**1. Introduction**

Rice grain quality, including appearance, milling, cooking and eating, as well as nutritional qualities, determines the market value, and is getting more and more concern from rice researchers, producers and consumers [1,2]. Grain protein content (GPC) is not only one key factor determining

nutritional quality, but is also closely associated with cooking and eating qualities [3,4]. Generally, the increase of GPC may consequently lead to low eating quality.

Compared to the protein content of other cereal crops such as wheat and barley, GPC in rice is relatively low, with a mean about 8.0% and a range of 4.9% to 19.3% in the *indica* subspecies and 5.9% to 16.5% in the *japonica* subspecies [5]. As a typical quantitative trait, GPC in rice is easily affected by environmental conditions, especially the level of nitrogen fertilizer in the field, which makes it very difficult to manipulate in a traditional breeding program. Therefore, illuminating the genetic basis of GPC makes a lot of sense in constructing a molecular marker-assisted selection system to improve rice grain quality [6,7].

Quantitative trait locus (QTL) analysis is the main strategy for dissecting the genetic mechanism underlying a target quantitative trait. During the past two decades, hundreds of QTLs for GPC in rice were detected throughout the entire 12 chromosomes, using different mapping populations, including the recombinant inbred line (RIL) [3,8–13], double haploid population [14–17], chromosome segment substitution line [4,18,19] and the backcross-inbred population [20]. As GPC is sensible to environmental factors, QTLs controlling the GPC are difficult to be repeatedly identified in different populations, or in the same population under different environments [19]. Till now, only two QTLs, *qPC1* and *qGPC-10*, have been map-based cloned and functionally analyzed. *qPC1* was found on the long arm of chromosome 1, which encodes a putative amino acid transporter OsAAP6 and functions as a positive regulator of GPC in rice [6]. *qGPC-10* located on chromosome 10 encodes a glutelin type-A2 precursor, and is also a positive regulator of GPC [7]. Besides, another stably inherited QTL *qPC-1* that is nonallelic to *qPC1* was validated and delimited to a 41-kb region on the long arm of chromosome 1 [19]. Owing to the detection instability of GPC QTLs, it is important to confirm the genetic effect of the QTLs detected in the primary mapping before their map-based cloning and application in the improvement of rice nutritional quality.

Genotyping and phenotyping of the mapping population are two essential components for QTL analysis. With the development of next-generation sequencing, genotyping-by-sequencing becomes a feasible technique to rapidly identify a huge number of single nucleotide polymorphisms (SNPs) throughout the whole genome. Then, a high-density linkage map can be constructed with saturated SNP markers, while most current maps are low-density, and only contain hundreds of gel-based DNA markers, such as restriction fragment length polymorphism and simple sequence repeat (SSR) markers. The detection power of minor QTLs can be improved, and the confidence interval of the QTL can be reduced in a high-resolution linkage map [21]. Among the most common genotyping-by-sequencing methods, specific length amplified fragment sequencing (SLAF-seq) is acceptable as an efficient and high-resolution technology with a relatively lower sequencing cost [22]. With the availability of rice genome draft, genotyping-by-resequencing (GBR) has been applied in linkage mapping and genome-associated analysis to map QTLs for important agronomic traits. Nevertheless, for GPC, the information of QTL identified by GBR of a bi-parent population is still limited in rice.

GPC is traditionally measured by the Kjeldahl nitrogen determination (KND), which is time-consuming and needs a large amount of chemicals such as strong acid and alkali. Therefore, this KND method is difficult for the measurement of a massive number of samples, which is usually necessary in the map-based cloning of a target QTL. Compared to the KND method, near infrared reflectance spectroscopy (NIRS) is a promising technique that is fast and easy-to-use [23]. A lot of QTLs were identified, and two major QTLs have been successfully isolated through the NIRS method. However, it keeps unknown whether there is difference in the detection power of QTLs for GPC between the two methods.

In this study, we mainly completed the following research objectives. First, we analyzed the correlation between the GPC values measured by the KND and NIRS methods, and validated the feasibility of mapping QTLs for GPC using NIRS instead of KND. Second, we identified QTLs for GPC with a high-resolution genetic map containing 18,194 SNP markers in an RIL population, which was derived from a cross between an *indica* variety Huanghuazhan (HHZ) and a *japonica* accession Jizi1560 (JZ1560). Third, we compared the GPC QTLs identified with the high- and low-density map in the same HHZ/JZ1560 RIL population. Finally, one stably inherited major QTL (*qGPC1-1*) located on the short arm of chromosome 1 was validated using three secondary populations developed from three residual heterozygotes (RH) with the heterozygous genotype at the target interval.
