Marker development using SLAF-seq and whole-genome shotgun strategy to fine-map the semi-dwarf gene ari-e in barley

Semi-dwarf genes were crucial in the green revolution because they conferred semi-dwarfness, which reduced lodging and increased crop yields [1–3]. Most of the green revolution semi-dwarf varieties resulted from a gene mutation in the gibberellic acid (GA) pathways, e.g. rice sd1 gene, wheat DELLA genes and maize d8 [1–5]. For barley, the world's fourth leading cereal crop, semi-dwarf genes such as sdw1/denso, uzu1 and ari-e have been extensively explored and widely accepted in barley breeding programs in different regions. Barley sdw1/denso varieties carry a mutation in GA 20 oxidase-2, a GA biosynthesis gene that is homologous to rice sd1 [6–9], which has been used to develop short-stature cultivars in western USA, Canada, Australia and European countries [10–14]. The uzu1 mutant had a significantly reduced response to brassinolide resulting from a missense mutation in the BR receptor gene HvBRI1 and has been widely used in East Asia [14–17]. The ari-e gene from Golden Promise is a commonly-used semi-dwarf gene in spring barley in Europe, especially Scotland. A new barley variety with the ari-e gene, Hindmarsh, has been grown in nearly half of Australia. Besides the magnitude of the effect of this gene on reducing plant height, ari-e has significant effects on some agronomic traits, such as grain size, grain composition, malting quality and yield [18, 19]. Golden Promise is a model cultivar for barley genetic transformation [20, 21]. Recently, some studies have reported that ari-e is the only semi-dwarf gene to improve salt tolerance [19, 22–24]. Genetic analysis showed that ari-e was a recessive semi-dwarf gene and was mapped on barley chromosome 5H as a quality character or quantitative trait locus (QTL) using traditional markers and next-generation sequencing technologies [19, 22, 23, 25, 26]. However, the ari-e gene is yet to be identified.

Map-based cloning is the most promising approach for molecular isolation of target genes. The first map-based cloning gene in barley was the powdery mildew resistance gene mlo [27]. A growing number of genes have been isolated from barley based on genetic information. Eight genes are involved in disease resistance against powdery mildew (mlo, Rar1, mla and Ror2) [27–30], stem rust (Rpg1 and Rpg4/Rpg5) [31, 32], barley yellow mosaic virus (rym4/rym5) [33] and leaf strip (Rdg2a) [34]. Besides resistance genes, some of the genes identified by map-based cloning contribute to complex agronomic traits such as photoperiod sensitivity (Ppd-H1) [35], vernalization requirement (Vrn3) [36], row type (Vrs1) [37], tolerance to boron toxicity (Bot1) [38], covered/naked caryopsis (nud) [39], anthocyanin pigmentation (ant2) [40], floret closing (Cly1) [41], awn length (lsk2) [42], increased number of internodes (mnd) [43], leaf color variation (HvSGRA) [44] and branch development (Cul4) [45]. A larger number of molecular markers are the basis of map-based cloning for genes related to complex agronomic traits. There are thousands of molecular markers (RFLP, AFLP, SSR, STS, SNP and DArT) available for barley genomes (http://wheat.pw.usda.gov/GG2/index.shtml, http://bioinf.scri.ac.uk/barley_snpdb/index.html, http://www.diversityarrays.com/), allowing an efficient and rapid localization of genes at low resolution. Further narrowing of the target region and distinguishing which gene is responsible for the interested traits is limited by the size of the segregating population and molecular marker density.

Recently, the ongoing revolution in sequencing techniques has significantly reduced costs and advanced technologies to provide new strategies for discovering thousands of markers. Next-generation sequencing (NGS) facilitates the rapid identification of genomic variants for marker development in most species. For example, NGS identified 416,856 markers in wheat [46] and 11,805 SNP markers in cotton [47]. In barley, 1,391 high confidence SNPs generated from genotyping by sequencing (GBS) were used to construct a linkage map of doubled haploid (DH) Golden Promise × Morex (GPMx) RILs population [26]. Aligning genomic DNA sequences between two sequenced barley varieties, Morex and Barke, identified 436,640 InDel markers [48].

Bulked segregant analysis (BSA) combined with NGS technologies offers new strategies for marker discovery associated with genes/QTLs. Specific-length amplified fragment sequencing (SLAF-seq) is considered an efficient and high-resolution genotyping method [49], which has identified major QTLs in maize [50], soybean [51], cucumber [52], rice [53] and candidate gene isolation in cucumber [54] and barley [44]. In this study, traditional marker assays and the combination of BSA with SLAF-seq were used to detect a genomic region in barley harboring the semi-dwarf gene ari-e. SNPs and InDel markers were developed by SLAF-seq and whole-genome shotgun sequencing technology, respectively, to fine-map ari-e. The current study exemplifies the use of SLAF-seq for SNP discovery and whole-genome shotgun sequencing for InDel development in the target region, which may pave the way for map-based cloning of the ari-e gene and unraveling the molecular mechanisms of the semi-dwarf phenotype.

Traditional methods for developing markers are tedious, time-consuming and expensive [57]. Limited number of markers in the barley genomic region is a major obstacle in the fine-mapping of barley genes, which has encouraged the development of new marker systems that have a greater degree of polymorphism. SLAF-seq is a relatively new experimental method which is considered an efficient and high-resolution strategy for large-scale genotyping using an enhanced reduced representation library (RRL) sequencing method [49]. It is an effective, low-cost technology for constructing high-density linkage maps, which has successfully facilitated the identification of major QTLs related to complex traits in plants [50, 51, 53]. Using the SLAF-seq approach, an SNP-based saturated genetic map was constructed and a major fruit length QTL, which explained 44.6 % of the phenotypic variance in cucumber, was detected [52]. Combining SLAF-seq with BSA identified chromosome regions related to 1000 grain weight in rice and the gene that control inflorescence meristem function in maize [50, 53]. Xu et al. [54] reported that fruit flesh thickness in cucumber was fine-mapped by two SLAF markers in the 0.19 Mb genomic region. After anchoring the markers to the cucumber 9930 reference genome, 20 genes were predicted, and Csa2M058670.1 was identified as a good candidate gene due to its high expression level in thick fruit flesh varieties. In some cases, the interval identified by SLAF-seq is too large to isolate the target gene. The mapping region of barley HvSGRA (Stage Green-Revertible Albino Gene) defined by SSR and SLAF-seq, was insufficient to pin the candidate gene in barley [44]. Therefore, new markers based on the barley physical map were designed to further fine-mapping and its candidate gene was isolated according to its annotation and a nonsense mutation. In the present study, 319,656 SLAF tags were developed by high-throughput sequencing, with 82,284 SNPs identified (Table 3). Associated analysis with the SNP-index obtained semi-dwarf gene ari-e related candidate regions on chromosome 5H with a size of 181.03 Mb (Fig. 2). The large candidate region of ari-e by SLAF-seq was possibly due to the low recombinants between semi-dwarf and high pools in the target region. After combining with primary mapping, the SNP markers converted from polymorphic SLAFs provide useful data for restricting the candidate-associated regions into 7.2 cM interval, delimiting an approximate 3.6 Mb region in 119 Dash × VB9104 DH lines. Therefore, SLAF-seq is a highly-efficient strategy for mapping the candidate gene, but further work needs to be done to refine the target interval to isolate the candidate, especially in barley with such a large haploid genome.

InDel markers are a gel-based molecular marker which has played a major role in genetic studies and QTL mapping in rice and Arabidopsis [58–60]. The recent release of barley draft assembly sequences provides a new approach to finding insertion/deletion polymorphisms in barley varieties at the DNA level. A total of 436,640 InDels were identified between two sequenced barley varieties, Morex and Barke, based on genome-wide alignment, and 1,140 InDel markers were integrated with the barley consensus map [48]. Taking the same approach, we developed InDel markers in the target region; the newly-designed polymorphic InDel markers were linked with ari-e (Fig. 1b, c). Of these, InDel-17 marker co-segregated with ari-e and we narrowed the ari-e gene to the region flanked by InDel-16 and DGSNP21 in the Dash × VB9104 DH population (Fig. 1b). Most of the developed markers were co-segregated, possibly due to the small population size. Furthermore, we validated the mined markers to another mapping population (Hindmarsh × W1 DH population), and the ari-e gene was flanked by InDel-7 and InDel-17 (Fig. 1c). Combining both mapping results, ari-e can be narrowed between InDel-16 and InDel-17. As a result, alignment of the whole-genome shotgun sequence with the released barley genome sequence enabled the discovery of a large number of DNA markers. This study is an example of the efficient exploitation of targeted InDel markers developed by the whole-genome shotgun strategy in barley. Different alleles at the ari-e locus such as such as ari-e.1, ari-e.119, ari-e.156 and ari-e.228 described by Kucera et al. [61], were induced by ionizing radiations (neutrons and X-ray) and by chemical mutagens (propyl methanesulfonate and N-methyl-N-nitrosourea). Allelism to the ari-e mutant showed that ari-e.1 and Golden Promise mutant are alleles [62]. The variety Golden Promise (a gamma-ray mutant of the variety Maythorpe) with ari-e is famous for its desirable agronomic traits such as earliness, short stiff straw, reduced awn length and better salt tolerance compared with other semi-dwarf varieties [23, 63]. Considering its importance to barley breeding, ari-e has been genetically characterized in numerous genetic backgrounds. Thomas et al. [25] used the phenotypic markers in both F2 populations, and a doubled monoploid population to genetically map the ari-e locus to chromosome 5H of barley near the centromere. In the AFLP and SSR marker map constructed for the Derkado × B83-12/21/5 DH population, the ari-e gene was flanked by SSR marker Bmag337 (3 cM) and Bmag357 (7 cM) [19]. After genotyping with the 1,536-SNP Illumina GoldenGate oligonucleotide pool assay in a three-way barley cross, Malosetti et al. [64] identified that ari-e was close to SNP marker 2_1239 which has been included in the barley integrated map of Aghnoum et al. [65] in 5H bin 6. Based on Illumina’s oligo pool assays (OPAs) (the same SNP genotyping), ari-e is associated with SNP markers 1_1198 to 2_0449 in 5H bins 2 to 9 in BW043, which resulted from the introgression ari-e into the background of cv. ‘Bowman’ [66]. All these results confirm that barley 5H bin 6 encompasses the major semi-dwarf gene ari-e. Using GBS data of GPMx RILs population, ari-e is closely linked with GBS marker MR_47526P1793R57, which is flanked by MR_335403P1239R45 and MR_1560792P1192F41 [26]. These markers defined a 7.2 cM interval on the GPMx GBS map. According to the barley IBSC physical map, the interval defined at least 46 Mb and contained an estimated 397 genes [26]. The same interval spanned a total sequence length of approximately 10 Mb on the barley POPSEQ map [26]. The authors explained that both the relatively low recombination and lack of detected polymorphism resulted in the low-resolution map in the ari-e region of the GPMx RILs population. In our initial mapping, ari-e was defined at a 9.1 cM interval, which corresponds to 17.5 Mb on the barley POPSEQ physical map. Although the genetic distance was similar, the physical map interval of ari-e defined by M47526 and InDelHvces8 in the Dash × VB9104 DH population was larger than that delimited by MR_335403P1239R45 and MR_1560792P1192F41 in the GPMx RILs population. The target region decreased to 6.8 cM covering the physical map from 487,890,066 bp to 490,942,368 bp intervals in the Dash × VB9104 DH population. After we had validated the polymorphic markers to the Hindmarsh × W1 DH population, ari-e was flanked by InDel-7 and InDel-17 and the genomics DNA interval of 482,623,538 to 488,465,184 bp on chromosome 5H. Based on the barley physical map, its physical position and combining the mapping results of both populations, the candidate region of ari-e declined to an interval of 0.58 Mb, which represents significant progress toward identifying the functional gene compared to previous studies [26].

Ellis et al. [19] reported that recessive dwarf mutants in barley indicate a loss of function. It was confirmed in both barley recessive semi-dwarf mutants, uzu1 and sdw1/denso, which resulted from the loss of function of the BR receptor gene HvBRI1 and gibberellin biosynthesis gene HvGA20ox ₂ , respectively [8, 15]. Genetic analysis showed that ari-e mutants also carried a recessive allele [25]. We deduced that the semi-dwarfism of ari-e was possibly due to the mutation of one of the functional genes. According to the most current POPSEQ barley map [55], there are five predicted genes in the candidate region (Table 6). Considering the semi-dwarf phenotype of ari-e and the gene annotation of the five genes, MLOC_66038 predicted as an E3 ubiquitin-protein ligase may be a candidate gene for ari-e. E3 ubiquitin ligase members are involved in the regulation of some biological processes including hormonal control of vegetative growth, plant reproduction, light response, biotic and abiotic stress tolerance, and DNA repair [56]. For example, both Arabidopsis sly1 and rice gid2 encode the F-box subunit of SCF E3 ubiquitin ligase that regulates GA responses and results in the dwarf phenotype [67, 68]. Unfortunately, no nucleotide sequence variants of MLOC_66038 or the other four genes were present between the ari-e variety Hindmarsh and the reference variety Morex.

In barley, semi-dwarf gene sdw1/denso affects not only plant height but also multiple agronomic traits, because sdw1/denso encodes GA 20-oxidase and is involved in gibberellin biosynthesis [6–8]. Therefore, sdw1/denso is expressed all of the tissues and affects many developmental processes through GA level [7, 8]. ari-e also reduces plant height and is related with lots of agronomic traits, such as erect juvenile growth habit, earliness, compact inflorescene, short awns, small seed size and cell size reduction in leaf blades [19, 24, 25]. This indicated that ari-e may be involved in different developmental processes. Furthermore, reduced expression of HvGA20ox ₂ was identified in various organs in denso semi-dwarf mutant compared with tall variety [7]. Rieu et al. [69] reported that AtGA20ox ₁ (also named as semi-dwarf gene ga5) was showed different expression in different tissues, and at different developmental stages. The difference of plant height between ari-e semi-dwarf and tall varieties reaches the maximum at the heading time and the uniform leaf samples can easily be collected. Therefore, leaves of VB9104, Dash and Golden Promise were collected to detected relative expression levels of all predicted genes at the heading stage. But expression levels of the five genes were not consistent with ari-e. The unexpected results might be due to the limited resolution of the current barley physical map while the non-coding region may play a key role in inducing semi-dwarfness. Owing to the large size of the barley genome (5.1 Gb) and the high repeat content (80 %), the current physical map comprises 9,265 contigs with a cumulative size of 4.9 Gb representing 96 % of the physical length of the barley genome [70]. Chromosomal assignments between the POPSEQ and IBSC maps agree in 97.6 % of the cases and discordant contig placements mostly occurred in the genetic centromere due to the severely reduced recombination frequency [55, 68]. Moreover, most of the map-based cloned barley genes have been located in the distal regions of the chromosome. Considering that ari-e occurred near the centromere of barley 5H [25], the repetitive nature of the barley genome [55, 70, 71] and limited resolution of the barley physical map, there may be other unanchored candidate genes for the ari-e semi-dwarf gene.

In summary, next-generation sequencing combined with the current barley physical map served as a hub for marker development in the desired region and the candidate gene will be isolated with the ongoing construction of the high-resolution barley physical map.

Traditional marker assays and SALF-seq in conjunction with BSA were conducted to map the semi-dwarf gene ari-e. SNPs and InDel markers were developed by SLAF-seq and whole-genome shotgun sequencing technology, respectively, to fine-map ari-e. The ari-e gene was mapped between two developed markers InDel-16 and InDel-17 with an interval of 0.58 Mb. The current study exemplifies the use of SLAF-seq for SNP discovery and whole-genome shotgun sequencing for InDel development in the target region, which may pave the way for map-based cloning of the ari-e gene and unraveling the molecular mechanisms of the semi-dwarf phenotype.