Genetic Dissection of Rice Ratooning Ability Using an Introgression Line Population and Substitution Mapping of a Pleiotropic Quantitative Trait Locus qRA5

Ratooning ability is a key factor that influences ratoon rice yield, in the area where light and temperature are not enough for second season rice. In the present study, an introgression line population derived from Minghui 63 as the recipient parent and 02428 as the donor parent was developed, and a high-density bin map containing 4568 bins was constructed. Nine ratooning-ability-related traits were measured, including maximum tiller number, panicle number, and grain yield per plant in the first season and ratoon season, as well as three secondary traits, maximum tiller number ratio, panicle number ratio, and grain yield ratio. A total of 22 main-effect QTLs were identified and explained for 3.26–18.63% of the phenotypic variations in the introgression line population. Three genomic regions, including 14.12–14.65 Mb on chromosome 5, 4.64–5.76 Mb on chromosome 8, and 10.64–15.52 Mb on chromosome 11, were identified to simultaneously control different ratooning-ability-related traits. Among them, qRA5 in the region of 14.12–14.65 Mb on chromosome 5 was validated for its pleiotropic effects on maximum tiller number and panicle number in the first season, as well as its maximum tiller number ratio, panicle number ratio, and grain yield ratio. Moreover, qRA5 was independent of genetic background and delimited into a 311.16 kb region by a substitution mapping approach. These results will help us better understand the genetic basis of rice ratooning ability and provide a valuable gene resource for breeding high-yield ratoon rice varieties.


Introduction
As one of the important staple crops, rice is very important for humans worldwide, especially for people in Asia. In the last three decades, rice yield has sharply increased due to the use of the semi-dwarf gene and heterosis [1]. However, because of the explosive increase in the world's population, the rice yield must be largely increased in the next half-century to meet the needs of the world's population. This rice yield increase depends on a higher yield and more frequent harvests on the existing land [2]. Since the agricultural acreage is hard to increase, we have to increase rice production by improving the rice yield or cropping index per unit [3]. Due to the increasing cropping index, double-season rice could largely increase the total yield per unit. However, this double-season system requires The background recovery rates of ILs were demonstrated in Figure 1. The average background recovery rate of the IL population was 92.84%, with a range of 12.39-99.98%, and the background recovery rates of more than 70% ILs were above 90%, indicating that the genetic backgrounds of most of the ILs were similar with the recipient parent (MH63).

Performance of Ratooning Ability
As shown in Table 1, the two parents exhibited obvious differences in all traits, except GYR, in both years. MH63 had more tillers and panicles, and a higher yield in both the first and ratoon seasons in two years, resulting in MH63 having a higher ratooning ability than 02428. All traits showed continuous distributions and transgressive segregations in the IL population, suggesting that all nine traits were quantitative traits and controlled by multiple genes.

Performance of Ratooning Ability
As shown in Table 1, the two parents exhibited obvious differences in all traits, except GYR, in both years. MH63 had more tillers and panicles, and a higher yield in both the first and ratoon seasons in two years, resulting in MH63 having a higher ratooning ability than 02428. All traits showed continuous distributions and transgressive segregations in the IL population, suggesting that all nine traits were quantitative traits and controlled by multiple genes.
The correlations among different traits in the IL population were very similar between 2018 and 2019 ( Table 2). In either the first season or ratoon season of each year, the maximum tiller number had a significantly positive correlation with the panicle number, and they were also significantly positively correlated with grain yield, suggesting that individuals with more tillers would have more panicles and higher grain yield. There were significantly positive correlations between two seasons within the same traits, indicating plants with more tiller and panicles as well as higher grain yield in the first season would have more tillers and panicles as well as higher grain yield in the ratoon season. The correlation coefficients between TNR and PNR, between TNR and GYR, and between PNR and GYR were all significant, indicating that varieties with higher TNR would have higher PNR and GYR. As expected, TNR, PNR, and GYR had significantly negative correlations with FTN, FPN, and FGY, respectively, in the first season, while all of them were significantly positively correlated with FTN, FPN, and FGY in the ratoon season, except between TNR and RGY in 2019, indicating that the ratooning ability was more correlated with the ratoon season than the first season.

QTL Mapping of Ratooning Ability
A total of 22 M-QTLs for nine ratooning-ability-related traits were identified on chromosomes 1-3, 5, 7-9, 11, and 12 in 2018 and 2019. Among them, six QTLs were detected in both years, and the others were identified in one year (Table 3; Figure 2). In the first season, a total of 9 QTLs were identified, including four, four, and one QTLs for FTN, FPN, and FGY, respectively (Table 3; Figure 2). Among them, 3 QTLs (qFTN8, qFTN11, and qFPN12) were detected only in 2018, accounting for 8.69-16.71% of the phenotypic variations, while qFTN7, qFPN3, and qFGY11 were identified only in 2019, explaining 8.11-10.66% of the phenotypic variations. Three QTLs (qFTN5, qFPN5, and qFPN8) were stably expressed in both years with mean phenotypic variation of 13.33%, 15.26%, and 7.76%, respectively. MH63 alleles at all QTLs increased traits.
Five QTLs were identified in the ratoon season, including two, two, and one QTLs for RTN, RPN, and RGY, respectively (Table 3; Figure 2). They were distributed on chromosomes 1-3 and 11. Among them, three QTLs (qRTN11, qRPN3, and qRGY11) were identified in 2018 and explained 3.26-9.52% of the phenotypic variations. The remaining two QTLs (qRTN1 and qRPN2) were found only in 2019, accounting for 4.35% and 4.27% of phenotypic variation, respectively. MH63 alleles at all QTLs increased traits.
A total of eight QTLs were identified for ratooning ability, including three, three, and two QTLs for TNR, GYR, and PNR, respectively (Table 3; Figure 2). Among them, three QTLs (qTNR5, qPNR5, and qGYR5) were identified in both years with the mean phenotypic variations of 13.95%, 9.41%, and 11.59%. qTNR9, qGYR9, and qGYR12 were expressed only in 2018 and accounted for phenotypic variation of 5.89-11.20%. The other two QTLs (qTNR11 and qPNR11) were found only in 2019 and explained 14.00% and 9.73% of phenotypic variations. MH63 alleles were associated with increased trait values at qTNR9, qTNR11, qGYR5, and qGYR12, while associated with decreased trait values at the rest of the four QTLs. In the first season, a total of 9 QTLs were identified, including four, four, and one QTLs for FTN, FPN, and FGY, respectively (Table 3; Figure 2). Among them, 3 QTLs (qFTN8, qFTN11, and qFPN12) were detected only in 2018, accounting for 8.69-16.71% of the phenotypic variations, while qFTN7, qFPN3, and qFGY11 were identified only in 2019, explaining 8.11-10.66% of the phenotypic Among the 22 QTLs detected in this study, three regions had pleiotropic effects on two or more traits (Table 3; Figure 2). For example, the region of 14.12-14.65 Mb on chromosome 5 harbored qFTN5 for FTN, qFPN5 for FPN, qTNR5 for TNR, qPNR5 for PNR, and qGYR5 for GYR; the region of 4.64-5.76 Mb on chromosome 8 harbored qFTN8 for FTN and qFPN8 for FPN; and the 10.64-15.52 Mb region on chromosome 11 harbored qFTN11 for FTN, qFGY11 for FGY, qRTN11 for RTN, qRGY11 for RGY, and qPNR11 for PNR.

Validation of qRA5
The region of 14.12-14.65 Mb on chromosome 5 had a pleiotropic effect on FTN, FPN, TNR, PNR, and GYR in both 2018 and 2019. Thus, it was called pleiotropic QTL qRA5 (ratooning ability 5). An IL (DQ146) was selected and used for validation of the gene effect of qRA5 (Table 4). DQ146 had 82.5% genome recovery of MH63 and did not carry any other M-QTL, except qRA5 in the whole genome, so it could be considered a near-isogenic line of MH63. The FTN, FPN, and GYR of DQ146 were 18.50, 8.67, and 0.58, respectively, significantly lower than 21.53, 11.33, and 0.71 for MH63, while the TNR and PNR of DQ146 were 1.24 and 1.90, respectively, significantly higher than those of 0.95 and 1.65 of MH63. The above results indicated that qRA5 was truly inherited and had pleiotropic effects on FTN, FPN, TNR, PNR, and GYR.

Genetic Background Effect on qRA5
To identify the genetic background effect on qRA5, one IL (DQ285) was also developed using 02428 as the recipient parent and MH63 as the donor parent, with a background recovery rate of 88.9%. It did not contain any other M-QTL, except qRA5 in the whole genome, and could be also considered as a near-isogenic line of 02428. The FTN and FPN of DQ285 were significantly higher than those of 02428, while its TNR and PNR were significantly lower than those of 02428 ( Table 5). The results suggested that qRA5 was consistently expressed in indica (MH63) and japonica (02428) backgrounds, independent of genetic background.

Substitution Mapping of qRA5
An F 2 segregation population derived from DQ146 and MH63 was used for substitution mapping of qRA5. Among the 11 SSR markers in the region of qRA5, four markers (RM18361, RM18366, RM5705, and RM18378) had polymorphism between MH63 and 02428. Two hundred and eighty-eight individuals in the F 2 segregation population were genotyped using the four polymorphic markers. Four recombinants were identified in the interval of RM18361-RM18378, and their homozygous progenies were used for phenotypic evaluation (Figure 3). Both R1 and R3 showed significantly lower FTN, FPN, and GYR as well as significantly higher TNR and PNR than those of MH63, suggesting qRA5 was delimited in the region of approximately 311.16 kb, flanked by RM18366 and RM5705.

Discussion
Ratoon ability is a very important factor that influences yield in ratoon season, and it is controlled by multiple genes and largely affected by the environment. In the present study, a total of 22 QTLs were identified for nine ratooning-ability-related traits in an IL population derived from MH63 as the recipient parent and 02428 as the donor parent; some QTLs were located in the same or adjacent regions of QTLs [7,10,12], with genes [25][26][27] previously reported, as shown in Table 3. In particular, qRA5 in the region of 14.12-14.65 Mb on chromosome 5 was located near qETN5 [10], qRA5 [10], and qRAT5 [7]. Most previously reported QTLs were identified in large regions and have not been validated. Moreover, some previously cloned genes for tiller number or panicle number remain to be identified for their effects on ratoon ability. Allelic correspondence of the QTLs for ratooning ability, as identified in the present study along with previously reported ones, should be further investigated by fine-mapping and cloning. Besides, the only validated QTL, qRAT5, was substitution mapped between InDel 3 and RM249 in a large region of 5.29 Mb. In the present study, the qRA5 was validated by an IL (DQ146) and has been fine-

Discussion
Ratoon ability is a very important factor that influences yield in ratoon season, and it is controlled by multiple genes and largely affected by the environment. In the present study, a total of 22 QTLs were identified for nine ratooning-ability-related traits in an IL population derived from MH63 as the recipient parent and 02428 as the donor parent; some QTLs were located in the same or adjacent regions of QTLs [7,10,12], with genes [25][26][27] previously reported, as shown in Table 3. In particular, qRA5 in the region of 14.12-14.65 Mb on chromosome 5 was located near qETN5 [10], qRA5 [10], and qRAT5 [7]. Most previously reported QTLs were identified in large regions and have not been validated. Moreover, some previously cloned genes for tiller number or panicle number remain to be identified for their effects on ratoon ability. Allelic correspondence of the QTLs for ratooning ability, as identified in the present study along with previously reported ones, should be further investigated by fine-mapping and cloning. Besides, the only validated QTL, qRAT5, was substitution mapped between InDel 3 and RM249 in a large region of 5.29 Mb. In the present study, the qRA5 was validated by an IL (DQ146) and has been fine-mapped by a substitution mapping approach in a smaller region, with 311.16 kb flanked by RM18366 and RM5705. There were nine newly identified QTLs (qFTN8, qFPN3, qFPN8, qFPN12, qRTN1,  qRPN1, qTNR9, qGYR9, and qGYR12) in the present study. With increasing attention paid to this aspect, more consistent QTLs in different studies would be identified, and the genetic basis of ratoon ability would be clearly revealed.
There were 47 candidate genes in the region of qRA5 (http://rice.plantbiology.msu. edu/; accessed on 1 July 2021). Among them, 26 genes encode transposon protein, 10 genes encode expressed protein, and 2 genes encode hypothetical protein. Besides, LOC_Os05g24930 and LOC_Os05g25060 encode PPR repeat-containing protein, LOC_Os05g24684 encodes structural constituent of ribosome, LOC_Os05g24760 encodes Harpin-induced protein 1 domain-containing protein, LOC_Os05g24770 encodes reticulon domain-containing protein, LOC_Os05g24780 encodes OsCML21-calmodulin-related calcium sensor protein, LOC_Os05g24880 encodes dehydrogenase, LOC_Os05g24890 encodes tetratricopeptide-like helical, and LOC_Os05g24970 encodes LSM domain-containing protein. In previous studies, a lot of PPR proteins were reported to be associated with the biosynthesis of chloroplast [28][29][30], and many white or yellow color leaf mutants exhibited a decreased panicle number [31,32], so they may affect tiller number and ratooning ability. A gene (OsALDH2B1) encoding aldehyde dehydrogenase was reported to control tiller number [33], thus, genes encoding dehydrogenase may also control ratooning ability. Therefore, LOC_Os05g24930, LOC_Os05g24880, and LOC_Os05g25060 were more likely the candidate genes. Among the three candidate genes, there was only one non-synonymous SNP (G/T), in the fourth exon of LOC_Os05g25060. The MH63 allele encodes Arg, while the 02428 allele encodes Met. Moreover, according to the CREP database (http://crep.ncpgr.cn/crep-cgi/home.pl; accessed on 1 March 2010), the LOC_Os05g25060 was expressed highly in the stem during the tilling stage. Thus, LOC_Os05g25060 was the most likely candidate gene of qRA5. Another region of 10.64-15.52 Mb on chromosome 11 is important, affecting FTN, FGY, RTN, RGY, and PNR, and involving grain yield both in the first season and ratoon season. Next, we will develop a large segregating population to fine-map the important gene for rice ratoon ability inside this region.
Increasing ratoon rice yield is a very important breeding goal, in the area where light and temperature are not enough for second season rice. However, the rice ratooning ability is a complex quantitative trait controlled by multiple genes and largely affected by the environment. Thus, identifying QTL and introgressing them into elite rice varieties is an effective way to improve ratooning ability. In many previous studies, the genetic background effect in QTL expressions was identified to be very large for appearance quality [34] and other traits. Besides, pleiotropic QTL also had a large value for improving ratooning ability, as breeders could only introduce the pleiotropic QTL instead of pyramiding multiple loci and save a lot of labor, money, and time. To date, many yield-related genes with pleiotropic effects on different traits have been identified and cloned, such as Ghd7 [35] and Ghd7.1 [36]. In the present study, three genomic regions, including 14.12-14.65 Mb on chromosome 5, 4.64-5.76 Mb on chromosome 8, and 10.64-15.52 Mb region on chromosome 11, simultaneously controlled two or more ratooning-ability-related traits and would be very valuable in molecular breeding for improving rice ratooning ability. Moreover, the region of 14.12-14.65 Mb on chromosome 5 was validated to be independent of genetic background. Although MH63 alleles decreased TNR and PNR in the region of 14.12-14.65 Mb on chromosome 5 and PNR in the region of 10.64-15.52 Mb region on chromosome 11, they could increase FTN, FPN, FGY, RGY, and GYR. The MH63 alleles of these three regions could be pyramided, simultaneously increase FTN and FPN, and, finally, increase grain yield in both the first and ratoon seasons. Besides, novel alleles increasing ratooning ability at important QTLs such as qRA5 should be mined from rice germplasms in the future, enriching gene resources for the molecular breeding of high-yield ratoon rice varieties.

Development of IL Population
MH63 is the restorer parent of Shanyou 63, a widely adapted indica hybrid variety in China with high ratooning ability, while 02428 is a typical japonica variety, with wide compatibility and low ratooning ability. MH63 was used as a female parent to cross with 02428, F 1 was backcrossed with MH63 to generate a BC 1 F 1 population with about 80 plants, and BC 1 F 1 individuals were randomly used as male parents to cross with MH63 to produce a BC 2 F 1 population with about 200 individuals. Then, the BC 2 F 1 progenies were successively self-crossed for 6 generations with no selection from BC 2 F 1 to BC 2 F 7 generation. Finally, a set of BC 2 F 7 ILs population containing 226 lines was successfully developed.

Genotyping of the IL Population
Young leaves of about 20 individuals of each line and two parents were harvested, and DNA was extracted using the DNeasy Mini Kit (Qiagen). Their genomes were resequenced using an Illumina Genome Analyzer IIx, as described by Huang et al. [37]; then, the genotypes of ILs were determined based on the SNPs generated by whole-genome sequencing. The whole genomes were obtained as 5,336,108,154 bp and 5,562,905,674 bp of data for MH63 and 02428, respectively, while 5,062,106,567 bp (96.57%) and 5,278,080,725 bp (94.03%) of consistent sequences were obtained and aligned to the Nipponbare sequence (IRGSP 1.0) [38]. There were 58,936 polymorphic SNPs between the two parents. These SNPs were used for constructing a bin map with 4568 bins, as described by Xie et al. [39]. A bin was defined as coverage of SNPs that have the same genotypes along a chromosome. Field experiments were conducted using a randomized complete block design with two replications. All field management followed local farmers' practices.

Field Experiment and Trait Evaluation
At the tillering stage in the first season, ten individuals with uniform growth in the middle rows of each line were measured for their tiller numbers every week, to determine the maximum tiller number per plant (FTN). At the heading stage, heading dates were recorded. At the maturity stage, panicle number per plant (FPN) was investigated, then harvested 30 days after flowering. The yield of plots was measured and the grain yield per plant (FGY) was calculated by the ratio of the yield of a plot to plant number. The stubble of each plant was maintained at the height of about 40 cm, as described by He et al. [3].
In ratoon season, individuals measured in the first season of each line were measured for their tiller numbers every 3 days, to determine the maximum tiller number per plant (RTN). Heading date and panicle number per plant (RPN) were investigated at the heading and maturity stages, respectively. Then, all measured plants were harvested about 25 days after flowering. The yield of plots was measured, and the grain yield per plant (RGY) was determined by the ratio of the yield of a plot to plant number.
The ratooning ability was measured as three traits: tiller number ratio (TNR), determined by the ratio of RTN to FTN; panicle number ratio (PNR), determined by the ratio of RPN to FPN; and grain yield ratio (GYR), determined by the ratio of RGY to FGY.

Data Analysis and QTL Mapping
Statistical description and correlations among different traits of the IL population and the difference of traits between ILs were analyzed by Statistica 5.5 [34]. Main-effect QTL was identified, using the inclusive interval mapping (ICIM) function with the bi-parent population (BIP) module in IciMapping 4.0 [40], and LOD thresholds were set as 3.0.

Validation of qRA5
qRA5 affecting five ratooning-ability-related traits was identified in the region of 14.12-14.65 Mb on chromosome 5 in both years. To confirm qRA5, one IL (DQ146) was selected from the IL population because it carried only one QTL (qRA5), without genetic noise resulting from any other QTL for ratooning-ability-related traits and a higher recovery rate of the MH63 genome. DQ146 and MH63 were planted in the spring to autumn seasons of 2020 on the experimental farm of Yangtze University in Jingzhou, China, with eight rows and 10 individuals per row. Twenty individuals with uniform growth in the middle rows of each line were selected, and all their ratooning-ability-related traits (FTN, FPN, FGY, RTN, RPN, and RGY) were evaluated; thus, TNR, PNR, and GYR were calculated, following the method described above. Differences were tested for TNR, PNR, and GYR between DQ146 and MH63 using the Ducan t-test at the threshold of p ≤ 0.05.

Detection of Genetic Background Effect on qRA5
One IL (DQ285) with MH63 allele of qRA5 at 02428 background was developed [34]. It was also planted in the spring to autumn seasons of 2020 on the experimental farm of Yangtze University in Jingzhou, China, with eight rows and 10 individuals per row. About 20 individuals with uniform growth in the middle rows were selected and all their ratooning-ability-related traits were evaluated following the method as validation of qRA5. Finally, the differences between DQ285 and 02428 were tested using the Ducan t-test at the threshold of p ≤ 0.05.

Substitution Mapping of qRA5
DQ146 was crossed with MH63 and generated an F 2 segregating population. The F 2 population comprising 288 plants was grown in the summer season of 2020 on the experimental farm of Yangtze University in Jingzhou, China. Eleven simple sequence repeat (SSR) markers (RM18359, RM18360, RM18361, RM7449, RM18366, RM18368, RM18371, RM18374, RM5705, RM18376, and RM18378) in the region of 14.12-14.65 Mb on chromosome 5, according to the Gramene database (https://www.gramene.org/markers; accessed on 1 January 2005), were used to evaluate the genotypic polymorphism between MH63 and 02428. PCR and electrophoresis followed the method described by Jiang et al. [41]. PCR was conducted in a 20 µg reaction mixture containing 50 ng genomic DNA, 10 mM Tris-HCl, 50 mM KCl, 1.5 mM MgCl 2 , 2 mM dNTP, 0.1 mM of each primer, and 1 unit Taq polymerase. The template PCR was initially denatured at 95°C for 5 min, followed by 35 cycles performed at 95°C for 30 s, 55°C for 30 s, and 72°C for 1 min, and, then, a final extension at 72°C for 5 min. The PCR products were then electrophoretically resolved on a 4% polyacrylamide gel. Recombinant individuals with different genotypes of the flanking markers were selected by the polymorphic markers flanking the QTL. About 100 progeny plants of each recombinant were planted in the spring to autumn seasons of 2021 on the experimental farm of Yangtze University in Jingzhou, China. All of them were genotyped by the polymorphic markers to select the homozygous recombinant plants. About 20 homozygous progenies of each recombinant were evaluated for their ratooning ability, following the method as validation of qRA5. MH63 and DQ146 were used as the control. They were tested for the differences between homozygous recombinants and MH63, using the Duncan t-test at the threshold of p ≤ 0.05.

Conclusions
QTL mapping affecting nine ratoon-ability-related traits was performed using a bin map containing 4568 bins in an IL population derived from MH63 as the recipient parent and 02428 as the donor parent. Nine ratoon-ability-related traits showed continuous distributions and transgressive segregations. A total of 22 main-effect QTLs were identified for nine ratoon-ability-related traits. Three genomic regions, including 14.12-14.65 Mb on chromosome 5, 4.64-5.76 Mb on chromosome 8, and 10.64-15.52 Mb on chromosome 11, were identified to simultaneously control two or more ratooning-ability-related traits. Of them, a pleiotropic QTL qRA5 was validated, independently of genetic background, and mapped into a 311.16 kb region flanked by RM18366 and RM5705. MH63 alleles at QTLs in the above three regions could be used for molecular breeding to improve rice ratooning ability.

Conflicts of Interest:
The authors declare no conflict of interest.