*2.3. Validation and Delimitation of qGPC1-1 Using RH-derived F2 Populations*

To confirm the genetic effect and location of *qGPC1-1*, three RH individuals were selected from one F8 RIL line with the heterozygous genotype covering the target marker interval of JD1006–JD1007. Three RH-derived F2 populations, named as WB01, WB02 and WB03, were developed from the three plants with sequential heterozygous segments extending from JD1006 to JD1007, respectively. Based on the sequence differences between the parents HHZ and JZ1560 identified by 30-fold whole genome re-sequencing, additional six InDel markers were developed and used to genotype the three populations (Table S2). GPC was continuously distributed and ranged from 8.33% to 11.90%, 8.27% to 12.19% and 8.32% to 10.47% in WB01, WB02 and WB03 populations, respectively (Figure S1).

Three segmental linkage maps were constructed for WB01, WB02 and WB03, respectively (Figure 5). Combined the genotype and phenotype information, *qGPC1-1* was identified in WB01 and WB02 populations, with the JZ1560 allele always increasing GPC (Table 4; Figure 5). This QTL explained 26.00% and 27.40% of phenotypic variations with the similar additive effects of 0.36% and 0.39% in WB01 and WB02, respectively. No QTL for the GPC was detected in the WB03 population. Therefore, *qGPC1-1* should be located within the common segregating regions of WB01 and WB02, but outside the segregating region of WB03. As shown in Figure 5, *qGPC1-1* was delimited to the interval between JD1006 and JD1075 (~862 kb) with a common segregating region JD1068–JD1075 and one flanking cross-over region JD1006–JD1068.


**Table 4.** QTLs for the GPC detected in the three residual heterozygote-derived F2 populations.

**Figure 5.** Genotypic compositions of the three residual heterozygote-derived F2 populations in the segregating regions.

#### **3. Discussion**

Elucidating the genetic mechanism of GPC accumulation is very important for regulating rice grain qualities in breeding. In the present study, we characterized the genetic basis of GPC and identified a total of 14 QTLs using the high-resolution map in the HHZ/JZ1560 RIL population. Although GPC is sensitive to environmental conditions and the QTLs for GPC are difficult to be repeatedly identified in different environments, the majority of the 14 QTLs have been reported in the previous studies. On the long arm of chromosome 1, QTLs for GPC have been reported in some studies including *pro1* between RM226 and RM297 [15], *qPC-1* between R888 and R1485 [13], *qPC1* between RM472 and RM104 [6], *qPC-1* between RM1196 and RM302 [19] and *TGP1b* between RM1297 and RM1067 [4]. The *qGPC1-2* and *qGPC1-3* were located in the adjacent chromosome regions with these reported QTLs, and the *indica* variety HHZ allele decreased the GPC. Two minor QTLs, *qGPC2* and *qGPC11-1*, were also mapped at the similar locations with *pro2* between RM6 and RM112 and *pro11* between RM209 and RM229 [15]. On the chromosome 3, we detected three QTLs including *qGPC3-1*, *qGPC3-2* and *qGPC3-3*, and the enhancing alleles were all from the *japonica* rice JZ1560. The *qGPC3-1* with the largest effect in 2016 and *qGPC3-2* were repeatedly detected as *qPC-3.1* in the interval XNpb212–G1318 and *qPC-3.2* in the interval R758–XNpb15, respectively [13]. The *qGPC3-3* was located in the overlapping confidence interval with the QTLs for protein content in several previous reports [13,14,18]. These results indicated that there are multiple genetic factors controlling GPC on chromosome 3. The *qGPC4* was detected within the *qPC-4* region between RG214 and RG620 [12]. We still noted that *qGPC5* was repeatedly identified as *qRPC5* for rice protein content in the interval RG435–RG172a using a doubled haploid population [16]. Although *qGPC7* was only detected in 2016, four QTLs were located in the same or adjacent regions as reported by previous studies [3,4,8,17]. On the chromosome 8, *qGPC8* showed overlapping intervals with *cp8.1*, *qPC-8a* and *TGP8*, which have been identified using different populations in different environments [4,18,24]. The *qPC11-2* with minor effect was mapped near to *qPC11* between RM202 and RM206 [25]. No QTL for GPC has been mapped in the region of *qGPC10* on chromosome 10 before, therefore *qGPC10* might be newly detected in this study. Unlike many previous studies, we did not detect the QTL for GPC near to the *Wx* locus on chromosome 6 [4,8,12,25]. Over all, the locations of QTLs for GPC showed a significant similarity between our studies and the previous findings.

Of the 14 QTLs identified in this study, the *qGPC1-1* was a stably expressed QTL with a relatively large effect and it was repeatedly detected in both years. This QTL was also identified in the similar chromosome region in different populations and environments, implying that *qGPC1-1* plays an important role in controlling GPC [4,11,18]. Based on the primary mapping result, *qGPC1-1* was further validated and delimited in the interval JD1006–JD1075, corresponding to the 6.0–6.8 Mb region on the short arm of chromosome 1 in the Nipponbare genome [26]. The *japonica* rice JZ1560 allele contributed to the increase of GPC in the RH-derived F2 population (Table 4). Dissecting the genetic mechanism underlying GPC is important for the improvement of rice grain quality, and the main obstacle to date is the absence of key genes/QTLs regulating GPC. Primary mapping leads to a large confident interval and poor repeatability of target QTLs, which makes it difficult to find tightly linked markers for marker-assisted selection. Validation and delimitation of *qGPC1-1* contributed to the facilitation of marker-assisted selection in rice breeding for high nutritional quality. Furthermore, based on these results, fine mapping and map-based cloning of *qGPC1-1* is under way.

Mapping and isolation of QTLs need a high efficiency method to measure GPC. Cloning of QTLs controlling the natural variation of GPC is the important step toward uncovering the regulatory mechanism underlying this quantitative trait. However, map-based cloning of a QTL needs phenotype and genotype information of a massive number of samples, suggesting a rapid and easy operation method for GPC is necessary. Using the NIRS system, the GPC value can be directly measured once the brown rice is grinded into flour.

Compared with the NIRS, the KND method for GPC in rice needs further lengthy operation, which is time-consuming and laborious. More importantly, high correlation between the GPC values determined by the NIRS and KND methods was observed in the HHZ/JZ1560 RIL population across

two years. Only two minor QTLs (*qGPC10* and *qGPC11-1*) were detected using the single measurement method of GPC, the remaining twelve QTLs were identical to be identified by both methods in 2016 or 2017 (Figures 2 and 3, Table 3). Comparative analysis between the two methods suggested that NIRS could be a feasible strategy for the mapping and map-based cloning of QTLs for GPC instead of the KND method. In recent years, NIRS has been successfully employed in the isolation and characterization of two major QTLs for GPC, *qPC1* and *qGPC-10* [6,7].

Accompanied with the development of DNA sequence techniques, the sequencing cost decreases continuously and more and more high-density genetic maps have been constructed to detect QTLs for different traits through genotyping-by-sequencing [21,27,28]. In the present study, QTLs for GPC were mapped simultaneously using a high-density genetic map with 18,194 SNP markers identified by GBR and a low-density map with 208 gel-based SSR and InDel markers in the same RIL population. Although the same number of QTLs were identified using the different resolution genetic maps, 7 of 14 and 3 of 14 QTLs were repeatedly detected across two years in the high- and low-density genetic maps, respectively. Identification of more stably expressed QTLs might be attributed to the increased detection power resulted by the saturated SNP markers in the high-density genetic map.
