The rice accessions used in this study included 400 hybrid combinations (F₂) and 80 parental lines, all of which were frontier breeding materials from the College of Agronomy, Hunan Agricultural University (China) and the Yuan Longping High-tech Agriculture Co., Ltd. (China). In addition, 12 representative materials, which were collected from the Hunan Hybrid Rice Research Center (China), were used for correlation analysis between sensory evaluation and taste value determined by the rice taste analyzer. Among these 12 materials, Meixiangzhan 2, Xiangwanxian 13, Xiangwanxian 17, Xiangyaxiangzhan, and Yuewangsimiao (A1–A6), are representative indica conventional rice with high eating quality in South China. Luliangyou 996, Zhuliangyou 819, Cliangyou 343, Yliangyou 1, Jingliangyou 534, and Tianyouhuazhan (A7–A12) are six representative indica hybrid rice varieties with high yields but different eating qualities in South China. All varieties were planted in the same field on an experimental farm in Chunhua town, Changsha, China (28°2592" N, 113°2311" E). The planting season was from June 3 to October 3. Each hybrid rice line was planted in six rows per plot and six plants per row. The seeds were harvested from 20 mature plants in the middle of each plot, thoroughly mixed, and allowed to air-dry at room temperature (15–25°C) for 2 months.

The mature seeds were de-husked with a rice huller (SY88-TH, Korea) and milled with a grain polisher (Kett, Tokyo, Japan). A portion of the polished rice samples was stored in sealed bags at 4°C for later use in taste evaluations. The remaining polished rice was ground into flour in a mill (FOSS 1093 Cyclotec Sample Mill; Foss A/S, Hillerød, Denmark) and passed through a 100-mesh sieve.

A rice taste meter (model STA1B, Sasaki Company) was used to determine the taste values of all hybrid rice and parental lines. First, 30 g of milled rice was placed in a stainless steel sink containing water and soaked for 30 min. The rice was rinsed three times in cold water, and sufficient water was replenished to achieve a 1:1.4 mass ratio of rice to water. The rice was cooked for 45 min, kept warm for 10 min, and cooled to room temperature for ≤2 h. The rice taste meter included three lines of measurement: Japanese rice, China japonica rice, and China indica rice. As all varieties were indica hybrid rice or indica rice, we chose the China indica rice measuring line for the taste evaluation. The taste value mainly reflects the appearance (glossiness and whiteness) of cooked rice: <50, very poor; 50–60, poor; 60–70, mediocre; 70–80, good; 80–100, very good. The high-quality indica rice line Yuzhenxiang and indica hybrid rice line Jingliangyou 534 were used as controls. Each sample was measured three times, and the average was calculated and recorded.

The sensory evaluation was conducted according to GB/T 15682-2008 published by the Ministry of Agriculture of China. The sensory evaluation team comprised eight people of different genders and ages who could professionally identify taste. Four samples were evaluated at a time, including one reference sample (Huang Huazhan, with moderate eating quality) and three samples to be evaluated. The tasting evaluation included fragrance, appearance, palatability (viscosity, hardness, and elasticity), taste, and cold rice texture. The comprehensive scores were as follows: ≤50, very poor; 51–60, poor; 61–70, average; 71–80, relatively good; 81–90, good; and >90, very good.

Grain shape and chalkiness were measured with a ScanMaker grain appearance analyzer (WSeen SC-E; Hangzhou WSeen Detection Technology Co. Ltd., Hangzhou, China). The apparent amylose content and gel consistency were measured according to the method of (25). The protein content was measured by near-infrared spectroscopy. The viscosity was determined using an RVA (Newport Scientific PTY Ltd., Warriewood, Australia) (26). The cooked rice texture was evaluated with a texture analyzer (SATAKE RHS1A; Satake Corp., Hiroshima, Japan). The thermal properties were measured by differential scanning calorimetry (DSC200F3; Netzsch Instruments NA LLC, Burlington, MA, USA) (27). All tests were performed in triplicate.

Eighteen intact milled hybrid rice grains were immersed in 1.7% (w/v) KOH and incubated at 30°C for 24 h. Grain dispersity was then observed, and the ASV for each hybrid combination was calculated according to Mariotti et al. (28).

Purified rice starch was debranched with isoamylase (EC3.2.1.68, E-ISAMY; Megazyme, Bray, Ireland) and dissolved in dimethyl sulfoxide. The relative molecular weight distribution of the debranched starch was determined by GPC in a PLGPC 220 system (Polymer Laboratories Varian, Inc., Amherst, MA, USA). GPC data were transformed, and molecular weight distribution curves were plotted as described by Zhang (29). All tests were performed in triplicate.

Genomic DNA was extracted from the fresh leaves of all parental lines using a modified CTAB method. The Wx and ALK allelic variations were analyzed by KASP genotyping (24) and Sanger sequencing. KASP genotyping: the allele-specific primers were designed to carry the standard FAM (5-GAAGGTGACCAAGTTCATGCT-3) and HEX (5 GAAGGTCGGAGTCAACGGATT 3) tails and the targeted SNP at the 3 end. KASP 1-Wx^a/Wx^b (F1: 5′-GAAGGTGACCAAGTTCATGCTTTCATCAGGAAGAACATCTGCAAGG-3; F2: 5 GAAGGTCGGAGTCAACGGATTTTCATCAGGAAGAACATCTGCAAGT-3; R1: 5′- GCCCAACACCTTACAGAAATTAGCA-3′); KASP 2-Wx^lv/Wx^a (F3: 5′-GAAGGTGACCAAGTTCATGCTCTGGAGGAACAGAAGGGCC-3′; F4: 5 GAAGGTCGGAGTCAACGGATTCTGGAGGAACAGAAGGGCT-3′; R2: 5-GAAGAACGATCTGGACGTCCTC-3′). Assays were carried out in 384-well formats and 10-μL reactions (20–30 ng/μL DNA, 5 μL 1 × KASP master mixture, 0.14 μL KASP assay mix, and 4.86 μL water). PCR was conducted using the following protocol: hot start at 94°C for 15 min, ten touchdown cycles (94°C for 20 s, initial touchdown at 61°C, and then a decrease of −0.6°C per cycle for 60 s), and 26 additional cycles of annealing (94°C for 20 s, 55°C for 60 s). Finally, the PCR product with fluorescent labeling was scanned using a Roche Light Cycler 480 (37°C for 1 min) (24). Sanger sequencing: A 529-bp sequence containing the polymorphic sites was amplified using the primer pair 5- GGGCAGAAAGGTGTGGACATCAT-3 and 5- ACCATTGGTACTTGGCCTTGACA-3. PCR amplification was as follows: initial DNA denaturation at 95°C for 4 min; 30 cycles of denaturation at 95°C for 30 s, annealing at 58°C for 30 s, and extension at 72°C for 30 s; and final extension at 72°C for 5 min. After gel purification, the PCR products were sequenced by TsingKe Biology Technology (Beijing, China).

The experiments were carried out in triplicate, and the data were reported as mean values and standard deviations (±SD). One-way ANOVA and a Tukey's multiple-comparison test were used to determine significant differences among the mean values by using the SPSS 16.0 statistical software program.

The taste values of 400 hybrid rice and 80 parental lines were measured using the indica rice standard of the rice taste analyzer. The comprehensive taste value scores of the hybrid combinations ranged between 50–90. All hybrid combinations were categorized into five taste value levels with proportions of 13%, 17%, 22%, 27%, and 21% (Figure 1A). A series of conventional indica and hybrid rice varieties were used for correlation analysis between the sensory evaluation and taste value. The overall trend of the sensory evaluation was consistent with that of the taste value, with a correlation of 0.86 (Figure 1B). Thus, the rice taste analyzer can feasibly assess hybrid rice eating quality.

We determined the AAC, GC, GT, RVA viscosity parameters, ASV, protein content, cooked rice hardness and stickiness, and grain shape and chalkiness of the hybrid rice and parental lines. Supplementary Table 1 shows the statistical values of the physicochemical properties of all lines.

We analyzed the correlation between the physicochemical properties and taste values of hybrid rice. Supplementary Table 2 shows that the taste value of the hybrid rice was negatively correlated with the AAC, protein content, setback value (SBV), final viscosity of the RVA profile (CPV), and hardness of the cooked rice grain. Moreover, it was positively correlated with the gel consistency, breakdown value (BDV), peak viscosity of the RVA profile (PKV), and stickiness of the cooked rice grain. These findings were consistent with those reported for conventional rice in previous studies (7, 30, 31). However, there were no obvious correlations between taste value and grain shape or chalkiness.

We found that the hybrid rice grains had distinct and separate ASVs (Figure 2A). Thus, DSC was used to analyze GT variation in the hybrid rice. Figure 2B shows two-peak thermal curves for most of the hybrid rice varieties, especially when there were obvious differences in GT between the parental lines of the hybrid combination. This finding was consistent with the ASV results. ALK (SSIIa) is the major gene controlling GT, and allelic variation in ALK is responsible for natural variations in GT (32, 33). Varieties carrying ALK^c have higher GT values than those carrying ALK^b (34). Analysis of the allelic variations in ALK among the varieties indicated that ALK^c and ALK^b are common in hybrid rice. Therefore, we examined the ASVs and DSC thermograms of hybrid rice with the allelic combinations ALK^b/ALK^b, ALK^c/ALK^c, ALK^b/ALK^c, and ALK^c/ALK^b. The ASV analysis revealed no clear differences in gelatinization among hybrid rice grains with the same ALK allelic combinations (ALK^b/ALK^b and ALK^c/ALK^c); however, obvious differences in gelatinization among hybrid rice grains with different allelic combinations (ALK^b/ALK^c and ALK^c/ALK^b) were seen (Figures 2A1–A4). The DSC thermograms showed that hybrid rice with the same ALK allelic combinations (ALK^b/ALK^b and ALK^c/ALK^c) exhibited a single peak, whereas those with different allelic combinations (ALK^b/ALK^c and ALK^c/ALK^b) presented distinct two-peak thermal curves (Figures 2B1–B4). The relationship between the hybrid combinations and parental lines in terms of peak temperature (T_p) was also evaluated. Supplementary Table 3 shows that most hybrid combinations with single thermal curves had a relatively lower T_p when both parents carried ALK^b. In contrast, they had a comparatively higher T_p if both parents carried ALK^c. For most hybrids (ALK^b/ALK^c and ALK^c/ALK^b) with two-peak thermal curves, those with parents carrying the ALK^b allele showed peak 1 ≥ T_p. However, peak 2 ≥ T_p was observed for those with parents carrying the ALK^c allele. As the GTs and ASVs of the hybrid rice combinations could not be accurately measured, correlation analysis between taste value and pasting temperature of RVA were analyzed. However, we detected no obvious correlation between these parameters.

In this study, certain hybrid combinations with different taste values were obtained by crossing the same sterile line with various restorer lines. Hence, five sets of materials were selected to analyze the physicochemical relationship between the hybrid combinations and their parental lines (Figure 3). Each group included a sterile line, 22 restorer lines, and 22 hybrid combinations; the restorer lines were the same for all five groups. The taste values and physicochemical properties of all hybrids and parental lines are shown in Appendix 1.

As the sterile line for the first group, 1109S presented the lowest taste value and viscosity but the highest AAC (26%). Most hybrid combinations derived from it had comparatively lower taste values, breakdown values, and cooked rice stickiness but higher setback values, AAC, and cooked rice hardness. Sterile lines 601S and Tian S had relatively higher taste values and RVA breakdown but showed lower RVA setback values and AAC (13%–16%), except for some combinations, which were bred by crossing the same sterile line and certain restorer lines with higher setback values and AAC but very low breakdown values and taste values. All other hybrid combinations derived from the aforementioned sterile lines presented relatively high breakdown values and cooked rice stickiness but comparatively low setback values, AAC, and cooked rice hardness. The sterile line of the fifth group, 211S, exhibited intermediate taste value, AAC (18%), RVA profile breakdown, and setback values. The hybrid combinations derived from 211S had roughly equal numbers of every taste grade. The sterile 388S line had relatively lower taste values and breakdown values but comparatively higher setback values and AAC (23%). In these respects, it resembled the 1109S line; however, the taste values were relatively higher for most hybrid combinations in this group. A comparison of the physicochemical properties of the two sterile lines revealed that 1109S showed a slightly higher AAC but a lower RVA curve, which was similar to that of the indica cultivar Q11 carrying the Wx^lv allele (35). In general, the above results indicated that the sterile lines determined the overall eating quality characteristic levels of hybrid rice. Therefore, in future quality breeding programs of hybrid rice, sterile lines similar to 1109S should be removed, while those similar to Tian S and 601S should be used effectively.

Since each group included a sterile line, 22 restorer lines, and 22 hybrid combinations, we performed Pearson correlation analysis on the physicochemical properties of the hybrid combinations and corresponding restorer lines in each group (Table 1). For the same sterile line among these groups, the hybrid combinations and their restorer lines were positively correlated in terms of taste value. Except for the 1109S line, there were some correlations between the hybrid combinations and restorer lines in terms of AAC, RVA breakdown, and setback values. Supplementary Figure 1 shows the RVA profile between the parental lines and hybrid combinations with different taste values. Overall, the hybrid combinations with higher taste values had higher RVA breakdown and lower RVA setback values than the hybrid combinations with lower taste values. This was consistent with the corresponding restorer lines in the same group.

We analyzed the allelic variation of Wx and ALK among all parental lines (Supplementary Figure 2). Three Wx (Wx^a, Wx^b, and Wx^lv) and two ALK (ALK^b and ALK^c) allelic variations were found in the parental lines. Furthermore, the genetic effects of different allelic combinations of Wx or ALK on the eating quality traits of hybrid rice were analyzed. Figure 4 reveals that the hybrid combinations with Wx^b/Wx^b had the highest overall taste values, stickiness, breakdown values, and peak viscosity, while the hardness, AAC, setback values, and final viscosity were lower than those of the other allelic combinations. However, the Wx^lv/Wx^b hybrid combinations showed the lowest overall taste values, stickiness, breakdown values, and peak viscosity as well as the highest hardness, AAC, and setback values. The genetic effects of Wx^a / Wx^a were similar to those of Wx^lv/Wx^b. The overall level of eating quality traits of the Wx^b/Wx^a or Wx^a/Wx^b hybrid combinations were intermediate. As Wx^b was the main allelic variation of the parental lines, we further analyzed the genetic effects of different allelic combinations of ALK in the genetic background of Wx^b/Wx^b. Hybrid combinations with different allelic combinations of ALK showed no obvious difference in the overall taste value, hardness, stickiness, and AAC (Figure 5). However, hybrid combinations with ALK^c/ALK^c showed a comparatively higher breakdown value and final viscosity with a comparatively lower setback value, while hybrid combinations with ALK^b/ALK^b showed the opposite effects.

In total, 21 hybrid combinations with different taste values and 18 parental lines were used to analyze the relative molecular weight distributions (Supplementary Table 4). All hybrid combinations and parental lines showed typical trimodal molecular weight distributions with low, mid, and high peaks corresponding to the short-branch amylopectin chains (AP1), long-branch amylopectin chains (AP2), and amylose (AM) fractions, respectively (36). Supplementary Figure 3 shows that the hybrid combinations and parental lines with higher taste values had relatively fewer AM fractions and greater numbers of amylopectin chains in the AP1 fraction than the hybrids with lower taste values. The AM and AP1 fractions mainly accounted for the substantial differences in taste values among the hybrid rice lines. Correlation analyses between the molecular weight distributions and physicochemical properties in both hybrid combinations and parental lines revealed that AP1 fractions were positively correlated with the taste value, BDV of the RVA profile, and stickiness of the cooked rice grain. However, they were negatively correlated with the SBV of the RVA profile, hardness of the cooked rice grain, and AAC. The AM fractions were negatively correlated with the taste value, BDV of the RVA profile, and stickiness of the cooked rice grain but were positively correlated with the SBV of the RVA profile, hardness of the cooked rice grain, and AAC (Table 2). Correlation analyses also revealed that AP2 fractions were correlated with BDV, SBV of the RVA profile, and texture properties. Moreover, AP2 showed a positive correlation with the taste value in hybrid combinations (r = 0.54, p < 0.01), but no obvious correlation was observed between these in parental lines. Furthermore, AP1/AP2 were positively correlated with the taste value in parental lines but not in the hybrid combinations.

In this study, we revealed positive correlations between the hybrid combinations and their restorer lines in terms of their AM and AP1 molecular weight distributions (r = 0.61, p < 0.01; r = 0.66, p < 0.01).