A QTL was located in a 2.5-cM interval between markers RM1246 and RM7102 on the long arm of chromosome 12 (Figure 4). This QTL was linked to the SSR marker RM1047 with LOD scores of 7.8 and 7.9, explaining 17.9% and 18.4% PVE (phenotypic variance explained) of the variation in pollen and spikelet fertilities, respectively. We also used two three-way cross populations in 2014, as described above, to confirm the location of this QTL.

For this locus, the ‘Chenghui9348’ genotype exhibited higher fertility than the 9311 genotype. In 2013, the average pollen and spikelet fertilities of Chenghui9348-genotype plants at RM1047 were 30.7% and 21.2%, respectively, which are significantly higher than those of the 9311 genotype (17.6% and 13.1%, respectively). In 2014, the average pollen fertility of Chenghui9348-genotype plants was 69%, also significantly higher than that of the 9311 genotype (45%). All the differences were highly significant as determined by the t-test (

Table 3. Pollen and spikelet fertilities averaged for the F12-C locus.

RM1047	Pollen Fertility (%) (2013)	Spikelet Fertility (%) (2013)	Pollen Fertility (%) (2014, Nipponbare)	Pollen Fertility (%) (2014, Balilla)
0 a	17.6 **	13.1 **	45 **	44 **
2 a	30.7 **	21.2 **	69 **	62 **

^a Genotype 0 denotes an allele from 9311. Genotype 2 denotes an allele from Chenghui9348. ** Denotes a significant difference between two genotypes at 1% level (t-test).

).

3.4. Mapping the F12-C Locus to a 630 kb Interval

To narrow down the genomic region of F12-C, we chose 481 hybrid plants from two three-way cross populations with extreme phenotypes (pollen fertility lower than 30% or higher than 70%) for further mapping. Sixteen recombinant individuals were found by using RM1246 and RM7102 to genotype the extreme individuals. To obtain more polymorphic markers between Chenghui9348 and 9311 in the region of RM1246-RM7102, we compared the sequence of Chenghui9348 with that of 9311 in this region to seek the InDel markers, and fifteen divergent InDel markers were chosen for further mapping (Supplemental Table S2). Thus, eighteen markers, including three SSR markers, were available to analyze the recombination individuals. The analysis revealed one recombinant event between D1101 and F12-C, and two recombinant events between D1164 and F12-C. In addition, there are six markers that co-segregated with F12-C, including RM1047 (Supplemental Table S2). Thus, we narrowed down the genomic region containing the F12-C locus to a region approximately 630 kb in length by InDel markers D1101 and D1164 (

Table 4. Molecular marker genotypes of partial recombinants.

Markers	Sterile Individuals	Fertile Individuals
	n30-02	n79-09	n36-05	n27-04	n78-04	n60-09	n79-07
RM1246	0	2	2	2	2	2	0
D1047	0	2	2	2	2	2	0
D1071	0	2	2	2	2	2	0
D1081	0	2	2	2	2	2	0
D1101	0	2	2	2	2	2	0
D1114	0	2	2	2	2	2	2
RM1047	0	2	2	2	2	2	2
D1150	0	2	2	2	2	2	2
D1164	2	0	2	2	2	2	2
D1176	2	0	0	0	2	2	2
D1265	2	0	0	0	0	0	2
RM7102	2	0	0	0	0	0	2

Genotype 0 denotes an allele from 9311. Genotype 2 denotes an allele from Chenghui9348.

). NCBI Map Viewer (http://www.ncbi.nlm.nih.gov/mapview/, accessed on 14 November 2013) showed that there are six BAC clones in this region, according to the Nipponbare genome (OSJNBa0017A21, OJ1111_F12, OSJNBa0016A01, OJ1312_H10, OSJNBb0090H23, OJ1112_B07) (Figure 4).

3.5. Candidate Genes in the Location Region

Gene prediction analysis of the 630-kb region from Nipponbare using the Rice Genome Annotation Project (RGAP, http://rice.plantbiology.msu.edu/, accessed on 6 February 2013) showed that there were 29 predicted ORFs in these six BAC clones (Supplemental Table S3). Because F12-C has a similar function to Sn 5, which has a 136-bp deletion compared with Si 5 and Sj 5, we sequenced the predicted ORFs in this region to find the target gene by comparing the genomic sequences of Chenghui9348 and Nipponbare. Finally, two ORFs, ORF1 and ORF2, were identified. ORF1 is a gene with a transcript of 1.04 kb with three exons encoding a 232-amino-acid hypothetical protein, and ORF2 has five exons with a transcript length of 1344 bp encoding a 447-amino-acid hypothetical protein. Using the genomic sequence of Nipponbare as the reference sequence, we found that the coding sequence of ORF1 in Chenghui9348 has 16 SNPs and a deletion of 46 bp in the third exon, which resulted in a frameshift mutation (Figure 5a). ORF2, however, has 44 SNPs and an insertion of 12 bp in the first exon (Figure 5b).

3.6. Allelic Sequencing of F12-C Candidate Genes

We investigated the diversity of the coding sequences of ORF1 and ORF2 among 38 rice cultivars, including 16 indica cultivars and 22 japonica cultivars, to deduce their sequence features (Supplemental Table S4). The results suggested that ORF1 and ORF2 can be classified into three and four types according to their sequences, respectively. Specifically, the re-sequencing results of ORF1 showed that all cultivars who had the same genotype as Chenghui9348 just have one genotype: Type3 (Figure 5a). In contrast, the ORF2 alleles that had a 12-bp insertion similar to Chenghui9348 had two genotypes, Type2 and Type3, indicating the diversity of Chenghui9348 alleles in ORF2 (Figure 5b).

3.7. mRNA Expression of Candidate Genes of F12-C

The expression patterns of ORF1 and ORF2 in parents (9311, Chenghui9348) and F₁ hybrids (9311/Nipponbare, Chenghui9348/Nipponbare) were analyzed by RT-PCR. The results suggested that ORF1 was expressed constitutively in many organs such as the mature root, mature stem, mature leaf, and panicles. Unfortunately, we did not observe any significant difference between the F₁ of the Chenghui9348/Nipponbare and9311 × Nipponbare hybrids in various tissues (Figure 6). ORF2, however, did not show any expression in all examined tissues.

4. Discussion

We have identified a new wide-compatibility locus, F12-C, from Chenghui9348 in this study. Correspondingly, we designate the genotype of Chenghui9348 at this locus as F12-C, that of 9311 as F12-9, and that of Nipponbare as F12-ni. Consequently, individuals with F12-C/F12-9 and F12-C/F12-ni genotypes produced fertile pollen and panicles, while individuals with the F12-9/F12-ni genotype produced semi-sterile pollen and panicles due to allele interaction. There were no wide-compatibility genes reported in this region before; we considered F12-C as a new wide-compatibility locus that can improve the fertility of spikelets by improving pollen fertility in rice inter-subspecific hybrids.

A notable feature of the F12-C locus observed in this study is the obvious effect of male fertility on spikelet fertility. Just like the F12-C locus, there was only one f5-Du locus that played a similar role in a previous study [20,21]. f5-Du, which is a neutral allele, could increase pollen fertility by more than 50% and spikelet fertility by over 20%, and up to 79% and 35%, respectively [20]. F12-C, however, has an advantage compared with the f5-Du locus because it can increase spikelet fertility to approximately 70%. These results indicated that the F12-C locus has more potential in rice breeding programs. Apart from the F12-C, f5-Du, and S5-n loci, there are also some other wide-compatibility loci that have been found in different varieties, such as S1-g, S7-n, S8-n, S9-n, S15-n, S16-n, S29-n, S30-n, and S32-n, all of which were just found by using the position from a testcross [15,22,23,24,25].

Two predicted candidate genes were identified within the F12-C locus by comparing their sequences between parental lines. The deletion/insertion of these two genes may lead to loss of function in protein–protein interactions, resulting in normal fertility in hybrids with indica and japonica. Nowadays, several WCGs have been cloned and all of them have a deletion or insertion compared with the indica or japonica allele. The wide-compatibility allele S5-n has a 136-bp deletion at the N-terminus of the predicted protein, resulting in the loss of the signal peptide and therefore the mislocalization of the protein [11,12]. A second major locus, Sa, has been cloned in chromosome 1 and Dular carries a neutral allele at this locus. Sequence analysis of the Sa locus showed that Dular carries a 6-bp insertion compared with SaM in the indica and japonica allele, which is considered responsible for its fertility-neutral function [26,27]. Another neutral S4-n allele, namely f5-Du, has a 30-bp insertion near the 3’UTR of ANK-3 in Dular (AK105314) [21,28,29]. All of these studies suggested that the deletion/insertion of WCGs is responsible for their fertility-neutral function.

We found a high degree of polymorphism in the ORF1 and ORF2 sequences. The ORF1 allele in Chenghui9348 was conserved compared with ORF2; however, we could not confirm that ORF1 is F12-C, regardless of ORF2. It is possible that these two candidate genes that we deduced from the sequence characteristics of WCGs might not be F12-C. However, the candidate gene should be in the region between the D1101 and D1164 markers according to the results of the map-based cloning. To further clone the F12-C locus, larger mapping populations and smaller location regions, as well as complementation tests, are necessary.

The finding of F12-C might have significant implications in rice breeding programs. The wide-compatibility gene S5-n has been widely applied in many rice breeding programs to overcome the sterility of inter-subspecific hybrids. However, it has been frequently found that improvements in embryo sac fertility by the S5-n gene alone are not sufficient for producing indica/japonica hybrids with normal fertility [24,30,31,32]. So, it is necessary to explore other wide-compatibility genes, such as f5-Du and F12-C, to overcome hybrid sterility by increasing pollen fertility [20]. Nowadays, this can be easily achieved with marker-assisted selection using the markers identified in the present study.

5. Conclusions

In this study, we identified a wide-compatibility locus, F12, in inter-subspecific hybrids of rice and developed a pair of InDel markers for map-based cloning of F12. Moreover, two candidate genes, ORF1 and ORF2, were identified at the F12 locus.