Relationship between QTL for grain shape, grain weight, test weight, milling yield, and plant height in the spring wheat cross RL4452/‘AC Domain’

Kernel morphology characteristics of wheat are complex and quantitatively inherited. A doubled haploid (DH) population of the cross RL4452/‘AC Domain’ was used to study the genetic basis of seed shape. Quantitative trait loci (QTL) analyses were conducted on a total of 18 traits: 14 grain shape traits, flour yield (Fyd), and three agronomic traits (Plant height [Plht], 1000 Grain weight [Gwt], Test weight [Twt]), using data from trial locations at Glenlea, Brandon, and Morden in Manitoba, Canada, between 1999 and 2004. Kernel shape was studied through digital image analysis with an Acurum® grain analyzer. Plht, Gwt, Twt, Fyd, and grain shape QTL were correlated with each other and QTL analysis revealed that QTL for these traits often mapped to the same genetic locations. The most significant QTL for the grain shape traits were located on chromosomes 4B and 4D, each accounting for up to 24.4% and 53.3% of the total phenotypic variation, respectively. In addition, the most significant QTL for Plht, Gwt, and Twt were all detected on chromosome 4D at the Rht-D1 locus. Rht-D1b decreased Plht, Gwt, Twt, and kernel width relative to the Rht-D1a allele. A narrow genetic interval on chromosome 4B contained significant QTL for grain shape, Gwt, and Plht. The ‘AC Domain’ allele reduced Plht, Gwt, kernel length and width traits, but had no detectable effect on Twt. The data indicated that this variation was inconsistent with segregation at Rht-B1. Numerous QTL were identified that control these traits in this population.


Introduction
Wheat (Tritcium aestivum L.) is an allohexaploid species (2n = 6x = 42) comprised of A, B, and D sub-genomes totalling~17 Gbp. Along with other important cereal crops, it has been subject to artificial selection for increased grain size since the early stages of its cultivation [1]. Size a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 capture of the sample (i.e. grain) and neural network analysis. Both average and standard deviation values for grain shape traits were calculated. A plot-wise analysis of grain traits with the Acurum1 grain analyzer permitted calculation of average values (for all of the grains per plot) that were used for detecting QTL. Standard deviation values for grain shape were included to study variability in grain size and/or shape within grain samples (i.e. possibly from tillers or fertile tertiary florets).

Plant height, grain weight, test weight, and flour yield (Plht, Gwt, Twt and Fyd)
Data on Plht was obtained from field trials at Glenlea (1998Glenlea ( , 1999Glenlea ( , and 2000 and Morden (1998Morden ( , 1999Morden ( , and 2000 in Manitoba, Canada. Gwt and Twt measurement were carried out using grain harvested from trials at Glenlea (1999 and2000) and Morden (1999 and2000) as described in McCartney et al. (2005). LS means for Gwt and Twt were used for QTL detection. Similarly, data for Fyd was collected and previously reported in McCartney et al. (2006). Grain samples were milled into straight-grade flour with a Buhler laboratory automatic-pneumatic mill (Model 202, Buhler AG, Uzwil, Switzerland) after tempering to 16.5% moisture. Flour yield was calculated based on total recovered products.

Statistical analyses of trait data
Analysis of variance (ANOVA) was conducted with the GLM procedure of SAS1 9.3 (SAS Institute Inc., Cary, North Carolina, USA) with environments, replicates, and genotypes as random effects. Heritability was calculated on an entry mean and per plot basis with the ANOVA mean squares and the expectations of mean squares. Genotype line means were calculated for the agronomic traits with the LSMEANS statement of the MIXED procedure, which calculates least-square means. In this case, genotypes were considered fixed effects, while with environments and replicates were random effects. An overall mean dataset was generated for all traits by averaging trait data over all replicates. Correlation analysis was used to investigate potential genetic relationships between the traits. Pearson's correlation coefficients were estimated between the agronomic, milling, and seed shape traits with procedure CORR of SAS1 using the DH line means from the overall mean dataset. Table 1. Grain shape traits measured on wheat grain samples with the Acurum1 grain analyzer.

Abbreviation
Trait Description

Linkage mapping and QTL analyses
Linkage and QTL mapping procedures for this experiment have been previously detailed [39]. In brief, an initial of 12,351 polymorphic markers (SSR, SNP, Diversity Arrays Technology [DArT], and ESTs) of an Illumina wheat 90K Infinium Custom beadchip [42] were screened on 193 DH progeny of the RL4452/'AC Domain' population. A total of 12,202 informative markers were used for linkage mapping with MapDisto1 [43]. Linkage groups were identified using a minimum LOD score of 4, and a maximum recombination fraction of 0.25. Recombination fractions were converted into centiMorgan (cM) map distances using the Kosambi mapping function. The RL4452/'AC Domain' linkage map is reported in S1 Table. More than one linkage group was obtained for chromosomes 1B, 2B, 3D, 5A, 5D, 6D, 7B, and 7D. For instance, there were two linkage groups for chromosome 1B. Linkage group 1B.1 consisted of the short arm and most of the long arm, and linkage group 1B.2 consisted of the distal end of the long arm. The most informative marker per linkage bin was utilized for QTL analyses (i.e. 1,055 markers were retained). QTL IciMapping software version 4.1.0.0 was used to test for additive effect and epistatic QTL from multi-year trial datasets using inclusive composite interval mapping (ICIM) [44]. Additive effect QTL were detected by ICIM (QIC module) with a walk speed of 0.1 cM. LOD thresholds were based on 1,000 permutations. The confidence interval was determined by one LOD drop-off, which approximates a 96.8% confidence interval [45]. Epistatic QTL were identified via a two-dimensional scan for mapping digenic epistasis using ICIM-epistasis (QICE module) with default LOD scores of 5.0, coupled with walk speeds of 2 cM. QTL were deemed significant if For agronomic and milling traits, QTL were reported when the peak LOD score exceeded the significance threshold determined by the permutation analyses in two or more environments. For seed shape traits, QTL were reported when the peak LOD score exceeded the significance threshold determined by permutation in a minimum of three combinations of shape traits by environment (Glenlea 2000, Brandon 2004, or meaned over both years). The phenotypic variation explained due to respective QTL was derived from marker-trait regression (r 2 ) values.

Physical locations of SNP markers
The physical locations of SNP markers were obtained with a BLASTN search against the IWGSC Chinese Spring RefSeq v1.0 database (https://urgi.versailles.inra.fr/blast_iwgsc/blast. php). The best BLAST hit for a SNP marker was reported for the chromosome to which it mapped in the RL4452/'AC Domain' DH population. The BLAST hits are reported in S1 Table. with seed length (AmaL_M, DMax_M, SzLn_M), but less so relative to kernel width. Plht and Twt were positively correlated with kernel width, but were not significantly correlated with kernel length. Flour yield (Fyd) was not strongly correlated with any trait, although statistically significant correlations were identified. Fyd was most strongly correlated with Sphericity (Sphr; r = -0.328). Rectangularity (Rect) had a strong positive correlation with Sphericity (Sphr; r = 0.986), and a strong negative correlation with Roundness (Rndn; r = -0.993). Kernel Area (Area_M) was correlated with all traits, except Rndn and Sphr. Kernel Perimeter (Per_M) was correlated with all traits, except Twt, Rect, and Rndn. Seed area-perimeter ratio (ArPe_M) was correlated with all traits to a certain degree.
ICIM-epistasis (QICE module) identified a small number of epistatic QTL, which are reported in S4 Table. For seed shape traits, epistatic QTL were detected for Area, ArPe, DMen, DMin, Rect, SzWd, variability of Area, variability of AmiL, and variability of DMen. Epistatic QTL were also detected for Plht, Gwt, Twt, and Fyd. The epistatic interaction between chromosome 1D at 38 cM and chromosome 6B at 84 cM was detected more consistently than the others. This epistatic interaction involved Gwt, Area, ArPe, and DMen, which intuitively should be correlated. Interestingly, additive effect QTL were not detected on chromosome 1D at 38 cM or chromosome 6B at 84 cM using the QIC module for any trait. The remaining epistatic interactions were not consistently identified in different seed shape traits and/or between different datasets (i.e. environments).

Chromosome 3A
A strong QTL for variability in Sphericity (Sphr) was consistently detected at 65.9 cM. The 'AC Domain' allele increased variability in this trait. No other QTL were detected on chromosome 3A.

Chromosome 4B
On chromosome 4B, a 4.5 cM interval (51.4-55.9 cM) coincided with the LOD peak locations of a major QTL for Plht (QPlht.crc-4B), Gwt (QGwt.crc-4B), and Fyd (QFyd.crc-4B) (Fig 1). In addition, a significant QTL was detected in this region of chromosome 4B for each of the 14 seed shape traits, and for variability in AMaL, AMiL, Area, ArPe, Asym, DMax, DMen, DMin, Per, SzLn, and SzWd. When considering 1 LOD drop-off confidence intervals for these QTL, this narrow genetic region corresponds to a very large~601 Mbp physical region in the IWGSC Chinese Spring RefSeq v1.0 (S1 Table). The interval includes the centromere and appears to be an area of low recombination.  (Tables 2 and 3).

Chromosome 6D
QTL for a number of grain shape traits had LOD peaks mainly located within a 2.

Chromosome 7D
The major flour yield QTL QFyd.crc-7D was not coincident with QTL for Gwt, Twt, or seed shape, but was coincident with a major maturity QTL QMat.crc-7D previously identified in this population [27]. QFyd.crc-7D was a broad QTL with the main peak located at 16.2 cM and secondary peaks located at 31.9 and 43.1 cM based on interval mapping (IM-ADD, S1 Fig).
Based on these data, it is possible that QFyd.crc-7D is the result of two or more linked genes. Interestingly, the maturity QTL QMat.crc-7D has a single peak at 19.6 cM and no secondary peaks (S1 Fig). No seed shape QTL were identified on chromosome 7D.

Discussion
The objectives of our study were to identify significant grain shape and agronomic trait QTL, and determine their interrelationships. SNPs from a wheat 90K Infinium Custom Beadchip were previously used to generate a high density linkage map of a RL4452/'AC Domain' DH population [39], which in turn was used to identify QTL for the above traits. QTL were identified on chromosomes 1A, 1B, 1D, 2B, 2D, 3A, 3B, 3D, 4A, 4B, 4D, 5B, 5D, 6B, 6D, 7A (grain shape); 4B, 4D (Plht); 2B, 3B, 3D, 4A, 4B, 4D, 6B (Gwt); 1D, 2A, 2B, 2D, 3B, 3D, 4D, 7A (Twt); and 1B, 3B, 3D, 4B, 7D (Fyd). The most significant QTL for grain shape, Plht, and Gwt were detected on chromosomes 4B and 4D. The most significant Twt QTL were identified on 3B and 4D. The most significant Fyd was located on chromosome 7D, but another important Fyd QTL coincided with the Twt QTL on chromosome 3B. A major QTL for grain shape, Gwt, Fyd, and Plht mapped to a narrow genetic region on chromosome 4B, which corresponds to the centromere and a very large portion of the chromosome. 'AC Domain' alleles contributed to a reduction in Plht, Gwt, Fyd, and grain size, in addition to negative additive effect values for grain shape traits. The same Gwt QTL was identified on 4B, and is associated with SSR marker wmc238 [27], located at 51.9 cM. In our study, wmc238 was located 0.5 cM from Tdurum_contig5562_441, positioned at 52.4 cM on chromosome 4B [39]. Further, the Plht QTL of both studies mapped essentially to the same position based upon the linked SSR marker gwm513 that co-segregates with TA003708-0300. Markers gwm513 and TA003708-0300 were 0.6 cM apart from the grain shape and Gwt QTL peak SNP marker Tdurum_contig5562_441 of our study. Therefore, all SNP and SSR markers within this narrow region on 4B may be useful for MAS of grain size, Gwt, and Plht traits in germplasm and breeding material. QPlht.crc-4B, QGwt.crc-4B, QFyd.crc-4B, and the many seed shape QTL in this region overlap with the 'QTL region 15' in the cross ND705/PI 414566, which affects Twt, Gwt, kernel area, and kernel length [10]. The grain shape, Fyd, and Gwt QTL in the RL4452/'AC Domain' are likely the result of pleiotrophy of the reduced plant height 'AC Domain' allele at QPlht.crc-4B.
The plant height QTL QPlht.crc-4B was previously believed to be the result of segregation at the Rht-B1 locus [27]. However, the improved SNP-based linkage map revealed that QPlht.crc-4B mapped proximal of the expected location of Rht-B1 (possibly on 4BL), which could not be resolved based on the older SSR map [27]. It is important to note that the RL4452/'AC Domain' mapping population was also monomorphic for the KASP marker wMAS000001, a diagnostic marker for Rht-B1. Rht-B1 is also physically located outside the confidence interval for these QTL. QPlht.crc-4B and the other linked/pleiotropic QTL must not be the result of segregation at the Rht-B1 locus. Based upon the BLAST locations of the SNP markers in the 4B linkage map and 1 LOD drop-off confidence intervals for these QTL, this region contains 2,979 high confidence genes in the IWGSC Chinese Spring RefSeq v1.0. Additional research is needed to identify candidate genes responsible for these QTL.
The genetic interval on chromosome 4D flanked by SSR markers wmc617 and wmc48 was found to carry the most significant QTL for seed shape, Plht, Gwt, and Twt. Rht-D1 mapped to this centre of this region as indicated by the diagnostic SNP marker wMAS000002. RL4452 carries the dwarfing allele Rht-D1b, which reduced Plht, Gwt, Twt, grain width, and Area, but had no detectable effect on Fyd. Rht-D1b also negatively impacted the grain shape traits Rect and Sphr, and the net result of these changes in kernel shape was a reduction in Twt. Our findings are in agreement with those previously reported for the same RL4452/'AC Domain' population mapped using 369 SSR markers [27,46]. Based on these results, it is likely that the variation in seed shape near Rht-D1 is due to its pleiotropic effects.
Chromosome 4A is particularly interesting in this population. Three QTL regions were identified in this study affecting seed size and shape. QTL for grain length (AMaL, DMax, SzLn), Per, Rect, and Sphr mapped to 38 cM on chromosome 4A. These QTL were supported by the identification of the same QTL region (Twt, Gwt, kernel area) in the cross ND705/PI 414566 (Twt, Gwt, kernel area) [10] and in Synthetic/Opata (vertical perimeter) [7]. QTL for grain width (AMiL, DMin) and variability for kernel parameters within samples were detected at 63 cM. Likewise, a QTL for length-width ratio (QKLWR.ndsu.4A.1) was detected in the same region [10]. Finally, a QTL for grain weight QGwt.crc-4A mapped to 90 cM along with numerous grain shape parameters in the RL4452/'AC Domain' population. This was also supported by a second length-width ratio QTL (QKLWR.ndsu.4A.2) in the ND705/PI 414566 population [10]. Surprisingly, there were no QTL for Twt detected on chromosome 4A in the RL4452/'AC Domain' population.
Another interesting locus in the RL4452/'AC Domain' population is located on chromosome 7D (linkage group 7D.2). The most important Fyd QTL QFyd.crc-7D mapped to a large interval with predicted locations at 14.4, 23.8, and 43.3 cM based on ICIM. Interval mapping showed multiple peaks in each environment (S1 Fig). Also in this region is a major, days to maturity QTL at 19.6 cM [27,39], which did not have any secondary peaks (S1 Fig). The presence of multiple peaks for Fyd suggested that multiple genes controlling the trait could be located in this region. One of the genes affecting Fyd could be the maturity QTL itself. It was hypothesized that this maturity QTL was responsible for a falling number QTL in this region since the parental allele contributing the beneficial additive effect varied between growing environments [39]. Presumably the weather conditions affecting pre-harvest sprouting (i.e. rain and high humidity) varied in different growing seasons. In addition, QFyd.crc-7D overlaps with QTKW.ndsu.7D, QTW.ndsu.7D, QKW.ndsu.7D, and QKLWR.ndsu.7D [10]. Additional research is needed to clarify the genetic control of these traits at this point in the genome.
A number of other QTL detected in the RL4452/'AC Domain' population were also in common with grain shape and size QTL detected in the cross ND705/PI 414566 [10]. QFyd.crc-1B overlaps with QTL region 2 (QKW.ndsu.1B, QLWR.ndsu.1B.2) [10]. The QTL QTwt.crc-2D, QRndn.crc-2D, and QSphr.crc-2D likely overlap with the QTL QTW.ndsu.2D.1 and QKLWR. ndsu.2D.1. QGwt.crc-3D likely overlaps with the thousand kernel weight QTL QTKW.ndsu.3D and the kernel area QTL QKA.ndsu.3D. Grain width QTL (QDMin.crc-5B and QSzWd.crc-5B) overlapped with kernel area and length QTL. Grain length (AMaL, DMax, SzLn), Asym, and Per QTL of our study mapped to approximately 149 cM on the 7A linkage group, which is consistent with QTL for Gwt, kernel length, width, and area [10]. Similarities between QTL on chromosomes 4A, 4B, and 7D in these two populations have already been discussed in the preceding paragraphs. These similarities likely result from shared parentage. 'AC Domain' has the pedigree ND499/RL4137//ND585. ND499, ND585, and ND705 are all wheat lines developed by North Dakota State University, so 'AC Domain' and ND705 are likely to share many of the same alleles. In addition, the kernel shapes of 'AC Domain' and ND705 are similar (short plump kernels), while the kernels of RL4452 and PI 414566 are relatively longer.

Conclusions
This study identified significant QTL for grain morphology, plant height, grain weight, test weight, and flour yield. Previous QTL studies to identify grain shape have utilized SSRs, DArTs, and other PCR-based markers in segregating populations. In deploying a combination of wheat 90K Infinium SNPs and landmark SSRs, we have been successful in enhancing the marker density on the RL4452/'AC Domain' linkage map, and in defining QTL relative to this improved genetic map and the Chinese Spring pseudomolecules. The association between Plht, Gwt, Twt, Fyd, and grain shape QTL confirmed past findings. Genetic analysis of kernel image analysis showed promise, and uncovered additional variation for Gwt, Twt, and Fyd. The per plot heritability estimates were higher for the grain shape traits than Gwt and Fyd, and grain shape QTL were identified that were not associated with Gwt, Twt, and Fyd. Our results should also provide a consensus on the location of linked SNPs and landmark SSRs across maps, which in turn might enable validation of these grain shape QTL in other populations. SNP markers associated with the above traits might also be useful for MAS, and in the identification of candidate genes from rice or other monocots.