Phenotypic evaluation of 11 yield-related traits
We cultivated Kitahonami and Shunyou plants in the field for three growing seasons and compared 11 yield-related traits and grain number between them. Since the three field blocks did not significantly differ during the three seasons (p > 0.05), we did not correct trait values. Kitahonami headed significantly later than Shunyou during the three seasons (Fig. 1) and tended to have higher SNS, GNS, and TGW, although the difference was not significant in one or two growing seasons. The central and basal parts of the Kitahonami spike contained more, whereas the apical part contained less grains than the Shunyou spike. Since GWS was dependent on both GNS and TGW, the higher GNS and TGW contributed to the higher GWS, and also probably to the higher grain yield in Kitahonami.
We calculated correlation coefficients of the 11 traits using 188 DHLs for three growing seasons to determine relationships between the 11 traits measured in this study (Figs. 2 and S2). Many traits significantly correlated with DH. Correlations were moderately positive and negative between SNS and DH and between TGW and DH, respectively, for the three growing seasons. The mean values of DH and GWS significantly and negatively correlated only during the 2017-2018 season. The 2017-2018 season showed different correlations between DH and the other traits than in the other seasons. No significant correlation was found between DH and GNS during the 2017-2018 season. The GWS significantly and positively correlated with GNS and TGW, although GNS and TGW significantly and negatively correlated. Unlike GNS, GNSP at each part of the spike did not significantly correlate with TGW.
Heritability estimates of traits
We calculated marker-based estimates of heritability for the 11 traits (Table 1). Among the yield-related traits, the heritability of SNS and TGW was the highest (0.87). The heritability was higher for GNS (0.71) than DH (0.56), and that of GWS and DH was similar (0.57). The grain number was more heritable at the basal, than that at the central and apical parts of the spike.
QTL analysis of yield-related traits
We analyzed the QTLs of the 11 traits using inclusive composite interval mapping of additive (ICIM-ADD). Table S3 and Table 2 respectively show and summarize all QTLs with significant LOD scores (p < 0.05) and mean values for the three growing seasons.
Grain number per spike (GNS)
We identified QTLs for GNS on chromosomes 2A, 2D, 3B, 4A, 5A and 6D. Seven QTLs were identified when QTLs with overlapping confidence intervals were the same QTL. The phenotypic variation explained (PVE) ranged from 3.8% to 14.3 %, and the QTL on chromosome 2A had the highest PVE. No QTLs were detected during all seasons, and the QTL on chromosome 2A was the only significant QTL that was detected in two seasons.
Grain number at the basal, central and apical parts of spikes (GN-B, GN-C, GN-A)
Compared with GNS, QTLs for total grain number three parts of five spikes were abundant (23, 14, and 17 QTLs for total grain numbers at the basal [GN-B], central [GN-C] and apical [GN-A] parts, respectively). We detected QTLs with the highest PVE for GN-B, GN-C and GN-A on chromosomes 7A, 2A and 5A. QTLs on chromosome 2A were detected for GN-B, GN-C, and GN-A. The QTL for GN-C was detected in this region for three growing seasons, whereas the those for GN-A and GN-B in this region were found for one or two growing seasons. The QTL for GN-B that was identified on chromosome 7A had the highest PVE among the QTLs for GN-B (10.8% to 17.0%). The Kitahonami allele of this QTL increased GN-C and GN-B (negative additive effect). The QTLs for GN-A identified on chromosomes 5A and 7A were detected for three growing seasons, and the QTL for GN-A on chromosome 5A had the highest PVE (8.8% to 12.9%). These QTLs were named qGN-A-5A and qGN-A-7A, respectively. The Shunyou alleles of the QTLs on chromosomes 5A and 7A were increased in GN-A.
Spikelet number per spike (SNS)
In total, 26 QTLs for SNS were found on chromosomes 2A, 2D, 4A, 4 B, 5A, 6A, 7A, and 7D. Thirteen QTLs with overlapping confidence intervals were in fact the same QTL. The PVE ranged from 2.6% to 21.4% and was the highest (16.3% to 23.1%) on chromosome 2D. The QTLs located on chromosomes 2A, 2D, and 7A were detected during all growing seasons. The Shunyou allele of the QTL on chromosome 2A and the Kitahonami allele of the QTLs on chromosomes 2D and 7A increased SNS. The QTLs located on chromosomes 2A and 2D were discovered in the Ppd-A1 and Ppd-D1 regions, respectively, as the QTL for DH (see below).
Grain number per spikelet (GNSP)
To evaluate grain number without being affected by spikelet numbers, we also analyzed GNSP. Nineteen QTLs for GNSP-A, 11 for GNSP-C, and 14 for GNSP-B were detected on 13 chromosomes. The highest PVE of the QTLs for GNSP-C (24.0% to 43.3%) and GNSP-B (14.9% to 23.0%) was chromosome 2A, which was near GNI-A1. The QTLs for GNSP-C and GNSP-B on chromosome 2A were detected during all three growing seasons. The highest PVE of the QTL for GNSP-A was chromosome 5A (7.5% to 15.3%). The QTLs for GNSP-A on chromosomes 5A and 7A were also detected during all three growing seasons. The Kitahonami allele of the QTLs for GNSP-C and GNSP-B on chromosome 2A increased GNSP, whereas the Kitahonami allele of the QTL for GNSP-A on chromosome 5A decreased GNSP.
Thousand grain weight (TGW)
In total, 33 QTLs for TGW were identified on chromosomes 1A, 2A, 2D, 3A, 4A, 5D, 6D, 7 B, and 7D. Twelve QTLs with overlapping confidence intervals were in fact the same QTL. The PVE ranged from 3.0% to 13.1%, with the highest being for the QTL on chromosome 6D. The QTLs for TGW on chromosomes 1A, 2A, 4A, 5D, 6D, and 7D were identified for the three growing seasons. Three QTLs on chromosomes 4A, 6D, and 7D with a high PVE were named qTGW-4A, qTGW-6D, and qTGW-7D, respectively. The Shunyou allele of qTGW-4A and qTGW-6D and the Kitahonami allele of qTGW-7D increased TGW.
Grain weight per spike (GWS)
Four QTLs for GWS were found on chromosomes 2B and 7A. The PVE of these QTLs ranged from 6.5% to 8.7%, and that located on chromosome 2B was the highest. Although the QTLs located on chromosome 2B were detected during two growing seasons, and the positions of the QTLs slightly differed between the growing seasons. In addition, two QTLs with opposing additive effects were located near each other on chromosome 2B during the 2017-2018 season.
Days to heading (DH)
We detected 29 QTLs for DH on chromosomes 2A, 2D, 4A, 4D, 5A, 5D, 6A, 6D, and 7D. Twelve QTLs with overlapping confidence intervals were in fact the same QTL. The PVE ranged 1.6% to 39.1%, and QTLs with the highest PVE were located on chromosomes 2A and 2D, which were discovered in the Ppd-A1 and Ppd-D1 regions, respectively. The QTLs located on chromosomes 2A and 2D exerted positive and negative additive effects, respectively, and those located on chromosomes 2A, 2D, 4A, 6A, 6D and 7D were detected during all growing seasons.
Quantitative trait loci located near known genes associated with yield
Significant QTLs for various traits were detected in the chromosomal region near genes that are associated with yield (Figs. 3, S3 and S4). The QTLs for grain number were found mainly in the GNI-A1 region. Kitahonami had the 105Y allele of GNI-A1, which produces a higher GNS (Sakuma et al. 2019). The KASP marker, which distinguished the 105Y and 105N variants of GNI-A1, revealed that Shunyou had the 105N allele of GNI-A1 (data not shown). The Kitahonami allele of GNI-A1 increased GN in the three parts of the spike, especially in the central part. The QTLs for GNSP-A and GNSP-C were detected in the GNI-A1 region, indicating that the higher grain number in Kitahonami was due to increased GNS by the 105Y allele of GNI-A1. Shunyou and Kitahonami had the WAPO-A1a and WAPO-A1b alleles, respectively (data not shown). The WAPO-A1b allele increased SNS and GN-B, but decreased GNSP-A. The QTLs for SNS, GN-B, TGW and DH were found in the Ppd-A1 and Ppd-D1 regions. Kitahonami and Shunyou had the Ppd-A1a and Ppd-D1a alleles, respectively, which conferred early heading phenotypes.
Candidate genes for qGNSP-A-5A and qGNSP-A-7A
We mapped qGNSP-A-5A between snp2947 and tarc0637 (Table S3). We respectively assigned snp2947 and tarc0637 to physical positions of 688.32 and 689.92 Mbp in the CS reference genome sequence (Table S2). However, the range of candidate genes for qGNSP-5A possibly extended to the distal end of chromosome 5A. The 21.45 Mbp region between snp2947 and the distal end contained 302 high-confidence genes (Table S4). We mapped qGNSP-A-7A between snp4426 and wmc83, which were respectively assigned to the physical positions of 84.75 and 89.97 Mbp in the CS reference genome sequence. The 5.22 Mbp region contained 55 high-confidence genes (Table S5) among which, TraesCS7A02G136600 was specifically expressed in the spike (Fig. S6). TraesCS7A02G136600 encodes a COBRA-like protein, which is homologous to OsBC1L5 (Os06g0685100).
Pyramiding effect of QTLs for grain number and TGW
To analyze the pyramiding effects of favorable alleles of GN and TGW QTLs on GNS, we compared GNS, TGW and GWS among lines with different number of favorable alleles for GNS and TGW (Fig. 4). Three GN QTLs (GNI-A1, WAPO-A1 and qGNSP-A-5A) with the highest PVE in the three parts of the spike, were selected as the QTLs for grain number. Both GNS and GWS increased as pyramiding favorable alleles for GNS, whereas TGW did not significantly differ among the number of favorable alleles for GNS. More QTLs were found for GN-A than for GN-C and GN-B. Pyramiding of the two favorable alleles for GN-A significantly increased GN-A and GNSP-A (Fig. S5). None of GNSP-C, GNSP-B and TGW significantly differed among the lines with different number of favorable alleles for GN-A. As the QTLs for TGW, qTGW-6D (tarc1335), qTGW-4A (Wx-B1) and qTGW-7D (tarc0372) with a high PVE were selected. As pyramiding favorable alleles for TGW, TGW increased but GNS decreased. The GWS did not significantly differ among lines with one or more favorable alleles for TGW. We compared GNS, TGW and GWS among the four characteristic combinations of the two respective QTLs for GNS and TGW to determine their effects (Fig. 5). The TGW did not significantly differ between lines with various allele combinations of GNS and the same allele combination of TGW. In contrast, GNS was higher in lines with a favorable, than an unfavorable allele combination of TGW QTLs. The minimum value of GWS in the eight lines with favorable alleles of both GNS and TGW QTLs (2.22 g) was much higher than that of lines with the other allele combinations (1.09–1.78 g).
Relationships between DH and the other traits
Correlations between DH and the other grain yield traits were significant during the three growing seasons (Figs. 2 and S2). The SNS and TGW were the most positively and negatively correlated with DH, respectively. Five chromosomal regions, including chromosomes 4A, 6D and 7D, coincided with QTLs for both DH and TGW (Fig. S4). However, these QTLs had different influences on DH and TGW. For example, the QTL for DH on chromosome 2D (Ppd-D1) had the highest PVE, but Ppd-D1 had a low PVE for the TGW QTL (Table S3). Lines with two or more favorable alleles for TGW headed earlier than those with one or more favorable alleles for TGW (Fig. S7). The SNS also decreased as the QTLs for favorable alleles for TGW increased, although this did not reach statistical significance. The QTLs for DH, the Ppd-A1, Ppd-D1 and qDH-7D (same as qTGW-7D, tarc0372) with a high PVE were selected. The DH and SNS significantly decreased (Fig. S8), and GNS also decreased as alleles conferring early heading accumulated, although the effect on GNS was lower than that of DH and SNS. In contrast, TGW significantly increased as alleles that confer early heading accumulated.