To develop CSSLs and HCSSLs, Dongnanihui 810, a restorer indica cultivar, was used as the recipient, and ZhangPu wild rice was used as the donor. The F₁ plants were generated from Dongnanihui 810 as female and ZhangPu wild rice as male. The F₁ plants were back-crossed with Dongnanihui 810 to produce the BC₁F₁ generation. These BC₁F₁ plants were backcrossed to Dongnanihui 810 to produce BC₂F₁. In the same way, 156 BC₃F₁ individuals were obtained. MAS with a whole-genome survey of 156 BC₃F₁ lines, which were selected at random by taking three from each line, identified 213 plants in which most genomic regions were homozygous for Dongnanihui 810 alleles. However, more than four heterozygous substituted segments from ZhangPu wild rice were excluded.

A lot of 302 SSR markers were selected from dense rice microsatellite maps (McCouch et al., 2002), and 206 InDel markers were developed using Primer Premier 5.0 software according to the publicly available rice genome sequence comparisons between Nipponbare and 9311². Of the 506 markers, 149 (29.4%) displayed better polymorphisms between the two parents.

Plant DNA was extracted from the frozen leaves of rice plants using the CTAB method (Murray and Thompson, 1980) with minor modifications. The extracted DNA was dissolved in ddH₂O. DNA amplification was performed by PCR with the following parameters: 5 min at 94°C, 35 cycles of 1 min at 94°C, 1 min at 60°C (for Indel) or 55°C (for SSR), and 50s at 72°C, with a final extension of 10 min at 72°C. For PCR amplification of markers, each 20 μL reaction mixture contained 50 ng DNA, 5 μmol of each primer, 10× PCR buffer [100 mM Tris (pH 8.3), 500 mM KCl, 15 mM MgCl₂, 2 μg gelatin], 250 μM of each dNTP and 0.5 U of Taq polymerase. Amplified PCR products were separated through electrophoresis on a 6% non-denaturing polyacrylamide gel, and bands were revealed using a silver staining procedure.

The substituted segment was determined based on its location on the rice microsatellite map (McCouch et al., 2002). The length of substituted segment in CSSLs was estimated based on graphical genotypes (Young and Tanksley, 1989; Zhu et al., 2009). A chromosome segment flanked by two markers of donor genotype (DD) was considered to have a 100% donor genotype, while a chromosome segment flanked by two markers of recipient genotype (RR) was considered to be 0% donor genotype. A chromosome segment flanked by one marker of donor type and one marker of recipient type (DR) was considered as 50% donor genotype. Therefore, the length of DD plus the length of two half DR was considered to be the estimated length of a substituted chromosome segment.

Dongnanihui 810, ZhangPu wild rice, and 146 CSSLs were grown in a paddy field under natural conditions at the experimental farm of Fujian Academy of Agricultural Sciences (Fuzhou, China), in early of 2015. The field experiment was designed in randomized plots with one plot per genotype. For parents and each CSSL, 60 plants were planted in six rows, and 12 plants in the center of each plot were selected to investigate the characters of purple apiculus. Plant height was the mean value of 6 plants from the middle section of each plot, and QTLs were identified on the basis of significant differences between parents and each CSSL, as determined by t-test. All plants were grown according to standard commercial practices, with spacing of 13.3 cm between plants within each row and 26.4 cm between rows. Field management essentially followed normal agricultural practices, and the amounts of N, P₂O₅ and K₂O applied were 127.5 kg/hm², 45.0 kg/hm², and 30.0 kg/hm², respectively.

According to standard commercial practices, 146 CSSLs were grown in a paddy field under natural conditions at the experimental farm of Fujian Academy of Agricultural Sciences (Fuzhou, China). All materials were planted in the same field, the level height of field was consistent, N, P₂O₅ and K₂O were applied in the following amounts: 127.5 kg/hm², 45.0 kg/hm², and 30.0 kg/hm², respectively, and the amount and time of fertilization was uniform. Plant height, panicle length, number of effective panicles, spikelet number per panicle, seed setting rate and 1000-grain weight were all measured at maturity in 2015.

The physical map of the qPa-6-2 gene was constructed by bioinformatics analysis using the published sequences of BAC and P1-derived artificial chromosome (PAC) clones of cv. Nipponbare released by the International Rice Genome Sequencing project (IRGSP³).

Polymorphisms were detected by SSR and InDel markers between the two parents, and were used in this study to survey the polymorphisms between the two parents (Table 1). 302 SSR markers were selected from dense rice microsatellite maps (McCouch et al., 2002). A total of 206 InDel markers were developed using Primer Premier 5.0 software according to the publicly available rice genome sequence comparisons between Nipponbare and 9311⁴. Of the 506 markers, 149 (29.4%) displayed better polymorphisms between the two parents (Supplement Table S1). The length of the interval between two polymorphic markers ranged from 0.1Mb to 12.5 Mb, with an average of 2.53 Mb on the rice physical map (Figure 1; Table 1). The polymorphic markers were utilized further for the further development of CSSLs and HCSSLs.

The CSSL and HCSSL development procedure was summarized in Figure 2. In total, we identified 213 BC₃F₁ plants in which the majority of genomic regions were homozygous for Dongnanihui 810 alleles, and 31 BC₃F₁ plants that had one substituted segment were self-pollinated to produce thirty-one BC₃F₂ lines. Then, 21 plants from each BC₃F₂ line were genotyped with the polymorphic markers on the target-substituted segments to select plants that had one homozygous substituted segment from ZhangPu wild rice. Heterozygous substituted segments from ZhangPu wild rice were also selected. As a result, 42 CSSLs were obtained, and 36 HCSSLs were obtained. The remaining 182 BC₃F₁ plants, which had two to four substituted segments, were back-crossed with Dongnanihui 810 to produce the BC₄F₁ lines, and these plants were self-pollinated to produce BC₄F₂ lines. Twenty-one plants from each line were genotyped on the target region to select plants that had one homozygous substituted segment. Forty-five plants from each line were genotyped on the target region to select plants that had two homozygous substituted segments. Ninety-three plants from each line were genotyped on the target region to select plants that had three homozygous substituted segments. At the same time, heterozygous substituted segments from ZhangPu wild rice were selected. As a result, 104 CSSLs were obtained, and 208 HCSSLs were selected. Using these techniques, a total of 146 CSSLs and 244 HCSSLs were selected. Of these CSSLs, eight carried two substituted segments and one carried three substituted segments. Nine HCSSLs, which contained the target region, were planted. Then nine CSSLs were obtained, but only one carried two substituted segments.

According to the physical locations and genotypes of the 149 molecular markers in the 146 CSSLs, the lengths and locations of the substituted chromosome segments were estimated. Then, using the results of the estimation, a physical map of the 146 CSSLs was constructed (Figure 3). In addition, based on 149 molecular markers in the 244 HCSSLs, a physical map, which contained the 244 HCSSLs, was also constructed (Figure 4).

From the physical map constructed with molecular markers, the 146 CSSLs carried 147 homozygous substituted segments. Only one CSSL on chromosome 1 carried two substituted segments and the remaining 145 carried only one substituted segment (Table 2). In the 146 CSSLs, the length of substituted segments ranged from 550 kb to 24.75 Mb, but averaging 7.79 Mb. The length of 81 substituted segments was shorter than 7.0 Mb and 15 were longer than 16.0 Mb (Figure 5). The average number of substitution segments per chromosome was 12.2, but the distribution of the segments was not random among the 12 chromosomes. Different substituted frequencies existed in different chromosomes. For example, 17 existed on chromosome 4, 14 existed on chromosome 1, 4, and 6, while only six existed on chromosome 11 (Table 2).

The physical map constructed with molecular markers indicated that the 244 HCSSLs carried 244 heterozygous substituted segments and one homozygous substituted segment. Of these, only one HCSSL carried two substituted segments, which contained one heterozygous and one homozygous segment, while the remaining 243 HCSSLs carried only one heterozygous substituted segment (Table 2). The length of heterozygous substituted chromosome segments in the 244 HCSSLs ranged from 300 kb to 23.45 Mb, but averaging 6.90 Mb. Overall, 152 heterozygous segments were shorter than 7.0 Mb, and 21 heterozygous segments were longer than 16.0 Mb (Figure 6). The average number of substitution segments per chromosome was 20.3 in the 244 HCSSLs. However, different introgressed frequencies were identified among the 12 chromosomes in that 38 heterozygous segments existed on chromosome 3, while 12 heterozygous segments existed on chromosome 12 (Table 2).

The physical map indicated that the average length of substituted segments per chromosome was 95.47 Mb in the CSSLs, ranging from 44.30 Mb on chromosome 10–144.80 Mb on chromosome 4. The total length of substituted segments in the CSSLs was 1145.65 Mb, which was 3.04 times the total length of the rice genome, and there were different in 12 chromosomes, ranging from 1.65 times on chromosome 11–4.11 times on chromosome 4. All of the chromosomes had 100% coverage in both cases (Table 3).

The physical map showed that the average length of substituted segments per chromosome was 140.31 Mb in the HCSSLs, ranging from 74.35 Mb on chromosome 10–340.10 Mb on chromosome 3. The total length of substituted segments in the HCSSLs was 1683.75 Mb, which was 4.47 times the total length of the rice genome, and different frequencies in rice chromosome, ranging from 2.81 times on chromosome 2 to 9.02 times on chromosome 3. The average coverage of substituted segments per chromosome was also 100% (Table 3).

In order to evaluate the potential advantages of the CSSLs for QTL detection, phenotypic variations of plant height were observed in 146 CSSLs.

The two parents showed highly significant differences in plant height. Using six CSSLs, two QTLs were identified, in which qPH-1-1 was mapped in the marker intervals between Ind 1–20 and Ind 1–23 (7.1 Mb) on rice chromosome 1 (Figure 7A), while qPH-7-1 was mapped between Ind7-1 and RM427, which spanned 2.0 Mb in genetic distance on rice chromosome 7 (Figure 7B).

For purple apiculus, ZhangPu wild rice displayed purple apiculus, while the indica variety, Dongnanihui 810, displayed green apiculus. Using four CSSLs, one major QTL for purple apiculus was identified, in which qPa-6-2 was located between Ind6-1 and RM276, which comprised 5.7 Mb in physical distance on rice chromosome 6 (Figure 8).

The two segregating populations, HCSSL132 and HCSSL136, which were heterozygous in the target region and showed purple apiculus (Table 4), were grown in a paddy field under natural conditions at the experimental farm of Fujian Academy of Agricultural Sciences (Fuzhou, China), in late 2015. Segregating plants were therefore recorded as either purple or green apiculus. The segregation of purple to green apiculus based on 4932 F₂ plants fitted a ratio of 3:1(χ² = 0.330 ∼ 0.688, P > 0.05) (Table 4), indicating that the gene for purple was dominant over green color and was controlled by a major gene.

To map the gene to a smaller region, 1241 recessive individuals were identified from the two HCSSL populations (Table 4). A higher precision map was constructed using published markers⁵ in the region between Ind6-1 and RM276 (Figure 9A). All recombinants were genotyped using seven polymorphic markers. The results showed that the qPa-6-2 gene was mapped between molecular markers RM19551 and RM19590 on chromosome 6 and that the physical distance between the two markers was 539 kb with a physical distance (Figure 9B; Table 5).

To fine map the qPa-6-2 gene, five polymorphic InDels were selected from 12 new InDels (Table 5). The InDel markers were designed from the publicly available rice genome sequences, and the likelihood of detecting polymorphism between ZhangPu wild rice and Dongnanihui 810 was predicted by comparing sequences from Nipponbare⁶ and Indica cultivar 93-11⁷. Firstly, the BAC clone’s sequences of japonica and indica were aligned, then primers were designed using Primer premier 5.0 based on the polymorphism region between the two rice subspecies and the polymorphic markers were used for gene mapping. Recombinant screening with five markers (S6-2, S6-3, S6-6, S6-9 and S6-11), which were more internal to the qPa-6-2 locus, detected four, two, one, three and eight recombinants, respectively. Thus, the qPa-6-2 gene was precisely mapping in an 88 kb region by S6-3 and S6-6 (Figure 9C).

Oryza Rufipogon Griff., a wild rice, is the progenitor of the cultivated rice O. sativa L. As such, it is well recognized as a natural gene bank that conserves many specific genes. During the course of domestication from common wild rice to cultivated rice, profound changes in morphology and molecular genetic profile occurred via natural and artificial selection. Over the years, polymorphisms have been identified between wild and cultivated rice, including some that affect yield, disease resistance, or insect resistance (Tian et al., 2006; Jena, 2014).

Major resistance genes have been transferred to cultivated rice from O. nivara and O. longistaminata (Khush, 1977; Khush et al., 1990). A new gene, Xa-23, which showed resistance to bacterial blight was discovered and cloned from common wild rice of Guangxi (Wang et al., 2014, 2015). The new resistance gene for Magnaporthe oryzae, Pi9, was identified from wild rice (Zhuang et al., 1998; Jiang et al., 2012). The major gene for Magnaporthe oryzae, Pi40, was discovered from the wild rice O. australiensis (Jeung et al., 2007). The new genes, Bph14 and Bph15, which showed resistance to the brown planthopper, were identified from the wild rice (Oryza officinalis Wall) (Yang et al., 2004; Du et al., 2009). Meanwhile, Li et al. (2002) identified two high-yielding QTLs from Dongxiang common wild rice of China. Tan et al. (2008) cloned the PROG1 gene from common wild rice, and prog1 variants identified in O. sativa disrupt prog1 function and inactivate prog1 expression, leading to erect growth, greater grain number and higher grain yield in cultivated rice. At the same time, the flowering time QTL dth3, was detected from wild rice species (Bian et al., 2011), and the seed dormancy QTL was also isolated in the genetic populations derived from wild rice species (Gu et al., 2004).

Chromosome segment substitution lines were a series of NILs in which the substituted segments of containing the entire information of the donor, and each CSSL carried one or more donor chromosome segments. Since the main characteristic of CSSLs is that the substituted segment of each CSSL was homozygous and stable, they were useful for genetic studies and fine mapping of QTLs for genome-wide target traits. In the present study, we developed 146 CSSLs in the genetic background of rice restorer Dongnanihui 810. The 146 CSSLs had different agronomic traits, including, for example, long grain, plant height, panicle length, effective panicle, and long flag leaf. Using these lines, we mapped two QTLs for plant height. The qPH-1-1 mapped in this study was probably an allele of sd1 (Sasaki et al., 2002; Ye et al., 2015), while qPH-7-1 might be a novel QTL. The qPH-1-1 mapped between the markers Ind1-20 and Ind1-23 in this study, which contained the sdl locus, a gene known to play a significant role in height in rice. These results demonstrated that the uniformity of the genetic background of the lines facilitated the identification of a number of QTLs by direct comparison between each line and the recipient parent. For further fine mapping and positional cloning of interesting QTLs, secondary F₂ populations could be derived from a further back-cross between the target CSSL and the recurrent parent (Frary et al., 2000; Yano et al., 2000). Therefore, CSSLs simplified the process of data analysis and increased the accuracy of the results. In this study, using the 146 CSSLs, one major QTL for purple apiculus was identified. Here, the qPa-6-2 was located between two markers Ind6-1 and RM276 on rice chromosome 6. More importantly, the qPa-6-2 gene was precisely mapped to an 88 kb region, which contained the rice OsC1 gene (Saitoh et al., 2004). The occurrence of varying degrees of apiculus coloration (also shows purple apiculus) due to anthocyanin pigmentation, which was reported to be caused by a series of alleles at the C locus, and the C gene appears to be the rice homolog (OsC1) (Saitoh et al., 2004). In our study, we identified the qPa-6-2 for purple apiculus, which was also located in the same region. Meanwhile, the Pa-6 gene was also mapped to the same region with a genetic distance of 41.7 kb (Liu et al., 2012). It is well known that male sterile lines play very important in hybrid rice. Meanwhile, purple apiculus is helpful to increase outcrossing rate of male sterile lines in rice. Therefore, fine mapping and positional cloning of purple apiculus are of great significance in hybrid rice breeding.

Plant breeding, which aimed to improve the genetic basis of new varieties of crops with increased productivity and quality, combines art with science (Xi et al., 2006). In a general way, traditional breeding was predominantly based on phenotypic assays. Facilitated by the recent developments in genome sequencing, molecular markers and bioinformatics, plant breeding systems entered an era of molecular breeding, shortening the breeding period, improving efficiency, and overcoming the traditional shortcoming of low accuracy with broad applications. Thus, it became the standard in plant breeding programs to combine phenotype and genotype (Xu et al., 2010).

However, this approach only targeted the QTLs with very few traits for genetic improvement. On the other hand, CSSLs, which selected at the level of the whole genome with multi-trait breeding objectives, have expanded targets and thus become important in improving the properties of plants (Xu et al., 2010). Therefore, CSSLs for rice breeding were considered as a key tool in the ongoing technological innovation of plant breeding (Xu et al., 2010).

The indica cultivar Dongnanihui 810, used as the recipient in this study, has been planted on a large scale as an excellent restorer. Ganyou 810, derived from the CMS line Ganxiang A and a restorer line Dongnanihui 810, was certified and released for commercial production in 2014 in Fujian province. The ZhangPu wild rice, used as the donor in the present study, was the only wild rice in Fujian province. Interestingly, eight elite CSSLs in this wide population had a genetic background similar to that of Dongnanihui 810, but their agronomic traits were better than those of Dongnanihui 810 (Table 6). The effective panicle, each panicle contains at least five seeds (including 5) in per plant, was one of the important indicators of yield traits. For example, CSSL15 had lower plant height and more effective panicle than Dongnanihui 810. CSSL42, CSSL102 and CSSL123 had higher seed setting rate, and CSSL53 had more effective panicle. CSSL77 and CSSL78 had longer panicle and more spikelets, while CSSL102 and CSSL138 both had higher 1000-grain weight. None of these eight CSSLs had poor agronomic traits. Therefore, they could be used to create new varieties with direct marketing applications and, as a parent, create new hybrids. Furthermore, favorable alleles could be combined through MAS to improve the production of superior rice varieties.

HCSSLs showed technological innovation in rice genetic research. Up to now, after developing CSSLs in rice, many HCSSLs, which carried one or more donor heterozygous chromosome segments, were discarded. However, for further fine mapping and positional cloning of interesting QTLs, secondary F₂ populations were reconstructed, which should be derived from a further back-cross between the target CSSL and the recurrent parent. Obviously, this not only increased experimental cost, but also reduced efficiency. In the present study, 244 HCSSLs were produced on the basis of estimates of lengths and locations, and the total length of substituted segments among HCSSLs was 1683.75 Mb, or 4.47 times the total length of the rice genome.

It was convenient to perform gene mapping and analysis of genetic characteristics for agronomic traits using the HCSSLs. In this paper, using the 146 CSSLs, the major qPa-6-2 for purple apiculus was identified. To confirm this, further analysis showed that both HCSSL132 and HCSSL136 were heterozygous in the target region and showed purple apiculus. Meanwhile, HCSSL132 and HCSSL136 were planted. Further observation of these two populations showed that the gene for purple apiculus was dominant over green and was controlled by a major gene. Importantly, without reconstructing secondary F₂ populations, the qPa-6-2 was precisely mapping in an 88 kb region using the two HCSSLs populations.

It was easy to improve the development of CSSLs carrying overlapping chromosome segments of the whole rice genome using HCSSLs. From a genetics perspective, the number of heterozygous plants, as a separate population, was nearly twice that of homozygous plants. Regrettably, only homozygous substituted plants were selected, while many heterozygous substituted plants were discarded. As a result, CSSLs could barely cover the whole genome of donor parents. However, incomplete coverage of a donor genome in the CSSL population might, in turn miss some QTLs responsible for useful traits (Subudhi et al., 2015). In this paper, with the help of HCSSLs, CSSLs carrying one substituted segment could be obtained. At the early stage of constructing CSSLs, although 146 CSSLs were obtained in this paper, eight of these CSSLs carried two substituted segments and one carried three substituted segments. Using nine HCSSLs containing the target region, nine CSSLs were obtained, which only one CSSL carried two substituted segments, the rest of the eight carried one substituted segment.