Introduction

Barley (Hordeum vulgare L.) is the second most important crop species in Germany and ranks fourth in area cultivated worldwide. In barley, many monogenic traits as well as quantitative trait loci (QTL) have been mapped based on bi-parental mapping populations (for overview see Friedt and Ordon 2008), and the respective molecular markers are used for marker-assisted selection (MAS) in barley breeding. Since QTL detected in bi-parental populations are often not transferable to different genetic backgrounds, their general applicability for MAS is limited. In addition, in many cases QTL mapping is not performed in adapted breeding populations, so that breeders in general question its relevance for applied barley breeding (Jannink et al. 2001). Marker-trait association analysis is an alternative to QTL mapping in bi-parental populations (Jun et al. 2008). While linkage studies can only use family data, association studies can also use data from unrelated individuals. Broader genetic variation in a more representative genetic background can be included in association studies (Neale and Savolainen 2004). Therefore, a higher proportion of molecular markers are polymorphic, thus association mapping may attain better genome coverage than any bi-parental map (Breseghello and Sorrells 2006).

One of the prerequisites of a marker-trait association analysis is knowledge of the degree of linkage disequilibrium (LD) in the population analysed. The level of LD varies according to the species and the loci investigated. For barley as a predominantly self-pollinating species, high levels of LD were estimated (Kraakman et al. 2004; Malysheva-Otto et al. 2006) compared to other species such as maize (Remington et al. 2001).

A major complication in LD mapping is the detection of false-positive marker-trait associations due to population structure. This is probably of particular importance in barley, as diversity studies have shown barley germplasm to be highly structured, predominantly due to spike morphology and vernalization requirement (Malysheva-Otto et al. 2006; Rostoks et al. 2006; Cockram et al. 2008; Comadran et al. 2009; Hamblin et al. 2009).

A genome-wide LD mapping approach requires an adequate marker density (Cardon and Bel 2001). However, the number of markers needed depends on the level of LD (Mohlke et al. 2010). Today, high-throughput marker technologies such as Illumina’s Golden Gate™ Assay (Close et al. 2009) facilitate the estimation of the genome-wide genetic diversity and its use in barley breeding by linking the genetic to the phenotypic diversity. The Illumina Golden Gate oligonucleotide assays can detect single nucleotide polymorphisms (SNPs) by scoring the presence or absence of specific DNA alleles in a defined genomic representation through hybridization to microarrays (Oliphant et al. 2002).

Genome-wide association studies have already been conducted in many species, e.g. rice (Agrama et al. 2007), wheat (Breseghello and Sorrells 2006), perennial ryegrass (Skøt et al. 2005) and maize (Remington et al. 2001). In barley, Igartua et al. (1999) detected marker-trait associations for heading date based on restriction fragment length polymorphisms (RFLPs) in a population consisting of 32 barley cultivars. In addition, Ivandic et al. (2003) found associations between simple sequence repeat (SSR) markers and water-stress tolerance and powdery mildew resistance, respectively, in 52 wild barley lines. Using an LD mapping approach Kraakman et al. (2004, 2006) identified linkage between SSRs as well as amplified fragment length polymorphism (AFLP) markers with several agronomically relevant traits in 146 two-rowed spring barley cultivars. Pswarayi et al. (2008) studied changes in allele frequencies of Diversity Arrays Technology (DArT®) markers linked to yield QTL in 188 barley accessions (landraces, old and modern cultivars). Population structure and genome-wide LD were investigated in 192 H. vulgare accessions using nuclear microsatellite and DArT markers to identify associations with powdery mildew (Comadran et al. 2009). Inostroza et al. (2009) measured grain yield and plant height of 80 recombinant chromosome substitution lines of barley and marker-trait associations were identified using SSRs. DArT markers were used to identify associations for malting quality within a collection of 91 elite two-rowed malting barley lines (Beattie et al. 2010). Roy et al. (2010) genotyped 318 diverse wild barley accessions with DArT markers and SNP markers from the Illumina Golden Gate Assay for association mapping of spot blotch resistance. About 490 UK barley cultivars were genotyped using Golden Gate Bead Array technology to associate them with historical phenotypic data collected during varietal registrations (Cockram et al. 2010).

Besides whole-genome association approaches, candidate gene-based studies are used to detect associations between SNPs located in or near candidate genes and traits of interest. Several candidate gene association studies have been conducted in barley; e.g. Caldwell et al. (2006) measured LD between and within gene loci of the chromosomal region harboring the hardness locus in three different gene pools of H. vulgare (64 cultivated, 23 landrace and 34 wild barley accessions). Stracke et al. (2007) analysed LD in the physical and genetic context of the barley gene Hv-eIF4E, which confers resistance to Barley yellow mosaic virus and Barley mild mosaic virus. Cockram et al. (2008) identified vernalization genes by analysing 429 spring and winter barley cultivars. Associations between SNP markers detected for the α-amylase gene amy1 and important malting quality parameters were discovered in a collection of 117 European spring and winter barley cultivars by Matthies et al. (2009a, b). Stracke et al. (2009) estimated associations between nucleotide diversity at three genes (Ppd-H1, HvCO1, and HvFT1) and variation in flowering time in a collection of 220 spring barley cultivars.

The aim of this study was to determine whether a whole-genome association mapping approach is suited to detecting marker-trait associations in a gene pool relevant for German winter barley breeding. Consequently, the German barley collection investigated represents the most important varieties from the period 1959–2003.

Materials and methods

Plant material

The association panel consisted of a collection of 109 German winter barley cultivars from 31 different breeding companies, grouped by ear type into 61 six-rowed and 48 two-rowed types. The cultivars have been released in Germany between 1959 and 2003 and were the most important cultivars in their decades of release. The complete list of cultivars used in this study, the year of release and the breeding companies are presented in Electronic Supplementary Material Table S1.

Field trials

Field experiments were conducted in cooperation with private breeding companies in the growing seasons 2003/2004, 2004/2005 and 2005/2006. The 109 winter barley cultivars were grown at 12 sites located throughout Germany. Each trial was arranged according to an alpha-lattice with two replications per site. The winter barley cultivars were grown according to local practice for sowing rate and other inputs.

For association studies, the following traits were considered: grain yield (t/ha), 1,000-kernel weight (g), grains per spike, spikes/m2, hectolitre weight (kg/hl), plant height (cm), lodging at flowering (rating 1–9), lodging before harvest (rating 1–9), heading date (days after 1 May), brackling (rating 1–9), necking (rating 1–9), and sieve fraction (2.8 mm). The number of ears/m2 was determined by counting the number of tillers in one meter of row. The number of grains per spike was determined by dividing the grain yield per plot by the 1,000-kernel weight and the number of spikes per plot. The visual ratings were based on a 1–9 rating scale, with 1 for minimum and 9 for maximum.

For the analysis of variance, Type I mean squares (MS) were calculated using the general linear models procedure in SAS (SAS software; SAS Institute 2004). Heritabilities (h 2) for all investigated traits were estimated as follows:

$$ h^{2} = \, \left( {{\text{MS}}_{\text{G}} - {\text{MS}}_{\text{GE}} - {\text{MS}}_{\text{GY}} {\text{ + MS}}_{\text{E}} } \right) / {\text{MS}}_{\text{G}} $$

where MSG = mean square of genotypes (cultivars), MSGE = mean square of genotype × environment, MSGY = mean square of genotype × year, MSE = mean square of error.

DNA extraction

Genomic DNA was extracted from leaf tissue according to a modified protocol of Stein et al. (2001).

SSR analysis

The cultivars were genotyped by means of 72 SSR markers covering the whole barley genome (Ramsay et al. 2000). SSR amplification was performed according to Ramsay et al. (2000) in a total volume of 10 μl. PCR products were separated on 6% polyacrylamide gels on a LI-COR DNA Sequencer 4200 (MWG Biotech AG, Ebersberg, Germany). The bands on all gels were scored in a binary format, 1 for the presence and 0 for the absence of bands at polymorphic sites, respectively.

Genetic similarity estimation and principal coordinate analysis

Genetic similarity (GS) between the 109 winter barley cultivars was calculated by the Numerical Taxonomy and Multivariate Analysis System (NTSYS) ver. 1.70 computer software (Rohlf 2000) based on the 1–0 data from the SSR analysis. GS was calculated using the Dice (1945) coefficient. Graphical representations of the similarities between cultivars were obtained by principal coordinate analysis (PCoA; Sneath and Sokal 1973).

Population structure

The SSR data were used to identify subpopulations within the barley collection by means of the software package STRUCTURE (ver. 2.2; Pritchard et al. 2000). Independent simulations were performed for a K setting between 2 and 10 using the admixture model and a burn-in of 500,000 followed by 1,000,000 iterations. The optimal number of subpopulations K was chosen on the basis of the second-order rate of change of the likelihood function with respect to K. Results were summarized in matrices of fractional subpopulation membership (Q matrices).

Genotyping with SNP markers

To investigate allelic diversity in the barley collection, whole-genome profiling using Illumina’s Golden Gate Bead Array technology (Illumina Inc., San Diego, CA, USA; http://www.illumina.com) was carried out.

A set of 1,536 genome-wide biallelic SNPs from a customized oligonucleotide pool assay (IPK-OPA, unpublished) was tested on the DNA of the 109 cultivars using the Golden Gate BeadArray technology by Illumina Inc. The IPK-OPA has been mainly built on a selection of markers from two pilot assays (pOPA1, pOPA2) that are polymorphic between the two barley cultivars Barke and Morex. Raw data were transformed to genotype calls, initially using Illumina’s GenCall software. The data from all 109 winter barley cultivars were visually inspected to manually set 1,536 archetypal clustering patterns. The cluster positioning was guided by knowledge that heterozygotes are nearly non-existent because all cultivars were regarded as pure lines. A small proportion of detected non-homozygous data points was treated as missing data. Genotype calls were exported as spreadsheets from GenCall to then parse to create input for statistics software.

LD analysis

LD was measured using the squared correlation coefficient r 2 between pairs of SNP markers (Pritchard and Przeworski 2001). As the majority of SNPs of the IPK-OPA have been included in a barley consensus map (Close et al. 2009), it was possible to compare LD at known genetic map distances. Values of r 2 were calculated using the TASSEL software 2.1 (http://www.maizegenetics.net) between all pairs of loci and plotted against the genetic distance in centimorgans (cM). A smooth line was drawn using a 1,000-period moving linear regression. Background LD was assessed as the 95% percentile of physically unlinked markers. To determine the distance at which LD decays below background LD, functions for r 2 measurements of genetic distances were calculated using a sliding-window approach with a window size of 500 adjacent pairwise SNPs. All calculations were done for the whole dataset and for six- and two-rowed cultivars, separately.

Association analysis

For the association mapping approach, only polymorphic markers with allele frequencies larger than 5% were taken into account. The K matrix was generated based on 72 SSRs using the kinship matrix function in TASSEL software. Based on the least square means for the cultivar performance and 833 Illumina SNPs, marker-trait associations were calculated in a mixed linear model (MLM) approach with population affiliations as covariates and ear morphology as a factor using TASSEL. The threshold for significance was based on a false discovery rate of 0.1% (P ≤ 0.001). For all associated markers, the average difference of the phenotypic values of both marker classes was estimated using the UNIANOVA procedure of SPSS (SPSS 19.0, SPSS Inc., Chicago, IL, USA). Association profiles were created by plotting −log10(P) values against chromosomal positions. The percentage of variance explained by each SNP (R 2) was obtained using the TASSEL software.

Results

Phenotypic variation

Summary statistics for each trait are presented in Table 1. The 109 winter barley cultivars phenotyped in this study in 3 years’ field trials showed a broad variation for all traits measured, e.g. grain yield of the six-rowed cultivars ranged between 6.28 and 8.98 t/ha. A broad variation was also detected between six- and two-rowed cultivars, implying that the complete sample represents the phenotypic diversity present within the German winter barley breeding pool.

Table 1 Agronomical traits of 109 winter barley cultivars estimated in 3 years’ field trials
Full size table

Variance components were calculated by SAS. The results confirmed that the genotypic variance was significant for all traits (P < 0.0001). Detailed results of the analysis of variance are presented in Electronic Supplementary Material Table S2. Heritabilities (h 2) were high and ranged from 0.46 for grain yield to 0.94 for grains per spike (Table 1).

Principal coordinate analysis

GS of 109 winter barley cultivars was assessed using 72 SSRs. A total of 404 loci were detected by this approach. The average number of alleles per locus was 5.4, ranging from 2.0 to 13.0. Dice’s GS within the two groups ranged from 0.290 (six-rowed) versus 0.333 (two-rowed) to 0.956 (six-rowed) versus 0.910 (two-rowed), with a mean of 0.590 (six-rowed) versus 0.575 (two-rowed).

Principal coordinate analysis was conducted to visualize the GS of the subpopulations. The samples formed two clusters separating two- and six-rowed cultivars based on GS (Electronic Supplementary Material Fig. S1).

Population structure

To avoid spurious associations, the genetic structure among all 109 barley cultivars was investigated on the basis of the 72 SSR markers randomly distributed over the genome. In a first set of runs, two groups separating the two- and six-rowed cultivars were detected. Because principal coordinate analysis also pointed to the existence of these two groups, the population structure of the six-rowed and the two-rowed cultivars was calculated separately and, as a result, four subgroups could be distinguished within each of the two groups (Fig. 1).

Fig. 1
figure 1

Calculated population structure based on 72 SSR loci. The population was partitioned into color-coded subpopulations. Each bar represents a single cultivar, and the colored segments within each bar reflect the proportional contribution of the subpopulations to that cultivar. Cultivars are not ordered (A six-rowed, B two-rowed winter barley cultivars). (Color figure online)

Full size image

Linkage disequilibrium

The squared allele-frequency correlations (r 2) representing LD were assessed for all pairs of Illumina SNP markers. Electronic Supplementary Material Fig. S2 shows pairwise r 2 values in relation to the genetic distance (cM). A critical value of r 2, beyond which LD is likely to be caused by genetic linkage in the absence of population structure, was estimated at 0.13 for all winter barley cultivars (Table 2). Similar estimates of background LD were calculated for the six-rowed (0.11) and two-rowed (0.14) cultivars. LD differed markedly for the different chromosomes, e.g. within the six-rowed cultivars LD ranged from 3.03 (2H) to 17.78 cM (3H). A rapid LD decay below the critical r 2 was estimated for distances >7.35 cM within the complete set of winter barley cultivars. A slower LD decay was observed in the subgroups than in the whole set of genotypes.

Table 2 Estimates of background LD and distances at which LD decays below background LD for six- and two-rowed cultivars and the complete set of genotypes
Full size table

In a few cases high r 2 values between theoretically unlinked markers (>50 cM) were observed (Electronic Supplementary Material Fig. S2).

Association mapping

The main objective of this study was the detection of marker-trait associations in the German winter barley breeding gene pool. Table 3 summarizes the results, and details are shown in Electronic Supplementary Material Table S3.

Table 3 Summary of all marker-trait associations found in a collection of 109 barley cultivars
Full size table

In summary, 91 significant associations were detected using the MLM approach with population structure as a covariate. The 91 marker-trait associations correspond to 66 different SNPs with a maximum of four significant associations for one SNP (marker 649) with different traits.

Most of the marker-trait associations were detected for plant height (25), lodging at flowering (12), lodging before harvest (11), and grain yield (9). Only one SNP could be detected for grains per spike, which showed a significant association within the 109 winter barley cultivars.

For all associated SNPs the map position was known, e.g. SNP 1462 significantly associated with hectolitre weight (P = 0.0006) is located on chromosome 1H at 101.45 cM. Most of the associated markers are mapped on chromosome 5H (43 SNPs) and marked 15 genomic regions. No marker-trait associations were detected on chromosome 7H.

The associated SNPs explained between 0.2% (e.g. SNP marker 2184 associated with spikes/m2) and 15.4% (e.g. SNP marker 4876 associated with plant height) of the phenotypic variance for individual traits. A high number of the significant SNP markers proved to be closely linked and map at same positions, e.g. four markers associated with plant height are located on chromosome 5H at 132.63 cM. In total, the 66 associated SNPs represented 48 genomic regions (Table 3).

The identification of associations was supported by association profiles, i.e. from plots of P values [–log10(P)] against chromosomal positions. The significance of the association between markers and phenotypic traits was tested with a value of –log10(P) = 3, which indicates a probability value of 0.001. As an example, Fig. 2 shows the associations of Illumina SNPs with grain yield, 1,000-kernel weight, and spikes/m2, respectively. Most of the associated markers for grain yield were mapped on chromosome 5H in a region between 50 and 60 cM. Only five SNPs were associated with 1,000-kernel weight mapped on chromosomes 2, 5 and 6H. Seven Illumina SNPs located on three different chromosomes were associated with spikes/m2.

Fig. 2
figure 2

Association profiles showing −log10(P) of correlation between Illumina SNPs and grain yield, 1,000-kernel weight, and spikes/m2, respectively, against the position of the markers on all seven barley chromosomes. SNP markers were arranged on the x axis by their position in cM

Full size image

Discussion

In the first step of our study, phenotyping was performed in 3 years’ field trials at 12 locations in Germany. The 109 winter barley cultivars used in this study showed a broad phenotypic variation for all traits measured, i.e. yield, yield components and lodging. High values of heritability indicate that these traits can be improved by breeding and that QTL can be detected using such an experimental design.

Microsatellite (SSR) markers evenly distributed over the seven barley linkage groups were effectively used to identify population structure among the 109 winter barley cultivars. SSRs are very well suited for the estimation of genetic diversity due to their multiallelic nature. In addition, SSRs have a higher polymorphism information content (PIC) than other markers, which indicates the higher efficiency of these markers in showing the existence of diversity among the barley genotypes (Chaabane et al. 2009). Our results show that estimation of population structure using 72 SSR markers yielded clusters/subpopulations according to the ear type of the barley cultivars. In Ordon et al. (2005), the 109 winter barley cultivars were genotyped with at first 30 of the 72 SSRs and PCoA carried out on these data already separated two- and six-rowed cultivars. Grouping by ear type has been noted in other studies for spring (Brantestam et al. 2007; Hayes and Szucs 2006; Cockram et al. 2008; Hamblin et al. 2009; Stracke et al. 2009; Zhang et al. 2009) and winter barley cultivars (Stracke et al. 2007; Cockram et al. 2008; Hamblin et al. 2009). The ear morphology is encoded by the major gene v (Vrs1) which is located on chromosome 2H (Komatsuda et al. 2007). The strong separation between two- and six-rowed barley cultivars could be explained by the fact that breeders tend to work preferentially within each of these two groups and exploit crosses only to a lower extent between different row-types (Zhang et al. 2009).

Population structure may have a major impact on the outcome of association studies, resulting in false-positive associations (Pritchard et al. 2000). Many species have highly structured populations, e.g. maize (Thornsberry et al. 2001), wheat (Breseghello and Sorrells 2006), rice (Agrama et al. 2007) and barley (Comadran et al. 2009). Population structure is obviously the consequence of multiple events, e.g. modalities of domestication as shown in maize (Matsuoka et al. 2002) or geographical regions of origins as demonstrated in rice (Garris et al. 2005).

The set of 109 winter barley cultivars described in this publication was considered to be composed of two groups (based on ear type), each of them consisting of four subgroups.

Based on the information of 30 of the 72 SSR markers also used in this study, three populations were identified by Ordon et al. (2005), comprising the two-rowed cultivars and the six-rowed cultivars separately. Consequently, we could show that the implementation of more markers resulted in a better characterization of population structure. The observed population structure was taken into account during the identification of marker-trait associations.

A genome-wide LD mapping requires many markers, the number depending on the level of LD. LD decays faster with increasing genetic map distances. Therefore, a very large number of markers are required for effective whole-genome association genetic scans if the LD is low. LD is influenced by the recombination rate, allele frequency, population structure and selection (Flint-Garcia et al. 2003). Out-breeding species generally exhibit low LD, whereas self-fertilizing inbreeding species theoretically exhibit high LD. Therefore, the extent of genome coverage in different germplasm can vary from short distances of a few hundred base pairs up to genetic regions of tens of cM.

Our set of 109 German winter barley cultivars was explored using the Illumina Golden Gate assay, such that 918 polymorphic SNP markers could be used for determining LD. Mean LD values higher than r 2 = 0.13 extended up to 7.35 cM within the whole set of cultivars, which suggests that there is sufficient marker coverage for attempting a whole-genome association approach in barley. The LD decay turned out to be slower in the six-rowed and two-rowed cultivars than in the whole set of genotypes. Varying patterns of LD decay in two- and six-rowed barley cultivars probably reflect their different breeding histories (Flint-Garcia et al. 2003). A separate analysis of marker-trait associations within the six-rowed and two-rowed barley cultivars may therefore partly lead to different results, but was not conducted in this study due to the small sample size.

However, in comparison with other species the LD is high, e.g. in the out-breeding species maize LD diminished within a few hundred to 2,000 bp, depending on whether landrace or cultivars are analysed (Remington et al. 2001; Tenaillon et al. 2001; Palaisa et al. 2003). In sugar beet, LD extended up to 3 cM (Kraft et al. 2000). In contrast, in Arabidopsis as an inbreeding species, a higher LD of about 50 cM was found by Nordborg et al. (2002), but this research was done in populations with only a few genotypes and after extreme inbreeding. The most probable reason for the high level of LD observed in barley is the fact that it is an inbreeding species (Morrell et al. 2005). This observation is confirmed by Kraakman et al. (2004), who estimated LD over large distances >50 cM in some regions in related barley germplasm. Malysheva-Otto et al. (2006) reported on genome-wide LD, which extends up to 10 cM in a set of 953 cultivated barley accessions. Highly significant LD extended over >60 cM in a set of European spring and winter barley cultivars analysed by Rostoks et al. (2006). Comadran et al. (2009) used DArT markers to analyse 192 H. vulgare accessions and a lower degree of LD was estimated (extending up to 3.12 cM) due to a higher average density of markers per chromosome than in other association studies. Rostoks et al. (2006) estimated that approximately 1,000 evenly distributed markers are necessary to ensure good coverage of the barley genome in order to detect genetic linkage to QTL. For map-based cloning, low LD and a higher number of markers may be required, depending on factors such as genome location, which can influence the relationship between physical and genetic distance (Zhang et al. 2009). New high-density marker technologies such as Diversity Array Technology and Illumina Golden Gate BeadArray technology facilitate this. Zhang et al. (2009) genotyped 170 Canadian barley cultivars using 942 DArT loci to estimate LD up to 2.6 cM. In our study we used nearly the same number of SNP loci (918 Illumina), but the extent of LD was higher. Cockram et al. (2010) measured marker-trait associations between 1,111 Illumina SNPs and 500 barley varieties from the UK. They too detected high levels of long-range LD. One reason may be that there is a lower genetic diversity in the respective sets of barley varieties. It is known that LD would be expected to extend over much shorter distances when analysing a very diverse set of genotypes (ancestral populations).

In some cases in wild barley (Hordeum spontaneum) as well as in landraces, LD has low values in genes not subjected to the strong directional selection forces of plant breeding. Morrell et al. (2005) determined that LD of wild barley decays at rates similar to maize (Remington et al. 2001). On the other hand, Caldwell et al. (2006) identified LD over a region of about 200 kb encompassing the Ha locus controlling grain hardness in barley. Consequently, genes that have been under breeding selection pressure will be in higher LD.

In this study, we explored the possibility of mapping QTL in a set of barley cultivars instead of using a segregating population derived from a bi-parental cross. In summary, 66 significant SNP markers were selected using the software package TASSEL.

Many of the associated markers were found in a region where a candidate gene or QTL or a marker-trait association for one of the traits investigated has been reported previously, e.g. two SNP markers (147, 840) associated with plant height on chromosome 1H are mapped in a region where the photoperiod gene Ppd-H2 (Laurie et al. 1995) is located. The vernalization gene VrnH1 is located on chromosome 5H (Yan et al. 2005) where several SNPs associated with plant height (4,098, 2,184, 4,106, 4,874, 5,322, 70, 1,718, 2,180) were detected. Schmalenbach et al. (2009) and Qi et al. (1998) also reported the detection of plant height QTL within a region in which QTL/genes for flowering are located. They detected a high correlation between flowering time and plant height within a set of wild barley introgression lines and recombinant inbred lines, respectively. In our study, we estimated a significant correlation (r = 0.51) between these traits, too. This correlation may be due to pleiotropic effects of single genes or due to close linkage of several genes controlling these traits.

However, a couple of new loci were found for all traits measured. Coincidence of the detected marker-trait associations with known QTL positions might indicate a candidate gene locus, but the QTL were detected in different mapping populations using different marker types. Consequently, it is not certain that the associations identified in this study and known QTL indicate the same genes.

The work presented indicates that association mapping approaches can be a useful alternative to classical QTL approaches based on bi-parental crosses. The analysis was performed with barley cultivars which represent breeding progress in Germany during the last five decades, so the results are of great relevance for barley breeding.

Identification of marker-trait associations could be followed by the development of segregating populations, polymorphic for the involved loci, in which the associations are confirmed. Verified trait-associated markers can be used in MAS in combination with traditional field selection to enhance barley breeding. In addition, the isolation of the genes underlying the QTL by map-based cloning will be essential in unravelling the genetic basis of phenotypic variation among barley genotypes.