Mapping and Candidate Gene Prediction of qPL7-25 : A Panicle Length QTL in Dongxiang Wild Rice

: Panicle length (PL) is an important trait closely related to rice yield. More than 200 quantitative trait loci (QTL) for PL have been identified, but only a few can be used for breeding. Dongxiang wild rice contains many excellent genes, and mining favorable PL-related QTL from DXWR is helpful for rice variety improvement. Here, we report a QTL analysis for PL using a recombinant inbred line population consisting of 143 individuals derived from a cross between Dongxiang wild rice and indica cultivar Guangluai 4. A total of four QTL ( qPL1-37 , qPL4-26 , qPL7-25 , and qPL8-4 ) for PL were identified and located on chromosomes 1, 4, 7, and 8. Among them, qPL7-25 showed the largest F-value of 32.32 and 16.80, and the QTL explained 18.66% and 13.06% of the phenotypic variation of Dongxiang wild rice in Hangzhou and Hainan, respectively. QTL mapping was performed using a population of 1800 individuals derived from the crossing of NIL-qPL7-25 and GLA4. qPL7-25 was located between two InDel markers, InDel-24591 and InDel-24710, in a 119 kb region containing 14 predicted genes. Using Sanger sequencing and qRT-PCR analysis, we propose that LOC_Os07g41200 is probably a new allele of the well-known GL7 gene, which affects grain length and appearance quality in rice. These results provide new insights into the use of molecular marker-assisted selection for breeding high-yielding and high-quality rice varieties.


Introduction
Food security is a top priority for most countries. Rice is the staple food for more than 60% of the population in China, so improving rice production is crucial to ensuring national food security and is the goal of rice breeding. Rice yield is composed of the effective number of panicles per plant, the number of filled grains per panicle, the seed setting rate, and the kilogram grain weight. As a yield-related trait, panicle length (PL) directly determines the number and length of panicle branches in rice, which in turn affects the number of grains in the panicle [1,2]. Traditional rice breeding has made great strides, but there are also drawbacks of inefficiency, and molecular design breeding technology is beginning to show great potential for application, with mining for favorable alleles being the prerequisite for molecular design breeding.
Common wild rice (Oryza rufipogon Grff.) can provide favorable gene variants for many important agronomic traits in improved cultivated rice varieties [3][4][5][6][7][8][9][10]. Mining and exploiting favorable genes in O. rufipogon could be an effective way to overcome yield stagnation in cultivated rice. Dongxiang wild rice (DXWR) is the most northerly distributed O. rufipogon resource in the world, with many excellent characteristics and rich genetic diversity. Compared with other wild rice, DXWR is a close ancestor of cultivated rice, which is suitable for use as an excellent germplasm resource for cultivating rice genetic improvement. Previous studies have shown that DXWR could provide many favorable

Measurement of Traits
To find the relationships between PL and other agronomic traits, ten and five plants from two parents and 143 lines, respectively, were harvested at the maturity stage to measure the plant height (PH), PL, PBN, SBN, and SPP. PH was measured as the distance from the ground to the tip of the main panicle. PL was measured as the distance from the neck node to the tip of the main panicle using a ruler. More than 100 fully filled grains were used to measure seed length, width and length/width ratio using a seed phenotyping system (Wan Sheng, Hangzhou, China). The 1000-grain weight was measured by weighing 1000 full-filled grains with an electronic balance with three replicates. The mean values of all the characters measured were used for the analysis.

Linkage Map Construction and QTL Analysis
The SLAF sample preparation, sequencing, sequence comparison, polymorphic analysis, and associated marker identification were processed as previously reported [50]. Briefly, two restriction enzymes (Hae III and Hpy 166II) were selected for their uniform distribution and prevalence in the simulations of fragment alignments to the NPB reference genome. Arabidopsis thaliana was used as a control genome to verify the accuracy of the restriction digestion protocol. Fragments of 200-350 bp were isolated to be used as the SLAF tags. The fragments were sequenced on the Illumina HiSeq 2500 system. The SLAFs were grouped into 12 linkage groups with their positions in the reference genome. The modified logarithm of odds scores (MLOD) value between two adjacent markers was determined [51], and SLAFs with MLOD values less than 5 were filtered out. QTL analysis was performed using the QTL Network 2.0 software based on a mixed linear model with all the genotype data from the RIL mapping population. The contribution rates and additive effects of each QTL to the PL were calculated, and the detected QTL were named according to the method proposed by McCouch et al. [52]. If the additive effect value was positive, then the allele was from the DXWR parent, and if the additive effect value was negative, then the allele was from the GLA4 parent.

Construction of NIL
A VB431 line of RILs with the DXWR genotype was selected for backcrossing with the recurrent parent GLA4. The F 1 generation was then backcrossed with GLA4, combined with the PL phenotype of the progeny, and two InDel markers (InDel-24091 and InDel-25971) on either side of qPL7-25 were used for marker-assisted selection (MAS) of each generation. After three consecutive backcrosses and one self-cross, a BC 3 F 2 population containing 100 individuals with a GLA4 genetic background was isolated and used for primary mapping of qPL7-25 ( Figure S2A). The NIL-qPL7-25 contains 1.88 Mb of the qPL7-25 locus from a fragment of DXWR located on chromosome 7 between the InDel markers InDel-24091 and InDel-25971 ( Figure S2B). Rice variety P13 with the GL7 gene was used as the male, the restoring line Huazhan as the female, and the F 1 generation was then backcrossed with recurrent parent Huazhan. A pair of dominant functional markers  (NGL7-F: TGACACGCCACAGTCCAAGACGAGCAGT, NGL7-R: AAGGGAGTTGAGAG-TAGAAAAAA) was used for MAS of each generation. After four consecutive backcrosses and one self-cross, a BC 4 F 2 population was obtained. An NIL carrying a homozygous allele of P13 in the target QTL region between markers NGL7-F and NGL7-R, designated NIL-Huazhan-GL7 p13 , was also developed on a Huazhan background.

QTL Mapping and Statistical Analysis
QTL analysis was performed with the QTLNetwork2.0 software using the mixed linear models [53]. The VB431 line was crossed with the GLA4 variety to develop a mapping population. The F2 population was constructed by self-crossing of the F1 hybrid. A primary linkage of the QTLs for PL was obtained using 100 recessive plants from the F2 population. According to the resequencing and alignment results of GLA4 and DXWR, the InDel loci were identified on chromosome 7. Primers were designed using Vector NTI 11.5 biological software with the following design parameters: GC % 40-55%, base length 19-30 bp, and Tm value 55-60 • C. Seven new InDel markers were developed for fine mapping of the qPL7-25 locus (Table S1), and the qPL7-25 locus was mapped to the interval between InDel-24591 and InDel-24710 using a population of 1800 individuals derived from the crossing of NIL-qPL7-25 and GLA4. All statistical analyses were performed using Student's t-test (*, p < 0.05; **, p < 0.01).

RNA Extraction and QRT-PCR Analysis
Total RNA was isolated from 3 cm young panicles of DXWR and GLA4 plants according to the manufacturer's instructions of the AxyPerp Multisource Total RNA miniprep kit (Corning Life Sciences (WuJiang) Co., Ltd, Suzhou, China). DNAase-treated RNA (1 µg) was reverse transcribed using a ReverTra Ace qPCR RT master mix (FSQ-301, Toyobo, Osaka, Japan). qRT-PCR was performed using GL7 gene-specific primers, as described by Wang et al. [45], using the SYBR Green Real-Time-PCR master mix (QPK-201, Toyobo) and the Bio-Rad (Hercules, CA, USA) CFX96 Real Time-PCR system. Ubiquitin was used as an internal standard, and the results were calculated using the 2 −∆∆CT method [54].

Phenotypic Data of Panicle Traits of Two Parents and RIL Population
The phenotypes of the DXWR and GLA4 parents were relatively different. The PH of GLA4 in HZ and HN was 54.09% and 64.66% of that of DXWR, respectively. The PL of DXWR was 23.7 cm and 22.1 cm, while that of GLA4 was 21.2 cm and 20.0 cm in HZ and HN, respectively, and was significantly shorter than that of DXWR ( Table 1). The grain length and grain length/width ratio of DXWR were significantly greater than those of GLA4, but the grain width and 1000-grain weight were significantly lower than those of GLA4 ( Figure S1). In addition, the PBN and SBN of DXWR were significantly higher than those of GLA4, but there was no significant difference in the SPP between the parents ( Figure 1, Table 1). The skewness and kurtosis values of the PL in HZ and HN were both less than 1, indicating that the PL was normally distributed and was a polygene-controlled quantitative trait. The data obtained for the PL in the RIL population conformed to a continuous normal distribution with a wide range ( Figure 2). In addition, we identified significant positive correlations between PL pairs, the PBN, SBN, and SPP in both HZ and HN ( Table 2).   Means ± SD (n = 10 for parents, and n = 5 for RIL lines). * and ** indicate a significa between DXWR and GLA4 at the 0.05 and 0.01 levels according to the t-test, respectiv

QTL Analysis of PL
The 368 polymorphic SNP markers were evenly distributed among the 12 chromosomes and were used to construct a molecular linkage map with the SLAF markers. QTL analysis was performed using QTLNetwork2.0 software based on a mixed linear model, and p = 0.005 was used as the statistical detection threshold. A total of four QTL were detected for the PL and were distributed on chromosomes 1, 4, 7, and 8. Among them, qPL1-37, qPL4-26, and qPL7-25 were detected repeatedly in both HZ and HN, but qPL8-4 was detected only in HN ( Figure 3, Table 3). qPL1-37 was located in the marker M37-M38 on chromosome 1, with a physical distance of 1.0 Mb and an F-value of 17.13 and 9.42, and the QTL explained 8.99% and 7.31% of the phenotypic variation in HZ and HN,

QTL Analysis of PL
The 368 polymorphic SNP markers were evenly distributed among the 12 chromosomes and were used to construct a molecular linkage map with the SLAF markers. QTL analysis was performed using QTLNetwork2.0 software based on a mixed linear model, and p = 0.005 was used as the statistical detection threshold. A total of four QTL were detected for the PL and were distributed on chromosomes 1, 4, 7, and 8. Among them, qPL1-37, qPL4-26, and qPL7-25 were detected repeatedly in both HZ and HN, but qPL8-4 was detected only in HN ( Figure 3, Table 3). qPL1-37 was located in the marker M37-M38 on chromosome 1, with a physical distance of 1.0 Mb and an F-value of 17.13 and 9.42, and the QTL explained 8.99% and 7.31% of the phenotypic variation in HZ and HN, respectively. qPL4-26 was located in the marker M142-M143 on chromosome 4 with a physical distance of 1.0 Mb and an F-value of 18.77 and 10.28, and the QTL explained 12.92% and 7.48% of the phenotypic variation in HZ and HN, respectively. qPL7-25 was located in the marker M234-M235 on chromosome 7 with a physical distance of 1.0 Mb and an F-value of 32.32 and 16.80, and the QTL explained 18.66% and 13.06% of the phenotypic variation in HZ and HN, respectively. qPL8-4 was located in the marker M242-M243 on chromosome 8 with a physical distance of 1.2 Mb and an F-value of 10.25, and the QTL explained 6.17% of the phenotypic variation in HN. The enhanced alleles qPL1-37, qPL7-25, and qPL8-4 were from the DXWR parent (Table 3). As the PBN and SBN of DXWR were significantly different from GLA4, we also identified other QTL related to the panicle traits, and three QTL (qPBN1-8, qPBN2-4, and qPBN7-25) for the PBN and five QTL (qSBN1-2, qSBN1-5, qSBN1-6, qSBN5-22, and qSBN7-20) for the SBN were also detected. qPBN1-8 and qPBN2-4 were repeatedly detected in both HZ and HN, but other QTL were only detected in HZ or HN. Notably, qPBN7-25 was mapped to the same interval as qPL7-25 on chromosome 7 ( Figure 3, Table 3). 37, qPL7-25, and qPL8-4 were from the DXWR parent (Table 3). As the PBN and SBN of DXWR were significantly different from GLA4, we also identified other QTL related to the panicle traits, and three QTL (qPBN1-8, qPBN2-4, and qPBN7-25) for the PBN and five QTL (qSBN1-2, qSBN1-5, qSBN1-6, qSBN5-22, and qSBN7-20) for the SBN were also detected. qPBN1-8 and qPBN2-4 were repeatedly detected in both HZ and HN, but other QTL were only detected in HZ or HN. Notably, qPBN7-25 was mapped to the same interval as qPL7-25 on chromosome 7 ( Figure 3, Table 3).

Fine Mapping of qPL7-25
As qPL7-25 showed the greatest potential to increase the PL, we focused on determining its underlying gene using a map-based cloning strategy. The VB431 line from the RILs with the DXWR genotype in the qPL7-25 region was selected to backcross with the recurrent GLA4 parent to construct the near-isogenic line (NIL-qPL7-25). The phenotypic characteristics were measured in the F 2 population, which included 100 individuals derived from a BC 3 F 1 line with a GLA4 genetic background ( Figure S2A). By combining the genotype and phenotype of individuals, the QTL was mapped primarily between the two insertiondeletion (InDel) markers, InDel-24091 and InDel-25971, at 1.88 Mb intervals ( Figure S2B). To delineate the genomic region of the qPL7-25, InDel markers were designed from the re-sequenced genome sequences and tested to predict the probability of polymorphism between the NIL-qPL7-25 line and the GLA4 cultivar. In the end, seven InDel markers were successfully developed (Table S1). Genotyping of all recombinant genes was performed using seven polymorphic markers, and the qPL7-25 was located between the two InDel markers, InDel-24591 and InDel-24710 on chromosome 7, with an interval of 119 kb, using a population of 1800 recessive plants derived from the crossing of NIL-qPL7-25 and GLA4 ( Figure 4A). The target region contains 14 predicted genes based on the Rice Genome Annotation Project website (http://rice.plantbiology.msu.edu/), accessed on 10 August 2020 ( Figure 4B, Table S2).

Candidate Gene Analysis of qPL7-25
Among these 14 predicted genes (Table S2), LOC_Os07g41100 encodes a conserved hypothetical protein, LOC_Os07g41110 encodes a retrotransposon protein, and LOC_Os07g41170 encodes an expressed protein. We used Genevestigator (https://genevestigator.com accessed on 10 August 2020) software to analyze the temporal and spatial

Candidate Gene Analysis of qPL7-25
Among these 14 predicted genes (Table S2), LOC_Os07g41100 encodes a conserved hypothetical protein, LOC_Os07g41110 encodes a retrotransposon protein, and LOC_Os07g41170 encodes an expressed protein. We used Genevestigator (https://genevestigator.com accessed on 10 August 2020) software to analyze the temporal and spatial expression of the remaining 11 genes. LOC_Os07g41220 has no expression data in this software, LOC_Os07g41150 showed pollen-specific expression, LOC_Os07g41180, and LOC_Os07g41190 were highly expressed in root tips, and LOC_Os07g41230 was a cloned gene related to disease resistance in rice. It is unlikely that the above genes were candidate genes. LOC_Os07g41090, LOC_Os07g41120, LOC_Os07g41140, and LOC_Os07g41200 were highly expressed in the panicle, LOC_Os07g41200 (GL7/GW7) was a previously cloned major QTL, controlling the grain length and width [45,46], and its expression showed a positive correlation with the critical period of spike development, both in terms of timing and tissue specificity ( Figure S3A). In addition, NIL-Huazhan-GL7 P13 and NIL-qPL7-25 also showed increased grain length compared to Huazhan and GLA4, respectively ( Figure S3B). Therefore, we propose that LOC_Os07g41200/GL7 could be a candidate gene for qPL7-25.
To further confirm that LOC_Os07g41200/GL7 could be a candidate gene for qPL7-25, we sequenced and analyzed the coding sequence and promoter region of GL7 in the low GL7-expressing variety, Nipponbare (NPB) and the high GL7 expressing variety, landrace Ping13 (P13), DXWR, and GLA4. There were no SNP differences resulting in amino acid changes in the coding region of DXWR and GLA4, but there were SNP and InDel differences in the 5 untranslated region (UTR) and promoter, especially in the region from -111 to -101 (transcription start site, +1), with an 11 bp deletion in the GL7 promoter of DXWR compared to GLA4, and P13 had an 8 bp deletion. In contrast, the sequence of GLA4 was similar to that of NPB ( Figure 4C, Figure S4). This 11 bp deletion of GL7 was expressed 2.4-fold higher than GLA4 in young panicles of DXWR, as determined by qRT-PCR assays ( Figure 5). These results suggest that this deletion could enhance the function of GL7 and thus affect the PL in rice. The GL7 gene from P13 was introduced into the indica rice variety "Huazhan" to construct NIL-Huazhan-GL7 P13 . The PH, PL, PBN, grain length, and grain length/width ratio of NIL-Huazhan-GL7 P13 were significantly increased compared to Huazhan (Table S3), which was consistent with the comparative trend between NIL-qPL7-25 and GLA4 phenotypes (Table S4).
We performed a haplotype network analysis of the GL7 gene promoter reg the 5′ UTR region according to the RiceVarMap V2.0 website (http:/ map.ncpgr.cn/v2/), accessed on 15 October 2020 [55,56]. The results showed tha type I, P13 is type II, DXWR is closest to type III, and GLA4 is closer to type IV (F There are 1047 rice varieties in type I, and these include mainly temperate Japo (547 haplotypes), some IndI (164 haplotypes), and AUS rice (106 haplotypes), a several other rice subspecies. Type II rice had 1010 haplotypes, including P13, Japonica rice (471 haplotypes), and mixed Japonica (214 haplotypes). Type III h haplotypes and contained 882 IndIII Indica haplotypes, 353 IndII Indica haploty 443 mixed Indica haplotypes. Type IV rice contained 673 haplotypes and was do by IndI rice with 393 haplotypes and mixed indica with 216 haplotypes. As see the promoter region and the 5′ UTR region of the GL7 gene are evolutionarily dist different SNP haplotypes were derived from each of the Japonica and Indica rice and are closely related to each rice subtype. The NPB and P13 belong to differen types, and the GL7 genes of both have significantly different regulatory roles length. DXWR and GLA4 also belong to different haplotypes, and their GL7 ge also have similar differences leading to changes in the PL, and this type of differe be widespread in indica rice with different typing. In conclusion, the LOC_Os0 gene in DXWR can be tentatively identified as the candidate gene for qPL7-25 and GL7 allele.  We performed a haplotype network analysis of the GL7 gene promoter region and the 5 UTR region according to the RiceVarMap V2.0 website (http://ricevarmap.ncpgr.cn/v2 /), accessed on 15 October 2020 [55,56]. The results showed that NPB is type I, P13 is type II, DXWR is closest to type III, and GLA4 is closer to type IV ( Figure 6). There are 1047 rice varieties in type I, and these include mainly temperate Japonica rice (547 haplotypes), some IndI (164 haplotypes), and AUS rice (106 haplotypes), as well as several other rice subspecies. Type II rice had 1010 haplotypes, including P13, tropical Japonica rice (471 haplotypes), and mixed Japonica (214 haplotypes). Type III had 1964 haplotypes and contained 882 IndIII Indica haplotypes, 353 IndII Indica haplotypes, and 443 mixed Indica haplotypes. Type IV rice contained 673 haplotypes and was dominated by IndI rice with 393 haplotypes and mixed indica with 216 haplotypes. As seen above, the promoter region and the 5 UTR region of the GL7 gene are evolutionarily distinct, and different SNP haplotypes were derived from each of the Japonica and Indica rice species and are closely related to each rice subtype. The NPB and P13 belong to different haplotypes, and the GL7 genes of both have significantly different regulatory roles in grain length. DXWR and GLA4 also belong to different haplotypes, and their GL7 genes may also have similar differences leading to changes in the PL, and this type of difference may be widespread in indica rice with different typing. In conclusion, the LOC_Os07g41200 gene in DXWR can be tentatively identified as the candidate gene for qPL7-25 and is a new GL7 allele.

Discussion
PL is an important trait related to rice yield, and breeding long and large panicle rice varieties is an effective method to achieve higher yields. Therefore, the study of the PL QTL has important and far-reaching implications for improving rice yield and quality. The QTL of PL are controlled by multiple genes and influenced by environmental factors, which are difficult to map in individual plants [57]. Although a number of QTL affecting the rice PL have been localized by previous studies, few of them have been cloned, especially those located on chromosome 7. qPL6, LP1, qPL7, and qpl9 are some finely mapped QTL directly related to PL [30,58,59]. In this study, qPL7-25 was localized on chromosome 7 between InDel-24591 and InDel-24710 in a 119 kb interval ( Figure 4A), which did not overlap with the previously reported QTL/gene position of PL. Therefore, qPL7-25 should be considered as a new PL QTL and as an enhanced allele from the DXWR parents. Previous studies have shown that the change in PL can be accompanied by a change in other panicle traits. Loss of function of SP1, SP3, and DEP3 not only shortened the PL but also reduced the number of branches and SPP [26,35,38]. DEP1, OsAPC6, and PT2 mutations result in reduced PL and grain size [21,25,60]. In our study, qPBN7-25, a QTL for PBN, was mapped to the same interval as qPL7-25 on chromosome 7, and both were potential alleles of the QTL from DXWR (Figure 3). Phenotypic analysis showed that both NIL-qPL7-25 and NIL-Huazhan-GL7 P13 had more PBN than their parents (Tables S3 and S4). Our results were consistent with previous reports, and we propose that these two QTL were single-causal multiple effects of the same QTL/gene. Compared with other common wild and cultivated rice, DXWR has more abundant rare allelic variants (including SNPs and InDels), but many QTLs coexist with unfavorable genes and are masked. Gene editing technology can be used to quickly and efficiently break the tight linkage between favorable and unfavorable genes [61], thus enabling better and faster use of the favorable genes from DXWR.
There were 14 candidate genes in the fine mapping interval of qPL7-25 ( Figure 4B, Table S2); LOC_Os07g41200 is a well-known cloned gene of GL7, which controls the grain length. Previous studies have shown that overexpression of GL7 in NPB and ZF802 (Zhefu802) backgrounds, and SLG7 in the NPB background, can increase the grain length and length/width ratio, and upregulation of GL7/GW7/SLG7 expression increases the cell length and decreases the cell width for epidermal cells of the outer and inner glumes [45][46][47]. We hypothesized that this cell division pattern is consistent with panicle elongation in the direction that leads to increased PL, and thus increased GL7 expression promotes increased PL. NIL-SLG7 showed a higher level of expression in the panicle and produced longer panicles (+14.2%) compared to NIL-slg7 [47], confirming our hypothesis. In addition, there were 14 bp and 36 bp deletions in the promoter of slg7 mutants constructed using CRISPR/Cas9, which resulted in an increased expression level of SLG7 and produced more slender grains compared with WYJ30 [62]. Sequence alignment analysis showed that there was also an 11 bp deletion in the GL7 promoter of DXWR compared with GLA4, and the expression level of GL7 was 2.4-fold higher in DXWR than in GLA4 performed by qRT-PCR analysis ( Figure S4 and Figure 5). Unfortunately, we subsequently lost the seeds of NIL-qPL7-25 due to improper storage, so we could not provide data on the expression level of the GL7 gene in the young panicles of NIL-qPL7-25. However, another experiment provided strong support for our analysis: when we transferred GL7 from P13 (a long grain, long panicle, GL7 high expressing variety) to the relatively short panicle, short-grained Huazhan variety by backcrossing using MAS, the NIL-Huazhan-GL7 P13 showed a significant increase in the PL, grain length, and grain length/width ratio, as well as a significant increase in the PBN compared to the Huazhan parent (Table S3). This showed a consistent trend with the comparison between the phenotypes of NIL-qPL7-25 and GLA4 (Table S4). Our results were consistent with the previous studies. In addition, DXWR and GLA4 belong to different haplotypes in the promoter and 5 UTR regions of the GL7 gene ( Figure 6). In view of the above, we propose that qPL7-25 is a new allele of GL7 with a greater breeding utilization value in improving rice quality rather than increasing rice yield. We are currently introducing GL7 into the promoted varieties to simultaneously improve the yield and quality of existing varieties through gene polymerization breeding.
In addition, the parents and RIL populations in this study were re-sequenced when the QTL was first located. Although the cost was high, a large amount of SNP and InDel information was obtained. Combined with the biological information analysis, it provides convenience for our subsequent future research and provides a good experimental basis for the further fine mapping or cloning of QTL genes on other chromosomes.

Conclusions
In this study, four PL QTL (qPL1-37, qPL4-26, qPL7-25, and qPL8-4) were detected using an RIL population consisting of 143 individuals derived from a cross between DXWR and GLA4. Among these, qPL7-25 showed great potential to increase the PL, which was further localized to a 119 kb region on chromosome 7 by a map-based cloning strategy using 1800 individuals derived from a cross between NIL-qPL7-25 and GLA4. Sequence alignment and qRT-PCR analysis suggested that qPL7-25 was probably a new allele of GL7, a major QTL regulating grain length. The introduction of qPL7-25 into the population varieties will help to improve the yield and quality of the existing varieties.