Barley grain (1,3;1,4)-β-glucan content: effects of transcript and sequence variation in genes encoding the corresponding synthase and endohydrolase enzymes

The composition of plant cell walls is important in determining cereal end uses. Unlike other widely consumed cereal grains barley is comparatively rich in (1,3;1,4)-β-glucan, a source of dietary fibre. Previous work showed Cellulose synthase-like genes synthesise (1,3;1,4)-β-glucan in several tissues. HvCslF6 encodes a grain (1,3;1,4)-β-glucan synthase, whereas the function of HvCslF9 is unknown. Here, the relationship between mRNA levels of HvCslF6, HvCslF9, HvGlbI (1,3;1,4)-β-glucan endohydrolase, and (1,3;1,4)-β-glucan content was studied in developing grains of four barley cultivars. HvCslF6 was differentially expressed during mid (8–15 DPA) and late (38 DPA) grain development stages while HvCslF9 transcript was only clearly detected at 8–10 DPA. A peak of HvGlbI expression was detected at 15 DPA. Differences in transcript abundance across the three genes could partially explain variation in grain (1,3;1,4)-β-glucan content in these genotypes. Remarkably narrow sequence variation was found within the HvCslF6 promoter and coding sequence and does not explain variation in (1,3;1,4)-β-glucan content. Our data emphasise the genotype-dependent accumulation of (1,3;1,4)-β-glucan during barley grain development and a role for the balance between hydrolysis and synthesis in determining (1,3;1,4)-β-glucan content, and suggests that other regulatory sequences or proteins are likely to be involved in this trait in developing grain.

During barley endosperm development, HvCslF6 and HvCslF9 are the most abundant transcripts of the HvCslF gene family. Expression of HvCslF9 at 3-8 DAP coincides with the initial appearance of (1,3;1,4)-β-glucan in endosperm cell walls 4,13 . However, the function of HvCslF9 is not entirely clear and it cannot be assumed that all members of the CslF gene family encode (1,3;1,4)-β-glucan synthases 19 . By contrast, HvCslF6 is well characterized as a (1,3;1,4)-β-glucan synthase that is expressed throughout grain development 12 . Variation in HvCslF6 transcript levels during late grain development coincide with differential accumulation of grain (1,3;1,4)-β-glucan in different barley cultivars 15 . Although it is unclear whether this variation is causative, it provides prima facie evidence that variation in HvCslF6 expression could lead to differences in (1,3;1,4)-β-glucan concentration in mature grain.
While several naturally occurring polymorphisms have been identified in non-coding regions of HvCslF6 [29][30][31] , only three are located within exons and only one affects the protein sequence (A590T). This non-synonymous mutation was identified in the betaglucanless (bgl) mutant 29 , where it is associated with reduced (1,3;1,4)-β-glucan content and poor agronomic characteristics. However, as reported in a wide range of barley accessions with phenotypic extremes for this trait 5,30 , the A590T SNP is not diagnostic for grain (1,3;1,4)-β-glucan content. Additionally, Taketa et al. 29 described three Sodium azide-induced alleles of bgl which are the result of non-synonymous mutations in highly conserved nucleotides and lead to a lack of (1,3;1,4)-β-glucan.
Two of the end user markets of barley have distinct preferences regarding grain (1,3;1,4)-β-glucan content. Barley cultivars with high (1,3;1,4)-β-glucan content are preferred by the food sector as this polysaccharide is not digested by humans and therefore contributes to total dietary fibre intake, acting against several human health conditions [32][33][34] . Conversely, the brewing and distilling sector require barley cultivars with low (1,3;1,4)-β-glucan content for an efficient malting process 35,36 . To date HvCslF6 is the main gene known to functionally contribute to the synthesis of grain (1,3;1,4)-β-glucan content. However, several studies 5,29-31 , have revealed a lack of variation in the HvCslF6 coding sequence (CDS), suggesting that another aspect of HvCslF6, such as the differential regulation of transcript abundance might contribute to variation in the observed phenotype. Previous quantitative trait loci (QTL) studies have suggested that in addition to HvCslF6, both HvCslF9 and HvGlbI could contribute to variation in the grain (1,3;1,4)-β-glucan 5,37 . To investigate this further we quantified transcript abundance of all three genes in a grain development series across several barley accessions with divergent grain (1,3;1,4)-β-glucan levels. In the same experiment we quantified (1,3;1,4)-β-glucan levels throughout grain development to assess the temporal deposition pattern of this polysaccharide. In parallel, to investigate sequence diversity of HvCsl6 in a wider collection of germplasm than previously examined, we used a combination of existing exome capture data and de novo sequencing of 3,000 bp of the HvCslF6 promoter.
The assessment of grain characteristics across 2-row cultivars (cvs): Dew, Imidis, Egmont and Gull used in the grain development series showed significant differences in average TGW, grain surface area and its components of width and length (p-value < 0.01) as shown in Fig. S1. The cvs Imidis and Egmont showed significantly higher average TGW (68.77 ± 0.20 and 59.95 ± 0.58, respectively) and larger grain area (26.57 ± 0.03 and 25.80 ± 0.10, respectively) compared to cvs Dew and Gull (p-value < 0.01, both). Significant differences in grain width (ranged 3.1-3.9 mm) and length (ranged 8.1-9.1 mm) were found across the four cvs (p-value < 0.01, both).
While HvCslF6 transcript was found at high levels during mid grain development stages (8-10 and 15 DPA), a genotype-dependent response was observed throughout grain development in the present study. In contrast, transcript abundance of HvCslF9 in all genotypes was extremely low and variable compared to HvCslF6 (Fig. 2b). HvCslF9 transcripts were mostly detected during early grain development (3-5 DPA) (cvs Dew, Imidis and Egmont) except in cv Gull where HvCslF9 expression peaked at 8-10 DPA (p-value 0.037 Tukey's test). From 15 to 38 DPA, HvCslF9 transcript abundance remained undetectable for most genotypes except for cv Dew which showed a relative increase towards late stages of grain development, although levels were very low compared to early grain development (3)(4)(5).
HvCslF6 has low levels of genetic variation in coding and non-coding regions within a diverse collection of barley. An exome capture sequencing dataset comprising 1,336 barley accessions (including wild, landrace, and elite barley) was used to survey HvCslF6 for sequence variation. Three polymorphisms (MAF < 5% cut off applied) were found within the HvCslF6 coding region: two synonymous SNPs (first and third exon) and a non-synonymous SNPs on the third exon A590T, also described by Taketa et al. 29 . 10 SNPs were identified in non-coding regions of HvCslF6, seven were located within the first intron and three towards the end of the second intron (Fig. 3). Based on the predicted HvCslF6 gene structure, detected SNPs within introns are not expected to affect splice junctions. Overall, the 13 SNPs found within HvCslF6 had a low frequency across the genotypes analysed (Table 2); alternate alleles were present in approximately 19% of the barley accessions included in this study except for two SNPs located in the first and second introns of HvCslF6 (46.3% and 39.9%, respectively).
Limited sequence variation in a −3,000 bp HvCslF6 putative promoter region. We identified 12 SNPs across the 3,000 bp promoter region of 35 barley genotypes described in Table S2. Two of these SNPs were located in the proximal promoter (defined as −500 bp upstream the start codon), which is generally known to Expression data was averaged from three biological replicates per genotype and normalized to α-tubulin, GAPDH and HSP70 housekeeping genes. qRT-PCR reactions were performed in triplicate. Bars indicate standard errors (n = 3). Letters above each bar indicate significant differences within each developmental stage determined by one-way ANOVA followed by Tukey's test. www.nature.com/scientificreports www.nature.com/scientificreports/ contain key cis-acting regulatory elements in plant species 38 , while the remaining polymorphisms were detected at least −500 bp upstream of the HvCslF6 start codon (Fig. 4). No correlation was found between the 12 SNPs identified within the region upstream of the HvCslF6 start codon and the natural variation in grain (1,3;1,4)-β-glucan content in the germplasm analysed (a subset of 25 elite barley cvs split in two groups with contrasting grain (1,3;1,4)-β-glucan content, 6.3% and 3.0% (w/w) respectively, Table S2). The HvCslF6 promoter sequences of Hordeum vulgare subsp. spontaneum 'Caesarea' and progeny lines, OSU105 and OSU127 contained six SNPs (five of them located in the distal and one in the proximal promoter region) that were unique compared with the other genotypes analysed ( Table 3). Two of these were present in a set of 12 polymorphisms previously reported in six Australian barley lines 15 . Both single nucleotide substitutions were detected in the distal promoter.
The dose-dependent role of HvCslF6 in (1,3;1,4)-β-glucan biosynthesis has previously been demonstrated in barley, where endosperm-specific over-expression led to a considerable increase in (1,3;1,4)-β-glucan concentration 12 . In addition, the functional characterization of HvCslF6 orthologs confirmed a similar role in wheat 44 and rice 45 . In the present study, HvCslF6 and HvGlbI mRNA abundance was examined in four elite barley cvs with contrasting grain (1,3;1,4)-β-glucan to understand if the combination of transcript abundance of genes encoding biosynthetic and hydrolytic enzymes could explain mature grain (1,3;1,4)-β-glucan content. Although transcript levels and profile varied between the cvs during grain development, the combined variation did correlate with mature grain (1,3;1,4)-β-glucan levels for some cvs, but not for others. It has been observed previously that just considering the expression profile of HvCslF6 does not describe grain (1,3;1,4)-β-glucan content in Australian cvs Sloop and Himalaya 13 , and more recently in cvs CDC Bold, Beka, Logan, Harrington and breeding lines 'TR251' , 'TR306' 15 . However, qRT-PCR showed that for some cultivars (Dew and Egmont) at least a balance between biosynthetic and hydrolytic transcript abundance, and hence activities, likely impacts final (1,3;1,4)-β-glucan content. For others (Imidis and Gull) the lack of a direct relationship between transcript abundance of HvCslF6 and HvGlbI, and grain (1,3;1,4)-β-glucan content raises several possibilities. One, that there may be other genes contributing to final grain (1,3;1,4)-β-glucan content, and two, that wholegrain transcript abundance may not necessarily correlate with protein amount and enzyme activity. Although the lack of polymorphisms in the HvCslF6 coding sequence (see below) argues against any cultivar-specific modification to enzyme activity, it is possible that in some cultivars, polymorphisms in non-coding regions could impact interactions with upstream transcription factors and this could contribute to variation in transcript abundance and grain (1,3;1,4)-β-glucan.
Based on the current grain development series, 2-row genotypes (cvs Dew, Imidis, Egmont, Gull) had highest levels of HvCslF6 expression at 8-10 DPA and 15 DPA during grain development. Similarly Nemeth et al. 44 observed that the wheat CslF6 ortholog, TaCslF6, was highly expressed during mid-endosperm development in cv Hereward, corresponding to the grain filling stage, showing maximum expression levels from 8-15 DPA. In a more recent study, the average TaCslF6 mRNA levels across 10 wheat cvs were abundant at 21 and 28 DPA although different expression patterns were found within this subset of wheat genotypes, showing significant differences in TaCslF6 transcript abundance at 21, 28 and 40 DPA 46 . These expression profiles are consistent with a function of (1,3;1,4)-β-glucan as a destination for energy storage in the form of metabolizable glucose during grain filling and germination 47,48 . As mentioned in Wong et al. 15 , differences in HvCslF6 expression levels during late grain development might impact (1,3;1,4)-β-glucan accumulation in mature barley grain. In our experiment cv Egmont [6.09% (w/w) ± 0.19] showed an increase in HvCslF6 expression at 38 DPA and no HvGlbI expression at this time point in contrast to low grain (1,3;1,4)-β-glucan cvs Dew [4.13% (w/w) ± 0.25] and Imidis [4.26% (w/w) ± 0.14]. Additionally, cv Egmont showed a 14.10% increase in (1,3;1,4)-β-glucan content from 32 DPA and 38 DPA coupled with an increase in HvCslF6 expression, but a lack of HvGlbI.
In the cvs included in the present study the transcriptional profile of HvGlbI has a peak in expression at 15 DPA and 24-26 DPA, suggesting a potential role for this enzyme in determining (1,3;1,4)-β-glucan content during the later stages of grain development. A particularly high level of HvGlbI expression was observed at  www.nature.com/scientificreports www.nature.com/scientificreports/ 15 DPA in cv Dew, which is characterized by a relatively low (1,3;1,4)-β-glucan content in mature grain compared to cvs Egmont and Gull (both with higher (1,3;1,4)-β-glucan content). In contrast, for other genotypes we observed concurrent high or low HvCslF6 and HvGlbI transcript, which could initially appear to be counterintuitive. Concurrent synthase and hydrolase activity may contribute to re-modelling of the nascent polysaccharide, to hydrolysis of the polysaccharide off the synthase, or provide a rapidly degradable source of glucose in the developing grain due to its unbranched structure compared to starch 20,47,49 . Additionally, potential expression differences in (1,3;1,4)-β-endoglucanase HvGlbII, other putative β-glucan glucohydrolases (HvExoI and HvExoII) or β-glucosidases that were not investigated in this study may also contribute to variation in grain (1,3;1,4)-β-glucan content. Recently, Betts et al. 28 reported expression differences of (1,3;1,4)-β-endoglucanases (HvGlbI and HvGlbII) in simulated malting conditions across malting and feed barley cvs. HvGlbI expression was three-fold higher than HvGlbII whose expression is known to be restricted to the aleurone under GA regulation 26 . However, since a direct relationship between transcript abundance and activity levels cannot be assumed, the quantification of (1,3;1,4)-β-glucanase activity in future studies could provide useful data towards understanding variation during grain development in the cultivars investigated here. Alternatively, this could indicate that the relationship between HvGlbI expression, isoenzyme EI activity, and consequently (1,3;1,4)-β-glucan content is nonlinear. It is possible that HvGlbI acts in combination with other remodelling and degrading enzyme partners to explain differences in grain (1,3;1,4)-β-glucan content.
Based on the grain development series used in the current study, the expression of HvCslF9 was predominantly detected from 3-5 DPA, except for a later expression peak in cv Gull (8-10 DPA). Burton et al. 13 also observed a peak of expression for HvCslF9 for cv Sloop at 8 DAP, while at other time points in both genotypes included in their study HvCslF9 transcripts were absent or at extremely low levels. Similarly, in the current study during mid and late grain development, and coinciding with endosperm differentiation and maturation, HvCslF9 expression was almost undetectable across all genotypes. While other independent HvCslF9 expression studies support a peak expression of this gene during endosperm differentiation stages 13 , HvCslF9 over-expression did not increase (1,3;1,4)-β-glucan content in transgenic barley grain 12 , with similar results obtained in wheat addition lines 50 . The fact that the genomic location of HvCslF9 and HvGlbI [1 H, 48.1 cM and 1 H, 54.4 cM, respectively 51 ] co-locates with association peaks based on two independent QTL and genome-wide association studies (GWAS) on grain (1,3;1,4)-β-glucan content 5,37 might suggest that variation in (1,3;1,4)-β-glucan is due to HvGlbI (1,3;1,4)-β-glucanase activity and not HvCslF9 biosynthetic activity.
In the current study, a survey of HvCslF6 sequence variation in 1,336 exome-captured barley accessions identified three SNPs within the HvCslF6 coding sequence, two synonymous and one non-synonymous SNP, A590T which doesn't explain differences in grain (1,3;1,4)-β-glucan content 5,15,29,30 . Based on the model of HvCSLF6 described by Schwerdt et al. 52 , the A590T substitution is proximal to a 55 amino acid insert that is specific to CSLF6 proteins of the grasses and absent in other CSLF proteins. While several amino acids under positive selection (non-synonymous to synonymous nucleotide substitution rate greater than 1) were described flanking the 55 amino acid insert and at other predicted transmembrane domains, low levels of sequence variation were found in CSLF6 unlike other CSLF proteins in grasses. The role of the CSLF6-specific amino acid insert, far from the active site, is not entirely clear however it contains charged aa residues and a conserved Cys residue which might facilitate the interaction with other protein/s 12,52 .
In the same exome-captured dataset, 10 SNPs were detected in non-coding regions of HvCslF6, mostly within the first intron, confirming that natural variation within this gene is rare. Recent work in Brachypodium distachyon identified a trihelix family transcription factor (BdTHX1) which binds to a GT-motif in the second intron of BdCslF6 53 . The same motif is found in the first intron of HvCslF6 and another putative GT-motif is present in the second intron. However, the 10 SNPs within HvCslF6 introns identified in this study did not affect the THX1 binding site or GT-motifs described by Fan et al. 53 .
Similar to the intronic regions of HvCslF6, low levels of variation were identified in the putative HvCslF6 promoter sequence. In a −3,000 bp region upstream of the HvCslF6 start codon, DNA sequence alignments identified only 12 SNPs across 35 barley accessions representing natural (1,3;1,4)-β-glucan variation. Polymorphisms were mostly found in the HvCslF6 distal promoter, in which two SNPs were found in common with a set of six Australian cultivars previously analysed for sequence variation in 5' and 3' UTR regions 15 . These data indicate that there is remarkably narrow variation within HvCslF6 promoter and enzyme coding sequences. None of these SNPs appeared to associate with variation in (1,3;1,4)-β-glucan content. Although genetic diversity will likely be lower when considering elite cultivars, Russell et al. 54 observed 745,815 SNPs in high confidence exons across 20,729 high confidence gene models, representing an average of 35.9 SNPs per gene model in a collection of 267 wild and landrace barleys. The low sequence diversity in HvCslF6 identified in our current study and by others 15,30,31 appears to highlight atypical conservation of the HvCslF6 nucleotide sequence, possibly due to its indispensable role in (1,3;1,4)-β-glucan synthesis and the importance of this polysaccharide in the grasses, or a common origin of the gene 55 . This role has been confirmed via the characterization of barley chemically-induced mutants that completely lack grain (1,3;1,4)-β-glucan and exhibit poor agronomic traits 29,56 , or partially functional HvCslF6 mutants with less severe phenotypic effects 31 . Thus far, evidence suggests that the basis for differences in grain (1,3;1,4)-β-glucan is not linked to any polymorphism within HvCslF6 and hence other regulatory sequences or proteins, acting independently or in combination, are likely to be involved.
Mature grain characteristics, including size and shape, were not obviously correlated with the amount of grain (1,3;1,4)-β-glucan or expression differences in genes that produce enzymes that synthesize or hydrolyze (1,3;1,4)-β-glucan. Hence, natural diversity for barley grain size seems unlikely to affect (1,3;1,4)-β-glucan concentration. This is despite recent work on the barley lys3a shrunken endosperm mutant showing that defects in hordein synthesis could potentially affect HvCslF6 transcript abundance 57,58 . Shrunken endosperm barley mutants typically exhibit defects in starch biosynthesis and embryo development, but the (1,3;1,4)-β-glucan content was not quantified 59 . Since (1,3;1,4)-β-glucan has been proposed to act as an alternative source of stored www.nature.com/scientificreports www.nature.com/scientificreports/ glucose 60 , knowing the (1,3;1,4)-β-glucan content of shrunken endosperm mutants may help to explain the role of this polysaccharide and the regulation of HvCslF6. Furthermore, by comparing Brachypodium and barley grain during development Trafford et al. 61 suggest that starch synthesis influences endosperm cell enlargement and as a consequence endosperm size. Therefore, although the relationship between starch and grain size is well understood, perhaps a wider screen of mutant germplasm 62 might reveal a link between grain size and shape, and (1,3;1,4)-β-glucan content.

Material and Methods
Plant material for grain development series. Four elite 2-row spring cvs: Dew, Imidis, Egmont and Gull were used in this experiment. Based on previous work 5 , these cvs represent phenotypic extremes of grain (1,3;1,4)-β-glucan concentration (2.1-6.7% w/w). Three biological replicates across six grain development stages: 3-5, 8-10, 15, 24-26, 32 and 38 days post anthesis (DPA) were collected from material grown in glasshouse conditions of 16 h light/8 h dark, until maturity at The James Hutton Institute, United Kingdom (Feb-May 2015). In the present grain development series, DPA were considered equivalent to days after pollination (DAP), which has been used in previous studies 4,13,15 . For each genotype, six developing grains from the central part of the spike were collected at each developmental stage and snap frozen in liquid nitrogen for further mRNA and (1,3;1,4)-β-glucan quantification analyses.
Isolation of mRNA and cDNA synthesis. For mRNA isolation, we collected three biological replicates from the time points described above, and for each time point included three technical replicates. Samples were ground into a fine powder (100 mg) in liquid nitrogen and mixed with 1 mL TRIzol (Thermo Fisher Scientific, Waltham, USA). Manufacturer's instructions were followed with several modifications: (1) After phase separation, supernatant was transferred to a fresh tube and mixed with 0.25 mL isopropanol followed by 0.25 mL 0.8 M sodium citrate/1.2 M sodium chloride (per 1 mL TRIzol used, both) to help removal of polysaccharides. Samples were incubated for 10 min, at room temperature. (2) After an ethanol wash (1 mL 75% ethanol per 1 ml of TRIzol used), RNA pellets were re-suspended in 400 µL RNase-free water on ice. (2) For RNA purification, 400 µL chloroform/isoamyl alcohol (Sigma Aldrich, St. Louis, USA) was added, mixed well and centrifuged (16,000 × g, 4 °C) for 10 min. The top phase (~300 µL) was removed to a clean 1.5 mL Eppendorf tube on ice and 100 µL RNase-free water was added to the original tubes, mixed well and centrifuged (same conditions) for 5 min. The top phase (~100 µL) was removed and added to the 300 µL already collected in tubes, remaining on ice. (3) For RNA precipitation, samples were mixed with 1 mL absolute ethanol and 40 µL 3 M sodium acetate and stored at −80 °C overnight. The next day, samples were centrifuged (16,000 × g, 4 °C) for 10 min and supernatants were removed. RNA pellets were washed in 1 mL 70% ethanol and re-suspended in 100 µL RNase-free water. cDNA synthesis (1 µg total RNA) was performed using the RNA to cDNA EcoDry ™ Premix (Takara, Kyoto, Japan) according to the manufacturer's instructions.
qRT-PCR. Quantitative real-time PCR (qRT-PCR) was performed in a StepOne Real-Time PCR machine (Thermo Fisher Scientific, Waltham, USA) using PowerUp SYBR Green Master Mix (Thermo Fisher Scientific, Waltham, USA). Three replicate qRT-PCR reactions were performed for each cDNA sample including three negative controls using RNAse-free water. Gene specific primers and qRT-PCR reaction conditions were used as described in previous studies 12,13 . Absolute mRNA quantification of HvCslF6, HvCslF9 and HvGlbI was performed using three housekeeping genes for normalization: α-tubulin, (α-tub) glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and heat shock protein (HSP70) (Table S3). Multiple control gene normalization was performed as described in Vandesompele et al. 63 . Normalized mRNA copies were calculated based on standard concentrations. Each gene standard (10 1 -10 7 mRNA copies/µL) for HvCslF6, HvCslF9, HvGlbI, α-tubulin, GAPDH and HSP70 was generated by HPLC at The University of Adelaide 13 .
Phenotypic assessment of grain characteristics. A collection of 153 2-row elite barley lines was used for grain phenotypic measurements (Table S2). A sample of approximately 100-150 bulked seeds per genotype was used to measure seed weight, roundness, length, width, thickness and overall seed area using a SeedCount SC4 (Seed Count Australasia, Condell Park, Australia) at the University of Adelaide Barley Breeding Program/ Laboratory, following manufacturer's instructions. These data where combined with previously published grain (1,3;1,4)-β-glucan data 5 .
HvCslF6 promoter resequencing. In  www.nature.com/scientificreports www.nature.com/scientificreports/ genotypes were chosen based on their introgressions from cv Caesarea (Hordeum vulgare subsp. spontaneum), which cover the genomic location of HvCslF6. Genomic DNA was isolated from seedlings using the Qiagen DNeasy Plant Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. PCR amplification of a −3,000 bp HvCslF6 upstream region was achieved by dividing the fragment into four overlapping PCR reactions (Table S4). PCR cycle details are as follows: 98 °C for 2 min; 5 cycles of 98 °C for 15 s, 60 °C for 20 s, 72 °C for 1 min with a touchdown on the primer annealing step of −1 °C/cycle; then 35 cycles of 98 °C for 15 s, 56 °C for 20 s, 72 °C for 1 min; followed by 72 °C for 2 min. PCR amplicons were purified using ExoStar ™ (GE Healthcare UK Ltd., Buckinghamshire, UK) following the manufacturer's instructions. Sanger Sequencing was performed using an ABI3730 DNA Analyzer (Applied Biosystems Inc., Foster City, USA) at The James Hutton Institute. Resulting sequences were aligned to cv Morex HvCslF6 reference sequence retrieved from the barley genome explorer, Barlex 51 and analysed with Geneious V.9 67 to identify polymorphisms.
Exome capture dataset. The genomic position of HvCslF6 (HORVU7Hr1G070010.3) on the physical map was retrieved using the Barlex database 51 , available at: http://www.barlex.barleysequence.org. Sequence variation of HvCslF6 was analysed in 1,336 exome sequenced barley accessions (Unpublished, The James Hutton Institute). This dataset comprises a georeferenced collection of exotic barley alleles including 340 landraces (Hordeum vulgare ssp. vulgare), the Spanish core collection 68 , 288 wild lines (Hordeum vulgare ssp. spontaneum) of which 80 correspond to the Barley1K collection from Israel 69 and 2-and 6-row collections of contemporary European barley cultivars evaluated in previous projects (WHEALBI and CLIMBAR) at The James Hutton Institute. Polymorphisms identified in barley accessions with suspected heterozygosity and minimum allele frequency (MAF) of ≤5% (cut-off) were removed from the analysis as described in Russell et al. 54 .