Three dominant awnless genes in common wheat: Fine mapping, interaction and contribution to diversity in awn shape and length

The awn is a long needle-like structure formed at the tip of the lemma in the florets of some grass species. It plays a role in seed dispersal and protection against animals, and can contribute to the photosynthetic activity of spikes. Three main dominant inhibitors of awn development (Hd, B1 and B2) are known in hexaploid wheat, but the causal genes have not been cloned yet and a genetic association with awn length diversity has been found only for the B1 allele. To analyze the prevalence of these three awning inhibitors, we attempted to predict the genotypes of 189 hexaploid wheat varieties collected worldwide using markers tightly linked to these loci. Using recombinant inbred lines derived from two common wheat cultivars, Chinese Spring and Mironovskaya 808, both with short awns, and a high-density linkage map, we performed quantitative trait locus analysis to identify tightly linked markers. Because this linkage map was constructed with abundant array-based markers, we converted the linked markers to PCR-based markers and determined the genotypes of 189 hexaploids. A significant genotype-phenotype correlation was observed at the Hd and B1 regions. We also found that interaction among these three awning inhibitors is involved in development of a membranous outgrowth at the base of awn resembling the Hooded mutants of barley. For the hooded awn phenotype, presence of the Hd dominant allele was essential but not sufficient, so B2 and other factors appear to act epistatically to produce the ectopic tissue. On the other hand, the dominant B1 allele acted as a suppressor of the hooded phenotype. These three awning inhibitors largely contribute to the genetic variation in awn length and shape of common wheat.


Introduction
The awn is a long needle-like structure formed on the distal end of the lemma in the florets of some grass species such as wheat, barley and rice. This extension of the lemma seems to be a modified leaf blade [1] and can serve to protect against animals [2]. The presence of silicified hairs on its surface facilitates seed dispersal by adhesion to animal fur [3]. The dispersal unit of wild wheat bears two pronounced awns that balance each unit as it falls, and the movements of the two awns, driven by the daily humidity cycle, propel the seeds into the soil [4].
Awns also contribute to the photosynthetic activity of the inflorescence in wheat and barley. In tetraploid wheat (Triticum turgidum L. subsp. durum) the total surface area of the awns can exceed that of the leaf blade [5], and more than half of the total number of stomata of wheat spikelets is present in the awns [6]. Because the pathway for assimilate movement from awns to the kernels is minimal, the awns can be considered as an ideal place for light interception and CO 2 uptake [7]. These observations may contribute to the observation that the presence of awns can double the net rate of spike photosynthesis [8]. Although awns have only a limited effect on yield in wet climates, their contribution is significantly higher than in awnless or deawned wheat varieties in drier conditions [9].
During domestication of rice, humans have selected awnless varieties because of the convenience of seed collection and the ease of seed storage. In contrast to barley and wheat, rice awns do not seem to contribute to the photosynthetic activity of the panicle, since they have no chlorenchyma [10]. Many genes involved in awn formation and elongation have been identified in rice, such as a basic helix-loop-helix transcription factor named Awn-1 (An-1) [11], a YABBY transcription factor (DROOPING LEAF [DL]) [12], an auxin responsive factor, OsETTIN2 [12], LONG AND BARBED AWN1 (LABA1), encoding a cytokinin-activating enzyme [13], and REGULATOR OF AWN ELONGATION 2 (RAE2), which is a member of the EPIDERMAL PATTERNING FACTOR-LIKE (EPFL) family [14].
In barley, short awn 2 (Lks2), located on chromosome 7H and encoding a SHORT INTERNODES (SHI) family transcription factor, has been identified as a causal gene for short awns in Eastern Asian accessions [15]. The dominant Hooded mutation (K) is also known in barley, transforming the awns into an extra flower of inverse polarity on the lemma, which is caused by a 305 bp duplication in the homeobox gene HvKnox3 [16,17].
In hexaploid common wheat (Triticum aestivum L.), three dominant inhibitors of awn development are known as Hooded (Hd), Tipped1 (B1) and Tipped2 (B2) [18,19], which are respectively located on chromosome arms 4AS, 5AL and 6BL [20]. In wheat varieties with the dominant Hd allele, awns are reduced in length and are curved and twisted near the base. In some cases, the awn is considerably broadened at the base and could possess membranous lateral expansions, resembling the Hooded mutants of barley. The dominant B1 mutation produces short awns at the base and middle of the spike, but the length increases toward the top of the spike and may reach 1 cm. These awn tips are usually straight and unbent at the base. In contrast, in B2 mutants, the awn length is nearly equal all along the spike, and the longest ones are found near the middle of spike. The B2 mutant awns are often gently curved but are never bent around themselves as in Hd, and do not produce membranous lateral outgrowths [18]. Wheat varieties homozygous for the three recessive hd, b1 and b2 alleles are fully awned, while those with two dominant inhibitors such as HdHdb1b1B2B2 and hdhdB1B1B2B2 are awnless [19]. Using a doubled haploid population derived from two common wheat varieties, Chinese Spring (CS) with the HdHdb1b1B2B2 awnless genotype and Courtot with the hdhdb1b1b2b2 fully awned genotype, quantitative trait locus (QTL) analysis was performed for awn length, and the chromosomal locations of Hd and B2 were determined [21]. B1 is a phenotypic marker on wheat genetic maps [22][23][24]. In a previous association study using a panel of hexaploid wheat mainly composed of UK varieties, the presence of the dominant B1 allele was found [25]. However, the prevalence of Hd and B2 alleles in hexaploid wheat varieties was unknown. In our previous study, an array-based genotyping system was developed and a high-density genetic map was constructed using recombinant inbred lines (RILs) derived from a cross between common wheat varieties CS and Mironovskaya 808 (M808) [26]. M808 presents a phenotype similar to that of the B1 mutant, with extensive short awns at the base and middle of the spike but longer awns at the top. Using this mapping population, we aimed to identify markers tightly linked to the three awning inhibitor loci by performing QTL analysis and to then convert the linked array-based markers to PCR-based markers for genotyping 189 hexaploid wheat varieties to predict the presence of the Hd, B1 and B2 alleles. We also performed QTL of the membranous lateral outgrowth formation (hooded phenotype). The observed genetic interactions among the awning inhibitors and their contribution to awn shortening and the hooded phenotype are discussed.

Plant materials and awn length evaluation
A mapping population of 210 RILs derived from a cross between two common wheat cultivars, CS and M808, developed by Kobayashi et al. [27], and 189 of the 192 hexaploid wheat lines (S1 Table) of the National BioResource Project (NBRP)'s hexaploid wheat core collection [28] (http://shigen.nig.ac.jp/wheat/komugi/) were used in this study. The plants were grown individually in pots arranged randomly in a field of Kobe University (348430N, 1358130E) in the 2012-13 season for RILs and the 2013-14 season for the hexaploid wheat core collection. Total DNA from each RIL and its parents was extracted from leaves using standard procedures.
The awn length of the three main spikes of one plant for RILs and two plants for the hexaploid collection was measured at the top and middle of the spike at flowering. We also evaluated the hooded phenotype (Fig 1C and 1D) as the presence (trait value of 1 for QTL analysis) or absence (trait value of 0) of a membranous structure at the base of the awn. Some RILs also presented a broadening of the base of the awn (Fig 1E), and they were also considered as hooded for QTL analysis.

QTL analysis
QTLs were analyzed by composite interval mapping using R/qtl package version 1.21-230, and the high-density linkage map had been previously constructed using the RIL population derived from CS and M808 [26]. First, the QTL genotype probability was calculated using the function calc.genoprob with a step size of 1 cM and the Kosambi map function. QTL analysis was performed with the cim function using three marker covariates and the Kosambi map function. The log-likelihood (LOD) score threshold was determined by computing a 1,000 permutation test for all traits. A multiple QTL model and epistasis were evaluated using the fitqtl function and the QTL locations were refined using the refineqtl function. The percentage of phenotypic variation (PV) explained by a QTL for a trait and epistasis were estimated using the fitqtl function in R/qtl.

Conversion of array-based markers to PCR-based markers
For conversion to PCR-based markers, the read or contig sequences of the array-based markers located around the peak of the QTLs were used as a query to perform a BlastN search against the survey sequence of common wheat (S1 Fig) [26,29]. The read or contig sequences of the array-based markers with the corresponding blast hit were aligned to search for putative PstI and/or BstNI sites. If no PstI or BstNI sites were observed, single nucleotide polymorphisms (SNPs) and indels (insetions/deletions) between CS and M808 were used to develop cleaved amplified polymorphic sequence (CAPS) markers. If neither PstI/BstNI sites nor SNPs were found, simple sequence repeat (SSR) motifs were searched for using SciRoKo ver. 3.4 [30]. Primers were designed using the Primer3Plus module (S2 Table) [31]. Electrophoresis of the PCR products and digested fragments was performed in 2% agarose gels for CAPS and indel markers, and 13% non-denaturing polyacrylamide gels were used for SSR markers. For polyacrylamide gel electrophoresis, the high efficiency genome scanning system (Nippon Eido, Tokyo, Japan) of Hori et al. [32] was used. Genotyping of RILs and the wheat core collection were performed using the markers developed. A linkage map was constructed using 161 RILs with predicted genotypes at the Hd, B1 and B2 loci and MapDisto ver. 1.7.7 software [33], and drawn by MapChart ver. 2.30 software [34].

QTL analysis of awn-related traits
CS showed a characteristic curved short awn ( Fig 1A), with an awn length of 6.69 ± 0.75 mm at the top and 4.06 ± 0.55 mm at the middle of the spike. In contrast, the awns of M808 were relatively straight and short at the middle of the spike (5.05 ± 0.95 mm) but relatively long (30.45 ± 7.79 mm) at the top of the spike (Fig 1B). In a RIL population derived from CS and M808, the awn length at the middle of spike ranged from 0.88 to 81.93 mm with an average of 15.30 ± 21.35 mm, and the length ranged from 0.98 to 78.42 mm at the top of spike with an average of 18.05 ± 19.76 mm. Many of the RILs showed an awn length 5 mm (Fig 1H).
We performed QTL analysis for awn length using a high-resolution map constructed in our previous study [26] and identified three main QTLs for awn length at the top and middle of the spike located on chromosomes 4A, 5A and 6B ( Table 1). The 4A QTL, corresponding to Hd locus, was located around 20 cM with a LOD score greater than 38 and explained over 23.1% of the phenotypic variation in the RIL population. The LOD score of the 5A QTL (the B1 locus), located at the telomeric region of the long arm (306 cM), was higher than 47 and explained more than 30.1% of the phenotypic variance. The 6B QTL, corresponding to B2 locus, was identified in the centromeric region with a LOD score of around 50, and around 30% of the phenotypic variation could be explained by this QTL. The CS allele at the 4A and 6B QTLs and the M808 allele at the 5A QTL decreased the awn length.
Significant QTL x QTL interactions (epistasis) were also observed among these three loci ( Table 1, Fig 2). Interaction between the 4A and 5A QTLs had the greatest effect on awn length at the top of the spike, with a LOD score of 19, and explained 8.3% of the phenotypic variation. 4A x 6B and 5A x 6B interactions exhibited a LOD score of 6.7 and 5.9, respectively, explaining 2.5 and 2.2% of the phenotypic variation. On the other hand, 4A x 5A and 5A x 6B interactions exhibited a great effect on awn length at the middle of the spike, with a LOD score higher than 21, and could explain more than 10% of the phenotypic variation.
Comparing the genotype data of flanking markers at the three QTLs with the phenotype data, we identified RILs with dominant Hd (CS allele at the 4A QTL), B1 (M808 allele at the 5A QTL) and B2 (CS allele at the 6B QTL) alleles and recessive wild type (WT) alleles (S3 Table). We found 17 RILs with WT alleles at the three loci (hdhdb1b1b2b2), nine RILs with only the dominant Hd allele (HdHdb1b1b2b2), 13 RILs with only the dominant B1 allele (hdhdB1B1b2b2) and 21 with only the dominant B2 allele (hdhdb1b1B2B2). In the case of RILs with two dominant awning inhibitor genes, 24 RILs had Hd and B1, 22 had Hd and B2, and 27 had B1 and B2 alleles. On the other hand, 28 RILs with all three awning inhibitor genes were found. There were no significant differences in awn length at the top of spike between RILs with two or three awning inhibitor genes, except for those containing Hd and B1, which had slightly longer awns ( Fig 3A). RILs containing only one inhibitor gene exhibited intermediate awn length at the top compared with WT and lines containing at least two inhibitors. In contrast, no differences were observed in awn length at the middle of the spike among individuals with two or three awning inhibitors (Fig 3B), or between individuals with HdB1 (HdHdB1B1b2b2) or B1 (hdhdB1B1b2b2) genotypes. RILs with Hd (HdHdb1b1b2b2) and B2 (hdhdb1b1B2B2) genotypes presented an intermediate awn length, as observed at the top of the spike. Based mainly on awn length at the top of the spike, the RILs can be grouped into WT, lines containing one inhibitor, and individuals with at least two inhibitors (Fig 3A and  3C).
In this RIL population, the hooded phenotype ( Fig 1C to 1E) was also observed in some individuals (31 out of 210 RILs) but not in the parents. For QTL analysis, we assigned a trait value of "1" to the RILs with a membranous structure (Fig 1C and 1D) or broadening of the base of awn (Fig 1E), and "0" to RILs without a membranous structure or broadening because it was difficult to quantify this trait. QTL analysis revealed that the three QTLs for awn length are also involved in development of the hooded phenotype (Table 1). These loci explained 31.2 (4A QTL), 14.6 (5A QTL) and 13.0% (6B QTL) of the phenotypic variation. CS alleles at the three QTLs contributed to appearance of the hooded phenotype. Significant epistasis was also observed among these QTLs (Table 1, Fig 2G to 2I), for which the effects of 4A x 5A and 4A x 6B interactions were higher than 5A x 6B interaction. There were no differences in the allelic effects of the 5A and 6B QTLs when the genotype of 4A QTL was fixed to the M808 allele, and the hooded phenotype was not expressed under these conditions. Therefore, this phenotype might only be observed when an individual contains a CS allele at the 4A QTL, and also the 5A and 6B QTLs. Indeed, in the 161 RILs with known genotypes, 17 of the 22 hooded RILs contained homozygous Hd and B2 alleles (Fig 3D), and formed a membranous outgrowth (Fig 1C  and 1D). The remaining five hooded RILs presented HdHdB1B1b2b2 or HdHdB1B1B2B2 genotypes and only a broadening of the base of awns was observed, without formation of a Two-way interaction plots for the three identified QTLs. Two-locus genotypic effects for awn length at the top (A to C) and middle (D to F) and for the hooded phenotype (G to I) were plotted using the genotype data of the markers with the highest LOD score. Red lines represent a genotype homozygous for CS and blue lines represent a genotype homozygous for M808. Error bars are ± standard errors. https://doi.org/10.1371/journal.pone.0176148.g002 Three dominant awnless genes in common wheat membranous structure (Fig 1E). In a near-isogenic line (NIL) (HdHdb1b1b2b2) of common wheat cv. S-615 (hdhdb1b1b2b2) with Hd from CS (named Hd-S615), broadening of the base of awns could be observed (Fig 1F) and sometimes a membranous structure was also formed without the presence of B2 locus (Fig 1G).

Location of candidate genes on wheat genome
To test whether the genes known to be involved in awn development in rice and barley could be causal genes of the Hd, B1 and B2 loci, we performed a BlastP search against wheat protein sequences of the EnsemblPlants Triticum aestivum database (http://plants.ensembl.org/ Triticum_aestivum/Info/Index). An ortholog of DL was located on chromosome 4AS and an ortholog of RAE2 on chromosome 6BL (S4 Table). To check their location, contig sequences of the array-based markers were mapped on wheat genomic scaffolds and then the genes contained in these scaffolds were searched against proteins in the rice (http://rapdb.dna.affrc.go. jp/), barley (http://plants.ensembl.org/Hordeum_vulgare/Info/Index), or both databases using the Blast algorithm. Based on the synteny between wheat and the barley and rice genomes, the locations of these two genes were predicted on the wheat genome.
Nine rice genes were found as orthologs of the wheat genes contained in the genomic scaffolds, and DL seemed to be located around the Hd locus (Fig 4). Because only six barley orthologs were found, it was difficult to compare this genomic region between barley and wheat. Alignment of the DL homologs indicated that this gene on chromosome 4A is functional and that the differences in the protein sequence were no greater than for other wheat genomes and other species (S2 Fig). Similarly, the RAE2 location was predicted based on the synteny with barley using 12 orthologs (Fig 5). This result indicated that RAE2 was near but not at the B2 locus. Wheat and barley group 6 chromosomes are syntenic to rice chromosome 2 [35,36], but the ortholog of RAE2 (located on chromosome 8 of rice) in barley and wheat was located on the long arm of group 6 chromosome, distal to the B2 locus. We did not compare the results for rice because no orthologs of rice chromosome 8 were found in this B2 region.

Variation of awn length in the hexaploid wheat core collection
A hexaploid wheat core collection, established by NBRP-Wheat [28], included 161 common wheat (T. aestivum subsp. aestivum) varieties and the following subspecies of T. aestivum: compactum (n = 6), macha (n = 4), spelta (n = 13), sphaerococcum (n = 3) and vavilovii (n = 2), which were collected worldwide (S1 Table). The awn length at the top of the spike ranged from 0. 16  Of the six subspecies analyzed, aestivum, compactum and spelta showed wide variation in awn length (Fig 6). In contrast, macha, sphaerococcum and vavilovii exhibited shorter awns than other subspecies. In vavilovii, awn length at the top was longer than at the middle of the spike. On the other hand, awns at the middle of the spike were longer in macha. Hexaploid wheat varieties from Jordan, the USA, Iraq, India, Greece, Spain, Iran, Romania, Mexico and Egypt mostly presented a long awn at the top and middle of the spike (S4 Fig), and varieties from Bhutan, Italy, the UK, Australia and Syria presented short awns throughout the spike. The awn length of wheat varieties from Canada and Lebanon were short at the middle but long at the top of the spike. In contrast, varieties from Georgia presented relatively short awns at the top but long awns at the middle of the spike.

Conversion of array-based markers to PCR-based markers
To determine the genotype of the NBRP wheat core collection at the three QTLs, we attempted to convert the array-based markers to PCR-based ones. First, we selected 40 array-based markers around the peak of each QTL (21 markers from chromosome 4A, seven from 5A and 12 from 6B). Using the read or contig sequence of each marker, we performed a BlastN search against the hexaploid wheat genomic survey sequence (S1 Fig). No hits were found for two reads/contigs from chromosome 5A and one from 6B. Because the array-based markers were developed using genomic fragments with PstI sites at both ends and without any BstNI sites within the fragment [26], putative PstI sites were searched at both ends of the alignment of the read/contig sequence and genomic scaffold. CAPS markers were developed using the genomic scaffold with PstI sites. If PstI restriction sites were not found, SSR motifs were searched for in the genomic scaffold to develop SSR markers when the read or contig was derived from CS. For sequences derived from M808, we searched for SNP, indel and SSR motifs to develop PCR-based markers. Finally, we checked for polymorphisms in 31 of the markers developed (12 from chromosome 4A, two from 5A and 17 from 6B) of which 12 (four from chromosome chromosome 3 (top of right panel), we located DL near the Hd locus. On the other hand, this region of chromosome 4A is known to have an inversion with respect to the corresponding region of chromosome 4H (bottom of right panel). Therefore, the ortholog of DL in barley might be around the Hd locus, but the exact location could not be determined.
https://doi.org/10.1371/journal.pone.0176148.g004 Three dominant awnless genes in common wheat 4A, one from 5A and seven from 6B) allowed detection of polymorphisms between CS and M808. Using these markers, RILs derived from CS and M808 were genotyped and genetic maps were constructed to confirm their locations. All 12 markers were located near the awning inhibitor loci (S5 Fig). Prediction of Hd, B1 and B2 alleles in the wheat core collection Using the 12 markers developed and two publicly available SSR markers (gwm192 on chromosome 4A and gwm291 on 5A), 189 hexaploid wheat lines of the NBRP core collection were  Table). The SSR markers gwm291 and WABM229716 were discarded because of the presence of multiple alleles, and the band pattern was similar to CS or M808 in only a few accessions. Based on the identified genotypes, the average awn length was compared between varieties with the CS and M808 alleles. Awn length was significantly lower in varieties with the CS allele at the markers for chromosome 4A, except for WABM241105, for which the opposite effect was observed (S6 Table). Because a greater difference in awn length was associated with marker WABM233735, individuals with the CS allele at this marker were selected and their phenotype was compared with that of 161 RILs with known genotypes at the three awning inhibitor loci. The phenotype of 23 wheat varieties with the CS allele coincided with RILs containing at least one awning inhibitor gene, but not with that of WT (Fig 7A). Of the 23 varieties, six seem to have at least one additional inhibitor of awn development. The 23 hexaploids were mainly composed of Asian wheat varieties (Table 2), and the subspecies aestivum was the most abundant, with 17 varieties, followed by all four macha varieties, one of sphaerococcum and one of compactum (Table 3).
For markers linked to the B1 locus, only one could be used for genotyping of the wheat core collection, and a significant difference in awn length at the top of the spike was observed between lines with the CS and M808 alleles (S6 Table). Out of 49 wheat varieties, 24 with the M808 allele at marker WABM232824 presented a phenotype similar to WT (Fig 7B). We considered these lines as WT if the awn length at both the top and middle was greater than 60 mm based on the observation that a RIL with the B2 genotype had a respective length of 56.04 mm and 57.73 mm at the top and middle of the spike (the longest among the non-WT varieties). Six varieties with a WT-like phenotype but with a length less than 60 mm were considered to have an unknown genotype. Of the remaining 45 varieties, six seem to have at least one additional gene inhibiting awn development. The dominant B1 allele seems to be present mainly in hexaploid varieties from the UK and other countries such as Australia, Germany, Japan and Turkey (Table 2), and was found only in the subspecies aestivum and spelta (Table 3). Although one hexaploid presented the CS allele at WABM233735 of chromosome 4A and the M808 allele at WABM232824, the awn length was 45.43 mm at the top and 55.18 mm at the middle of the spike, a phenotype longer than expected.
In one marker linked to B2 on chromosome 6B (WABM125872) awn length at the top of the spike was significantly higher (P = 0.015) in individuals with the CS allele, which was the opposite of the expected effect (S6 Table). Thus, we were not able to predict the presence of the B2 dominant allele in the NBRP wheat core collection. Based on the criterion mentioned above, 61 varieties belonging to subspecies aestivum, six to spelta and three to compactum were classified as WT ( Table 2, Fig 7C). Most of the hexaploid varieties from the USA, Iran, Spain and Romania exhibited an awn length similar to the WT of the RIL population (Table 3). In total, the allele composition of 71 varieties with short awns could not be explained using the markers analyzed, including most of subspecies sphaerococcum and vavilovii (Table 3, Fig 7D). Most of the hexaploid varieties from Japan and USSR remained categorized as as unknown genotype.

Discussion
Although important roles of awns in spike photosynthetic activity and seed dispersal have been suggested in wheat, the genes involved in awn development have not been identified. In this study, the dominant inhibitors of awn development, Hd, B1 and B2, were mapped to a high-density genetic map of a RIL population derived from CS and M808 [26], which indicated that the genotypes of CS and M808 are respectively HdHdb1b1B2B2 and hdhdB1B1b2b2. Sourdille et al. [21] suggested that the genotype of CS is HdHdB1B1B2B2 based on the observation that the CS deletion line 5AL-10 was slightly awned (bearded). This deletion line lacks the telomeric region of the long arm of chromosome 5A, where the break point is located between two SSR loci, Xgwm156 and Xgwm617 [37] (S6 Fig). However another CS deletion line, 5AL-17 (which has the break point between Xcfa2163 and Xcfa2155), showed an awnless phenotype [21]. If CS contains the B1 allele, these earlier results indicate that the B1 locus should be located between the break point of 5AL-10 and 5AL-17 (between the two SSR loci, Xgwm156 and Xcfa2155). We found that B1 is located distal to the break point of 5AL-23, near the telomere, as previously reported [38]. Our results indicate that CS has the b1 homozygous allele, as previously reported [39].
QTL analysis for awn length indicates that the Hd locus has a large effect at the top of the spike and the B1 locus at the middle of the spike. Significant genetic interactions among the three awning inhibitors were also observed, mainly between Hd and B1 for awn length at the top of the spike and between B1 and other two loci at the central part of the spike. These observations imply that the awn length of individuals with two or more inhibitors cannot be explained only by the additive effect of each dominant allele, and that the combination of awning inhibitor alleles at different loci potentiates the suppression of awn elongation. A stable awnless phenotype can only be observed when all three dominant alleles are present (Fig 3).   Because the linkage map used in this study was constructed using array-based markers, we selected markers tightly linked to the three loci and attempted to convert them into PCRbased markers. Using these PCR-based markers, 189 hexaploid wheat varieties with a wide variation in awn length were genotyped. A significant association was observed between the genotype of the marker WABM233735 (tightly linked to Hd) and the phenotype (S5 Fig, S6 Table). The phenotype of hexaploid wheat varieties with the CS allele overlapped the phenotype of RILs with at least one awning inhibitor (Fig 7A). Many of these hexaploids were cultivated in Asia (Table 2), and this is consistent with the report that some Chinese varieties present curved awns with a reduced length (classified as hooded bearded) or very short hook-shaped awns (hooded beardless) [18].
Although WABM232824 was not necessarily tightly linked to B1, hexaploids with the M808 allele at this marker tended to have shorter awns than those with the CS allele. Around 50% of these wheat lines with the M808 allele presented a WT-like phenotype (Fig 7B), indicating that there is no strong linkage disequilibrium between WABM232824 and B1. In a previous study, association between the B1 locus and a short awn phenotype was observed in a panel of 64 wheat varieties of predominantly UK origin [25], which is consistent with our prediction that all of the UK varieties with short awns contained the B1 allele. Many markers tightly linked to B2 were also found on the linkage map. However, we could not observe any significant association between phenotype and genotype in the hexaploid wheat core collection. Because B2 is located near the centromeric region, where the recombination rate is low [40], many markers appeared to be closely linked to B2 in our mapping population. In a population derived from a biparental cross, QTL mapping is performed based on recent recombination events. However, historic recombination is used to assess phenotype-genotype correlation in a natural population. Therefore, recombination events might have occurred between B2 and the B2-linked markers analyzed in the hexaploid core collection.
At the subspecies level, many T. aestivum subsp. aestivum varieties were found to have B1 or Hd alleles. However, 62 of the 100 varieties with short awns remained classified as of unknown genotype. In the subspecies spelta, around half of the lines with short awns were predicted to contain the B1 allele. However, we could not determine the genotype of vavilovii accessions or of many accessions of sphaerococcum and compactum. Other genes may be involved in the reduced awn length in the hexaploids with unknown genotype. In addition, there are no reports on Hd alleles in compactum, macha or sphaerococcum. This indicates that the marker WABM233735 is not tightly linked to Hd. Therefore, markers that represent strong linkage disequilibrium to the Hd, B1 and B2 loci should be developed to precisely determine the genotypes of comprehensive collections of hexaploid wheat accessions.
In our mapping population, RILs with the hooded phenotype were also observed. We found that in addition to the Hd locus, B1 and B2 are involved in the development of the hooded phenotype. The dominant B2 allele seems to be important for the formation of the membranous outgrowth at the base of the awn in individuals with the dominant Hd allele (Figs 2H and 3). Although CS contains the Hd and B2 alleles, it did not develop this phenotype, indicating the involvement of another locus or loci with M808 allele in the RIL population that could not be identified in our QTL analysis. In contrast, in the NIL of S-615 with the Hd allele (HdHdb1b1b2b2 genotype), the membranous structure was also observed in some cases ( Fig  1G). Moreover, Watkins and Ellerton [18] stated that hoodedness is considerably exaggerated in late tillers. These observations suggest that Hd is essential for expression of hoodedness, and that other genetic factors (such as B2) and developmental stage-dependent factors potentiate the stable development of this membranous structure (Fig 8). To identify these genetic factors, the hooded phenotype must be better quantified, since we only used trait values of "1" and "0" for QTL analysis. The hooded phenotype, in sensu stricto, is the formation of a membranous lateral expansion, but we also considered the broadening of the base of the awn as a hooded trait based on our hypothesis that this is an intermediate phenotype of hooded. However, other intermediate phenotypes or different degrees of phenotype expression might exist. On the other hand, the dominant B1 allele seems to act as a suppressor of membranous outgrowth formation (Figs 2G and 3), and only a broadening of the base sometimes occurs when the Hd allele is present. This relationship is similar to Kap (Hooded) and Lks2 or suK (suppressor of Kap) in barley, where Lks2 and suK act as suppressors of the Hooded phenotype [41,42].
Several genes involved in awn development have been reported in rice and barley [11][12][13][14][15]17]. Although DL and RAE2 are respectively located near the Hd and B2 loci, neither seems to be a causal gene in wheat. In barley, the Hooded mutation leads to ectopic expression of HvKnox3 at the tip of the lemma, producing an extra floret instead of an awn [17]. However, its ortholog in wheat, Wknox1a, is located on the long arm of chromosome 4A (S4 Table; [43]), not on the short arm, where the Hd locus is located. Although ectopic expression of Wknox1 was confirmed in the lemma of the Hd NIL of S-615 [43], no structural mutation related to the hooded phenotype has been identified at the Wknox1a locus of CS and no Three dominant awnless genes in common wheat functional or transcriptional differences were found among the three Wknox1 homoeologs of CS [44]. Thus, Wknox1a on chromosome 4A appears not to be the causal gene of Hd.
These observations indicate that the awning inhibitors in wheat are not orthologs of known genes involved in awn development. Further study is needed to identify the causal genes. Using a high-resolution genetic map, we found markers tightly linked to the three main genes involved in inhibition of awn development in common wheat. Although there was incomplete linkage between these markers and the awning inhibitor genes, we found a significant correlation between genotype and phenotype at the Hd and B1 loci in the hexaploid wheat core collection. We also found that the dominant Hd and B2 alleles are also involved in the stable development of membranous structures at the base of awns and that the B1 allele can suppress the hooded phenotype. The cloning of these genes might clarify the molecular mechanisms of awn formation and elongation and the development of the hooded phenotype in wheat.

S1 Fig. Schematic of conversion of array-based markers to PCR-based markers.
First, BlastN searches of reads/contig sequences of the markers were performed against the wheat survey genome sequence. The aligned sequences were searched for PstI restriction sites for development of CAPS markers. If no PstI sites were found, SSR motifs were searched and primers were designed to amplify these motifs. If no SSR motifs were present in the genomic sequence and if the read/contig sequence was derived from M808, SNPs and indels were searched because the genome sequence was derived from CS. In total, 12 markers were developed and used for further analyses.   Table. Primers used in this study. Ã These two SSR markers have been previously described (http://wheat.pw.usda.gov/GG3/). (PDF) S3 Table. Identification of RILs with the dominant Hd, B1 and B2 alleles at the three QTLs. The presence of the Hd, B1 and B2 alleles were estimated based on the genotypes of markers at each QTL. Alleles are indicated in yellow when the genotype and phenotype were consistent, and green indicates an allele estimated by phenotype. The LOD peak of QTLs for awn length (top and middle) was observed in markers indicated by red. The first to third columns indicate the chromosome, marker name and position in cM of each marker, respectively. (XLSX) S4