A common allohexaploid wheat line TAA10, the ETW, an Ae. tauschii accession TQ18 and the resynthesized allohexaploid wheat XX329 were grown with three replicates (two rows/replicate) in Shangzhuang, Beijing in the autumn of 2014. Seeds were well-distributed in rows that were 1.5 m long and 0.3 m apart with a sowing rate at 20 seeds per row, which is the same with the following field experiments.

An F₂ population consisted of a number of F₂ individuals as the progenies of the F₁ individual by the cross between TAA10 and XX329, and an F_2:3 population contained all the derived F_2:3 lines from the corresponding individuals in the F₂ population. For QTL analysis, a total of 328 F₂ individuals generated by the cross between TAA10 and XX329 were grown in the greenhouse of China Agricultural University in 2014, and the derived 328 F_2:3 lines were grown in Shangzhuang, Beijing in the spring of 2015. In addition, the other six F₂ populations by the same cross and two parental lines were grown in the experimental field under six different environments at three locations (Shangzhuang, Beijing, E116°, N40°; Shijiazhuang, Hebei, E114°, N38°; and Linfen, Shanxi, E111°, N36°) in the spring of 2 years (2015 and 2016). These six F₂ populations contained 377 (Beijing2015), 206 (Beijing2016), 265 (Heibei2015), 216 (Heibei2016), 198 (Shanxi2015), and 260 (Shanxi 2016) F₂ individuals, respectively. The parental lines TAA10 and XX329 were grown with three replicates (two rows/replicate) in each environment.

Field trials for evaluating the plot yield of TAA10 and XX329 were performed in a randomized complete block design with three replicates (20 rows/replicate) at two locations (Beijing and Hebei) in the autumn of 2015.

The measurements of thousand grain weight (TGW), grain length (GL), grain width (GW), grain area (GA), and grain perimeter (GP) were performed using the grain analysis program developed by Wanshen Science and Technology Ltd. (Hangzhou, China; Cheng et al., 2015). For TAA10, ETW, TQ18, and XX329, seeds were harvested with 30 plants (10 random plants in each replicate). For each F₂ and BC₄F₂ population, every individual was harvested and measured for these grain traits, and for the F_2:3 progeny, all traits were described by the mean values of 20 plants for the corresponding line from each F₂ individual.

In plot-yield trials of two parental lines TAA10 and XX329, spike length (SL), spikelet number per spike (SLN), spike number per plant (SN), grain number per spike (GN), and TGW were evaluated with 30 plants per replicate, and the data from SL, SLN, and GN were collected from the main tillers. SL was measured from the base of the rachis to the tip of the terminal spikelet, and SLN contained the fertile spikelet number and sterile spikelet number per spike. All grains in a single plot were collected for measurements of yield per plot (YPP).

The statistical analyses containing Student's t-test, variance analysis and correlation analysis were performed with SPSS version 20.0 (SPSS, Chicago, USA).

Wheat 660K SNP array was designed by Chinese Academy of Agricultural Sciences and synthesized by Affymetrix (http://wheat.pw.usda.gov/ggpages/topics/Wheat660_SNP_array_developed_by_CAAS.pdf). Two parental lines TAA10 and XX329 were genotyped with wheat 660K SNP chip by Compass Biotechnology Company (Beijing, China). The genetic position information of SNP markers was provided by Compass Biotechnology Company (Unpublished data).

Genomic DNA was extracted from leaf tissues in the seeding stage using the cetyltrimethyl ammonium bromide (CTAB) method (Allen et al., 2006). The enough leaf sample of each plant was collected and stored at −20° to ensure obtaining DNA with high quality for genotyping analysis. Extracted DNA was dissolved and preserved in TE buffer, and the quality of DNA was assessed using 1% agarose gel electrophoresis.

The genomic sequence of Ae. tauschii is a useful reference for marker development on the D genome of wheat. The scaffold sequence published at ATGSP (http://aegilops.wheat.ucdavis.edu/ATGSP/)was used to identify the SSR region containing at least 10 dinucleotide or trinucleotide repeats, and primers for this region were designed in the flanking sequence using Primer3 web (version 4.0.0; http://bioinfo.ut.ee/primer3/). The synteny analysis of these SSR markers, Ae. tauschii markers and rice genes referenced to Luo et al. (2013), and the information of rice genes were obtained from the International Rice Genome Sequencing Project (Matsumoto et al., 2005). Published primer sequences for the SSR markers are available at http://wheat.pw.usda.gov/. The PCR was conducted in a 10 μL reaction system containing 50 ng genomic DNA, 1 μL 10 × reaction buffer, 0.2 μL 10 mmolL⁻¹ dNTPs, 2.0 μL primer, 1 U rTaq DNA polymerase (Takara, Dalian) and 4.7 μL ddH₂O. The PCR program was set to denature the template DNA at 94°C for 5 min, followed by 35 cycles at 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s, and finally extend the PCR products for 10 min at 72°C. The length polymorphism of the SSR markers was identified using 8% non-denatured polyacrylamide gel electrophoresis (PAGE; Marklund et al., 1995).

For quick scanning of the entire genome to find best possible QTL, single marker analysis by a simple linear regression model were performed with Windows QTL Cartographer version 2.5 (Wang et al., 2012). The trait values and genotypic data of each marker in different populations were used for single marker analysis, and a significance level of 0.05 was used to declare that the marker was associated with the corresponding trait. Genetic linkage maps were constructed with the program JoinMap 4.0 (Van Ooijen, 2006). Markers with <5% missing data were employed and organized into linkage groups with the LOD thresholds ranged from 4 to 10. The order of markers in each linkage group was arranged using a regression mapping algorithm (Stam, 1993). The map distances were calculated based on recombination frequencies using the Kosambi mapping function (Kosambi, 1943). The QTL analysis using the composite interval mapping (CIM) method were performed with Windows QTL Cartographer version 2.5 (Wang et al., 2012). Model 6 with forward and backward regression was employed to QTL mapping. Five markers as cofactors and a 10-cM scanning window were chosen for the detection of QTL. The LOD threshold was set via 1000 permutations at P ≤ 0.05 (Yan et al., 2006). Considering the fact that phenotypic data of each genotype in F₂ populations may be inaccurate for the quantitative trait, significance for QTL was defined as a LOD value higher than 2.0. The identified QTL were named according to McIntosh et al. (2011).

To develop the NILs of the three QTL regions 2DS, 2DL, and 7DS, the F₁ individuals derived from TAA10 and XX329 were backcrossed with the recurrent parent (TAA10). Marker-assisted foreground selection was performed in each backcross generation. The BC₄F₁ individuals which exhibited heterozygosity in each QTL region were self-pollinated to produce their corresponding BC₄F₂ populations. In addition, a total of 60 SSR markers on the D genome were used for background selection of the BC₄F₁ individuals. Finally, the numbers of BC₄F₂ populations for QTL verification on 2DS, 2DL, and 7DS were 1 (BC₄F₂-2DS), 2 (BC₄F₂-2DL-1 and BC₄F₂-2DL-2) and 2 (BC₄F₂-7DS-1 and BC₄F₂-7DS-2), respectively. The population sizes of BC₄F₂-2DS, BC₄F₂-2DL-1, BC₄F₂-2DL-2, BC₄F₂-7DS-1, and BC₄F₂-7DS-2 were 213, 231, 186, 150, and 153, respectively. These BC₄F₂ populations were grown in Shangzhuang, Beijing in the spring of 2017.

The ETW had lower TGW, GL, GW, GA, and GP compared with TAA10, and these values of the resynthesized wheat XX329 were higher than those of TAA10 (Figure 1; Table 1). TGW, GL, GW, GA, and GP variations of the two parents (TAA10 and XX329) and the segregation populations were evaluated in six environments (Appendix B in Supplementary Material). XX329 consistently showed higher values by at least 6.65 g TGW, 1.17 mm GL, 0.13 mm GW, 2.58 mm² GA, and 1.65 mm GP compared with TAA10 in all environments (Table 2). The frequency distributions of the investigated traits revealed continuous variations in the F_2:3 population, suggesting that the phenotypic data of TGW, GL, GW, GA, and GP are normally distributed and these traits are controlled by multiple loci (Figure 2). Additionally, correlation coefficients among the TGW, GL, GW, GA, and GP traits in the F_2:3 population were calculated. All five traits showed significant positive correlations with each other (Table 3). The strongest positive correlation was observed between GL and GP, followed by GA and GP. However, GL had a weak positive correlation with GW.

In the plot-yield trials, XX329 had higher TGW than TAA10 in the two environments. Conversely, grain number per spike of TAA10 was significantly higher than that of XX329. No significant differences were observed in spike length, spikelet number per spike, and spike number per plant. The grain yield per plot of XX329 increased by 17.43–30.36% compared with that of TAA10 in two environments (Table 4).

To evaluate the genetic difference between TAA10 and XX329, wheat 660K SNP chip with 630,517 makers was employed for analysis (Appendix D in Supplementary Material). Of 594,299 SNP markers with genotype data, 48,753 (8.20%) markers exhibited polymorphism between TAA10 and XX329. Based on genetic position information provided by Compass Biotechnology Company, 392,088 of 594,299 SNP markers have been mapped to wheat genomes, and the numbers on A, B, and D genomes were 148,425, 179,667, and 63,996, respectively. Accordingly, 30,121 of 48,753 polymorphic SNP markers between TAA10 and XX329 could be mapped to wheat genomes, and the numbers on A, B, and D genomes were 5,114, 3,417, and 21,590 markers, respectively ((Table S1). Taken together, the genetic similarities of A, B, and D genomes between TAA10 and XX329 were 96.55, 98.10, and 66.26%, respectively. Notably, 2,617 of 5,114 polymorphic SNP markers on A genome were located on chromosome 1A, among which the majority (1988, 75.96%) was located in the telomere region on the long arm of chromosome 1A.

A total of 436 D-genomic SSR markers were screened, and 81 polymorphic markers between TAA10 and XX329 were used for single marker analysis of TGW, GL, GW, GA, and GP by genotyping 328 individuals from the F₂ population grown in 2014 and using the phenotypic data of the 328 F_2:3 lines grown in 2015. The results showed that two markers Xbarc11 and Xcfd2 on chromosome 2D were significantly associated with all five traits, and four markers (Xgwm261, Xcfd53 on chromosome 2D and Xwmc702, Xbarc260 on chromosome 7D) were significantly linked with GL, GA, and GP (Table 5).

To further conduct QTL analysis for traits related to grain size and shape on chromosomes 2D and 7D, 31 SSR makers were developed using the referential sequence from Ae. tauschii, and the information of developed SSR markers for linkage map construction was shown in Appendix C in Supplementary Material. The collinearities of these markers with the Ae. tauschii markers and rice genes were shown in Figure 3. These polymorphic markers were used for the linkage analysis by genotyping 328 individuals from the F₂ population grown in 2014 (Appendix A in Supplementary Material). The resulting linkage maps of chromosomes 2D and 7D consisted of 25 and 34 SSR markers, spanning 129.78 and 198.88 cM in length, respectively (Figure 3).

Using the phenotypic data from the 328 F_2:3 lines grown in 2015, 11 QTL located on chromosomes 2D and 7D were identified for the five grain traits (Table 6). One QTL associated with TGW was located on the long arm of chromosome 2D, which explained 13.91% of the phenotypic variation, and it was designated QTgw.cau-2D. XX329 contributed effects for increased TGW at the QTL. Additionally, one QTL for GW (QGw.cau-2D) was detected in the same QTL interval on chromosome 2DL and explained 18.1% of the GW variation. The QTL associated with GL, GA and GP were all identified in each of the three genomic regions, chromosome 2DS (QGl.cau-2D.1, QGa.cau-2D.1, and QGp.cau-2D.1), chromosome 2DL (QGl.cau-2D.2, QGa.cau-2D.2, and QGp.cau-2D.2) and chromosome 7DS (QGl.cau-7D, QGa.cau-7D, and QGp.cau-7D). Three QTL for GL explained the proportions of phenotypic variation ranging from 10.50 to 20.13%, accordingly from 4.65 to 25.12% for GA and from 10.05 to 12.40% for GP. XX329 contributed favorable alleles at all of the identified QTL (Figure 3; Table 6).

To further investigate the environmental stability of the 11 QTL, another six F₂ populations derived from TAA10 and XX329 were grown at three locations (Beijing, Hebei and Shanxi) during 2 years (2014 and 2015). Three SSR markers (Xgwm455, Xgwm261, and Xcfd53) on chromosome 2DS, four markers (SSR-2225, SSR-2336, SSR-2129, and Xbarc11) on chromosome 2DL and six markers (Xwmc702, SSR-7074, SSR-7252, SSR-7176, Xbarc260, and Xwmc488) on chromosome 7DS in and nearby the QTL intervals were used for linkage map construction by genotyping individuals from each F₂ population. The QTL that could be detected in five or more environments were regarded as “environmentally stable QTL.” According to this criterion, nine of them were environmentally stable QTL. In the QTL interval on chromosome 2DS, QGl.cau-2D.1, QGa.cau-2D.1, and QGp.cau-2D.1 were detected in all six F₂ populations, which explained from 14.84 to 32.78%, from 3.70 to 20.03%, and from 8.91 to 27.43% of phenotypic variation, respectively. Remarkably, QTgw.cau-2D, QGl.cau-2D.2, QGw.cau-2D, QGa.cau-2D.2, and QGp.cau-2D.2 in the region on chromosome 2DL were all environmentally stable in the six F₂ populations, and the QTL for TGW, GL, and GW explained the proportion of phenotypic variation ranging from 4.28 to 13.72%, from 5.93 to 21.92%, and from 3.59 to 10.48%, respectively. In addition, among the QTL on chromosome 7DS, QGl.cau-7D was the environmentally stable QTL detected in five environments and explained from 3.80 to 12.30% of the GL variation (Table 7).

To further verify the QTL on chromosomes 2DS, 2DL, and 7DS, five NIL populations were developed. In the process of backcross, the SSR markers in the three QTL regions were used for foreground selection: five markers (Xcfd36, Xgwm455, Xgwm261, Xcfd53, and SSR-2092) for the 2DS region, six markers (SSR-2225, SSR-2336, Xcfd2, SSR-2129, Xbarc11, and SSR-2133) for the 2DL region and six markers (Xwmc702, SSR-7074, SSR-7252, SSR-7176, Xbarc252, and SSR-7148) for the 7DS region. Finally, five BC₄F₁ individuals (BC₄F₂-2DS, BC₄F₂-2DL-1, BC₄F₂-2DL-2, BC₄F₂-7DS-1, and BC₄F₂-7DS-2) were self-pollinated to produce their corresponding BC₄F₂ populations, which exhibited heterozygosity in corresponding QTL region and 93.97 to 98.31% similarities in genetic background with the recurrent parent (Appendix E in Supplementary Material).

To determine whether the identified QTL affect the traits of grain size and shape in each BC₄F₂ population, we compared TGW, GL, GW, GA, and GP traits between two homozygous groups, that is, TAA10 homozygotes and XX329 homozygotes (Table 8; Appendix F in Supplementary Material). In the BC₄F₂-2DS population, the mean values of GL, GA and GP of the XX329 homozygotes showed significantly higher than those of TAA10 homozygotes, whereas TGW and GW had no significant differences between the two homozygous groups. In the two BC₄F₂-2DL-1 and BC₄F₂-2DL-2 populations, the XX329 homozygous groups exhibited significantly higher TGW, GL, GW, GA, and GP than TAA10. In addition, the significant differences of GL, GA, and GP were identified between TAA10 and XX329 homozygous groups in the BC₄F₂-7DS-1 population, and the significant differences of GL and GP were found between two homozygous groups in the BC₄F₂-7DS-2 population (Table 8). Collectively, these results provided further evidence that these QTL on chromosomes 2D and 7D significantly affected on grain size and shape, which were in agreement with the results by the F₂ and F_2:3 population.

Resynthesized allohexaploid wheat XX329 (AABBDD) was produced by crossing ETW (AABB) and the Ae. tauschii subsp strangulate (TQ18, DD). ETW contains an AABB genome from natural allohexaploid bread wheat donor (TAA10) by nine cycles of backcrossing (Kerber, 1964). Theoretically, the genome (AABB) of XX329 should be >99.8% identical to the AABB subgenomes of its bread wheat donor (TAA10) after the ninth backcross (Zhang et al., 2014). To verify this hypothesis, wheat 660K SNP chip was firstly employed for analysis in this study. Based on the information of SNP markers with genetic position, we found that the A and B genomes of TAA10 and XX329 showed 96.55 and 98.10% genetic similarities, respectively, which provided molecular evidence that the genome (AABB) of XX329 was near-identical to that of TAA10. Interestingly, we observed that the numbers of polymorphic SNP markers are not uniform across the A genome. Of 2617 polymorphic SNP markers on chromosome 1A, 1988 (75.96%) were located in the telomere region of the long arm. This phenomenon may be attributed to genetic recombination or genomic variation during the process of extracting the AABB genomes, which is needed for further investigation.

Allohexaploid common wheat (AABBDD) evolved by natural hybridization of emmer wheat (AABB) and Aegilops tauschii Coss. (DD; Nesbitt and Samuel, 1995; Petersen et al., 2006). Growing evidence revealed that a few Ae. tauschii's intraspecific lineages contributed to the evolution of common wheat, which resulted in relatively narrow genetic variation on the D-genome in wheat (Dvorak et al., 1998; Dubcovsky and Dvorak, 2007). Consistently, Jin et al. (2016) and Cui et al. (2017) reported the construction of high-density genetic maps using the wheat 660K SNP chip based on recombinant inbred line populations derived from common allohexaploid wheats, and the number of polymorphic SNP markers on the D genome was 3,905 and 13,820, respectively, which were much lower than that of TAA10 and XX329. Remarkably, the resynthesized allohexaploid wheat XX329 manifested obvious different phenotypes at multiple growth/developmental stages relative to natural allohexaploid wheat TAA10 (Zhang et al., 2014). Considering the higher genetic similarity of A and B genomes between TAA10 and XX329, we proposed that the observed phenotypic variation was mainly induced by differences on the D genome between TAA10 and XX329.

Grain size and shape are important traits in wheat and represent a classical example of a trait with variations that arose after polyploidization and domestication. A long, thin primitive grain was transformed into a wider, shorter modern grain during wheat domestication, indicating that grain shape became rounder during wheat domestication (Gegas et al., 2010). A notable aspect of polyploid wheat evolution is genomic asymmetry in the control of grain shape, and that is the predominant control of grain shape by the A genome (Feldman et al., 2012). However, the wide variation in grain size and shape observed among Ae. tauschii genotypes is retained in the synthetic allohexaploid wheat using natural tetraploid species as the AB genome donor, suggesting that the D genome partially affects grain size and shape of allohexaploid wheat (Röder et al., 2008; Okamoto et al., 2013; Rasheed et al., 2014). Here, we found that the D genome could lead to drastic change in grain size and shape of allohexaploid wheat. First, the grain length, width, and size of the extracted allotetraploid wheat (ETW; AABB) are significantly reduced compared with the donor allohexaploid bread wheat (TAA10). Second, the grain length, width and size of the resynthesized allohexaploid wheat (XX329) obtained by crossing ETW and Ae. tauschii subsp strangulate (TQ18) are much higher than that of TAA10. Remarkably, although the increased kernel weight of XX329 is at the expense of reducing grain number per spike, the grain yield per plant of XX329 is significantly enhanced compared with TAA10. Collectively, these data indicated that synthetic allohexaploid wheats with the D genome from Ae. tauschii is a potentially useful resource for genetic improvement of yield in wheat breeding.

Allohexaploid common wheat was produced by natural hybridization of emmer wheat and Ae. tauschii Coss. When compared with the A and B genome, relatively narrow genetic variation was detected on the D genome, which was partially attributed to a limited number of Ae. tauschii involved in the evolution of common wheat (Dubcovsky and Dvorak, 2007). Synthetic allohexaploid wheats (AABBDD) provide potentially novel genetic variations associated with the D genome of Ae. tauschii. To date, some useful genes/QTL controlling desirable traits have been identified on the D genome of synthetic allohexaploid wheat, including disease resistance, abiotic stress tolerance, suitable quality and anti-sprouting (Tadesse et al., 2007; Imtiaz et al., 2008; Li et al., 2012; Ilyas et al., 2015).

Grain weight is an important component of grain yield. Recently, Simmonds et al. (2014) reported that the effect of yield QTL, located on chromosome 6A, was driven primarily by increased grain weight, suggesting that the enhancement of grain weight could contribute to the genetic improvement of wheat yield. Several studies were conducted to identify beneficial QTL for grain weight from the diploid D donor of common wheat (Liao et al., 2008; Röder et al., 2008). For example, an environmentally stable QTL QGw.caas-3D was identified on chromosome 3D and the synthetic allohexaploid wheat Am3 contributed effect for increased grain weight (Liao et al., 2008). In this study, QTL mapping for grain weight was conducted using the F₂ and F_2:3 populations derived from TAA10 and XX329. One major QTL (QTgw.cau-2D), with the synthetic allohexaploid wheat XX329 contributing favorable alleles, was consistently detected on chromosome 2DL under different environments, which was linked to the marker SSR-2336. In the QTL region of QTgw.cau-2D, QTL for GL, GW, GA, and GP were also identified under different environments, and these co-localized QTL shared similar confidence intervals and had tightly linked QTL peak positions, which are indicative of potential pleiotropy among the traits. To the best of our knowledge, no QTL for grain weight has been detected on the long arm of chromosome 2D from synthetic allohexaploid wheats. Therefore, this QTL detected in our populations represent a novel loci controlling grain weight from the diploid D donor, which may be a more desirable target for genetic improvement in wheat breeding. However, grain weight is generally negatively correlated with grain number per spike and spike number per plant. Thus, it is necessary to develop lines carrying introgression segments with this region and to further clarify the function of QTgw.cau-2D to yield and its components in different genetic backgrounds.

Grain shape is a complex quantitative and important agronomic trait. To date, many studies have identified QTL controlling grain shape in common wheat cultivars, and these QTL were assigned to various chromosomes (Dholakia et al., 2003; Breseghello and Sorrells, 2006; Sun et al., 2009; Williams and Sorrells, 2014). Recently, genetic loci controlling the differences in grain shape between common wheat and synthetic hexaploids have been investigated (Okamoto et al., 2012; Yu et al., 2014). Glumes tenaciously enclose grains in the synthetic allohexaploid wheat lines, whereas modern cultivars are free-threshing wheat. Interestingly, one pleiotropic locus on chromosome 2DS significantly contributed to the determination of wheat grain shape, which corresponded to that of Tg-D1 (Dvorak et al., 2012). Consistent with the result, we found that a major QTL with an LOD score of higher than 9.8 was located on the short arm of chromosome 2D and the allele from synthetic wheat at the QTL produced longer grains. The molecular markers Xbarc95 and Xcfd53 flanked this QTL at an interval of 12.02 cM, which was located in the same position as Tg-D1 (Figure 4). Notably, no QTL was detected for grain width and weight on chromosome 2DS. Collectively, our data supported the notion that wheat grains were rapidly improved to the smaller, rounder phenotype that accompanied the formation of free-threshing wheat, because the domestication from Tg1Tg1 to tg1tg1 occurred at an early phase after allohexaploid wheat speciation (Kerber and Rowland, 1974).

In the pleiotropic QTL region associated to GL, GA and GP on chromosome 7DS, QGl.cau-7D was the environmentally stable locus with the highest LOD value under different environments. When comparing previous results, we found that the region on chromosome 7DS was similar to the QTL interval reported by Okamoto et al. (2013) (Figure 4). Analogously, the QTL on chromosome 7D identified in that study significantly contributed to the variation of grain size and shape in the F₂ population derived from two synthetic allohexaploid wheat lines, and further analysis was needed to clarify the relationship of these QTL. Several QTL contributing to important agronomic traits were fine mapped based on orthologous regions across several grass species (Chen et al., 2007; Handa et al., 2008; Somyong et al., 2011). In the present study, we increased markers' saturation in the QTL interval on 7DS using the referential sequence from Ae. tauschii, and the comparative analysis demonstrated that the peak region of QGl.cau-7D was syntenic to rice chromosome 6 at 7.21–8.70 Mb. Dense spike-ar (dsp.ar) in barley controlling spike density and morphology was identified on chromosome 7H, and the genomic region exhibited highly conserved synteny with part of rice chromosomes 6 and 8 (Shahinnia et al., 2012), which is similar to our QTL interval on chromosome 7DS (Figure 4). However, the QTL in the present study may not correspond to the same gene governing the spike and grain development because there was no significant difference in spike density decided by spike length and spikelet number per spike between TAA10 and XX329.

With orthologous genes controlling similar phenotypes across many grass species, including wheat, rice, barley, and sorghum, comparative genomics has shown collinearities and provided a powerful tool for gene discovery in wheat (Valluru et al., 2014). In the present study, the genomic region of the QTL on chromosome 2DL harboring stable QTL for TGW, GL, GW, GA, and GP exhibited good collinearity with the genomic region of rice chromosome 4 (Figure 3). Furthermore, the peak region of QTgw.cau-2D was syntenic to rice chromosome 4 at 18.49–23.76 Mb, and one cloned gene GIF1 (Os04g0413500) encoding a cell-wall invertase required for carbon partitioning during early grain filling, which was located at 20.44 Mb on rice chromosome 4 (Wang et al., 2008). In addition, the peak region of QTL for GL, GA and GP on chromosome 7DS was syntenic to rice chromosome 6 at 7.21–8.70 Mb in which one cloned gene PFP_β(Os06g0247500) regulating carbon metabolism during grain filling was located at 7.65 Mb on rice chromosome 6 (Figure 3; Duan et al., 2016). These results suggest that the two QTL for grain size and shape identified could be orthologous genes to GIF1 and PFP_β, which deserve further research.

The F₂ and F_2:3 populations were extensively utilized for identification of QTL for a number of agronomic traits such as yield-related traits (Lu et al., 2011; Zhang et al., 2012; Wang et al., 2015). The advantage of the F₂ and F_2:3 mapping populations was that the process of construction was convenient and fast. However, the F₂ and F_2:3 populations were temporary separation populations and required large population size to ensure the precision of mapping (Zhang et al., 2010). A trade-off would be to further validate these QTL by the secondary separation populations such as NIL and residual heterozygous line (RHL) populations. As expected, in the present study, the identified QTL associated with grain size and shape based on the F₂ and F_2:3 populations could be verified using the NIL populations. Notably, the linkage maps based on six different F₂ populations derived from the same cross using the SSR markers exhibited different genetic distance, which is normal in consideration of different population sizes and limited numbers of markers. However, the identified QTL from different F₂ populations for the same QTL shared similar QTL intervals and peak positions. Thus, the identified QTL based on the F₂ and F_2:3 populations could be regarded as the same QTL. Additionally, the same QTL detected in different F₂ and F_2:3 populations exhibited different values of additive effect and contribution, which could be partially caused by environmental factors such as the climate and field conditions under different environments.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.