The formation of hulled seeds represents a distinctive characteristic of barley since no other crop of the Poaceae family shows this character. To dissect its genetic architecture, we performed GWAS on a panel of spring barley landraces comprising 222 2-rowed and 303 6-rowed accessions. The results revealed that loci on four different chromosomes 2H, 3H, 6H, and 7H were associated with the adherence of husks to the caryopsis. Marker-trait associations for the first three chromosomes turned out to be subpopulation-specific, while associations on 7H were detected across all subpopulations (Table 1).
Nud is a major locus for the naked caryopsis
The 7H locus was detected in Lr_all as well as all subpopulations. Hence, the 7H locus represents the major and geographically most widespread locus influencing the adhesion of hulls to the mature seeds.
The regulation of cutin biosynthesis pathways by Nud leads to a highly permeable lipid layer on the caryopsis responsible for hull-caryopsis fusion. The deletion or a low expression level of the Nud gene entails a naked caryopsis phenotype [37]. High associations were observed on 7H in a chromosomal region ranging from 70.8 to 86.6 cM. Since Nud-specific SNPs are not included on the SNP array employed in this study, the position of the gene was estimated to be at 78.33 cM (MLOC_59305.1) by comparison of the SNP data from this study and to the reference sequence available [38]. This position is in good agreement with the approximate location (75–80 cM) given by [39]. As expected, GWAS within the entire panel revealed the highest association on 7H for marker BOPA2_12_30301, mapping at 79.8 cM, most closely to the putative genetic position of the Nud gene at (78.33 cM or 546.6 Mbp, [38]) (Additional file 3: Table S4). The same result was observed for the subpanels Lr_6, and Lr_6Eur, while the subpanel Lr_6Eth peaked at 78.1 cM (BOPA2_12_30996) even closer to the genetic position of the causal gene (Additional file 3: Table S5). Altogether, this supports the hypothesis that in these populations naked seeds are the result of genetic variation in Nud. By contrast, the 2-rowed subpanels Lr_2Eth and Lr_2Eur revealed the highest associations at 84.4 cM and 70.8 cM, respectively.
The wide range of associations flanking the putative locus of the Nud gene up to 7.5 and 8.3 cM to both sides reflects the complex pattern of linkage disequilibrium among the corresponding markers (Additional file 3: Figure S6a) corresponding to previous findings of by [34].
Novel associations identified by analysis of subpopulations
Until now, Nud on chromosome 7H is the only cloned gene reported to control the naked barley caryopsis trait, which supports a monophyletic origin of domesticated naked barley [13]. However, other authors have suggested more than one origin of domesticated naked barley [3, 40]. Pasam et al. [21] reported a low level of genetic relatedness (0.49) between naked barley from Ethiopia and other geographical origins and further suggested at least two evolutionary lineages of the naked barley, both of which probably originated from the Eastern Fertile Crescent. In this regard, the detection of additional loci indicates a complex evolutionary history of naked barley.
In addition to the 7H locus, GWAS of Lr_all revealed the presence of a second hitherto unknown association on chromosome 2H at 91 cM (Fig. 3, Additional file 3: Table S4). Since spike type is a major determinant of population structure, a separate analysis of the 2- and 6-rowed panels was performed to investigate whether the new locus would be confined to one of the two subpopulations. This approach revealed that the 2H locus is only present in 6-rowed accessions. When separately analyzing naked barley lines from Ethiopia and Eurasia, this QTL could be traced back to Lr_6Eur. A second locus on 2H at ~ 58 cM was dected in both six-rowed subpanels Lr_6Eur and Lr_6Eth. While in Lr_6Eth the number of associated SNPs increased spanning a genetic interval from 52.3–58.8 cM, only two SNPs of that locus were significant in Lr_6Eur. The panel of Lr_6Eth revealed a third locus on 2H at 76.8 cM (Table 1). The highest associations in the Ethiopian subpanel were found on 2H at about 58–59 cM in the centromeric region containing the flowering time gene HvCEN. The extent of LD between associated SNPs on 2H shows a complex pattern indicating the possibility that this region might harbor several loci and not only one (Additional file 3: Figure S6b). The co-localization with a flowering time gene might arise from the fact that more hulled barley accessions originate from Europe, while the majority of naked accessions comes from the region West Asia and North Africa (WANA). Therefore, the QTL in the vicinity of HvCEN might arise from the underlying population stratification regarding European origin (hulled) vs WANA (majority of naked), while the role of the other two QTL on 2HL remains unclear.
Similarly, analysis of Lr_2 yielded a novel locus on chromosome 3H (Fig. 3). Further subdivision of the naked barley lines in Lr_2Eth and Lr2_Eur demonstrated that the association is confined to naked accessions from Eurasia (Fig. 4). LD analysis of all significant SNPs detected on 3H in Lr_2 revealed a high level of LD between the four markers at positions 45.4 and 46 cM, while the SNP at position 51.1 cM shows only moderate LD to the other SNPs (Additional file 3: Figure S6c). However, when only Eurasian naked barleys were considered in the analysis (Lr_2Eur), all 3H SNPs were in high LD with each other and with the Nud locus (Additional file 3: Figure S6d).
Finally, one association on 6H was detected only in 6-rowed accessions from Ethiopia (Fig. 4, Additional file 3: Table S5). It is marked by two SNPs at 65.9 cM, which are in very high LD (Additional file 3: Figure S6e).
In a recent association analysis of a large germplasm set comprising 2417 accessions fingerprinted with the same iSelect array, significant associations were detected with SNPs located around the Nud locus as well as with unmapped SNPs [34]which might correspond to additional loci as in our study. However, inspection of the updated positions of these unmapped SNPs revealed their location on 7H, close to the Nud locus. A possible reason why no footprints of selection were detected in this study is the low frequency of naked accessions in the corresponding panel (9% vs 17% in the present study). The analysis of subpopulations, as it was performed in the present study, proved crucial for the detection of these loci, likely due to the increase in allele frequencies and the reduction of confounding effects. The detection of multiple, subpopulation-specific associations might be the result of two different scenarios. In the first scenario, this reflects a complex genetic architecture of the naked caryopsis trait, giving rise to several QTL involved in trait expression. Alternatively, the observed QTL represent footprints of the selection of additional traits in naked barley following a geographic (Ethiopia vs Eurasia) or morphological pattern (2-rowed vs 6-rowed).
PCR analysis reveals Nud as the main locus for hull-caryopsis adherence
Regarding the first scenario, PCR results and sequence data revealed, that Nud is the main locus involved in hull-caryopsis adherence. This is supported by our GWAS results and confirms the hypothesis of [13] that Nud is completely deleted in all naked accessions. Nevertheless, one landrace originating from Ethiopia (Entry 1365) yielded an amplicon of the diagnostic size for covered barley. This is an indication that, although this accession is hull-less, the entire 17 kb fragment is not deleted as would have been expected. Also, the phenotype of this accession showed increased hull-caryopsis adherence (Fig. 6). Whether this is due to the presence of a second locus other than Nud or owed to a different mutation of Nud, not resulting in a size polymorphism, needs to be further investigated. The primer pair, wF2, and tR2, which amplified an approximately 850 bp fragment in covered barley, binds at a far distance (around 2400 bp) off the genetic position of the Nud gene. Hence, the PCR assay does not reflect the allelic status of the gene itself. On the other hand, the presence of additional allelic variation at the Nud locus or even a mutation in another gene giving rise to naked caryopses would be anything but unexpected, given the large number of genes involved in the biosynthesis of lipids [41] and the presumption that there could be a quantitative variation in the formation of the cuticular lipid covering the pericarp hence making the naked trait a quantitative rather than a qualitative trait [34].
Novel associations indicate footprints of selection
Munoz-Amatriain et al. [34] predicted the presence of additional QTL other than Nud as a result of a difference in the degree of hull adhesion and suggested the presence/absence of husks be a quantitative rather than a qualitative trait.
As we could not see any phenotypic differences between naked accessions except one accession (entry 1365) and the deletion at the Nud locus is pervasive in naked barley except for entry 1365, it is deemed unlikely, that the associations detected outside the Nud locus are functionally related to the formation of a naked caryopsis. Thus, we suspect that the observed associations on chromosomes 2H, 3H, 6H instead represent footprints of selection for hitherto unknown adaptive or end-use traits. Since, seed protein content is a well-known determinant of the nutritional value of seeds, we investigated, if the presence of the identified footprints of selection was correlated with seed protein content.
Compared to hulled barley, naked barley has an increased protein content [35, 36]. Also, this study revealed differences in protein content between naked accessions of different origin (Eurasian having a lower protein content than Ethiopian landraces).
The QTL region on 3H (45.4–51.1 cM) detected in the 2-rowed Eurasian panel harbors Epb1 (http://www.barleyhub.org/barley/breeding/), a major barley cysteine proteinase involved in endosperm protein degradation [42]. The difference in protein content between naked barleys from Eurasia and Ethiopia could a result of unconscious or directed selection towards different end-uses, especially, as naked barley is one of the primary food sources in Ethiopia. However, since the 3H QTL was restricted to the Lr_2Eur subpanel, it seems unlikely, that allelic variation in Epb1 would account for the observed differences in protein content of 2-rowed and 6-rowed naked barleys from the Ethiopian and Eurasian subgroups. Moreover, inspection of the recently published reference sequence of barley revealed two paralogues, Epb1 and Epb2 (GB accession no. U19359.1 and U19384.1, respectively) which are very similar (94% sequence identity by comparing both sequences). By performing a BLAST analysis of both genes against genomes of different barley lines (three cultivars, a landrace and a wild barley line, M. Mascher personal communication), we found that both genes are mapping outside the QTL region 109 cM or 636 Mbp, spaced at a distance of only 20 kb.