Association Mapping and Haplotype Analysis of a 3.1-Mb Genomic Region Involved in Fusarium Head Blight Resistance on Wheat Chromosome 3BS

A previous study provided an in-depth understanding of molecular population genetics of European and Asian wheat gene pools using a sequenced 3.1-Mb contig (ctg954) on chromosome 3BS. This region is believed to carry the Fhb1 gene for response to Fusarium head blight. In this study, 266 wheat accessions were evaluated in three environments for Type II FHB response based on the single floret inoculation method. Hierarchical clustering (UPGMA) based on a Manhattan dissimilarity matrix divided the accessions into eight groups according to five FHB-related traits which have a high correlation between them; Group VIII comprised six accessions with FHB response levels similar to variety Sumai 3. Based on the compressed mixed linear model (MLM), association analysis between five FHB-related traits and 42 molecular markers along the 3.1-Mb region revealed 12 significant association signals at a threshold of P<0.05. The highest proportion of phenotypic variation (6.2%) in number of diseased spikelets (NDS) occurred at locus cfb6059, and the physical distance was about 2.9 Kb between umn10 and this marker. Haplotype block (HapB) analysis using a sliding window LD of 5 markers, detected six HapBs in the 3.1-Mb region at r2>0.1 and P<0.001 between random closely linked markers. F-tests among Haps with frequencies >0.05 within each HapB at r2>0.1 and P<0.001 showed significant differences between the Hap carried by FHB resistant resources, such as Sumai 3 and Wangshuibai, and susceptible genotypes in HapB3 and HapB6. These results suggest that Fhb1 is located within HapB6, with the possibility that another gene is located at or near HapB3. SSR markers and Haps detected in this study will be helpful in further understanding the genetic basis of FHB resistance, and provide useful information for marker-assisted selection of Fhb1 in wheat breeding.


Introduction
Fusarium head blight (FHB), or scab, caused mainly by Fusarium graminearum Schwabe (Giberella zeae Schw. and Petch), is a devastating worldwide fungal disease of wheat (Triticum aestivum L.). FHB infection is favored by warm humid conditions during the flowering and early stages of kernel development [1]. FHB reduces yield and grain quality through shrivelled kernels, and contaminates the grains with mycotoxins such as deoxynivalenol [1][2][3][4] making the grains unsuitable for human or animal consumption [5].
The use of host resistance is an economically and environmentally effective strategy for controlling FHB. So far, only a few highly resistant wheat cultivars have been identified from different geographic regions, including Asia, South and North America, and Europe [6][7][8]; for example, spring wheats from Asia including Sumai 3 (Funo/Taiwan Wheat) and its derivatives, spring wheats from Brazil and winter wheats from Europe. However, wheat breeding programs worldwide have relied heavily on Sumai 3derived FHB resistance with a risk of rapid overcome of the resistance. The utilization of novel resistance sources is needed to diversify the genetic basis of FHB resistance and to increase the level of resistance through pyramiding of resistance genes that tend to act additively [9].
Association mapping, a powerful approach to unravel the genetic architecture of complex traits in crops [27,28], has been used in few studies to identify the relatedness between molecular markers and FHB resistance in wheat natural populations [29,30]. Moreover, haplotype association is likely to be more powerful in the presence of LD [31]. A haplotype is a set of closely linked intra-chromosome genetic markers that tend to be inherited together [32]. Following genetic diversity and linkage disequilibrium studies on a 3.1-Mb genomic region on chromosome 3B in European and Asian bread wheat populations [33], the same set of accessions were verified on FHB-related traits in multiple environments. In this paper, our objective, as in the previous study, was to target loci significantly associated with Fhb1 in European and Asian wheats through association analysis in a candidate region in contig ctg954 (Genebank accession number: FN564434) on the short arm of chromosome 3B. Haplotype diversity and its relationship to FHB-related traits were also analyzed in the 3.1-Mb genomic region. Such an association analysis should provide useful information for marker-assisted selection of Fhb1 in wheat breeding. The ultimate aim in the study was to discover potential resistance sources for use in wheat breeding and genetic improvement.

Genetic Structure and Relative Kinship of Overall Wheat Accessions
Based on genotyping datasets of 70 genome-wide microsatellites, principal coordinate analysis of the 266 accessions revealed that the European and the Asian materials were independent subgroups ( Figure 1a). Re-evaluation of genetic structure for the test genotypes with STRUCTURE software further demonstrated two sub-groups, as revealed by the Evano criterion (Figure 1b, c). Figure 1d indicates that average LD decay could be higher than 500 Kb with r 2 ,0.2 at P,0.001 along the 3.1-Mb region for the two sub-groups. The European sub-group had stronger LD than the Asian sub-group. These results were consistent with the previous study on 376 accessions using the same genotyping dataset [33].
Besides evaluation of genetic structure of the test genotypes based on the 70 genome-wide SSRs with major allelic frequencies (MAF) .5% at the population level, the same sets of genotyping data were used to calculate relative kinship between pairs of individuals in the current study. This revealed the approximate identity between two given individuals over the average probability of identity between any two random individuals [34,35]. About 74.7% of the pairwise kinship estimates ranged from 0 to 0.05, among which the percentage of kinship values close to 0 reached as high as 57.1% ( Figure 2). This indicated unrelated of the accessions. Other ranges of pair-wise kinships showed an obviously declining tendency.

FHB Responses in Three Environments
Using the phenotypic data from three environments and their average values, comparisons were performed across environments for each FHB-related trait (Table 1). Obvious differences were observed for each of the five FHB-related traits in all environments. For example, in E1, the ranges of NDS, PDS, LDR, DS and DI were 1.00 to 8.53, 4.21 to 41.51, 0 to 6.97, 0 to 0.76 and 0 to 22.66, respectively. The same inner variances were also present in other environments. Highly consistent results were obtained among the three environments. No statistically significant differ-ence was detected in most tests, except between E1 and the others in PDS and LDR. Furthermore, FBH-resistance-related traits of the four control accessions also showed good consistency among the three environments (Table S1). With the help of phenotypic data and the kinship matrix, heritability (h 2 ) was calculated for each trait (Table 1). The h 2 ranged from 56.5% to 62.5% using average values of the three environments. The highest h 2 value was for DI (62.5%), indicating that this trait was less affected by environmental factors than the other four.
Pearson correlation coefficients between environments were calculated for each FHB-related trait using SPSS v12.0 (Table 2). There were highly significant positive correlations between environments for all five traits. The ranges in the environmental correlation coefficients without taking into account mean value for NDS, PDS, LDR, DS and DI were 0.498 to 0.676, 0.539 to 0.699, 0.530 to 0.675, 0.531 to 0.676 and 0.542 to 0.689, respectively. Moderately high correlations between environments indicated good repetitiveness and high heritability in the selected population. It was notable that the coefficients between the average and each environment were relatively higher than that between individual environments, indicating that the average value had a high representation in the test traits. Pearson correlation coefficients between traits were therefore estimated using the average values of the three environments (Table S2). There were significant (P,0.01) positive correlations between all FHB resistance traits. All pairs of phenotypic correlation coefficients ranged from 0.778 to 0.970, and the highest value was observed for the correlation between the PDS and DI. This was consistent with the interrelationship among these parameters at the statistical level.

Hierarchical Clustering of FHB Traits
In order to infer possible classifications for the test genotypes, hierarchical clustering (UPGMA) of all wheat accessions based on Manhattan dissimilarity matrix with DARwin v5 software using the five FHB-related traits was used ( Figure S1). The hierarchical tree showed that the genotypes could be divided into eight groups. As shown in Figure S1, the numbers of genotypes in each group were 51, 20, 51, 53, 28, 37, 20 and 6. The resistant controls, Sumai 3 and Wangshuibai, clustered with group VIII, whereas the susceptible controls, Ningmai 11 and Mianyang 11, were in groups III and VI, respectively.
A comparative analysis of the five traits between clusters was made based on the average values for the three environments (Table 3). Cluster VIII had the lowest mean values for all five traits, whereas II had the highest. This indicated that accessions in cluster VIII were the most resistant among the eight subgroups. The order of clusters, from resistant to susceptible, was the same for all five traits, i.e. VIII, VII, V, VI, III, IV, I, II. Significant differences between clusters further revealed that subgroup VIII, including the two positive controls, had significant (P,0.01) resistance compared to the other subgroups, indicating that these six accessions might therefore have potential as resistant genetic resources for improvement of FHB resistance in wheat breeding around the world. Detailed information for the six accessions in subgroup VIII is provided in Table S3.

Association Studies for Five FHB Traits along the 3.1-Mb Region
Given the strong population clustering of the test genotypes, the compressed mixed linear model (MLM) [36,37] of the Q+K model was used to identify association signals related to the five FHBrelated traits using mean values of three environments, viz. number of diseased spikelets (NDS), percentage of diseased spikelets (PDS), length of diseased rachides (LDR), disease severity (DS) and disease index (DI), with 42 molecular markers (32 SSR and 10 SNP) along the 3.1-Mb region on chromosome 3BS. The results of the association studies, including the Manhattan and quantile-quantile plots, are mapped in the Figure 3 and Figure S2. Twelve significant association signals were found at the threshold of P,0.05 (Table 4). No significant association was identified for LDR. The number of significant associations was 4, 4, 1 and 3 for NDS, PDS, DS and DI, respectively. Among the 12 associations, locus cfb6059 accounted for the highest amount of phenotypic variation (6.2%) for NDS, whereas cfp5062_S38 explained the lowest (1.6%). Most markers had significant associations with traits in a specific environment, and only three significant associations  (between cfb6110 and DI, between cfb6072 and PDS, and between cfb6072 and DI) occurred in two environments. In total, four loci including three SSR (cfb6110, cfb6072 and cfb6059) and one SNP (cfp5062_S38) were significantly associated with four FHB-related traits, and for each locus, there were at least two significantly associated traits. Locus cfb6059 was associated with four traits, indicating the relatedness between the traits or a situation of multiple traits determined by one gene.
After detection of molecular markers significantly associated with FHB response traits, the mean phenotypic values and genetic effects of alleles with frequencies of .0.05 for the four significantly associated markers were determined (Table 5). There were no significant phenotypic differences in mean values for the five scab traits between two random alleles at both the cfb6059 and cfp5062_S38 loci. Significant phenotypic differences (P,0.05) were found between the two alleles at cfb6072 only for DS. However, at cfb6110, significant differences (P,0.01) were detected between alleles cfb6110 289 and cfb6110 294 for traits NDS, PDS, LDR and DI. Accessions with cfb6110 289 included Sumai 3 and those sharing the same Hap with Sumai 3 had lower mean phenotypic values than those with cfb6110 294 for all four traits. Taking PDS as a trait example (Figure 4), there were negative genetic effects for the alleles 236 bp and 238 bp at cfb6059, 289 bp at cfb6110, 183 bp and 193 bp at cfb6072, and positive effects for 242 bp at cfb6059, 294 bp at cfb6110, and alleles 1 and 3 at cfp5062_S38. Accessions with negative alleles had relatively higher FHB resistance than those with positive alleles. Similar allelic effects were detected for the other four FHB-related traits ( Figure S3).

Haplotypes of 42 Markers along the 3.1-Mb Region and Their Phenotypic Effects
To decide the number of haplotype blocks (HapB) for 42 markers along the contig ctg954, HapB analysis was conducted by sliding window LD with 5 marker sets as the LD window size using TASSEL v3.0 software [37]. Six HapBs were found at r 2 .0.1 and P,0.001 between random closely linked markers in the sliding window LD ( Figure 5, Table 6). As shown in Table 6, the numbers of markers constituting HapBs varied from 2 to 6, and physical distances ranged from 21.5 Kb (HapB5) to 392.8 Kb (HapB6). Based on the allelic combinations for different markers in each HapB, the numbers of expected Haps for HapB1 to HapB6 were 24, 96, 16, 35, 4, and 1,584. However, the observed frequencies were 14, 21, 10, 19, 4 and 32, respectively. Therefore, except for HapB5, there were many fewer Haps than expected. Moreover, the frequency of Haps was also different within each HapB, and the dominant Hap frequencies for HapB1 to HapB6 were 39.7%, 29.0%, 59.9%, 30.2%, 58.2% and 70.6%, respectively. There was a large change in the Hap frequency for HapB5, the highest frequency of the major Hap reached 58.2%, although the observed and expected Hap numbers were consistent (Table 6).
Based on the observed Haps within each HapB at r 2 .0.1 and P,0.001, those Haps with frequencies .0.05 in the studied population, i.e. 6 in HapB1, 6 in HapB2, 3 in HapB3, 4 in HapB4, 3 in HapB5, 2 in HapB6, were selected for tests of association with the five FHB-related traits ( Table 7). The Haps carried by Sumai 3 were also included in the comparative analysis (asterisked in Table 7), although its frequency in HapB6 was less than 0.05. Almost no significance was detected between Haps in HapB1 and     (Table 8). There were six combinations between the two HapBs, and statistical significances at P,0.05 or lower occurred in all HapB3-2/HapB6-2 combinations. However, there were no statistical differences among the other five combinations. This indicated that there might be strong interactions between haplotypes HapB3-2 and HapB6-2 leading to increased FHB resistance.

Comparative analysis of genetic and physical maps on 3BS
Physical map of 42 molecular markers from sequencing the contig ctg954 (Genebank accession number: FN564434) on the short arm of chromosome 3B was constructed using Chinese Spring wheat (Figure 6a). On the 3.1-Mb genomic region, the interval from 2.2 to 2.8 Mb (between cfb6078 and cfb6061) might include the Fhb1 locus (B. Gill pers. comm.). Association study of this target region revealed that cfb6059 was significantly associated with FHB resistance. Importantly, this marker was very close to umn10 as a marker widely used for MAS in FHB-resistance breeding [26], and theirs physical distance was about 2.9 Kb between them ( Figure 6a).
As we all known, Fhb1 gene was originally located in the distal region of chromosome 3BS between SSR loci gwm493 and gwm533 [10]. In order to understand the position parallelism between markers used in our study and others related with FHB resistance, genetic map was constructed based on a recombinant inbred line (RIL) population (Nanda 24196Wangshuibai) (Figure 6b). Both cfb6061 and cfb6011 were mapped the interval between gwm493 and gwm533. Moreover, the physical distance between cfb6061 and cfb6059 was 376.3 Kb, while 373.4 Kb between cfb6061 and umn10. Comparative analysis of markers mapped in both genetic and physical maps revealed that cfb6059 significantly associated with FHB resistance was really located on the region contained Fhb1 locus, and also very close to the umn10 marker.

Efficiency of Association Analysis in the Target Genomic Region
During the past decade, numerous studies have focused on genetic mapping of FHB resistance in wheat, and knowledge on the genetic control of FHB resistance has continually increased with time [16,17]. Various studies locating QTL in chromosome 3BS are summarized as follows. Using a Sumai 36Stoa population, the type 2 FHB resistance QTL from Sumai 3 was mapped to chromosome 3BS with the linkage marker cdo981, and was designated as Fhb1 (syn. Qfhs.ndsu-3BS) [21]. In one of the first two published QTL mapping studies, Bai et al. [22] found one major QTL in a Ning 78406Clark population based on AFLP genotyping. Anderson et al. [10] confirmed the major QTL for resistance to fungal spread on chromosome 3BS (Qfhs.ndsu-3BS) linked with the marker gwm493. This QTL was verified by a series of mapping reports in bi-parental populations. Zhou et al. [23] revealed that the major QTL in a RILs population derived from Ning 78406Clarkwas also located in the same region on chromosome 3B between SSR markers gwm533 and barc147. At about the same time, Buerstmayr et al. [38] found a major QTL between gwm533 and gwm493 on 3BS using a large DH population of CM-820366Remus. Other intervals or linked markers with the major QTL in this genomic region were barc133 [39], gwm533 [11], between barc133 and gwm493 [40], between barc147 and gwm493 [41], between gwm533 and barc147 [42], between gwm533 and gwm493 [43], between sts3B.80 and sts3B.142 [13], between sts3B.189 and sts3B.206 [18] and umn10 [26]. As already mentioned, this major QTL was distally located on chromosome 3BS between microsatellite markers gwm493 and gwm533 [10] (Figure 6b).
In our previous study comparing LD and PIC value levels [33] the intensity of selection along the 3.1-Mb region was variable; for instance, in the interval from 2.2 to 2.8 Mb (between cfb6078 and cfb6061, Figure 6a), which could include the Fhb1 locus (B. Gill pers. comm.). In the present study, after diversity detection of SSR markers between the two parents, loci cfb6061 and cfb6011 were Table 4. Significance (P value) and genetic variation (R 2 ) explained by individual markers associated with FHB-related traits. mapped between gwm533 and gwm493 using a RIL population of Nanda 24196Wangshuibai (Figure 6b). Locus cfb6059 that explained the highest proportion of phenotypic variance (6.2%) was between cfb6078 and cfb6061, and very close (261 Kb) to the mapped marker cfb6061. Comparing the physical map of SSR markers constructed using Chinese Spring (Figure 6a) with the    genetic map (Figure 6b), all associated markers detected in this study were located in the genomic region indicated by numerous linkage studies. In addition, this significantly associated marker cfb6059 was very close to umn10 as a marker widely used for MAS [26], and theirs physical distance was about 2.9 Kb between them (Figure 6a), which indicates a complementation relationship between association study and linkage analysis using different mapping populations. To further dissect the relatedness of these two markers, we genotyped all our materials using the marker umn10 designed by  Liu et al. [26]. All accessions contained two types of alleles at this locus, i.e. 236 bp (susceptible control) and 239 bp (resistant control), and F-test based on our natural populations found that none of significant associations between unm10 and FHB-related traits existed in both the European and Asian subgroups (Table  S4), but five of six resistant germplasm resources ( Figure S1, Table  S3), except Liaochun 4 (Liaoning, China), had the same allele 239 bp at this locus with the famous resistant resource Sumai 3, indicating it was an effective diagnostic marker for MAS for gene Fhb1. Moreover, the linkage disequilibrium (LD) between cfb6059 and umn10 was r 2 = 0.28 (P,0.001) in despite of their closely linkage, revealing that there was a complementary relationship between these two markers in marker-assisted selection of FHBresistant breeding.

Complementary Relatedness between Markers and Haplotype Association Analysis
Association mapping based on linkage disequilibrium (LD), representing next-generation plant genetics [44,45], has become a powerful tool for dissecting complex agronomic traits and identifying specific alleles conferring target traits using natural crop populations [28,34,35,46]. Generally, population structure and genetic relatedness influence the success of association analysis, but may result in spurious marker-trait associations [47]. One effective strategy to address this problem is the compressed mixed linear model (MLM) suggested by Yu et al. [36] and Zhang et al. [37] based on the chosen Q-matrix derived from STRUCTURE and the kinship-matrix from SPAGeDi. A haplotype was described as a set of closely linked intrachromosome molecular markers that tend to be inherited together [32]. Haplotype association is likely to be more powerful in the presence of LD [31] and haplotype-trait association analyses are helpful for precise mapping of important genomic regions and location of superior or preferred alleles or haplotypes for breeding [48]. Recently, there has been an emphasis on haplotype (or haplotype block) analysis in many crops. For example, Malysheva-Otto and Röder [49] discovered novel haplotypes and analyzed their distribution and significance in the endosperm-specific bamylase gene Bmy1 of cultivated barley (Hordeum vulgare L.); Stracke et al. [50] showed that the linkage disequilibrium pattern and haplotype structure led to interplay between low recombination and recent breeding history surrounding a locus encoding Bymovirus resistance in barley; and Gore et al. [51] generated a first-generation haplotype map of maize to provide a foundation for uniting breeding efforts and for dissecting complex traits through genome-wide association studies.
With the increasing number of studies on association mapping using grouped molecular markers or several haplotypes, the efficiency of the two methodologies has greatly improved. In the present study, both marker association (Figure 3, Table 4) and haplotype association ( Figure 5, Tables 6, 7) were conducted for five FHB-related traits in a 3.1-Mb genomic region on wheat chromosome 3BS. Four markers and two haplotypes were significantly associated with FHB resistance. These four markers were located within or surrounding two haplotype blocks ( Figure 5); HapB3 and HapB6 each contained one haplotype significantly associated with FHB-related traits (Table 7). Further study revealed that these two haplotypes had strong interactive effects on FHB response (Table 8). Thus there was a good complementary relatedness between molecular markers and haplotype associations in the target chromosome region. As shown in Table S3, except Youzimai (Sichuan, China), other five materials contained not only similar disease rating but also same haplotype comparing with Sumai 3.

Molecular Markers and Haplotypes Associated with FHBrelated Traits
Using compressed MLM, association analysis between five FHB-related traits and 42 molecular markers along the 3.1-Mb region detected 12 significant association signals at the threshold of P,0.05 (Figure 3, Table 4), and locus cfb6059 explained the highest proportion of variation (6.2%) for the trait NDS mapped in the interval between gwm493 and gwm533 (Figure 6a, b), indicating a good integration of association mapping and QTL mapping in the discovery of molecular markers associated with Fhb1. Association analysis also validates superior or preferred alleles in germplasm collections [52]. For example, Zhang et al. [44] found that allele Xgwm130 132 underwent very strong positive selection for 1,000 kernel weight during modern breeding. Through association mapping of dynamic development of plant height in common wheat, Zhang et al. [53] exposed different allelic effects of associated markers; gwm495-4B 155 was associated with reduced height of 211.2 cm under drought stressed and 215.3 cm under well watered conditions, whereas the 167 bp allele exhibited increased height effects of 3.9 and 8.1 cm, respectively. In the present study, different allelic effects of significant loci were also found for FHB response (Figure 4, Figure S3). Wheat accessions such as Sumai 3 with a 289 bp allele at cfb6110 had lower mean phenotypic values for NDS, PDS, LDR and DI than those carrying a 294 bp allele. Taking PDS as a trait example (Figure 4), negative genetic effects were detected for the 236 and 238 bp alleles, but there were positive effects when the allele 242 bp was present. Thus the identification of superior alleles will help in choosing parents for crossing programs, to ensure maximum numbers of superior alleles across sets of loci targeted for selection. Fixation of those alleles can then be targeted [54].
As suggested by Garner and Slatkin [31], the presence of LD is the premise of haplotype association for precise mapping of important genomic regions. In this study, the identification of haplotype blocks was performed by a sliding window LD of 5 markers as the LD window size at the level of r 2 .0.1 and P,0.001 between random closely linked markers. Six HapBs were detected in the 3.1-Mb genomic region ( Figure 5, Table 6). HapB3-2 and HapB6-2 conferred the lowest phenotypic values in five FHBrelated traits. In addition, two landmark FHB resistance germplasms, Sumai 3 and Wangshuibai, carry these two subhaplotype blocks. Interestingly, HapB6 included the significantly associated loci cfb6059 and cfp5062_S38, suggesting that the Fhb1 locus might be included in this ,392 Kb haplotype block. Based on analyses of genes identified in the sequenced contig Ctg954 of wheat chromosome 3B [55], the 392 Kb region was the interval that included the highest number of genes, and marker cfb6059 was located between genes ctg0954b.00390.1 and ctg0954b.00400.1 (F. Choulet pers. comm.). On the other hand, HapB3 did not involve any associated locus, suggesting a complementary relationship between haplotype analysis and simple association mapping. The locus cfb6110 significantly associated with scabrelated traits was very close physically to haplotype block HapB3. It can be predicted that Fhb1 should be within the HapB6, but we cannot exclude the possibility of another gene located around HapB3. This study therefore provides useful information and genetic markers for cloning Fhb1 and for marker-assisted selection in breeding.

Plant Materials
Two hundred and sixty six wheat accessions comprising 157 Asian and 91 European genotypes were used in this study. They were chosen according to their heading date in China from those included in the previous study on genetic diversity and linkage disequilibrium in 376 Asian and European bread wheat accessions [33]. In addition, four lines, Sumai 3 (Jiangsu) and Wangshuibai (Jiangsu), and Ningmai 11 (Jiangsu) and Mianyang 11 (Sichuan) were used as resistant and susceptible controls in phenotyping of FHB response, respectively. Their detailed information including pedigree data is given in Table S5. All accessions were evaluated for FHB response based on the single floret inoculation method. A mixed F. graminearum conidial suspension containing four virulent strains with a concentration of 5610 4 conidial/ml was produced as described by [15,20]. Inoculation was carried out at anthesis with 20 ul of conidial suspension inserted into the floral cavity between the lemma and palea of a single floret in the middle of a spike. Inoculated spikes were humidified in an enclosed plastic chamber for 2 days to induce infection. Ten plants were selected from each row for evaluation 21 days after inoculation. Number of diseased spikelets (NDS) and the length of diseased rachides (LDR) for each inoculated plant, and its corresponding spikelet number per spike (SPN) spike length (SPL) were investigated. Three phenotypic parameters were calculated according to the formula: PDS (percentage of diseased spikelets) = (NDS/SPN)6100%, DS (disease severity) = LDR/SPL, and DI (disease index) = PDS6DS. The final phenotypic data included five FHB response traits, viz. NDS, LDR (cm), PDS (%), DS and DI (%) evaluated in three environments. The mean values are listed in Table S6. Statistical calculations of correlations between environments and between traits were performed with SPSS v12.0.

Genotyping, Population Structure, and Linkage Disequilibrium
In the previous study, Hao et al. [33] used 42 molecular markers (32 SSR plus 10 SNP) from the contig ctg954 (Genebank accession number: FN564434) [55,56] (Table S7) and 70 SSR markers from the whole genome to genotype 376 wheat accessions including the 266 accessions selected for the current study. The four controls were genotyped using the 42 3B-specific markers based on the method described in Hao et al. [33], and the marker umn10 designed by Liu et al. [26] was also used to genotype all materials. Then, two sets of genotyping data in both the 3.1-Mb sequenced region and the whole genome were used in this study (Table S8).
Additionally, a recombinant inbred line (RIL) population (Nanda 24196Wangshuibai) was genotyped using 32 SSRs to make fine mapping of these markers in the genomic region.
Additional analyses included population structure (Q value) with STRUCTURE v2.2 software [57], estimation of the most appropriate number of sub-groups (Dk value) on the basis of the Evanno et al. [58] criterion using 70 genome-wide SSR, squared allele-frequency correlations (r 2 ) and significances of each pair of loci using the 32 polymorphic SSRs and 10 SNPs in the approximate 3.1-Mb region on chromosome 3BS with the dedicated procedure of the TASSEL v3.0 software [37]. All parameter settings for population structure analysis and linkage disequilibrium evaluation in the present study were based on the method of Hao et al. [33,59]. In addition, we also produced a hierarchical clustering tree from the Manhattan dissimilarity matrix using DARwin v5 software [60] based on five phenotypic traits, and principal coordinate analysis of the test material to reveal relationships among accessions using NTSYS-pc version2.1 software [61].
In order to define the degree of genetic covariance between pairs of individuals, the relative kinship matrix (K) was calculated using genotypic data of genome-wide SSR with SPAGeDi software [62]. Calculation of pairwise kinship coefficients was according to Loiselle et al. [63] with 10,000 permutation tests. Negative values between individual pairs were changed to 0, as this indicated a lower than expected relationship between two random individuals [36].

Marker-trait Association
The software program TASSEL v3.0 (http://www2. maizegenetics.net/) [37,64] was used to calculate associations between the 42 3B-specific markers and phenotypic traits of scab resistance, with the compressed mixed linear model (MLM) suggested by Yu et al. [36] and Zhang et al. [37] based on the chosen Q-matrix derived from STRUCTURE and the kinshipmatrix from SPAGeDi. For estimating associations those markers with the allelic frequencies less than 0.05 were filtered as rare alleles and deleted. We adopted the MLM options of optimum level and population parameters previously determined (P3D). The significance levels (P values) between markers and traits and phenotypic variation (R 2 ) explained by the markers associated with traits were obtained for further study. Markers were defined as being significantly associated with traits according to their P values (-LogP.1.30, P,0.05).
Based on the procedure of the TASSEL software, only using phenotypic data and the kinship matrix, the heritability (h 2 ) of each test trait in different environments, defined as the proportion of genetic variance over the total variance, was calculated according to the formula h 2 = s a 2 /(s a 2 +s e 2 ) with the MLM options of no compression and re-estimation for each marker. Here, s a 2 means genetic variance, and s e 2 indicates the residual variance. Allelic effects were evaluated in comparison to the ''null allele'' (missing plus rare alleles) for each locus [52].

Haplotype Mapping
To determine the number and types of haplotype blocks (HapB) along the contig ctg954, the sliding window LD, with 5 markers as LD window size, was managed through TASSEL v3.0 [37,64], and the markers were filtered for rare alleles with frequencies of less than 5% in the whole collection. For assignment of one HapB, r 2 .0.1 and P,0.001 between random closely linked markers were regarded as the threshold in the sliding window LD. Based on different allelic combinations of all loci in the same HapB, the haplotype (Hap) was statistically calculated for each HapB in Excel.