*2.2. QTL Analysis of Grain Protein Content in the HHZ*/*JZ1560 RIL Population*

Combining the high-density genetic map containing 18,194 SNP markers with the GPC means of each RIL, 14 QTLs were detected on the whole genome except for chromosomes 6, 9 and 12 with each QTL explaining 0.81%–18.59% of the phenotypic variations (Table 3, Figures 2 and 3). Among the QTLs, 12 were identified in 2016 and nine in 2017. The detailed description of each QTL including peak location, peak LOD value, additive effect and percentage of total phenotypic variations (*R*2) are showed in Table 3. Except for *qGPC2*, *qGPC8* and *qGPC10*, the enhancing alleles for GPC were derived from JZ1560 at the remaining 11 loci as the brown rice of JZ1560 contained significantly higher GPC (Table 1).

**Figure 2.** High-density linkage map based on genotyping-by-resequencing showing the most likely positions of QTLs for GPC measured by NIRS and KND methods in the HHZ/JZ1560 RIL population.


**Table 3.** QTLs for GPC based on genotyping-by-resequencing in the RIL population.

NIRS: near infrared reflectance spectroscopy; KND: Kjeldahl nitrogen determination; *A*: additive effect of replacing a maternal allele with a paternal allele; *R*2: proportion of the phenotypic variance explained by the QTL.

**Figure 3.** The identified QTLs for GPC measured by the NIRS and KND methods through analyzing the SNP genotypes and corresponding phenotypes of the 280 RILs. Red font indicates that QTLs were detected in both 2016 and 2017, and blue font indicates that QTLs were identified in either year 2016 or 2017.

In order to find the difference in the detection power of QTLs using different measurement methods of GPC, we further compared the QTLs for GPC determined by the NIRS and KND methods. In 2016, we identified 12 QTLs for GPC measured by NIRS and 11 QTLs for GPC measured by KND, which explained 55.07% and 48.25% of the total phenotypic variations, respectively. Eleven QTLs were commonly mapped using the two measurement methods, and one QTL (*qGPC10*) with a small genetic effect was only detected using the NIRS method (Figures 2 and 3, Table 3). Similar results were observed in 2017. Eight common QTLs were identified by both the measurement methods, and only one minor QTL (*qGPC11*) was mapped on chromosome 11 using the NIRS method. This indicated that QTLs for GPC were coincided between the two GPC measurement methods. Seven of the fourteen QTLs, *qGPC1-1*, *qGPC1-2*, *qGPC3-1*, *qGPC3-2*, *qGPC4*, *qGPC5* and *qGPC8*, were repeatedly identified in both years. The remaining seven QTLs were only detected in one year.

QTL analysis was also performed using a low-density genetic map containing 208 SSR and InDel markers and a total of 14 QTLs were detected in the same HHZ/JZ1560 RIL population (Figure 4, Table S1). Compared with the seven stably inherited QTLs identified in the high-density genetic map, only three QTLs including *qGPC1*, *qGPC3-1* and *qGPC5* were mapped at the same region and showed the similar effects for the two years using the low-density genetic map, suggesting that the high-density genetic map increased the detection power of QTLs for GPC.

**Figure 4.** Low-density linkage map containing 208 gel-based SSR and InDel markers showing the most likely positions of QTLs for GPC measured by NIRS and KND methods in the HHZ/JZ1560 RIL population.

The *qGPC1-1* was detected on the short arm of chromosome 1 across two years using the high-density genetic map and accounted for 9.14% to 11.85% of the phenotypic variations. The allele from JZ1560 at this locus increased GPC by 0.47%-0.58%. Corresponding to *qGPC1-1*, *qGPC1* was mapped at the same location with the flanking markers JD1006 and JD1007 using the low-density genetic map in both years

(Figure 4, Table S1). The *qGPC1* identified in this study contributed 11.78% to 13.33% of phenotypic variations with a relatively large additive effect ranging from 0.54% to 0.71% (Table S1).
