Dissection of Closely Linked QTLs Controlling Grain Size in Rice

Two small-effect QTLs, qTGW7.2a for grain width consequently affect grain weight was fine mapped to a 21.10-kb interval, and qTGW7.2b was limited within a 52.71-kb interval for grain length and width with opposite allelic directions, exhibiting little influence on grain weight. Abstract Grain size is a key constituent of grain weight and appearance in rice. However, insufficient attention has been paid to the small-effect QTLs on grain size. In the present study, residual heterozygous populations were developed for mapping two genetically linked small-effect QTLs for grain size. After genotyping and phenotyping of five successive generations, qGS7.1 was dissected into three QTLs and two were selected for further analysis. qTGW7.2a was finally mapped into a 21.10-kb interval containing four annotated candidate genes. Transcript levels assay showed that the expression of candidates LOC_Os07g39490 and LOC_Os07g39500 were significantly reduced in the NIL- qTGW7.2a BG1 . Cytological observation indicated that qTGW7.2a regulated grain width through controlling cell expansion. Use the same strategy, qTGW7.2b was fine mapped into a 52.71-kb interval, showing a significant effect on grain length and width with opposite allelic directions but little on grain weight. Our study provides new genetic resources for yield improvement and fine-tunes of grain size in rice.


Introduction
Rice (Oryza sativa L.) is one of the most important staple crops which feeding half of the world's population. Therefore, grain yield became a prime target for breeders.
Grain yield is characterized by three components: panicle number, filled grain number per panicle, and grain weight. Grain weight is mainly determined by grain size, which simultaneously affects appearance (Zuo and Li 2014). Thus, grain size is a primary target for yield improvement.
Grain length and grain width determine grain size, and both are complex traits controlled by quantitative trait locus (QTL). To date, 20 grain size related QTLs with large-effect have been cloned and characterized. Several signals and regulatory pathways controlling grain size have been identified in rice, such as the G-protein signaling pathway, the ubiquitin-proteasome pathway, the mitogen-activated protein kinase (MAPK) signaling pathway, the phytohormone signaling, and transcriptional regulators (Fan and Li 2019;Li and Li 2016). GS3 and DEP1 encode G-protein γ-subunits and regulate grain size and weight (Fan et al. 2006;Huang et al. 2009).
Small-effect QTLs also play important roles in regulating grain size and are widely utilized in commercial rice varieties (Kinoshita et al. 2017). Many QTLs with small-effect are responsible for quantitative genetic variation, these QTLs are often unexpected based on prior knowledge of the trait or correspond to computationally predicted genes (Mackay et al. 2009). Therefore, it is beneficial to validate these small-effect QTLs for breeding. In recent years, more than 400 small-effect QTLs for grain size and weight were reported (Huang et al. 2013). However, only a few were fine-mapped or cloned. DTH2 encodes a CONSTANS-like protein that promotes heading by inducing the florigen genes Hd3a and RFLT1 (Wu et al. 2013). qTGW1.2b regulates grain weight through encodes a VQ-motif protein OsVQ4 (Chan et al. 2020). A naturally varying QTL, qTGW12a, which encodes the multidrug and toxic compound extrusion (MATE) transporter, regulates grain weight in rice (Du et al. 2021).
The residual heterozygous method (Du et al. 2008) was mainly used for QTL mapping in this study. Residual heterozygote, which shows heterozygosity of the target region and high homozygosity in the background. The progeny population obtained by selfing is equal to the natural near-isogenic line (NIL)-F2 population, which applies to validating, resolving, and fine mapping of QTL. To date, a series of small-effect QTLs have been fine mapped using this method Wang et al. 2019b;Zhang et al. 2020;Zhu et al. 2019).
In a previous study, a grain size QTL, qGS7.1 has been identified on chromosome 7 . Then, qGS7.1 was dissected into two QTLs, named qTGW7.1 and qTGW7.2. In the present study, we aimed to fine map the qTGW7.2 using a set of backcross recombinant inbred lines between BG1 (Big Grain 1) and XLJ (Xiaolijing). Two independent QTLs named qTGW7.2a and qTGW7.2b regulate grain size were genetically dissected in the target region. Finally, qTGW7.2a was located into a 21.10-kb region controlling grain width and weight, while qTGW7.2b was mapped to a 52.71-kb interval affecting grain length and width, not the grain weight.

Plant materials
Five runs and a total of 23 residual heterozygous populations were used to map the target QTL in this study. The populations were derived from two BC4F6 individuals from the cross of XLJ/////XLJ////XLJ///XLJ//XLJ/BG1 (Fig.S1).
In the first run, two single plants with heterozygous regions of qGS7.1 were selected and developed two BC4F7 populations consisting of 137 plants (R7) and 142 plants (R8) used for QTL validation and mapping. New polymorphic markers were designed and used to test genotypes of these populations.
In the second run, six resultant BC4F8 populations, R9 to R14, consisting of 189, 193, 198, 151, 116, and 213 plants respectively were developed from six residual heterozygous BC4F7 single plants with updated target regions. Then, the BC4F9 population contains 3989 individuals derived from the R9 population was constructed and used for selecting recombinants.
In the third run for QTL validation and mapping, eleven single plants were selected from the BC4F9 generation to develop eleven BC4F10 populations, R15 to R25, totally consisting of 794 plants.
In the fourth run, three NIL populations with homozygous in the segregating region, namely N1 to N3, were developed to validate the QTL. Two single plants without qTGW7.2b target region were selected and selfed to develop populations named R26 and R27, made up of 209 and 223 plants, respectively. Meanwhile, a BC4F11 population including 6128 individuals derived from the R23 population was constructed and used for further mapping.
In the fifth run, two single plants were selected from BC4F11 plants in the XP7-12-XP7-23 interval to develop progeny populations consisting of 233 (R28) and 98 plants (R29) for validation and fine-mapping of qTGW7.2a.

Field experiments and traits measurement
Plants were grown at the field stations of the China National Rice Research Institute in Lingshui, Hainan province, and Fuyang, Zhejiang province. After harvesting, 300 dry seeds were randomly selected for measuring thousand-grain weight (TGW, g), grain length (GL, mm), grain width (GW, mm), and the ratio of grain length to width (RLW) using an automatic seed counting and analyzing instrument (Model SC-G, Wanshen Ltd., Hangzhou, China).

DNA extraction and molecular markers development
Total DNA was extracted from fresh leaf samples by the CTAB method (Murray and Thompson 1980). The PCR products were visualized on 8% non-denaturing polyacrylamide gels by silver staining. A total of 31 polymorphic DNA markers were used (Table S1).

RNA extraction and qRT-PCR
Total RNA was extracted from rice panicles using RNAprep pure Plant Kit (TIANGEN). Quantitative reverse-transcriptase polymerase chain reaction (qRT-PCR) was performed using SYBR Premix Ex Taq Ⅱ (TAKARA). Data analysis used the 2 -ΔΔCt method and the UBQ10 was used as the internal reference to normalize the gene expression (Livak and Schmittgen 2001). The qRT-PCR primers used in this study are listed in Table S1.

Cytological observation
During the heading stage, young spikelet hulls of NIL-qTGW7.2a BG1 and NIL-qTGW7.2a XLJ were fixed in 2.5% glutaraldehyde for 12 hours at 4℃ and then dehydrated in serial graded ethanol (30%, 50%, 70%, 80%, 90%, 95%, and 100%), and last preserved in 100% ethanol. The samples were dried in a Hitachi HCP-2 critical point drier, and cell length and width of the inner glumes were observed by scanning electron microscopy (Hitachi SU-8010). ImageJ software was used to measure cell numbers and cell size.

Data analysis
Three genotypes could obtain after genotyping this population. Two homozygous genotype plants which carried alleles from XLJ and BG1 were used to detect the phenotypic differences by student's t-test. We deduce there was a QTL when p<0.05.
Subsequently, the heterozygous individual harboring target QTL was used for developing a new residual heterozygous population.
All of the analysis data, including the additive effect (A) and the proportion of phenotypic variance explained by the QTL (R 2 ) were obtained from the Windows QTL Cartographer Version 2.5 software to estimate the genetic effects.

Validation and mapping of qGS7.1
We have identified a grain size QTL, qGS7.1, in the X7-9-RM351 interval on chromosome 7 (Fig.1a). To narrow down the target region, 12 polymorphic markers were designed based on the sequence differences between BG1 and XLJ. RM21758 became the new boundary when all plants were homozygous for it. R7 and R8 populations were derived from two segregated single plants selected from the R6 population ( Fig.1b) to validate qGS7.1 and exclude the non-target interval. Both were showed significant enhancement of GW, GL, and TGW from XLJ alleles. In the R7 population, the additive effects were -0.445g for TGW, -0.082mm for GL, and -0.011mm for GW, explaining 21.95%, 30.35%, and 8.31% of the phenotypic variance, respectively. In the R8 population, the additive effects were -0.439g for TGW, -0.075mm for GL, and -0.010mm for GW, having R 2 of 28.01%, 25.40%, and 8.91%, respectively (Table 1). The effects detected in the two populations were comparable indicated that qGS7.1 was located in the region between RM21758 and Chr07MM3011.

Dissection of qGS7.1 into three QTLs controlling grain size
To validate and narrow down the update region, six progeny populations with sequential segregating regions jointly covering the entire QTL region (Fig. 1c) were developed. In the R9, R10, and R11 populations, significant enhancements were discovered in XLJ alleles for TGW, GW, and GL. The additive effects for TGW were -0.477g, -0.670g and -0.531g, for GL were -0.051mm, -0.114mm and -0.063mm, for GW were -0.014mm, -0.029mm and -0.021mm, respectively. The additive effects for TGW in R10 and R11 populations were higher than that in the R9 population. Meanwhile, significant genotypic variances were detected for TGW and GW in the R13 and R14 populations that enhancing alleles derived from XLJ. The additive effects for TGW were -0.265g and -0.180g, for GW were -0.031mm and -0.018mm, respectively. There were no significant differences in the R12 population ( Table 2).
qTGW7.2 was selected for further analysis. Eleven populations (R15-R25) were developed from 11 heterozygous individuals in BC4F9 populations (Fig. 1d). In the R15 population, significant genotypic effects were detected in TGW and GW with the positive allele from XLJ. The additive effects for TGW and GW in the R15 population were -0.604g and -0.034mm, with the R 2 values of 42.48% and 49.12%. There were no significant differences in R16, R17, R18, R19, and R20 populations. Similarly, significant genotypic variances were detected for GL and RLW in R21 and R22 populations. The additive effects for GL were -0.050mm and -0.053mm, explaining 8.32% and 20.81% of the genotypic variance in both populations. For RLW, the additive effects were -0.030 and -0.018, having R 2 of 19.61% and 6.62%, respectively; in R23, R24, and R25 populations, significant genotypic effects were showed for TGW, GL, and GW. The additive effects for TGW were -0.558g, -0.505g and -0.504g, for GL were -0.087mm, -0.065mm and -0.095mm, for GW were -0.011mm, -0.020mm and -0.020mm, respectively (Table 3).
To sum up, qTGW7.2 was dissected into two separate QTLs (Fig. 1d). The first QTL, qTGW7.2a had considerable effects on TGW and GW within a 53.96-kb region spanning XP7-12 to XP7-16. The second QTL, qTGW7.2b, was located between Chr07MM2985 and RM21891, a 52.71-kb interval and affected GL and RLW but had little effect on TGW.
Three NIL populations (N1-N3) derived from the R15, R16, and R21 populations were developed to validate the function of qTGW7.2a and qTGW7.2b (Table S2); meanwhile, two progeny populations, R26 and R27, derived from two recombinants containing qTGW7.2a only were used in this study (Fig. S2). In the N1 population, significant genotypic effects were showed for TGW and GW, the additive effects for TGW and GW were -0.421g and -0.021mm, explaining 31.59% and 40.05% of phenotypic variance, respectively, and which was coincident with R26 and R27 populations. However, in the N3 population, highly significant genotypic effects were detected for GL, GW, and RLW, the additive effects were -0.070 mm, 0.020 mm, and -0.051, explaining 25.78%, 34.30%, and 54.26% of phenotypic variance, respectively, and which was not coincident with the results of the R21 population. There was no significant difference in the N2 population (Table S2).
qTGW7.2a mainly controls TGW through regulating GW, with enhancing alleles derived from XLJ. qTGW7.2b simultaneously affected GL and GW in opposite ways with no significant effect on TGW (Fig. S3). The former was selected for further analysis for the stable function and considerable effect.

Fine-mapping qTGW7.2a into a 21.10-kb region
For further mapping of qTGW7.2a, we constructed a BC4F11 population consisting of 6128 individuals. Two recombinants in the RM21871-XP7-23 interval were utilized to develop two progeny populations, R28 and R29. Highly significant phenotypic effects were detected in TGW and GW in the R28 population. The additive effects were -0.213g and -0.013mm, having R 2 of 11.29% and 9.70%, respectively. There were no significant differences in the R29 population (Table 4). According to the mapping results of the BC4F12 population, we mapped qTGW7.2a to the 21.10-kb interval between Chr07MM2954 and XP7-16 (Fig. 1e). qTGW7.2a increased TGW and GW with the allele from XLJ as compared grain size and weight between NIL-qTGW7.2a XLJ and NIL-qTGW7.2a BG1 (Fig. 2), was same as the R28 population.
Grain size is restricted by the size of the spikelet hull in rice, which is determined by both cell proliferation and expansion. Therefore, we compared the cell number and cell size of the outer glume epidermal cells between NIL-qTGW7.2a XLJ and NIL-qTGW7.2a BG1 (Fig. 3a). There was no significant difference in cell number or cell length between NIL-qTGW7.2a XLJ and NIL-qTGW7.2a BG1 (Fig. 3b, c). However, the cell width of NIL-qTGW7.2a XLJ was greater than NIL-qTGW7.2a BG1 (Fig.3d). These findings suggest that the grain size increase in NIL-qTGW7.2a XLJ is predominantly due to cell width expansion.

Candidate genes of qTGW7.2a
There are four ORFs located in the region spanning qTGW7.2a. LOC_Os07g39470 encodes a rice GRAS family protein, CIGR2, which suppresses cell death in rice inoculated with rice blast via activation of a Heat Shock Transcription Factor, OsHsf23 (Tanabe et al. 2016). LOC_Os07g39480 encodes WRKY78, a transcriptional factor that is involved in regulating plant height and seed size . LOC_Os07g39490 and LOC_Os07g39500 are unknown functional proteins (Table.S3).
Sequences of the coding domain sequence (CDS) in four genes between the NIL-qTGW7.2a XLJ and NIL-qTGW7.2a BG1 were compared (Fig. 4a). Two synonymous SNPs were detected in LOC_Os07g39470, indicating that there were no differences between the two alleles. For LOC_Os07g39480, there were four polymorphism sites, three of which were synonymous and one 3-bp deletion in the XLJ allele, resulting in a serine deletion. For LOC_Os07g39490, three SNPs include one synonymous and two non-synonymous resulting in two amino acids substituted; especially, a 2-bp deletion in NIL-qTGW7.2a BG1 resulting in NIL-qTGW7.2a BG1 producing an alternatively spliced protein, in which the terminal 62 residues were truncated. Finally, there were 18 SNP variations in LOC_Os07g39500 between two NILs, including thirteen non-synonymous mutations and a premature stop codon at T784C in the BG1 allele. These results suggest that either LOC_Os07g39490 or LOC_Os07g39500 is the candidate gene for qTGW7.2a.
Subsequently, the expression levels of four candidates in panicles of NIL-qTGW7.2a XLJ and NIL-qTGW7.2a BG1 were analyzed (Fig. 4b). The expression levels of LOC_Os07g39490 and LOC_Os07g39500 were significantly higher in NIL-qTGW7.2a XLJ than that in NIL-qTGW7.2a BG1 , while there were no significant differences in LOC_Os07g39470 and LOC_Os07g39480. These results repeatedly indicated that the candidate gene of qTGW7.2a was more likely to be either LOC_Os07g39490 or LOC_Os07g39500.

Discussion
Remarkable progress has been achieved by the discovery of large-effect QTLs affecting yield and quality in recent years, however, rarely small-effect QTLs have been cloned in rice (Chan et al. 2020;Du et al. 2021). In our study, two small-effect QTLs regulating grain size were identified and fine mapped. qTGW7.2a was limited between Chr07MM2954 and XP7-16 with a 21.10-kb interval, affecting grain width and weight. qTGW7.2b inversely affects the ratio of grain length to width was mapped into the 52.71-kb region between Chr07MM2985 and RM21891.
All the populations used in this study were derived from a single plant with the same background and were cultivated in Fuyang and Lingshui followed the chronological order. qTGW7.2a could be detected in both environments, but the effects on TGW and GW were not stable. The additive effects on TGW and GW increased by XLJ allele were in the range of -0.604 to -0.213g, and -0.034 to -0.013mm, respectively (Table 3, 4 and Table S2). Especially for qTGW7.2b, in R21 and R22 populations, qTGW7.2b regulates grain length, has little influence on grain width and grain weight, but in the N3 population, qTGW7.2b was detected affecting grain length and grain width with opposite allelic directions and had little effect on grain weight (Table 3 and Table S2). These results suggested that small-effect QTLs could be steadily detected using the residual heterozygous method, but the effects of QTL could be affected by environmental interaction.
In the present study, four annotated genes were found in the 21.1-kb interval covering qTGW7.2a. Firstly, LOC_Os07g39470 encodes CIGR2 belonging to the rice GRAS family, and members of this family encode transcriptional regulators with functions in a wide range of signaling mechanisms such as growth and development, hormone signaling, and plant defense (Tanabe et al. 2016). However, there were only two synonymous SNPs between the CIGR2 alleles. Secondly, LOC_Os07g39480 encodes a transcriptional factor, WRKY78, which was involved in regulating plant height and seed size. Knocking-down of WRKY78 led to a semi-dwarf and small seed phenotype by reducing cell length . However, except for three SNPs showing synonymous mutation, there was just one serine deletion in the CDS of NIL-qTGW7.2a XLJ . The expression level of WRKY78 was comparable between the two NILs. These results suggest that CIGR2 and WRKY78 may not be candidate genes for qTGW7.2a. Previous studies showed that introducing a premature stop codon and preventing transcription of a mature protein influencing grain size, such as GW2, GS3, qLGY3/GW3p6, WTG1, OsMAPK6, TGW6, and GL6 (Fan et al. 2006;Huang et al. 2017;Ishimaru et al. 2013;Liu et al. 2015a;Liu et al. 2018;Song et al. 2007;Wang et al. 2019a;Wang et al. 2019c). In our study, LOC_Os07g39490 and LOC_Os07g39500 encode hypothetical proteins. In its coding region, a non-synonymous mutation existed as a premature stop codon and preventing transcription of a mature protein in NIL-qTGW7.2a BG1 . Therefore, more studies in gene editing such as CRISPR/Cas9-targeted mutagenesis and gene overexpression need to be done to confirm the gene for qTGW7.2a.
Among QTLs with large effect, GW2, GS5, GW5/GSE5, and GW6, those regulate grain weight through controlling grain width Li et al. 2011;Shi et al. 2020;Song et al. 2007;Xu et al. 2015). In our study, qTGW7.2a increased grain width and weight. These suggest that qTGW7.2a could be used for yield improvement. For large-effect QTLs such as GL7/GW7, GW8, and GS9, those have similar effects with qTGW7.2b on grain length and width that regulate grain size (Wang et al. 2012;Wang et al. 2015a;Wang et al. 2015b;Zhao et al. 2018). qTGW7.2b in the XLJ allele increased grain length but decreased grain width, which has little effect on grain weight indicated that qTGW7.2b could be used to fine-tune grain size.

Conclusion
Two small-effect QTLs for grain size and grain shape, qTGW7.2a and qTGW7.2b, were fine-mapped in this study. qTGW7.2a was limited to a 21.10-kb region containing four genes. This QTL regulates grain width and weight, which has potential for yield improvement. qTGW7.2b which inversely regulates grain length and width was within a 52.71-kb interval. This QTL has potential for fine-tuning grain shape and grain appearance. These results provide a basis for QTL cloning and offer new resources for yield and quality improvement.  Comparison of thousand grain weight. c Comparison of grain length. d Comparison of grain width. e Comparison of the ratio of grain length to width. Data are given as mean ± SD. Student's t-test was used to generate P value.

Figure 3
Scanning electron microscopic observation and analysis of the glume. a Scanning electron micrograph of the outer glume epidermal cells between NIL-qTGW7.2a XLJ and NIL-qTGW7.2aBG1. Bar, 100 μm. b Cell number of outer epidermal cells. c Cell length of outer epidermal cells. d Cell width of outer epidermal cells. Data are given as mean ± SD. Student's t-test was used to generate P value; *, P < 0.05; **, P < 0.01.

Figure 4
The coding domain sequence alignment and transcript levels of annotated genes between NIL-qTGW7.2aXLJ and NIL-qTGW7.2aBG1 . a The red words represent variants, the red frame represent premature stop. +, the variant sites in the coding domain sequence. Bar,200bp. b The experiment was performed using panicles of 1 ≤ P < 3 cm (P3) and 5 ≤ P < 8 cm (P8) collected from NIL-qTGW7.2aXLJ and NIL-qTGW7.2aBG1. Data are given as mean ± SD. Student's t-test was used to generate P value; *, P < 0.05; **, P < 0.01.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Tables.pdf

FigureS.pdf
TableS.pdf