Dissection of three quantitative trait loci for grain size on the long arm of chromosome 10 in rice (Oryza sativa L.)

Rice materials

A total of 12 populations of rice (Oryza sativa L.) were used (Table 1). As illustrated in Fig. 1 and described below, these populations were derived from two residual heterozygous plants of the indica rice cross TQ/IRBB52, namely ZC9 and Ti52-1. ZC9 was an F₉ plant carrying four heterozygous segments, including RM6704–RM3773 on chromosome 10 and three others on chromosomes 1, 9, and 11 (Fig. 2). Ti52-1 was an F₇ plant carrying one heterozygous segment, that is, RM25766–RM228 on chromosome 10. The intervals RM6704–RM3773 and RM25766–RM228 are adjacent and jointly covered the RM6100–RM228 region previously mapped for qGS10 (Wang et al., 2017). Two populations were constructed and used for QTL mapping. One consisted of 203 S_1:2 families derived from ZC9, and the other comprised 171 S₁ plants produced from Ti52-1 (Table 1).

The other 10 populations used for QTL mapping were all derived from Ti52-1. Five F₉ plants with sequential heterozygous regions covering RM25766–RM228 were selected from progenies of Ti52-1. They were selfed to produce five new S₁ populations. The number of plants in each population ranged from 172 to 343 (Table 1). In each population, non-recombinant homozygotes were selected and selfed. Five NIL populations in F_10:11 were established, in which the numbers of NILs having the same genotype ranged from 28 to 38 (Table 1).

Field trial and phenotypic evaluation

The filed experiments were conducted at the experimental stations of the China National Rice Research Institute located in either Hangzhou, Zhejiang Province, or Lingshui, Hainan Province, China. The plants were grown with 16.7 cm between plants and 26.7 cm between rows.

The six S₁ populations were grown without replication and the plants were harvested individually. The remaining six populations, including one S_1:2 population and five NIL populations, were tested using a randomized complete block design with two replications. A total of 12 plants per line were planted in each row. The middle five plants were harvested. Measurement of TGW, GL, and GW in all the experiments followed the method of Zhang et al. (2016).

DNA marker analysis

For the ZC9 population, marker data and genetic maps have been available (Sun et al., 2018). For the other populations, a total of 12 DNA markers were used, including five simple sequence repeat (SSR) and seven InDel markers. The SSRs were selected from the Gramene database (www.gramene.org), while the InDels (Table 2) were developed based on the sequence differences between TQ and IRBB52. DNA extraction and PCR amplification followed the method of Zheng et al. (1995) and Chen et al. (1997), respectively.

Data analysis

Mapmarker/Exp 3.0 was performed to calculate the genetic distance which was presented in centiMorgan using the Kosambi function (Lander et al., 1987). For the six S₁ populations and the ZC9 S_1:2 population, QTL analysis was performed with the inclusive composite interval mapping (ICIM) using the BIP (QTL mapping in the biparental populations) functionality of QTL IciMapping 4.1 (Meng et al., 2015). LOD thresholds were calculated with 1,000 permutation tests (P < 0.05) and used to declare a putative QTL. For the five NIL populations, SAS procedure GLM (SAS Institute Inc, 1999) was used to determine phenotypic differences in each population. A mixed model GENOTYPE + LINE (GENOTYPE) + REP + GENOTYPE*REP was applied, in which LINE (GENOTYPE) was defined as a random effect and used as the error term to test GENOTYPE differences (Dai et al., 2008). Additive effect and the proportion of phenotypic variance explained (R²) were estimated with the same model. QTL were designated following the rule proposed by McCouch & CGSNL (2008).

QTL for grain size detected in the ZC9 and Ti52-1 populations

The ZC9 S_1:2 population was segregated in the interval RM6704–RM3773 that covered the upper portion of the qGS10 region, and in three other intervals (Fig. 2). All the four regions were found to have significant effects on one or two of the three grain-size traits analyzed (Table 3). In the interval RM6704–RM3773 on chromosome 10, the effects were significant on GL but non-significant on the other two traits. This QTL explained 9.00% of the phenotypic variance and the IRBB52 allele increased GL by 0.020 mm. The qGL1 and qGW1 associated with RM12210 on chromosome 1 had R² values of 32.71% and 5.84%, respectively, with the IRBB52 allele increased GL by 0.041 mm and decreased GW by 0.007 mm. The qTGW9 and qGW9 in the interval RM219–RM1896 on chromosome 9 had R² values of 19.15% and 7.42%, respectively, with the IRBB52 allele increased TGW by 0.16 g and GW by 0.007 mm. The qTGW11 and qGW11 in the interval RM224–RM5926 on chromosome 11 had R² values of 13.81 and 31.69%, respectively, with the IRBB52 allele decreased TGW by 0.14 g and GW by 0.016 mm.

The Ti52-1 population was segregating in the interval RM25766–RM228 only, which covered the lower portion of the qGS10 region (Fig. 2). This region was found to have significant effects on all the three grain-size traits analyzed (Table 3). The QTL effects contributed 20.15–36.58% to the phenotypic variance of each trait, with the TQ allele increasing TGW, GL, and GW by 0.59 g, 0.081 mm, and 0.024 mm, respectively. Since qGS10 was previously reported to affect all these three traits with the TQ allele always increasing trait values (Wang et al., 2017), it is concluded that qGS10 was located in the segregating region of the Ti52-1 population, and another QTL, qGL10.1, was located in an adjacent interval that was segregated in the ZC9 population (Fig. 3A).

Dissection of qGS10 into two QTL for grain size and grain weight

For further validation and dissection of qGS10, five S₁ populations with sequential segregating regions covering the interval RM25766–RM228 were established (Table 1; Fig. 3B). Results of QTL analysis for TGW, GL, and GW using these populations are shown in Table 4. Significant QTL effects were detected in G10-2, G10-4, and G10-5 but not in the other two populations. In G10-2, the effects were significant for all the three traits. The TQ allele increased TGW, GL, and GW by 0.19 g, 0.030 mm, and 0.004 mm, respectively. In G10-4 and G10-5, significant QTL effects were detected for GL only. The enhancing alleles were derived from IRBB52 in both populations, with similar additive effects of 0.019 and 0.023 mm, and similar R² values of 5.11% and 3.01%. As shown in Fig. 3B, the region segregated in G10-2 was separated from those segregated in G10-4 and G10-5, indicating that at least two QTL were responsible for the grain-size variations in these populations. One QTL was located in the segregating region of G10-2 and another QTL may be located in the common segregating region of G10-4 and G10-5. To validate these results, non-recombinant homozygotes of the five S₁ populations were identified, from which five NIL populations maintaining the original sequential segregating regions were constructed (Table 1; Fig. 3B).

Distributions of TGW, GL, and GW in the five NIL populations are shown in Fig. 4. The three traits were continuously distributed in all the populations, but differentiation between the two genotypic groups was observed in four populations. Concentration of the TQ and IRBB52 homozygous lines, respectively, in the high- and low-value regions was evident for TGW and GL in G11-2, and for all the three traits in G11-3. The same tendency, though less obvious, was found for TGW and GW in G11-4 and G11-5. Since the segregating region of G11-3 covered the segregating regions of G11-2, G11-4, and G11-5 (Fig. 3B), these results suggest that allelic differences of the two QTL, one located in the segregating region of G11-2 and the other in the common segregating region of G11-4 and G11-5, could be the main cause of phenotypic variations in the five NIL populations.

Results of two-way ANOVA for TGW, GL, and GW using the five NIL populations are presented in Table 5. In G11-1, no significant phenotypic difference was detected between the two genotypic groups, matching well to the detection of no QTL in its origin population G10-1. In G11-2, the two genotypic groups showed significant differences on each trait. The TQ allele increased TGW, GL, and GW by 0.21 g, 0.032 mm, and 0.004 mm, with R² values of 40.58%, 28.55%, and 3.99%, respectively. These results were in general agreement with QTL detected in the G10-2 population from which G11-2 was originated, confirming that a QTL affecting TGW, GL, and GW was located in the segregating region of G10-2 and G11-2. This QTL, designated qGS10.1, was mapped within the region flanked by InDel markers Te20811 and Te21018 (Fig. 3B), corresponding to a 207.1-kb region in the Nipponbare genome.

In G11-4 and G11-5 which were segregated in homozygous regions of G11-2, significant QTL effects were detected for TGW and GW with the TQ allele always increasing the trait values. The additive effects were 0.13 and 0.12 g for TGW, and 0.012 and 0.011 mm for GW. These results indicate that a QTL for TGW and GW was located in the common segregating region of G11-4 and G11-5. This QTL, designated qGS10.2, was mapped within the region flanked by SSR markers RM3123 and RM6673 (Fig. 3B), corresponding to a 1.2-Mb region in the Nipponbare genome. It is noted that the effect of qGS10.2 on TGW and GW was not detected in the G10-4 and G10-5 populations from which G11-4 and G11-5 were originated. For the other trait GL, a small effect with the enhancing allele derived from IRBB52 was detected in G11-5, which was in accordance with the effects detected in G10-4 and G10-5.

For the remaining NIL population, G11-3, significant QTL effects with the enhancing allele derived from TQ were detected for all the three traits. The additive effects and R² values estimated for TGW and GW were higher in G11-3 than in G11-2, G11-4, and G11-5. This is reasonable because qGS10.1 and qGS10.2 have the same allelic direction and the segregating region of G11-3 covered both QTL.