A High-Density Genetic Linkage Map of SLAFs and QTL Analysis of Grain Size and Weight in Barley (Hordeum vulgare L.)

Grain size is an important agronomic trait determines yield in barley, and a high-density genetic map is helpful to accurately detect quantitative trait loci (QTLs) related to grain traits. Using specific-locus amplified fragment sequencing (SLAF-seq) technology, a high-density genetic map was constructed with a population of 134 recombinant inbred lines (RILs) deriving from a cross between Golden Promise (GP) and H602, which contained 12,635 SLAFs with 26,693 SNPs, and spanned 896.74 cM with an average interval of 0.07 cM on seven chromosomes. Based on the map, a total of 16 QTLs for grain length (GL), grain width and thousand-grain weight were detected on 1H, 2H, 4H, 5H, and 6H. Among them, a major QTL locus qGL1, accounting for the max phenotypic variance of 16.7% was located on 1H, which is a new unreported QTL affecting GL. In addition, the other two QTLs, qGL5 and qTGW5, accounting for the max phenotypic variances of 20.7 and 21.1%, respectively, were identified in the same region, and sequencing results showed they are identical to HvDep1 gene. These results indicate that it is a feasible approach to construct a high-quality genetic map for QTL mapping by using SLAF markers, and the detected major QTLs qGL1, qGL5, and qTGW5 are useful for marker-assisted selection (MAS) of grain size in barley breeding.


INTRODUCTION
Barley is one of the most important cereal crops in the world, and widely used for animal feed and malting (Baik and Ullrich, 2008;Bond et al., 2015). Previous genome sequencing projects had indicated that the barley has a genome of 5.1 Gb, which is much larger than human genome of 3.3 Gb and rice genome of 389 Mb (International Barley Genome Sequencing Consortium et al., 2012). Due to the high repetitive sequences and complex structure, the sequence assembly of barley genome had been affected greatly, and the accuracy and completeness of the physical map need to be further improved (Mascher et al., 2017). Although barley is also a diploid species, the numbers of genes have been cloned are far less than rice and Arabidopsis, and the reverse genetics is usually used to carry out gene function studies (Fu et al., 2007;Sikdar et al., 2016;Holme et al., 2017). Therefore, it has become a common strategy to identify quantitative trait loci (QTLs) for important agronomic traits for marker-assisted selection (MAS) in barley (Romagosa et al., 1999;Miedaner and Korzun, 2012;Zhang et al., 2017;Fang et al., 2019).
Quantitative trait loci mapping has been widely used to identify genomic regions associated with target trait, which mainly subject to the sample size and molecular marker density (Doerge and Rebai, 1996). Due to the low density polymorphism of traditional molecular markers over the whole genome, the precise of QTL mapping was greatly limited (McCouch et al., 1988(McCouch et al., , 2002Olson et al., 1989). With the development of high-throughput genotyping and sequencing technology, the massive single nucleotide polymorphisms (SNPs) were extensively identified in different species, which usually used for high-density map construction, genome-wide association analysis, gene mapping, gene chip, MAS, etc. (Wang et al., 1998;Slate et al., 2009;Huang et al., 2010;Shao et al., 2015;Li et al., 2018). However, whole genome deep re-sequencing is still costly and not necessary for most studies. So, reduced representation genome sequencing (RRGS) was developed by DNA fragments sequencing of restriction enzyme digestion, which exhibits the advantages in identifying and genotyping SNPs, including simple steps, high effectivity, low cost, short cycle, and so on (Van Tassell et al., 2008;Hyten et al., 2010). Among them, specificlocus amplified fragment sequencing (SLAF-seq) is one version of RRGS based on special fragment-length, which mainly applied in high-density genetic map construction and gene mapping in many species Li et al., 2014;Hu et al., 2016;Yu et al., 2019;Zhao et al., 2019).
In the present study, a total of 12,635 SLAF markers with 26,693 SNPs were employed to genotype the recombinant inbred lines (RILs) derived from a cross between H602 and Golden Promise (GP), and a high-density genetic map spanned 896.74 cM was constructed. The QTL analysis of grain size and weight was subsequently performed, and three major QTLs qTGW5, qGL1, and qGL5 were identified with the max phenotypic variances of 21.1, 16.7, and 20.7%, respectively. The results will accelerate the QTL mapping of important agronomic trait loci and facilitate the MAS of grain size in barley.

Plant Materials and DNA Extraction
A RILs population of F8 generation was constructed via single seed descent, which derived from the cross of GP and H602 (a wild barley strain). The parents and developed 134 RILs were planted in the experimental fields of Hangzhou Normal University, Hangzhou, Zhejiang province (120 • 20 E, 30 • 27 N) with conventional field cultivation (row spacing of 20 cm). After harvest and drying in 2017-2019, the TGW, GL and GW were measured using an SC-G automatic seed analyzer (WSeen, China, n > 50). Total genomic DNAs of young healthy leaves were extracted from parents and 134 lines by CTAB method with some modification (Doyle and Doyle, 1987). The full-length genomic DNA of HvDep1 gene was amplified and sequenced with primers HvDep1-1, HvDep1-2, HvDep1-3, and HvDep1-4, respectively, which were listed in the Supplementary Table 1.

SLAF Library Construction and High-Throughput Sequencing
Specific-locus amplified fragment library construction was carried out following the description in detail by Sun et al. (2013) and the 5.1G barley genome was used as a reference genome. The Genome DNAs of parents and 134 RILs were digested by HaeIII (New England Biolabs, NEB, United States) restriction enzyme, and a single nucleotide (A) overhang was subsequently added to the obtained fragment. Then, through sequencing adapters ligating, polymerase chain reaction (PCR) amplifying, Agencourt AMPure XP beads (Beckman Coulter, High Wycombe, United Kingdom) purifying, pooling, 2% agarose gel electrophoresing, and the fragments ranging from 364-414 bp were collected and purified by a QIAquick gel extraction kit (Qiagen, Hilden, Germany). Through the Illumina HiSeqTM 2500 platform (Illumina, Inc.; San Diego, CA, United States), the finally products sequencing was carried out at Biomarker Technologies Corporation (Beijing, China).

SLAF-Seq Data Grouping and Genotyping
The reads obtained from sequencing were further distinguished and qualified reads with quality score more than 20e were distributed into each progeny based on duplex barcode sequence. According to over 90% sequence similarity, the reads were blasted by one to one alignment and the sequences clustered into the same group were defined as one SLAF. Based on the parental sequence depth more than 10×, the genotype of each SLAF marker was determined, which contained no less than 30% progeny information. Because barley is a diploid species, and one polymorphic SLAF marker can contain two to four alleles in the progenies, so more than four were got rid of as repetitive SLAFs. To genotype the polymorphic SLAF, parental genotypes were first determined, and then the offspring were also defined according to the consistency of the sequence with the parent. However, RIL is a permanent homozygous population, only the polymorphic SLAFs with the segregation type of aa × bb were adopted.

Construction of High-Density Genetic Map
According to the standard of modified logarithm of odds (MLOD), the polymorphic SLAF markers were classified and partitioned primarily into seven linkage groups (LGs) by the position of barley reference genome. Due to the massive SNP data, the HighMap software was employed to construct highdensity genetic map (Liu et al., 2014), in which the genotyping errors were corrected, the linear arrays of markers were ordered, and the genetic distances between two adjacent markers were computed by Kosambi mapping function in each LG. To ensure the quality of genetic map, the colinearity analysis between genetic maps and barley genomes were also evaluated.

Data and QTL Analysis
To reduce errors, all measurement values were gained from average value of three biological replicates. The frequency distributions of grain weight and size were analyzed in the 3 years, and statistical and correlation analysis were performed with SPSS 20.0 software. Based on the high-density genetic map, a powerful software qgene-4.3.10 (Joehanes and Nelson, 2008) was adopted to carry out QTL analysis to identify the related locus. The composite interval mapping (CIM) model was used to scan the whole seven chromosomes by the interval of one milliMorgan. When logarithm of odds (LOD) threshold more than 3, the statistical significance (P = 0.05) was considered, and the target interval was determined again with 1,000 permutations. The QTL parameters of chromosomes, marker names and intervals, Generalized R 2 , LOD values and additive effects were computed by qgene-4.3.10.

Analysis of SLAF-Seq Data and SLAF Markers
After SLAF library construction and high-throughput sequencing, 231.56 Gb data containing 1158.81 M reads was obtained. The average Q30 ratio of all samples was 94.87%, and guanine-cytosine (GC) content was 47.81% in average (Table 1), which indicated that the data quality is qualified. All the reads were filtered and then aligned by blast software, and more than 90% similarity were defined as one SLAF. A total of 746,752 SLAFs were developed and divided into polymorphic, non-polymorphic and repetitive types, in which 245,618 polymorphic SLAF markers were identified, accounting for 32.95% of the total SLAFs (Figure 1 and Supplementary  1 and Supplementary  Table 3). Finally, 99,182 polymorphic markers fell into aa × bb class, which was applied for genetic map construction.

Construction of High-Density Genetic Linkage Map
To improve the quality of genetic map and accuracy of QTL detection, the polymorphic SLAFs were screened again, and the retained SLAF markers were compared with reference genome of barley in order to observe the distribution of markers on each chromosome. Ultimately, 12,635 SLAF markers with 26,693 SNPs were mapped to seven LGs by HighMap software with average coverage depth 82.04-fold in GP, 76.57-fold in H602, and 23.36-fold in offspring (Figure 2 and Table 2 1 ). The total genetic distance of linkage map was 896.74 cM with an average distance of 0.07 cM between adjacent markers. The number of markers in each LG varied from 154 to 6,109, and the genetic length of each LG differed from 82.85 to 153.06 cM. The degree of linkage between markers was reflected by "Gap < 5" ranging between 98.69 and 100% with an average value of 99.77%, and the largest gap was mapped on chromosome 5H with 18.47 cM ( Table 3).

Grain Weight and Size Data Analysis
The GL, GW, and TGW of the parents and 134 RILs planted in Hangzhou were measured in consecutively 3 years, and the data showed that parent GP exhibited decreased GL and increased GW than parent H602 (Figure 3 and Supplementary Table 4). Moreover, the average GL and width of 134 RILs in 3 years were all fell in between two parents, and average grain weight in 2018 and 2019 was higher than the two parents (Supplementary Table 4). It is well know that grain weight is mainly determined by grain size, so their correlations should be positively correlated. Consistent with it, the correlation coefficients between GL versus TGW, and GW versus TGW exhibited positive correlation (P < 0.01) (Supplementary Table 5). In addition, the frequency distributions of grain size and weight of 134 individuals were also analyzed, and the results indicated that data were normally distributed, and suitable for QTL mapping (Figure 4).

QTL Mapping of Grain Size and Thousand-Grain Weight
Based on the high-density genetic map, a total of 16 QTLs related to grain size and TGW were detected on chromosome 1, 2, 4, 5, 6H in consecutive 3 years, and accounted for phenotypic variances ranged from 10.2 to 21.1% (Figure 5 and Table 4). Among them, three QTLs for GL, one QTL for GW, and two QTLs for TGW were located in 2017, four QTLs for GL and two QTLs for TGW were mapped in 2018, and two QTLs for GL and two QTLs for TGW were detected in 2019. The four QTLs, qGL1, qGL5, qTGW5, and qTGW6 were repeatedly detected in 3 years, in which qGL1 explained the phenotypic variance ranged from 14.9 to 16.7%, qGL5 from 17.6 to 20.7%, qTGW5 from 17.9 to     LG, linkage group; SLAFs, the total SLAF markers on each LG; SNPs, the numbers of SNP markers; Total distance, the total genetic distance of each LG; Average distance, the average distance between two flank markers; Gap < 5 cM, the proportion of gap less than 5 cM; Spearman, correlation coefficient between each LG and the physical graph, and the closer the values to 1, the better the collinearity between the two maps. 21.1%, and qTGW6 from 10.9 to 12.0%. Meanwhile, the average GL, width and weight of 3 years were also used to detect QTL, and the results indicated that the located QTLs and its intervals were similar to single-year data, except for qGL6 (Supplementary Table 6). In addition, the positions of qGL5 and qTGW5 were located in the same interval or neighboring to each other in 3 years, suggesting the two QTLs may be the same locus.

Candidate Regions of Major QTLs
Considering that the three major QTLs (qGL1, qGL5, and qTGW5) with high LOD scores and high phenotypic variations are valuable for further gene cloning and MAS in breeding, we carried out physical distances analysis of candidate regions.
Among them, the major QTLs qGL5 and qTGW5 were finally located in a 2.18 Mb interval between Marker4941182 and Marker6734745. In the candidate region, HvDep1, a gene controlling grain weight and grain size was found (Wendt et al., 2016). The gene sequencing showed a single base insertion in the second exon of HvDep1 was identified in parent GP, but there were no changes in parent H602 (Figure 6). Sequence analysis revealed that the gene was premature termination in GP, which is the same as reported HvDep1 gene mutation (Wendt et al., 2016). So, qGL5 and qTGW5 should be the identical gene to HvDep1. In addition, another major QTL, qGL1 was located in an 11.16 Mb region between Marker17833427 and Marker16397031 on chromosome 1H, and no related QTLs or genes have been reported in the interval.

DISCUSSION
In the modern breeding, MAS is significant to accelerate the selection process of target traits (Collard and Mackill, 2008;Bankole et al., 2017;Xu et al., 2018). However, important agronomic traits are usually controlled by QTL, and most of them have not been cloned or identified except for a few species. As an effective way for screening gene linkage, QTL analysis has been widely used to locate target trait gene and obtain linked markers, in which high-quality and high-density genetic map is a key for QTL mapping (Li et al., 2014;Wei et al., 2014). However, limited number of polymorphic markers is the main obstacle in the construction of high-density genetic map by traditional molecular markers (Hyten et al., 2010;Sun et al., 2013). SLAF-seq is an enhanced reduced representation library (RRL) sequencing technology with the advantage of massive SNP, high-resolution and low-cost, which has been successfully used to construct highdensity linkage maps, gene mapping and association analysis (Wei et al., 2014;Li et al., 2014;Xia et al., 2015;Wen et al., 2020).   Due to large genome, it is difficult to develop enough traditional polymorphism markers for covering the genome uniformly, which results in a limited number of markers used to construct high-density genetic maps in barley. Using 1,000 SSR and DArT markers, a high-density genetic map was developed from a DH population, which spanned 1,100.1 cM and exhibited an average distance of 0.91 cM (Hearnden et al., 2007). Another high-density consensus map comprising 2,935 loci (2,085 DArT and 850 other loci) and spanning 1,161 cM was conducted, which derived from seven DH and three RIL populations, and showed an average inter-bin distance of 0.7 ± 1.0 cM (Wenzl et al., 2006). However, the two maps were not dense enough, and the average genetic distances were more than 0.7 cM. With the completion of barley genome sequencing in 2012, the SNP has become an important molecular marker for genetic analysis due to the massive single base-pair changes. Applying the RAD-seq strategy, 12,998 SNP markers were developed for the construction of highdensity genetic map, which spanned 967.6 cM and displayed an average distance of 0.07 cM (Zhou et al., 2015). In 2017, a more high-quality barley reference genome was assembled and version IBSC_v2 was released, which increased the linear order of sequences and reduced the interference of repetitive elements (Mascher et al., 2017). In this study, we constructed a highdensity genetic map of 12,635 SLAFs by the IBSC_v2 reference genome, which spanned 896.74 cM and exhibited an average distance of 0.07 cM. In the seven LGs, average value of "Gap < 5" reached 99.77%, and only two gaps larger than 10 cM were existed on chromosome 3H and 5H, respectively. So, the genetic map we constructed was high-quality and high-density, and suitable to conduct genetic analysis of important agronomic traits.  In order to verify the validity and accuracy of the map, the QTL analysis of grain size and weight was conducted, and a total of 16 QTLs related to GL, GW, and TGW were detected, in which four QTL loci, qGL1, qGL5, qTGW5, and qTGW6 were repeatedly detected in 3 years. In view of the overlapped location interval, negative additive effects, and high correlation coefficients between GL and TGW, we speculated that qGL5 and qTGW5 should be the same QTL locus. Previous study showed that qGL5H was initially located in the position of 48.7-71.1 cM and fine mapped to a 1.7 Mb interval on chromosome 5, which situated in the 2.18 Mb section of qGL5/qTGW5 (Watt et al., 2019). In the region, HvDEP1, an AGG3-type subunit of G protein encoding gene that regulating GL and TGW was also identified, which indicated that it might be the candidate gene of qGL5, qGL5H, and qTGW5 (Wendt et al., 2016). The subsequent sequencing results showed that the same single base insertion was found in the CDS of parent GP, which revealed the identity of qGL5/qTGW5 and HvDEP1. In addition, through different populations, several major QTLs for GL have been mapped on chromosome 2, 3, 4, 5, 6, and 7H, and no major QTL was found on chromosome 1H (Walker et al., 2013;Zhou et al., 2016;Watt et al., 2019Watt et al., , 2020. So, qGL1 should be an unreported new QTL for GL. Although qTGW6 only explained the max phenotypic variance of 12.0% for grain weight, the QTL was also repeatedly detected in 3 years, indicated that qTGW6 is stably inherited.

CONCLUSION
Grain size and weight are important agronomic traits determinant yield. In this study, we constructed a high-density genetic map with 12,635 SLAFs, and identified two major QTLs involved in regulating GL and grain weight. Among them, an unreported new QTL, qGL1 accounted for maximum phenotypic variance of 16.7% and showed a negative additive effect on GL, which indicated the QTL from H602 played a promoting effect on elongating GL. Another major QTL locus, qGL5/qTGW5 exhibited maximum phenotypic variance of 20.7 and 21.1% in GL and TGW, respectively, and also displayed negative additive effect, which revealed the QTL from GP reduced the GL and TGW. These results indicated the two QTLs, qGL1 and qGL5/qTGW5 are useful for MAS in accelerating the breeding process of barley grain size and weight.

AUTHOR CONTRIBUTIONS
DX and HW designed the research, wrote the manuscript, and revised the manuscript. YF, XQZ, XZ, and TT performed the experiments, analyzed the data, and wrote the manuscript. ZZ, GW, LH, JZ, CN, HW, JL, and WW planted and collected the experimental materials. All authors contributed to the article and approved the submitted version.