Seven SSSLs carrying the Sc gene for hybrid male sterility in their chromosome substitution segments and seven SSSLs carrying the S5 gene for hybrid female sterility in their chromosome substitution segments were selected from the HXJ74-SSSL library (Supplementary Table 1). A set of indica and japonica varieties were used as testers to test the hybrid fertility. The genotypes of Sa, Sb, Sc, Sd, and Se loci for hybrid male sterility and S5 locus for hybrid female sterility have been identified in some of the testers. It was found that at these six loci, the indica variety Guang-lu-ai 4 (GLA4) carried the Sⁱ alleles, while the japonica variety Taichung 65 (T65) carried the S^j alleles (Zhang et al., 1994; Guo et al., 2016). All the study samples were planted from 2008 to 2019 at the farm of South China Agricultural University, Guangzhou (23°07′N, 113°15′E). These plants were planted in two cropping seasons each year, with the first cropping season (FCS) running from late February to mid-July and the second cropping season (SCS) running from late July to mid-November. Seeds were sown in seedbeds, and seedlings were transplanted into the field. Field management, including irrigation, fertilization, and pest control, followed normal agricultural practices.

The SSR markers were selected on the rice microsatellite maps (McCouch et al., 2002; Zhang et al., 2007). The functional markers of the S5 gene were selected to identify the genotypes at the S5 loci (Sundaram et al., 2010; Du et al., 2011; Yang et al., 2012; Guo et al., 2016). Markers linked with the Sa, Sb, Sc, Sd, Se, and S5 loci were selected from the published studies (Yang et al., 2004, 2012; Li et al., 2006, 2008; Chen et al., 2008; Long et al., 2008; Zhu et al., 2008). New molecular markers were developed in this study (Supplementary Table 2). The PCR products were separated into 6% non-denaturing polyacrylamide gels (Panaud et al., 1996; Li et al., 2006).

To check pollen fertility, nine mature flowers were collected from the upper third of panicles during the flowering stage and fixed in FAA solution. Pollens were stained with the 1% I₂-KI solution containing 0.1% (w/v) iodine and 1% (w/v) potassium iodide. Pollens were divided into normal pollens and sterile pollens, which were further divided into stained abortive pollens (stained but small size) and empty abortive pollens (small size and empty) (Zhang and Lu, 1989). Three panicles per plant and 10–12 plants per line were used to examine the spikelet fertility, and 20–40 plants per line were used to investigate the agronomic traits.

For statistical analysis, the percentage data were converted to the square root of the arcsine values. Student’s t-test was used to compare the data between the two groups. The Dunnett t-test was used to compare multiple groups with the control group. The least significance range (LSR) was used for the multiple range test among the multiple groups. The chi-square (χ²) test was performed to detect the distorted segregation of three genotypes in F₂ populations according to the Mendelian ratio of 1:2:1. SPSS statistics 23.0 and Origin Pro 9.0 were used for data analysis and charting¹.

To identify the genotypes of Sa, Sb, Sc, Sd, Se, and S5 loci associated with hybrid sterility, HJX74 was test crossed with T65, a japonica variety with S^j alleles at these six loci, and GLA4, an indica variety with Sⁱ alleles at these six loci (Zhang et al., 1994; Guo et al., 2016). The F₁ hybrids obtained from the cross of T65/GLA4 showed severe sterility, where the pollen fertility was only 20.51% and spikelet fertility was only 5.89%. In contrast, the F₁ hybrid of the HJX74/GLA4 cross showed normal pollen fertility and spikelet fertility of 92.39% and 93.39%, respectively. In the F₁ hybrids obtained from the cross of HJX74/T65, the pollen fertility was 72.89% and the spikelet fertility was 61.58%, which were significantly higher than those of T65/GLA4 and significantly lower than those of HJX74/GLA4 hybrids (Figure 1A). The results showed that the hybrid of HJX74/T65 exhibited partial pollen sterility and partial spikelet sterility.

The molecular markers linked to the Sa, Sb, Sc, Sd, Se, and S5 loci were used to investigate genotype segregation in the F₂ populations obtained from the crosses of HJX74/GLA4 and HJX74/T65. At the Sb, Sd, and Se loci, the genotype segregation of F₂ populations from both crosses fit the Mendelian ratio of 1:2:1. At the Sa, Sc, and S5 loci, distorted segregation of the genotypes was detected in the F₂ population of HJX74/T65 but not in the genotypes of HJX74/GLA4. At the Sa locus, the genotype ratios of Sa^HJX74/Sa^HJX74, Sa^HJX74/Sa^T65, and Sa^T65/Sa^T65 were 68:100:33, which significantly distorted from the Mendelian ratio of 1:2:1. At the Sc locus, the genotype ratios of Sc^HJX74/Sc^HJX74, Sc^HJX74/Sc^T65, and Sc^T65/Sc^T65 were 69:70:3, which significantly distorted from the Mendelian ratio. Distorted segregation was also detected at the S5 locus, where the genotype ratios of S5^HJX74/S5^HJX74, S5^HJX74/S5^T65, and S5^T65/S5^T65 were found to be 72:65:43 (Figure 1B). In addition, HJX74 was tested using a group of indica and japonica testers. The results showed that distorted segregation was detected only at the Sa, Sc, and S5 loci in the crosses of HJX74/japonica testers (Supplementary Table 3).

These results indicated that HJX74 carried Sⁱ/Sⁱ at the Sa, Sc, and S5 loci and Sⁿ/Sⁿ at the Sb, Sd, and Se loci (Figure 1C). At the Sa and Sc loci, the allele interaction between Sⁱ of HJX74 and S^j of japonica testers caused the abortion of male gametes carrying S^j in hybrids, resulting in the significant reduction of plants with S^j/S^j in the F₂ populations. At the S5 locus, the allele interaction between S5ⁱ of HJX74 and S5^j of japonica testers caused the abortion of female gametes carrying S5^j in hybrids, resulting in the significant reduction of plants with S5^j/S5^j in the F₂ populations. At the Sb, Sd, and Se loci, allele interaction between Sⁿ of HJX74 and S^j of japonica testers or Sⁱ of indica testers could not cause the abortion of any gamete in hybrids, and genotype segregation in the F₂ populations fit the Mendelian ratio of 1:2:1 (Supplementary Table 3). In addition, compared with the Sc locus, the Sa locus showed weak distorted segregation, where χ²_(1:2:1) = 34.00–62.27 in the five segregation populations of the Sc locus, while χ²_(1:2:1) = 9.36–12.19 in the three segregation populations of the Sa locus (Supplementary Table 3). The results showed that the hybrid male sterility caused by the interaction between Sⁱ and S^j at the Sa locus was weaker than that at the Sc locus.

To screen the Scⁿ gene, seven SSSLs carrying the Sc locus on the substitution segments obtained from different donors were selected from the HJX74-SSSL library (Supplementary Table 1). The pollen fertility of F₁ hybrids from the crosses between the SSSLs and HJX74 was over 90% (Figure 2A). The SSSLs were then tested with three indica testers and three japonica testers. Four SSSLs (01-03, 05-03, 06-03, and 14-03) and HJX74 showed significantly higher pollen fertility in their F₁ hybrids with indica testers than those obtained with japonica testers. In contrast, the other three SSSLs (11-03, 22-03, and 27-03) did not show a significant difference in the pollen fertility of F₁ hybrids between the crosses with indica and japonica testers (Figure 2B). Two SSSLs (06-03 and 27-03) were then selected to detect the segregation of Sc genotypes in F₂ populations obtained from the crosses with T65. In the F₂ population of the T65/27-03 cross, the Sc genotypes of T65/T65, T65/27-03, and 27-03/27-03 segregated in the ratios of 40:83:53, which fit the Mendelian ratio of 1:2:1. In contrast, in the F₂ population of the T65/06-03 cross, the genotype ratios of T65/T65, T65/06-03, and 06-03/06-03 were 23:64:69, which significantly distorted from the Mendelian ratio (Figure 2C). These results indicated that at the Sc locus, SSSLs 11-03, 22-03, and 27-03 carried the Sⁿ allele, while 01-03, 05-03, 06-03, and 14-03 carried the Sⁱ allele.

To screen the S5ⁿ gene, seven SSSLs carrying the S5 locus in the substitution segments obtained from different donors were selected from the HJX74-SSSL library (Supplementary Table 1). The genotypes of the S5 locus in the SSSLs were detected by functional markers. The results showed that in the substitution segments, three SSSLs (04-06, 13-06, and 14-06) carried S5ⁱ, one SSSL (10-06) carried S5^j, and the other three SSSLs (21-06, 23-06, and 27-06) carried S5ⁿ (Supplementary Table 4).

Five genotypes of the S5 locus were obtained from the F₁ hybrids crossed by seven SSSLs (Supplementary Table 5). The pollen fertility of hybrids was normal in all crosses, ranging from 93.04 to 94.42%. The spikelet fertility of S5ⁱ/S5ⁱ, S5ⁱ/S5ⁿ, S5ⁿ/S5^j, and S5ⁿ/S5ⁿ genotypes was normal (from 89.84% to 91.28%), but that of S5ⁱ/S5^j genotype from the crosses between 10-06 carrying S5^j/S5^j and SSSLs carrying S5ⁱ/S5ⁱ was only 68.34%, which was significantly lower than the spikelet fertility of the other four genotypes (Figure 3A and Supplementary Table 5). The segregation of S5 genotypes in F₂ populations obtained from three heterozygous genotypes, S5ⁿ/S5^j, S5ⁿ/S5ⁱ, and S5^j/S5ⁱ, was detected by using the functional markers of the S5 gene. Distorted segregation was detected in the S5^j/S5ⁱ segregation population produced from the crosses between 10-06 carrying S5^j/S5^j and SSSLs carrying S5ⁱ/S5ⁱ, but was not detected in the segregation populations of S5ⁿ/S5^j from 21-06/10-06 and of S5ⁿ/S5ⁱ from 21-06/13-06 (Figure 3B).

The three SSSLs with S5ⁿ, 21-06, 23-06, and 27-06, were tested for their wide compatibility by crossing with indica and japonica testers. The F₁ hybrids from all crosses showed high spikelet fertility, from 80.28% to 95.05%. As a control, the spikelet fertility of F₁ hybrids in HJX74/japonica testers was 70.13% (Figure 3C). In addition, distorted segregation of the S5 locus was detected in the F₂ population of T65/04-06, but was not detected in the F₂ populations of the other three crosses, that is, GLA4/04-06, T65/23-06, and GLA4/23-06 (Figure 3D). These results showed that the three SSSLs (21-06, 23-06, and 27-06) were compatible with indica testers and japonica testers in spikelet fertility as a result of their carrying S5ⁿ locus.

Three SSSLs (11-03, 22-03, and 27-03) with the Scⁿ gene and three SSSLs (21-06, 23-06, and 27-06) with the S5ⁿ gene were selected to pyramid the two Sⁿ genes in the HJX74 genetic background. Three SSSLs with Scⁿ were crossed with three SSSLs with S5ⁿ, respectively. In the segregating populations, the plants carrying Scⁿ and S5ⁿ loci were selected. Nine pyramiding lines were developed, which carried Scⁿ and S5ⁿ loci from different donors and Sbⁿ, Sdⁿ, and Seⁿ in the HJX74 genetic background (Figure 4A and Supplementary Table 6). Therefore, the nine pyramiding lines thus obtained were WCILs.

In the nine WCILs, the plant type was similar to HJX74 (Figure 4B). In addition, no significant difference between HJX74 and WCILs was found in the majority of the investigated traits, including heading date, plant height, width of flag leaf, length of flag leaf, grain length, grain width, and grain weight (Supplementary Table 7).

To evaluate the compatibility of nine WCILs, the WCILs were test crossed with six indica testers and five japonica testers (Supplementary Tables 8–11). When tested with indica tester group, F₁ hybrids of nine WCILs showed normal pollen fertility and spikelet fertility, with no significant difference when compared to HJX74. When tested with the japonica tester group, nine WCILs showed significantly higher F₁ pollen fertility and spikelet fertility when compared to HJX74 (Figure 5). These results indicated that the WCILs showed wide compatibility, producing high pollen fertility and spikelet fertility in their F₁ hybrids with both indica and japonica rice varieties.

Although the WCILs showed significantly higher F₁ pollen fertility when tested with japonica testers, the F₁ pollen fertility was still lower when tested with indica testers (Figures 5C,D). To identify the problem, the genotype segregation at the Sa, Sb, Sc, Sd, Se, and S5 loci in F₂ populations was examined with molecular markers linked with these loci. No distorted segregation at the Sb, Sc, Sd, Se, and S5 loci was found in all the detected F₂ populations, further confirming the fact that the alleles of the Sb, Sc, Sd, Se, and S5 loci in WCILs were Sⁿ. However, significantly distorted segregation was detected at the Sa locus in all four populations (Supplementary Table 12). These results verified that WCILs carried the Saⁱ gene in the HJX74 genetic background, and the interaction between Saⁱ from WCILs and Sa^j from japonica testers caused some male gametes with Sa^j to become abortive in F₁ hybrids obtained from the crosses of WCILs with japonica testers.

Three indica lines, GLA4 carrying the genotype of Saⁱ, Sbⁱ, Scⁱ, Sdⁱ, Seⁱ, and S5ⁱ, HJX74 carrying the genotype of Saⁱ, Sbⁿ, Scⁱ, Sdⁿ, Seⁿ, and S5ⁱ, and WCIL 2223 carrying the genotype of Saⁱ, Sbⁿ, Scⁿ, Sdⁿ, Seⁿ, and S5ⁿ, were selected to test their compatibility with eight japonica varieties of different ecotypes. In the F₁ hybrids of GLA4 with eight japonica varieties, pollen fertility was 13.24–90.68% with an average of 46.94%, and spikelet fertility was 5.89–92.95% with an average of 44.40%. In the F₁ hybrids of HJX74 with eight japonica varieties, pollen fertility was 75.19–95.74% with an average of 85.37%, and spikelet fertility was 58.34–93.90% with an average of 74.69%. On comparison of data, pollen fertility was 82.35–95.79% with an average of 88.82%, and spikelet fertility was 89.19–94.29% with an average of 91.73% in the F₁ hybrids of WCIL 2223 with the eight japonica varieties (Supplementary Table 13). The results showed that WCIL 2223 had higher and wider compatibility with japonica varieties than GLA4 and HJX74. The pollen fertility and spikelet fertility in F₁ hybrids of WCIL with various japonica varieties were normal or near normal.

In the past decades, the genetic basis of indica–japonica hybrid sterility has been understood. In indica–japonica hybrid sterility, the S5 locus was found to be responsible for female sterility, and the Sa, Sb, Sc, Sd, and Se loci were responsible for male sterility. Following the tri-allele pattern and the one-locus sporo-gametophytic interaction model, the allele interaction between Sⁱ and S^j leads to the abortion of male or female gametes carrying S^j, whereas the allele interaction between Sⁿ and Sⁱ or S^j does not lead to the abortion of any gamete (Zhang, 2020, 2022). Thus, the sterility or compatibility of hybrids between indica and japonica subspecies is a complex trait that is controlled by multiple genes. Due to the diversity of indica and japonica rice varieties, the genotypes of hybrid sterility vary greatly among different varieties, particularly modern varieties, resulting in different crossing combinations with different degrees of hybrid sterility. In addition, the effects of alleles obtained from different donors are quantitatively different, resulting in the continuous variation of fertility at a single locus (Zhang et al., 1993, 1994). The molecular basis of allele diversity has been revealed by the cloned genes of S5 (Chen et al., 2008; Yang et al., 2012), Sa (Long et al., 2008; Xie et al., 2017), and Sc (Shen et al., 2017). In this study, we found that HJX74, the recipient of SSSLs, carried the Sⁿ allele at the Sb, Sd, and Se loci but the Sⁱ allele at the S5, Sa, and Sc loci (Figure 1). In addition, the effect of Saⁱ was weaker than that of Scⁱ in HJX74 (Supplementary Table 3). The identification of genotypes that lead to hybrid sterility provided a prerequisite for improving the compatibility of HJX74.

Based on the tri-allele pattern and the one-locus sporo-gametophytic interaction model, the indica–japonica hybrid sterility can be overcome by developing ICJLs and WCILs (Zhang and Lu, 1999; Zhang, 2020, 2022). ICJLs can be developed by transferring the Sⁱ allele from indica to japonica rice. In hybrids between indica varieties having Sⁱ allele and ICJLs having Sⁱ allele in japonica genetic background, the Sⁱ/Sⁱ genotype cannot cause the abortion of any gamete. In a previous study, we transferred the Sⁱ allele from indica donors to the japonica T65 variety to develop ICJLs, which carry the Sⁱ allele at hybrid sterility loci in the japonica genetic background. The result was that ICJLs were compatible with indica but incompatible with japonica rice (Guo et al., 2016). In another method, WCILs can be developed by transferring the Sⁿ allele from donors to indica rice. In hybrids between WCILs having Sⁿ allele in indica genetic background and japonica varieties having S^j allele, the Sⁿ/S^j genotype cannot cause the abortion of any gamete. In this study, we pyramided the Sⁿ allele of SSSLs to develop WCILs, which carry Sⁿ allele in indica HJX74 genetic background. The result was that WCILs showed wide compatibility, which was compatible with both indica and japonica rice varieties (Figures 4, 5). These results showed that the breeding of ICJLs and WCILs is practicable and that the indica–japonica hybrid sterility could be overcome by using ICJLs and WCILs.

The development of WCILs requires pyramiding Sⁿ alleles of multiple hybrid sterility loci to improve compatibility. The breeding of WCILs is a challenging task because it is a time-consuming and laborious technique. First, the Sⁿ alleles of the S5, Sa, Sb, Sc, Sd, and Se loci need to be identified and selected from a wide range of genetic resources. Second, Sⁿ alleles of multiple loci need to be pyramided in an indica genetic background by MAS. In addition, WCILs need to have improved traits to be used as parents of indica–japonica hybrid rice. Over the past two decades, we have constructed a HJX74-SSSL library, which is used as a platform for rice design (Zhang, 2021). Using this platform, a series of CMS, maintainer, and restorer lines were developed (Dai et al., 2015, 2016; Luan et al., 2019). In this study, we identified Sⁿ alleles at the S5, Sb, Sc, Sd, and Se loci from the HJX74-SSSL library. Since HJX74, the recipient of SSSLs, carried the Sⁿ alleles at Sb, Sd, and Se loci, but the Sⁱ alleles at S5, Sa, and Sc loci, the SSSLs carrying S5ⁿ or Scⁿ alleles were selected from the HJX74-SSSL library (Figures 2, 3). The Scⁿ and S5ⁿ of the SSSLs were then pyramided in the HJX74 genetic background. Nine WCILs carrying Sⁿ alleles at the S5, Sb, Sc, Sd, and Se loci in the HJX74 genetic background were developed (Figures 4, 5). The results show that the HJX74-SSSL library is a powerful platform for developing WCILs possessing the complex trait of wide compatibility.

It is believed that inter-subspecific hybrids have stronger heterosis than intra-subspecific hybrids (Fu et al., 2014; Birchler, 2015). Therefore, the exploitation of inter-subspecific heterosis for the production of improved rice varieties has long been considered (Cheng et al., 2007; Zhang, 2020). The main obstacle in utilizing inter-subspecific heterosis in rice is the indica–japonica hybrid sterility. In this study, WCILs were developed using the HJX74-SSSL platform. The WCILs had compatibility with a wide range of japonica varieties (Figure 5 and Supplementary Table 13). Therefore, the development of WCILs is an effective approach to overcoming the problem of indica–japonica hybrid sterility in breeding practice. By further improving their fertility restoration ability, WCILs can be improved to produce wide-compatible indica restorer lines (WCIRLs). Using the HJX74-SSSL platform, a series of WCIRLs is being developed and will be used to develop indica–japonica hybrid rice by crossing with japonica male sterile lines. Therefore, it is expected that indica–japonica hybrid rice will be the rice of next generation (Zhang, 2022).