Introduction

Rice is one of the most important food crops, with more than half of the global population dependent on it as a staple food1. In recent years, with the increasing demand for high-quality rice, the quality-related traits of rice have gained increasing worldwide attention, and the improvement and enhancement of rice quality-related traits will help provide greater returns on investment for smallholder farmers,therefore, breeding new varieties of high-quality rice to meet market demand has become one of the main initiatives for rice breeders and molecular geneticists and a key objectives of rice research2,3.

Rice quality-related traits broadly include processing quality, appearance quality, cooking quality, and eating quality4. The processing quality of milled rice refers to the ability of the grains to withstand challenge and polishing without breaking, and it determines the final yield of edible rice5. Appearance quality directly defines the market value of rice and is closely related to grain yield and head rice production6. The cooking and eating quality of food affects the sensory perception of people during meals, and high-quality rice should be light, oily, slightly sweet, strong, soft, and sticky after steaming7.

Quality-related traits often require the measurement of many indicators, and it is difficult for breeders to balance these indicators because of the complex correlation between them8. Principal component analysis (PCA) is a method for simplifying datasets containing highly relevant and relatively complex information by extracting as much key information as possible from original information9,10. In previous rice studies, the use of PCA to simplify multiple agronomic trait indicators into principal components (PCs) as quantitative indicators has been validated11.

Quality-related traits are considered quantitative traits controlled by multiple genes. To date, many QTLs for quality-related traits have been identified by different researchers. Yun et al.12, Gao et al.13, Ponce et al.14, Qiu et al.15, and Arikit et al.16 identified QTLs for different quality-related traits using different genetic populations. However, it is difficult to use these QTLs to screen excellent lines and assess quality improvement given the many quality-related trait indicators and lines.

In this study, we used RILs constructed from a cross between the northern elite japonica cultivar SN265 and the southern indica cultivar LH99 as materials to perform PCA and QTL mapping of 18 rice quality-related traits, and we identified five lines with excellent quality. Based on the results of genetic and QTL mapping, these lines with excellent quality can be used as intermediate materials to achieve quality improvement of SN265.

Results

Phenotypic variation of the parents

There were significant differences in quality-related traits between SN265 and LH99 across both years (S1, Fig. 1). In terms of milling quality, SN265 had a higher BRR, MRR, and HRR than LH99. With regard to appearance quality, SN265 showed short round milled rice and high chalkiness, while with respect to eating quality, SN265 showed superior CA, CTS, and CCS. In the RVA spectral eigenvalues, SN265 exhibited a higher PKV, HPV, and BKV than LH99, a lower CPV and SBV than LH99, and similar PeT and PaT values as LH99.

Figure 1
figure 1

Phenotypes of quality traits between SN265 and LH99. * represents significant at P < 5% (Student’s t-test).

Full size image

Phenotypic variation in the recombinant inbred lines (RILs)

Quality-related traits in both years were slightly different, but the general trend was consistent. The recombinant inbred lines (RILs) differed considerably, with an approximately normal distribution overall and bidirectional transgressive segregation (S1, Fig. 2); these results suggest that these genetic characteristics involve quantitative traits, in line with QTL mapping requirements.

Figure 2
figure 2

Distribution of quality traits in the RILs population.

Full size image

Correlation analysis between quality-related traits

There is a general correlation between the various indicators of quality-related traits. The RILs exhibited excellent repeatability of the same quality-related traits across both years, with some correlations occurring between different quality-related traits (Fig. 3).

Figure 3
figure 3

Pearson correlation coefficient to analyze the relevance on RILs Quality-related traits. The lower triangle indicates the correlation coefficient for 2013, the upper triangle indicates the correlation coefficient for 2014, and the diagonal data indicates the correlation coefficient between two years for the same trait.

Full size image

PCA of the RIL population

In PCA, a dataset is considered to be representative when the cumulative proportion of variance of the principal components is greater than 80%. In this study, the eigenvalues of the top 8 principal components were all greater than 0.87, with proportions of variance of 22.09, 13.31, 12.46, 10.97, 6.41, 5.53, 5.19 and 4.84%. Their cumulative proportion of variance reached 80.80% (S2), meaning that the top 8 independent principal components represented 80.80% of the variation in the 18 quality-related traits.

The first principal component was characterized by higher positive loading for PaT (0.597) and higher negative loading for CA (−0.742), CCS (−0.469), PKV (−0.851), HPV (−0.825), BDV (−0.592), CPV (−0.542) and PeT (− 0.627); this PC is referred to as the peak and hot paste viscosity factor. The second principal component had a higher positive loading for CPV (0.496) and higher negative loadings for MRW (−0.432), CGP (−0.764), and CD (−0.740); this PC is referred to as chalky grain percentage and chalkiness degree; The third principal component was characterized by higher positive loadings for BRR (0.619), MRR (0.715), MRW (0.521), CPV (0.519), and SBV (0.422) and a higher negative loading for CA (−0.432); this PC is referred to as the brown and milled rice recovery factor. The fourth principal component had higher positive loadings for WLR (0.535), CGP (0.403), CD (0.438), and SBV (0.451) and higher negative loadings for HRR (−0.527), and BDV (−0.470); this PC is referred to as the width length rate factor. The fifth principal component had a higher positive loading for WLR (0.500) and higher negative loading for CTS (−0.439); this PC is referred to as the cooked taste score factor; The sixth principal component had a higher positive loading for HRR (0.538) and a higher negative loading for CCS (−0.459); this PC referred to as the head rice recovery factor. The seventh principal component had higher positive loadings for CTS (0.793) and MRW (0.307); it is referred to as the milled rice width factor. The eighth principal component had higher negative loadings for CCS (−0.426) and SBV (−0.410); it is referred to as the cooked comprehensive score factor (S2, Fig. 4).

Figure 4
figure 4

Loadings of top 8 principal components.

Full size image

QTL analysis of rice processing quality

Eighteen rice milling quality QTLs were detected in both years, on chromosomes 1, 2, 3, 5, 7, 8, 9, 11, and 12, with LOD values ranging from 2.00 to 4.73 and individual QTL contribution rates ranging from 4.79 to 15.56%. qHRR1, qHRR2, qHRR3 and qHRR9 were reproducibly detected over two years. Of these, the enhancing allele of qHRR1 was from LH99, and the enhancing alleles of qHRR2, qHRR3, and qHRR9 were from SN265 (S3, Fig. 5).

Figure 5
figure 5

Location on the genetic map of QTL loci detected for quality traits. Green indicates QTL loci detected in 2013, yellow indicates QTL loci detected in 2014 and red indicates QTL loci detected in both years.

Full size image

QTL analysis of rice appearance quality

A total of 28 QTLs for rice appearance quality were detected in both years on chromosomes 1, 2, 3, 5, 6, 7, 8, 9, 11 and 12. The LOD values ranged from 2.01 to 11.00, and the individual QTL contribution rates ranged from 4.08 to 32.59%. Six QTLs (qMRL9, qMRW5a, qMRW5b, qWLR6; qCGP5, and qCD5) were reproducibly detected across both years, including one QTL for grain length—qMRL9; two QTLs for grain width—qMRW5a and qMRW5b; one QTL for aspect ratio—qWLR6; one QTL for chalky grain percentage—qCGP5 and one QTL for chalkiness degree—qCD5. The quality-enhancing alleles of qMRL9 and qMRW5b were from SN265, and those of qMRW5a, qWLR6, qCGP5, and qCD5 were from LH99 (S3, Fig. 5).

QTL analysis for cooking and eating quality

Nine QTLs for rice cooking and eating quality were detected in both years on chromosomes 1, 3, 6, 7, 9 and 11. The LOD values ranged from 2.00 to 3.32, and the individual QTL contributions ranged from 6.62 to 26.03%. One QTL, qCS6, was detected in both environments, and the enhancing allele was from SN265 (S3, Fig. 5).

QTL analysis for RVA

A total of 23 QTLs, associated with characteristic values of rice RVA profiles were detected in both years on chromosomes 1, 3, 4, 6, 7, 8, 9, and 11. The LOD values ranged from 2.00 to 4.21, and the individual QTL contributions ranged from 3.22 to 13.78%. Three QTLs (qPKV7, qPKV8, and qBDV7) were detected in both years, including two peak paste viscosity QTLs (qPKV7 and qPKV8) and one breakdown viscosity QTL (qBDV7). The quality-enhancing alleles of qPKV7and qBDV7 were from SN265, and that of qPKV8 was from LH99 (S3, Fig. 5).

QTL analysis for PC

A total of 16 QTLs associated with rice PC scores were detected in both years on chromosomes 1, 2, 3, 4, 5, 6, 7, 9 and 11. The LOD values ranged from 2.68 to 5.18, and the individual QTL contributions ranged from 7.58 to 16. 62%. Two QTLs were detected in both years (qPC2-9 and qPC3-5), with PC2 having higher chalkiness degree and chalky grain percentage values and PC3 having higher cool paste viscosity and setback viscosity values. The quality-enhancing alleles of qPC2-9 and qPC3-5 were from LH99 (S3, Fig. 5).

Multiple QTLs are distributed in clusters

A QTL cluster is defined as an interval on a chromosome containing several QTLs, usually with cumulative effects, or one pleiotropic QTL with major effects. In this study, 59 QTLs were distributed in clusters on chromosomes 1, 2, 3, 5, 6, 7, 8, 9, and 11, accounting for 62.76% of the total number of QTLs. Six QTL clusters containing OTLs detected by PCA across both years were identified for the quality-related traits (Fig. 5). QTL cluster 1 was between the molecular markers STS6 and STS7 on chromosome 3 and covered a genetic distance of approximately 7.1 cM; this cluster contained qMRR3b, qHRR3, qMRL3a, qWLR3 and qPC6-3a. QTL-cluster 2 was between molecular markers RM413 and RM18062 on chromosome 5 and covered a genetic distance of approximately 12.8 cM; this cluster contained qBRR5, qMRW5a, qWLR5, qCGP5, qCD5, qPC2-5 and qPC3-5. QTL cluster 3 was between molecular markers RM6835 and RM455 on chromosome 7 and covered a genetic distance of approximately 4.9 cM; this cluster contained qMRR7b, qCCS7, qCA7, qPKV7a, qHPV7, qBDV7, qPeT7 and qPC1-7. QTL-cluster 4 was between molecular markers RM407 and RM6356 on chromosome 8 and covered a genetic distance of approximately 3.4 cM; this cluster contained qMRR8, qPKV8, qHPV8 and qPC4-3. QTL cluster 5 was between molecular markers RM566 and RM3700 on chromosome 9 and covered a genetic distance of approximately 4.6 cM; this cluster contained qMRL9, qMRW9, qWLR9a, qCCS9, qPC3-9 and qPC6-9. QTL cluster 6 was between molecular markers RM7424 and RM257 on chromosome 8 and covered a genetic distance of approximately 5.4 cM; this cluster contained qHRR9, qHPV9, qCPV9, qPeT9 and qPC2-9.

Selection and genetic analysis of excellent lines

The RILs were grouped together according to their weights based on the scores of the 8 principal components, with scores ranging from −1.67 to 2.07. The lines were classified according to their total scores and were divided into 5 types, which were ranked in order from excellent to poor based on the quality-related traits (Fig. 6). Five lines were in the top group in both years (Fig. 7a). The PC1 and PC2 values of these 5 lines were significantly higher than those of SN265 in both years, among which PC2 showed a larger increase, while the values for the other 6 principal components were not significantly different from those of SN265 (Fig. 7c). PC1 mainly regulates the peak paste viscosity and hot paste viscosity factor, and PC2 mainly regulates the chalkiness degree and chalky grain percentage. The Five excellent rice lines carrying six QTL clusters were analyzed (Fig. 7b). Genetic regions including QTL cluster 1 were found in 5 lines, while those for QTL cluster 2 were found in line-133 and QTL cluster 3 in line-25; all regions were from SN265, indicating that the QTLs contributing to the quality-related traits in these lines were all from SN265. Therefore, these QTL clusters in these specific lines could not be used for the genetic improvement of SN265. The genetic regions including QTL cluster 4 and QTL cluster 5 were found in 5 lines and those for cluster 3 were found in line-92; all regions were from LH99, indicating that the QTLs contributing to the quality-related traits in these lines were all from LH99. Therefore, these QTL clusters in the abovementioned lines could be used for the genetic improvement of quality in SN265. Notably, the QTL for PC2, qPC2-9, which was detected in both years, was located in QTL cluster 6. The chalkiness degree and chalky grain percentage of the five lines were significantly lower than those of SN265 and close to those of LH99 (S4). The region containing QTL cluster 6 was derived from LH99, which supported the accurate identification of qPC2-9 to some extent. Therefore, qPC2-9 has significant potential for improving the quality of SN265.

Figure 6
figure 6

Distribution of total score in the RILs population. TypeI, TypeII, TypeIII, TypeIV and TypeV correspond to number of the lines with total score in the interval n > 1.00, 1.00 ≥ n > 0.50, 0.50 ≥ n > 0.00, 0.00 ≥ n > -0.50 and n ≤ -0.50.

Full size image
Figure 7
figure 7

(a) Photographs of head rice from SN265 and five excellent quality lines. (b) Genetic performance of five excellent quality lines on six QTL clusters, where 0 represents inheritance from SN265 and 2 represents inheritance from LH99. (c) Comparison of the scores of SN265 and the five excellent quality lines on the eight principal components.

Full size image

Discussion

QTL clusters contribute to the mining of genes for quality-related traits

Quality improvement is one of the most important ways to adapt rice to the consumer market17. With the advancement of science and technology, mining new QTLs for quality-related traits, cloning relevant genes and applying quality improvement to actual production in combination with molecular marker-assisted breeding and transgenic and gene editing technologies is a feasible way to improve the quality of rice2. The genetic basis for quality-related traits in rice is complex, and QTLs acting on multiple traits in the same chromosomal region are common18. Ponce et al.14 used a multiparent advanced generation cross-population to localize QTLs for rice cooking and eating quality and identified 17 QTLs, of which 9 were in clusters that contained GSSI genes and 5 were in clusters that contained SSIIa genes. These QTLs and the markers highly associated with their underlying traits will be useful for breeding indica rice to improve cooking and eating quality. Yao et al.19 measured RVA and mapped 93 QTLs in four environments based on 151 RILs. These QTLs formed five clusters on chromosomes 1, 3, 6, 7, and 11, and the study found that AC and RVA traits were not influenced by indica-japonica subspecies differentiation, indicating that the RVA curve was mainly influenced by the Wx gene.

In this study, we identified 94 QTLs for quality-related traits, of which 59 QTLs were distributed in six different clusters. Compared with previous studies, the QTLs we identified quality-related traits partially corresponded to those found by Yao et al.19, Yan et al.20, and Nelson et al.21. Among them, the gene GW5, which significantly affects grain width and grain weight in rice, has been cloned within the interval of QTL cluster 222. Within the interval of QTL cluster 3 major quantitative trait locus, GLW7, which encodes the plant-specific transcription factor OsSPL13, was reported to positively regulate cell size in the grain hull, resulting in enhanced rice grain length and yield23. QTL cluster 4 contains the cloned gene OsSSIIIa, which affects the structure of amylopectin, amylase content, and physicochemical properties of starch granules24. GS9 is a cloned gene that has been found to regulate grain length and reduce grain chalkiness in rice; this gene is located within the interval of QTL cluster 625. In the intervals of QTL-cluster 1 and QTL-cluster 5, we found no genes for rice quality-related traits that have been cloned at present,however, these two intervals should be explored further.

Selection and identification of rice lines with high quality

Quality-related traits are complex traits that are influenced by a combination of multiple factors and genes and are susceptible to external environmental conditions, making it difficult to fully and truly evaluate the quality of rice with a single indicator26,27. PCA is a statistical analysis method that converts many indicators into a few composite indicators, making complex problems simple and their analysis intuitive through dimensionality reduction9,10. In this study, 18 rice quality indicators were simplified into 8 mutually independent principal components using PCA that better reflected the basic characteristics of rice quality indicators; the cumulative contribution of all PCs was 80.80%.

Stable RILs are valuable breeding resources for studying rice quality. On the one hand, the genetic background of RILs is clear and can be used for gene mapping, and on the other hand, the quality of RILs shows a normal distribution with bidirectional transgressive segregation, thus screening lines with high affinity and excellent quality could prevent the use of intermediate materials when aggregating excellent traits28. In this study, we calculated a total score by weighting the 151 RILs base on 8 principal components and classified them into 5 types from superior to inferior, 5 lines with excellent quality were selected. Compared with SN265, these 5 lines showed better performance regarding chalky grain percentage and chalkiness degree factor. In future research, we can use the genetic map and the traits identified in excellent lines to achieve quality improvement in SN265 backcrosses.

Materials and methods

Plant materials and cultivation

We established a population of 151 lines of isolated RILs (F8) by crossing the elite japonica cultivar SN265 with the indica cultivar LH99. The rice plants were grown in 2013 and 2014 in experimental fields at Shenyang Agricultural University with conventional water and fertilizer management. After maturity, the rice was threshed by plant row and the threshed seeds were kept in a cool and ventilated place for 3 months to measure their quality.

Measurement of rice quality

Traits measured in this experiment include milling, appearance, cooking and eating quality and RVA spectral eigenvalues. The experiment was conducted with three biological repetitions and t tests were performed for statistical analysis.

The milling quality includes brown rice recovery (BRR), milled rice recovery (MRR) and head rice recovery (HRR). The calculations are as follows:

BRR (%) = Weight of brown rice/Weight of rough rice × 100.

MRR (%) = Weight of total milled rice/Weight of rough rice × 100.

HRR (%) = Weight of head rice/Weight of rough rice × 100.

Appearance quality includes milled rice length (MRL), milled rice width (MRW), width length rate (WLR), chalky grain percentage (CGP) and chalkiness degree (CD). These traits were measured by an ES-1000 rice quality analyser29.

Cooking and eating quality includes cooked appearance (CA), cooked taste score (CTS), and cooked comprehensive score (CCS). The milled rice grains were washed and placed in distilled water in a 1:1.2 proportion for 1 h, followed by steaming them for 30 min, maintenance at that temperature for 10 min, and even stirring in the fume cabinet for 20 min. The rice grains were measured with a rice taste meter (SATAKE-STA1B) after 2 h at room temperature.

The RVA spectral eigenvalues, including peak paste viscosity (PKV), hot paste viscosity (HPV), breakdown viscosity (BDV), cool paste viscosity (CPV), setback viscosity (SBV), peak time (PeT) and pasting temperature (PaT), were measured by the American Association of Cereal Chemists Standard Method (AACC 61‐02) (1995) as described by Bao et al.30.

Principal component analysis (PCA)

Principal component analysis of the 18 quality-related traits in rice was performed using GraphPad Prism 9 software, with standardized method selection, and principal components were selected based on eigenvalues. The eigenvalues (E), proportions of variance (PV), cumulative proportions of variance (CPV), loadings and PC scores were obtained. Among them, PC scores were used for QTL mapping. The total score was calculated according to the following formula. In the formula, \({W}_{i}\) is the weight of the ith principal component, \({P}_{i}\) is the proportion of variance of the ith principal component, \({T}_{i}\) is the total score of the ith lines and Si is the PC scores of the ith lines.

$${W}_{i}={P}_{i}/{\sum }_{i=n}^{n}{P}_{i} (i=\mathrm{1,2},3,\dots ,n)$$
$${T}_{i}={\sum }_{i=n}^{n}\left({W}_{i}\times {S}_{i}\right) (i=\mathrm{1,2},3,\dots ,n)$$

Genetic and QTL mapping analysis

We used 144 polymorphic simple sequence repeat (SSR) and insertion/deletion (Indel) markers31,32 distributed among 12 chromosomes to establish a genetic map. The QTL analysis was based on inclusive composite interval mapping implemented by QTL IciMapping 4.033. The QTLs were named according to the guidelines described by McCouch34.

Consent for publication

All authors and associated institutes have consented to the publication of this work.

Compliance statements

The experimental studies conducted and the plant materials used in this paper comply with relevant institutional, national, and international guidelines and legislation.