The wide-grained CSSL-Z708 was used as the primary material. Z708 was developed by crossing the progeny (F₄) of the recipient parents Xihui18 and Z403 using marker-assisted selection (MAS). Z403 was found to contain 10 substitution fragments from the donor Jinhui35 in the Xihui18 background, which was developed from Xihui18 as the recipient parent and Jinhui35 as the donor parent by advanced backcrosses in combination with single-sequence repeat (SSR) MAS from the BC₂F₁ to BC₃F₇ generations. Specific construction methods used for the development of Z403 have been described previously (Xu et al., 2023). Xihui18 and Jinhui35 are indica rice restorer lines bred by Southwest University, China. The Z708 chromosome substitution segment was identified as described previously (Sun et al., 2022). The specific flow chart of the genetic material construction was showed in the flow chart ( Figure 1 ). The estimated length of the substitution segment was calculated as described previously (Paterson et al., 1991).

Because differences were found in only four chromosomal segments between its recipients Xihui18 and Z708, the material of the QTL mapping population was a secondary F₂ population of 150 individuals constructed using Xihui18/Z708.

In July 2020, Xihui18 was crossed with Z708 and the hybrid seeds were harvested at the experimental station (Xiema town, Beibei distinct at 106.38° east longitude and 29.76° north latitude) of Southwest University, Chongqing, China. In September of the same year, the hybrids were sown at the Lingshui base (109.86° east longitude and 18.42° north latitude) in Hainan Province, China, and the F₁ seeds were harvested. In March 2021, the parent seeds of Xihui18, Z708, Jinhui35, and the F₂ population were sown in a field at the Rice Research Institute of the Southwestern University of Chongqing, China. In April 2021, 30 seedlings of Xihui18, Jinhui35, and Z708 and 150 individuals from the F₂ population were transplanted to the same field, with 26.4 cm spacing between rows, 16.5 cm between hills, and 10 plants per row. In March and April 2022, 30 seedlings each of Xihui18 and Z708 and four F₂ individuals for the development of SSSLs were cultivated and transplanted in the same field in Chongqing. Further, an F₃ population was sown for the fine-mapping of qGW4, and all plants were transplanted into the same field in Chongqing. In March and April 2023, four SSSLs for QTL identification, as well as Xihui18 and Z708 were sown and transplanted (30 individuals for each line) in Chongqing. Additionally, four secondary heterozygous F₃ individuals for the development of SSSLs and one NCL (all lanes of markers were the same as Xihui18) were cultivated for overlapping substitution mapping of qGW4 (100 plants for each line). Field management was the same as local standard practices.

At the maturity stage, 10 plants each of Xihui18 and Z708 and 150 F₂ plants were harvested. The following eight yield traits for yield-related traits were measured: plant height (PH), panicle number per plant (PN), panicle length (PL), grains per panicle (GPP), 10-grain length (GL), 10-grain width (GW), 1000-grain weight (GWT), and yield per plant (YD). The ratio of length to width (RLW) was calculated by dividing the grain length by grain width as described by Hui (Hui et al., 2020). Finally, simple statistical analyses, including the determining the mean value of each trait and the standard deviation, Student’s t-test for comparison between eight traits between Xihui18 and Z708, frequency distribution analysis of grain width in the F₃ population, and the chi-square test, were performed using the statistical functions in Microsoft Excel 2016.

The DNA of the parental plants and 150 F₂ individuals used for QTL mapping was extracted using the CTAB method described previously (McCouch et al., 1988). PCR amplification, non-denaturing polyacrylamide gel electrophoresis, and rapid silver staining were performed, as described previously (Zhao et al., 2016). The Xihui18-type band was scored as ‘-1,’ the Z708-type band was scored as ‘1,’ the heterozygote was scored as ‘0,’ and a missing band was scored as’.’. The specific description of QTL mapping using the mixed linear model (MLM) method in SAS 9.3 software was the same as that described in a previous study (Zhao et al., 2016). The threshold for determining whether a QTL existed was set at p < 0.05.

Based on the QTL mapping information, four F₂ individuals carrying only one single target substitution segment and 0-1 heterozygous markers for each line were selected and planted as Z1229, Z1231, Z1232, and Z1234 in 2023, with 30 plants per line. Then, the leaves of 20 individuals were taken from each line to construct SSSLs using MAS with the target substitution markers and residual heterozygous markers. Finally, the homozygous SSSLs (S1, S2, S3, and S4) were screened.

At maturity, 10 plants each of Xihui18 and SSSLs (S1-S4) were harvested. Eight yield traits were measured: grain length, grain width, ratio of length to width, 1000-grain weight, panicle number per plant, panicle length, grain number per panicle, and yield per plant. The specific method has been described by Xu and Sun (Xu et al. 2023; Sun et al. 2022). For each SSSL (S1-S4), QTL were identified using one-way analysis of variance (ANOVA) and least significant difference (LSD) multiple comparisons with Xihui18 for each SSSL_i in IBM SPSS Statistics 25.0. At p-value < 0.05, a QTL for a certain trait was considered to exist in the SSSL_i. The genetic model in a certain environment was p₀ = μ₀ + ϵ for Xihui18 and p_i = μ₀ + a_i + ϵ for an SSSL carrying a specific QTL (p₀ and p_i represented the phenotypic value of any plant in a plot of xihui18 and the SSSL_i, μ₀ represented the mean value of the Xihui18 population, a_i represented the additive effect of the QTL). Thus, a_i was equal to half of the difference between p_i and p₀ (half was estimated as the genetic effect).

An F₃ population comprising 285 individuals developed from an F₂ recombinant plant of qGW4 was used for qGW4 inheritance analysis and fine mapping. Among them, 74 recessive individuals (narrow grains) and seven newly designed polymorphism SSR markers together with RM3276, were utilized to analyze the linkage with qGW4. Moreover, four individuals with different genotypes were screened to construct secondary SSSLs of qGW4 in the F₃ generation. Furthermore, the grain widths of all SSSL individuals and one NCL population (10 individuals) (all lanes of markers were the same as Xihui18) were measured to be utilized for the overlapping substitution mapping of qGW4. The specific method has been described previously (Yang et al., 2021). When the grain width differed significantly between a secondary SSSL and Xihui18, the QTL controlling grain width was located in the substitution segment of the SSSL. When multiple substitution segments in the SSSLs overlapped with the grain width, the QTL was mapped to the overlapping region.

Based on the previous breeding of Z708 with 236 polymorphic SSR markers between two parents selected from 429 markers covering the whole rice genome 16 SSR markers in four substitution fragments and 36 outside them from 12 chromosomes were used to detect the accuracy of the substitution segments and the purity of genetic background in ten plants of Z708. The substitution segments of the 10 individuals of Z708 were identical, and no other residual fragments from Jinhui35 were detected. These results confirm the accuracy of the genotype in CSSL-Z708. Z708 contained four substitution segments from Jinhui35 in the Xihui18 background with a total length of 14.08 Mb and an average substitution length of 3.52 Mb, which were distributed on chromosomes 4, 5, 6, and 12 ( Figure 2 ).

As only four substitution segments differed from Xihui18, Z708 was considered a near-isogenic line (NIL) relative to Xihui18. The Z708 plant type ( Figure 3A ) resembled that of Xihui18, but still exhibited a heavy-spike phenotype. The most attractive characteristics of Z708 were its large grain size ( Figures 3B–G ) and less grain number ( Figures 3H–L , Table 1 ). Compared to Xihui18, the grain width of Z708 increased significantly by 25.48% to 3.78 mm ( Figures 3B, C , Table 1 ), while the 1000-grain weight (39.38 g) of Z708 increased by 36.31% ( Figure 3G ). In contrast, the grain length (10.54 mm) ( Figures 3D, E ) and the ratio of length to width ( Figure 3F , Table 1 ) in Z708 were significantly lower, by 1.12% and 21.20%, respectively. Additionally, the number of primary branches (13.81), secondary branches (30.82), spikelets per panicle (170.18), and grains per panicle (145.75) of Z708 decreased significantly by 20.27%, 26.50%, 32.96%, and 33.79%, respectively, compared to those in Xihui18 ( Figures 3I–L , Table 1 ).

Using an F₂ segregated population (the specific parameters in Table 1 ) from Xihui18/Z708 for QTL mapping, nine QTL were detected on chromosomes 4 and 5, including six for grain size and three for grain number related traits ( Table 2 ). The wide grain of Z708 was controlled by qGW4 and qGW5, both with additive effects of 0.08 mm, which explained 17.17% and 15.32% of the phenotypic variation, respectively. The additive effects of qRLW4 and qRLW5 from Jinhui35 reduced ratio of length to width of Z708 by 0.09 and 0.07, respectively, explaining the variation of 24.94% and 16.37%, respectively. The additive effects of qGWT4 and qGWT5 from Jinhui35 alleles enhanced the 1000-grain weight of Z708 by 1.13 g and 1.15 g, explaining the variation of 16.23% and 16.86%, respectively. The less grain numbers of Z708 was resulted from qGPP5, qSPP5 and qSSR5 from Jinhui35, whose additive effect reduced 9.54 grains per panicle, 8.30 spikeletes per panicle and 1.25 pencent points of seed setting rate in Z708, explaining 4.04%, 2.68% and 2.11% of the according variation, respectively ( Table 2 ).

Four SSSLs (S1, S2, S3, and S4) were developed using marker-assisted selection (MAS) based on the QTL mapping results ( Figure 4A ). The single-segment substitution line S1 harbored the substitution segment RM3476–RM3276-RM5611–RM6057 of chromosome 4. S2 carried the substitution segment RM5874–RM3322-RM2422-RM169–RM6082 of chromosome 5. S3 contained the substitution segment RM5957–RM3567–RM6458 of chromosome 6. S4 harbored the substitution segment RM1246–RM1986-RM2855-RM5282–RM1264 of chromosome12. The seven QTL identified above were validated using S1 and S2. Further, nine QTL for yield traits were detected by S1, S2, S3, and S4, including qGPP4, qGPP6 and qGPP12 for grain number per panicle; qYD4, qYD5, qYD6 and qYD12 for yield per plant; and qPN4 and qPN6 for panicle number per plant ( Figure 4A ).

The grain width (3.11and 3.74 mm) of S1 carrying qGW4 (a = 0.05 mm) and S2 containing qGW5 (a =0.36 mm) was significantly wider than that of Xihui18 (3.01 mm), whereas S3 and S4 without QTL for GW showed no significant differences in grain width (3.02 and 3.02 mm) from Xihui18 (3.01 mm) ( Figure 4B ). The ratios of the length to width (3.36 and 2.82) of S1 harboring qRLW4 (a = −0.05) and S2 with qRLW5 (a = −0.32) were significantly less than those (3.46, 3.42 and 3.44) of the recipient parent Xihui18, S3, and S4 without QTL for RLW ( Figure 4C ). The 1000-grain weights (30.02 and 41.42 g) of S1 carrying qGWT4 (a = 0.42 g) and S2 with qGWT5 (a = 6.12 g) were significantly higher than those of Xihui18, S3, and S4 (29.18, 28.97, and 29.24 g, respectively) without QTL for GWT ( Figure 4D ). The grain length of S1-S4 without QTL for this trait was the same as that of Xihui18 ( Figure 4E ).

The panicle number per plant (5.4 and 5.1) in S1 qGW4 harboring qPN4 (a = 0.70) and S3 carrying qPN6 (a = 0.55) was significantly higher than that of Xihui18 (4.00) and those of S2 and S4 (4.50 and 4.50, respectively) without QTL for PN ( Figure 4F ). Grain numbers per panicle (204.97 and 215.30) of S3 containing qGPP6 (a = 9.09) and S4 carrying qGPP12 (a = 14.26) were significantly greater whereas those of S1 containing qGPP4 (a = -12.46) and S2 with qGPP5 (a = -8.86) (161.86 and 169.07, respectively) were significantly less than that (186.79) of Xihui18 ( Figure 4G ). The yield per plant of S1 with qYD4 (a = 2.13 g), S2 carrying qYD5 (a = 3.27 g), S3 containing qYD6 (a= 3.68 g), and S4 harboring qYD12 (a= 2.39 g) (26.56, 28.85, 29.66, and 27.08 g, respectively) was significantly higher than that of Xihui18 (22.30 g) ( Figure 4H ).

The donor parent Jinhui35 exhibited a wide grain type, whereas the recipient parent Xihui18 displayed a narrow grain type. Z708, which carried four substitution fragments from Jinhui35 in a Xihui18 background, exhibited broad grains. In the F₃ population consisting of 285 individuals constructed using a qGW4 recombinant plant, the frequency of grain width displayed a bimodal distribution, one peak for a narrow grain type (from 2.91 mm to 3.06 mm), with 74 plants, and the other for broad grain type (from 3.06 to 3.45 mm), with 211 individuals ( Figure 5 ). The chi-square test indicated that the broad-grain plants (211) and thin-grain individuals (74) fitted a separation ratio of 3:1 (χ² = 0.17 < χ² _{(0.05, 1)} = 3.84) (Table 3). These results suggested that qGW4 controlling wide grains from Jinhui35 in S1 displayed a single dominant gene action ( Figure 5 ).

Given the qGW4 linkage with RM3276 in S1 ( Figure 6A ), qGW4 was further fine-mapped using 74 recessive plants in the above F₃ population. The grain width of these recessive plants (3.01 mm) was not significantly different from that of Xihui 18 (3.01 mm) ( Figure 6B ). Within the largest substitution interval of RM17389 and RM17485 of S1, ten SSR markers were designed, seven of which showed polymorphisms between Xihui18 and Jinhui35. Among the seven markers, the bands of three in S1 were the same as those in Jinhui35, indicating that these markers were in the substitution segment, whereas the other four were the same as the recipient Xihui18, suggesting that they belonged to the genetic background of S1. Therefore, by linkage analysis of all markers in the substitution segment using 74 recessive plants, the genetic distances of qGW4 were 6.20, 2.64, 0.79, and 0.79 cM from RM17468, RM17450, RM7453 and RM3276, respectively. Thus, qGW4 was fine-mapped between RM17453 and RM3276, with a physical distance of 80 Kb ( Figure 6A ). We also developed a series of secondary SSSLs (S5–S8) for the overlapping substitution mapping of qGW4 in the F₄ generation ( Figure 6C ). The grain width (3.03 mm) of the negative control line (NCL), whose lanes of all markers in the substitution fragment were the same as those of Xihui18, displayed no significant difference from that of Xihui18 (3.01 mm). However, S5 carrying the RM17450-RM17453-RM3276-RM17468 substitution segment, S7 containing the RM3276 segment, and S8 harboring the RM17450-RM17453- RM3276 segment exhibited significantly wider grains (3.10, 3.59, and 3.55 mm, respectively) than those of Xihui18 (3.01 mm). Further, S6 carrying the RM17468 segment, showed no difference in grain width (3.05 mm) from that of Xihui18 (3.01 mm). Therefore, qGW4 should be in 155 kb of the estimated length, and 310 Kb of the largest substitution length of RM17453–RM3276–RM17468 ( Figure 6C ). These results were consistent. qGW4 should be in 80 Kb of the minimum distance between RM17453 and RM3276, 155 Kb of the estimated length, and 301 kb of the largest substitution length of RM17453–RM3276–RM17468 ( Figures 6A, C ). By predicting the candidate genes within 301 kb of the largest substitution length, four candidate genes of qGW4 were predicted to encode a MYB protein, DNA-containing protein, PAP fibrinogen family protein, and DUF domain protein (Table 4).

Rice grain yield is determined by three major “visible” morphological traits: grain weight, grain number per panicle, and panicle number (Li et al., 2021). These yield traits exhibit complex inheritance patterns that are often controlled by several minor QTL. Chromosomal segment substitution lines are ideal materials for the genetic dissection and pyramiding of favorable QTL (Ebitani et al., 2005; Ando et al., 2008; Balakrishnan et al., 2019; Nagata et al., 2023). Here, we characterized a less-grain and wide-grain rice CSSL, Z708, which contained four substitution fragments from Jinhui35 in a Xihui18 background, with an average substitution length of 3.52 Mb.

Although Z708 has some favorable alleles from Jinhui35 for wide and large grains, Its direct use in rice breeding is challenging because of the presence of multiple QTL. To construct more suitable materials for the direct breeding and functional analysis of key genes, eight SSSLs containing various QTL were developed from Z708. As each SSSL has only one difference of a single substitution fragment from its recipient parent, they can identify QTL accurately, which can be verified by the fact that S1-S4 detected more QTL (16) for yield traits than those (7) detected using the F₂ population of Xihui18/Z708. Many studies have confirmed the high efficiency of QTL detection (Ando et al., 2008; Teng et al., 2012; Ookawa et al., 2016; Zhang et al., 2021a; Hui et al., 2020). Notably, S1, S2, S3, and S4 displayed a significant increase in yield per plant compared to that of Xihui18. For instance, S1 which harbors qPN4 (a = 0.70), qGW4 (a = 0.05 mm), qGWT4 (a = 0.42 g), and qGPP4 (a = -12.46) exhibited a 17.01% increase in yield per plant relative to Xihui18 (22.30 g); S2 carrying qGW5 (a = 0.36 mm), qGWT5 (a = 6.12 g), and qGPP5 (a = -8.86) showed a 29.4% increase in yield; S3 containing qPN6 (a = 0.50) and qGPP6 (a = 9.09) achieved a 33.00% increase, while S4 with qGPP12 (a = 14.26) increased the yield per plant by 21.40% than that of Xihui18 (22.30 g) ( Figure 4 ). Notably, both Xihui18 and Jinhui35 are strong restorer lines carrying restorer genes Rf1 and Rf3 on Chr.1, Rf2 on Chr.2, and Rf4 on Chr.10 (Akagi et al., 2004; Itabashi et al., 2011; Cai et al., 2013), Thus, some of them could be used as restorer lines to breed new hybrid cultivars by combining with sterile indica lines. Therefore, these SSSLs have promising prospects for improving rice yield in breeding by design.

In total, 16 QTL for yield traits were detected in both the F₂ population derived from Xihui18/Z708 and the dissected SSSLs. We compared these QTL with previously reported genes located within the corresponding substitution intervals, qGW4, qRLW4, qGWT4, qPN4, qGPP4, and qYD4, all linked to RM3276. Within the substitution interval, OsAGO2 regulates the distribution of cytokinins by activating BG3, thereby increasing rice grain weight, but has little effect on grain width (Yin et al., 2020). Thus, OsAGO2 may be a candidate gene for qGWT4. Determining whether OsAGO2 is a candidate gene for qGWT4 requires genetic complementarity studies in the future. Finally, qGW4 was fine-mapped to 80 Kb of the minimum interval and 301 Kb of the largest substation interval, and four candidate genes were predicted. MONOCULM 3 encoding WUSCHEL, is a key gene involved in the formation of rice tiller buds (Lu et al., 2015). Thus, MOC3 may be a candidate gene for qPN4. qGW5, qRLW5, qGWT5, qGPP5, and qYD5, all linked to RM2422. OsGSK2 was found in the substitution interval. OsGSK2, a homolog of Arabidopsis BIN2, affects cell proliferation and expansion to negatively control grain size by phosphorylating substrates, including OFP3 (Tong et al., 2012; Xiao et al., 2020). Thus, OsGSK2 may be a candidate gene for qGW5, qRLW5, qGWT5. Again, qGW5 and qGWT5 were also identified as qGW5 and qGWT5-2 by Sun et al. using CSSL-Z431 (Sun et al., 2022). However, they came from different donors (Huhan 3), and whether they are different alleles requires further study. qPN6, qGPP6, and qYD6 were linked to RM3567. Within the substitution interval of S3, MOC1/GNP6 regulates tiller and grain number development (Zhang et al., 2021b). Thus, MOC1/GNP6 is a candidate gene for qPN6, qGPP6, and qYD6. qGPP12 and qYD12 were linked to RM1986. Within the substitution interval of S4, GNP12 encodes RGH1A protein and regulates rice yield by regulating panicle length, grain number per panicle, and grain length (Pan et al., 2022). OsVIL2 improves biomas and grain production by suppressing OsCKX2 chromatin (Yang et al., 2019). Thus, GNP12 and OsVIL2 may be candidate genes for qGPP12 and qYD12. However, whether these genes are candidate genes for the related QTL requires further verification using genetic complementarity experiments.

In conclusion, qGW4, qGPP4, qYD4, qGPP5, and qYD5, which have not been previously reported, provide a good foundation for the functional analysis of these QTLs. Further, S1-S4 carrying favorable alleles, such as qPN4, qGW4, qYD4, qGW5, qGWT5, qYD5, qPN6, qGPP6, qYD6, qGPP12, and qYD12 could be used as new restorer lines for breeding improved hybrid varieties.