Genome-wide identification of sucrose nonfermenting-1-related protein kinase (SnRK) genes in barley and RNA-seq analyses of their expression in response to ABA treatment


 Background: Sucrose nonfermenting-1 (SNF1)-related protein kinases (SnRKs) play important roles in regulating metabolism and stress responses in plants, providing a conduit for crosstalk between metabolic and stress signalling, in some cases involving the stress hormone, abscisic acid (ABA). The burgeoning and divergence of the plant gene family has led to the evolution of three subfamilies, SnRK1, SnRK2 and SnRK3, of which SnRK2 and SnRK3 are unique to plants. Therefore, the study of SnRKs in crops may lead to the development of strategies for breeding crop varieties that are more resilient under stress conditions. In the present study we describe the SnRK gene family of barley (Hordeum vulgare), the widespread cultivation of which can be attributed to its good adaptation to different environmentsResults: The barley HvSnRK gene family was elucidated in its entirety from publicly-available genome data and found to comprise 50 genes. Phylogenetic analyses assigned six of the genes to the HvSnRK1 subfamily, 10 to HvSnRK2 and 34 to HvSnRK3. The search was validated by applying it to Arabidopsis (Arabidopsis thaliana) and rice (Oryza sativa) genome data, identifying 50 SnRK genes in rice (four OsSnRK1, 11 OsSnRK2 and 35 OsSnRK3) and 39 in Arabidopsis (three AtSnRK1, 10 AtSnRK2 and 26 AtSnRK3). Specific motifs were identified in the encoded barley proteins, and multiple putative regulatory elements were found in the gene promoters, with light-regulated elements (LRE), ABA response elements (ABRE) and methyl jasmonate response elements (MeJa) the most common. RNA-seq analysis of the roots of seedlings grown with and without ABA treatment showed that many of the genes responded to ABA, some positively, some negatively and some with complex time-dependent responses.Conclusions: The barley HvSnRK gene family is large, comprising 50 members, subdivided into HvSnRK1 (6 members), HvSnRK2 (10 members) and SnRK3 (34 members), showing differential positive and negative responses to ABA.

A SnRK1 cDNA, initially called RKIN1, was rst cloned from a rye (Secale cereale) endosperm cDNA library and shown to restore SNF1 function in yeast (Saccharomyces cerevisiae) snf1 mutants [4]. In fact, the gene usually referred to as SnRK1 in plants actually encodes a catalytic subunit of a heterotrimeric protein. AMPK also comprises three subunits, a catalytic α subunit and accessory β and γ subunits, while the yeast protein comprises SNF1 (equivalent to AMPKα), SNF4 (equivalent to AMPKγ) and one of three related proteins, SIP1, SIP2 and GAL83 (related to AMPKβ) [2]. For this reason, the SnRK1 gene is sometimes referred to as SnRK1α.
SnRK1 controls metabolism in part through the phosphorylation of enzymes such as 3-hydroxy-3methylglutaryl coenzyme A reductase (HMG-CoA reductase) and sucrose phosphate synthase, leading to their inactivation. It also phosphorylates nitrate reductase, trehalose-phosphate synthase, and 6phosphofructo-2-kinase/fructose-2,6-bisphosphatase, but these enzymes also require the binding of a 14-3-3 protein for inactivation (see [2] for review). Another important metabolic enzyme, adenosine diphosphate (ADP)-glucose pyrophosphorylase, is regulated by SnRK1 through modulation of its redox state [5]. SnRK1 also acts through the regulation of gene activity, causing changes in gene expression in response to nutrient starvation [6] and herbivory [7]. SnRK1 has also been shown to channel carbon into the starch biosynthetic pathway in potato (Solanum tuberosum) tubers through modulation of sucrose synthase and ADP-glucose pyrophosphorylase gene expression [8,9], and there is evidence that it plays a similar role in rice (Oryza sativa), sorghum (Sorghum bicolor) and maize (Zea mays) grain [10,11] as well as sorghum and barley (Hordeum vulgare) pollen [11,12]. Interestingly, both SNF1 and AMPK are required for storage carbohydrate synthesis in their respective systems, although the storage carbohydrate is different (glycogen instead of starch), the glucose donor is different (UDP-glucose instead of ADPglucose) and the mechanisms of regulation are different [13,14]. Paradoxically, SnRK1 is also required for the expression of a-Amy2 (α-amylase), a sugar-repressed gene that is involved in starch breakdown during germination [15].
The fact that SnRK1 phosphorylates and inactivates nitrate reductase provides a route through which SnRK1 can affect nitrogen metabolism. A second route is via asparagine synthetase gene expression, which has been shown to be regulated by SnRK1 in Arabidopsis (Arabidopsis thaliana) [6]. In wheat (Triticum aestivum), asparagine synthetase gene expression in the developing embryo increases in response to sulphur de ciency, and SnRK1 has been implicated in regulating this response [16,17].
Cereals contain a class of SnRK1 genes that are expressed endosperm-speci cally [18,19] and these have been called SnRK1b [1]. A third class that is closely related to SnRK1b but is expressed in the embryo as well as the endosperm has recently been identi ed and called SnRK1b* [16]. It was this class that was suggested to be regulating asparagine synthetase gene expression in the wheat embryo [16].
There is also growing evidence that SnRK1 plays an important role in the defence against pathogens (see [20,21] for reviews). Over-expression of SnRK1 in tomato (Solanum lycopersicum), for example, has been found to enhance viral resistance, while antisense suppression has the opposite effect [22]. SnRK1 overexpression has also been shown to confer broad-spectrum disease resistance in rice [23,24], enhancing the defence response mediated by jasmonate [24]. In wheat, SnRK1 has been shown to interact with a protein, TaFROG, that enhances wheat's resistance to deoxynivalenol (DON), a mycotoxin produced by pathogenic Fusarium fungi [25]. DON treatment increases SnRK1 phosphorylation/activation and activity, while down-regulation of SnRK1 gene expression using virusinduced gene silencing (VIGS) increases the damaging effects of DON on wheat spikelets [25]. SnRK1 will phosphorylate peptides containing a serine residue (much preferred to threonine) with hydrophobic residues at positions − 5 and + 4 with respect to the serine, and a basic residue at position − 3 or, less favourably, -4 [1,26]. However, plant extracts generally contain two activity peaks that will phosphorylate this target site, in addition to the one accounted for by SnRK1 [26][27][28][29], and these came to be called SnRK2 and SnRK3 [1]. It was soon apparent that a SnRK2 cDNA had already been cloned from wheat and named pkABA1, the name being assigned because pkABA1 mediates ABA-induced changes in gene expression in response to cold, dehydration and osmotic stress [30,31]. A very similar SnRK2 gene, SAPK2, confers ABA sensitivity and drought tolerance in rice [32]. The SnRK2 gene family is unique to plants and has expanded and diversi ed to comprise, for example, 10 members in Arabidopsis [2]. SnRK2s are much smaller than SnRK1, with molecular weights typically around 40 kDa compared with 58 kDa for SnRK1, and do not complement the snf1 mutation of yeast [1]. Indeed, SnRK2s are only large enough to include a protein kinase domain and a short, truncated, acidic C-terminal domain. They are associated with a variety of abiotic stress responses [33][34][35][36][37], and are integral to the abscisic acid (ABA) response pathway [34,[38][39][40], becoming active in response to ABA and phosphorylating transcription factors of the ABA-response-element-binding protein class (AREBPs). These transcription factors have been shown to be substrates for SnRK1 and SnRK3 as well [41], making them potential convergence points for multiple signalling pathways.
Halford and Hardie [1] divided SnRK2s into two subclasses, SnRK2a and SnRK2b, based on whether their acidic C-terminal domain contained an aspartic acid-or glutamic acid-rich patch. Subsequently, Kobayashi et al. [33] identi ed three subclasses, I, II and III, with subclass I equivalent to SnRK2b (glutamic acid-rich C-terminal patch) and subclasses II and III representing subdivisions of SnRK2a (aspartic acid-rich patch). These subclasses have been shown to respond differently to ABA, with subclass I not activated by ABA, subclass II not activated or activated very weakly by ABA, and subclass III strongly activated by ABA [42][43][44][45]. ABA is, of course, strongly associated with stress responses (see [46] for review), but it is also a key player in driving grain maturation and, in particular, the switch from the grain lling to maturation phases of grain development [47]. ABA has a very different effect on SnRK1 than SnRK2, promoting SnRK1 degradation in wheat roots, for example, while promoting SnRK2 phosphorylation/activation [48]. It has been suggested that the ABA-driven degradation of SnRK1 together with the activation of SnRK2 could be the trigger that pushes developing grain into the maturation phase [49].
The rst SnRK3 gene to be characterized was also from wheat, and was initially named wpk4 [50,51]. The SnRK3 gene family is even larger than the SnRK2 family, with the Arabidopsis genome containing 25 SnRK3 genes [2]. Unlike SnRK1 and SnRK2, SnRK3 is believed to be calcium-dependent, because it interacts with a calcium-binding protein, calcineurin B-like protein (CBL) [36,[52][53][54]. This has led to SnRK3s being more popularly known as CIPKs (CBL-interacting protein kinases). One of the most extensively characterized SnRK3/CIPKs is SOS2 (salt overly sensitive 2), which is involved in the response to salt stress [52,[55][56][57][58]. Sodium ions are just one of many species of ions that plants take up from the soil, of course, and, while many are important nutrients, some can be toxic at high concentrations. SnRK3s/CIPKs are now considered to be involved generally in enabling plants to adapt to changing ionic conditions. Their roles have recently been reviewed in depth by Tang et al. [59] and Ma et al. [60].
SnRKs represent one of the largest families within the plant kinome, and their positioning at the interface between metabolic, stress and ABA signaling has made them the focus of studies in many plant species.
As whole genome data have become available for many species, along with powerful bioinformatics tools, the identi cation of entire SnRK gene familes at the genome level has become possible, for example in Arabidopsis, Brachypodium distachyon, Chinese cabbage (Brassica rapa), soybean (Glycine max), and oilseed rape (Brassica napus) [36,[61][62][63][64]. However, the role of ABA in driving developmental changes in cereal seed development, the differential effects of ABA on the different SnRK families, and the role of SnRKs in regulating processes that affect grain composition, such as starch accumulation and asparagine biosynthesis, make the study of SnRKs in cereals particularly important. Barley was the rst plant species in which a characterization of the SnRK1 gene family was begun, with three different members of the family, called BKIN2, BKIN9 and BKIN12 at that time, shown to have different gene structures and to be differentially-expressed, with BKIN12, subsequently assigned to the SnRK1b class, expressed speci cally in the endosperm tissue of the grain [18,19]. Here we report the analysis of the entire SnRK gene family of barley, made possible by the recent publication of high quality barley genome data. It is the rst comprehensive analysis of the SnRK gene family of a commercial cereal species.

Genome-wide identi cation of the SnRK gene family members in barley
In total, 50 HvSnRKs were found to be encoded in the barley genome (Table 1), of which six were classed as SnRK1s by Pfams analysis and called HvSnRK1.1 -HvSnRK1.6. Each of these contained a kinase associated domain KA1 (Pfams reference PF02149), and all except one, HvSnRK1.2, also contained an ubiquitin-associated domain (UBA) (PF00627). Thirty-four proteins were assigned to the SnRK3/CIPK group based on the presence of a NAF domain (PF03822), a 24 amino acid domain required for binding to CBLs [65]. The remaining 10 proteins were assigned to the SnRK2 group. The proteins comprised between 232 and 797 amino acid residues and had molecular weights ranging from 25.9 to 89.9 kDa (see Table 1). The smallest of these, one of the SnRK3s/CIPKs, is not large enough to include a complete kinase domain and is unlikely to be functional.  [66].
To validate our screen and to compare and contrast the SnRK families across species, we also identi ed SnRKs from rice and Arabidopsis. The rice family comprised a total of 50 OsSnRKs, made up of four OsSnRK1s, eleven OsSnRK2s and 35 OsSnRK3s, while the Arabidopsis family comprised 39 AtSnRKs, including three AtSnRK1s, 10 AtSnRK2s and 26 AtSnRK3s (Additional le 1). The Arabidopsis family included one additional SnRK3 compared with those identi ed by Halford and Hey [2].

Phylogenetic Analyses
A rooted phylogenetic analysis was conducted for all the SnRKs identi ed from barley, rice and Arabidopsis. The phylogenetic tree that was generated ( Fig. 1) con rmed the three major groups of SnRK1, SnRK2 and SnRK3, and the assignation of each barley SnRK to the groups con rmed the Pfams analysis, although one of the rice genes that had been classi ed as a SnRK2 by Pfams appeared in the SnRK3 cluster in the phylogenetic analysis and was named OsSnRK3. 35. Note that SnRK1 is more closely related to SNF1 of yeast and AMPK of mammals than it is to SnRK2 and SnRK3 [1]. While SnRK2 and SnRK3 are plant-speci c and presumably diverged from SnRK1 after the separation of animals, plants and fungi, they must have evolved rapidly as they took on different functions.
A second phylogenetic tree of the barley HvSnRKs alone was produced using different software. The phylogenetic tree that was generated ( Fig. 2A) con rmed the clustering into the HvSnRK1, HvSnRK2 and HvSnRK3 families. As before, the SnRK1 family could be further divided into subclasses, corresponding to SnRK1a (SnRK1.1 and 1.2), SnRK1b (SnRK1.3 and 1.4) and SnRK1b* (SnRK1.5 and 1.6). Similarly, the SnRK2 subgroup was divided into two subclasses, corresponding to SnRK2a and SnRK2b as de ned by Halford and Hardie [1] and SnRK2 subclass II/III and SnRK2 subclass I as de ned by Kobayashi et al. [33] (subclasses II/III did not separate clearly in this analysis).

Gene Structure, Conserved Motifs And Promoter Analysis
There were 11 exons in all the HvSnRK1 genes except for HvSnRK1.5, and 9 exons in all of the HvSnRK2 genes except for HvSnRK2.7. In contrast, the exon/intron structure of the HvSnRK3s varied greatly, ranging from 1 exon to 16 (Fig. 2B), indicating that the genes in the HvSnRK1 and HvSnRK2 subfamilies are more conserved in structure than those in the HvSnRK3 subfamily.
The motif analysis of the HvSnRK proteins (Fig. 2C) showed that all the HvSnRK1s had seven motifs, 1, 2, 3, 4, 5, 6 and 10, apart from HvSnRK1.5, which lacked the rst three motifs. All of the HvSnRK2s had motifs 1, 2, 3, 4 and 6; in other words they lacked motifs 5 and 10 of the HvSnRK1s. They are, of course, readily distinguishable from the SnRK1s on the basis of size, being much smaller (around 40 kDa compared with around 58 kDa). That is not true of the SnRK3s, which vary considerably in size. However, the SnRK3s were characterized by the presence of the rst six motifs identi ed in the SnRK1s but also by three additional motifs, 7, 8 and 9, that were not present in the SnRK1s, and the lack of motif 10. The exceptions were HvSnRK3.9, which lacked the rst ve motifs, HvSnRK3.10, which lacked the rst four motifs and motif 7, HvSnRK3.27, which lacked the rst motif, and HvSnRK3.27, which had additional motifs 1, 2 and 9 before the shared nine motifs. Interestingly, motif 9 was also present in SnRK2.1, but motifs 7 and 8 were not. The presence/absence of these motifs could allow for the simple allocation of SnRKs from other species into the correct subgroup.
Considering that HvSnRKs play important roles in many biological processes, the promoters of the genes were analyzed to identify potential regulatory elements that could give clues as to how the genes are regulated and the different stimuli they respond to. Of the 50 HvSnRK genes, two (HvSnRK3.9 and HvSnRK3.10) lacked data for the promoter regions, but the analysis identi ed 18 different types of ciselement in the promoter regions of 47 of the other 48 HvSnRK genes (Fig. 3, Additional le 2). Apart from the one promoter (that of HvSnRK3.31) lacking any of these elements, each promoter contained between 2 and 9 of the different types of cis-element, with the promoters of HvSnRK1. 6

Chromosomal Location, Gene Duplication And Gene Synteny Analysis
Chromosomal location analysis showed the 50 HvSnRK genes to be distributed across all seven barley chromosomes (Fig. 4). Four of the six HvSnRK1 genes were located to Chromosome 3, with HvSnRK1. Gene duplication analysis of the HvSnRK genes showed that there were two segmental duplication events between different chromosomes and two tandem duplication events (Fig. 4), while there was only one interval gene in the cluster of HvSnRK1.3, HvSnRK1.4 and HvSnRK1.5. The KaKs calculation showed that Ka/Ks values of all the duplication pairs were less than 1, especially the segmental duplication pairs ( Table 2). The results also suggested that the HvSnRK genes related to those duplications were conserved, with purifying selection (the selective removal of alleles that are deleterious). In total, 41 of the 50 HvSnRK genes were found to be expressed in the roots, and 30 were affected signi cantly (p < 0.05) by the ABA treatment (Fig. 6 down-regulated at 1 h after ABA treatment, then up-regulated at 3 h and 6 h. In all, half of the HvSnRK2 genes and more than two thirds of the HvSnRK3 genes responded to ABA, showing a complex picture of differential regulation. This is consistent with the HvSnRK2 and HvSnRK3 genes playing important roles in ABA responses [59,60]. In contrast, of the 4 HvSnRK1 genes that were expressed, none showed a signi cant change in expression in response to ABA. ABA has been shown to in uence SnRK1 in wheat roots but at the post-translational rather than transcriptional level [48].

Discussion
SnRKs play important roles in linking stress and ABA signalling with metabolic signalling in plants. Barley is the fourth most important cereal crop in the world in terms of production, and its widespread cultivation can be attributed in part to its good adaptation to different environments. This makes the elucidation of the SnRK gene family in this species particularly important. In this study, a total of 50 barley HvSnRK genes were identi ed, providing a basis for studying their roles individually, as subfamilies and as an entire gene family.
BLAST searches or HMM searches or both are usually used for identifying gene homologues and families of homologues, and both types of search were adopted for this study. The methodology was validated by applying it to rice and Arabidopsis and showing that it identi ed all of the SnRK genes in those species. Halford and Hardie separated the SnRKs into three subgroups: SnRK1, SnRK2 and SnRK3, with the molecular weight of SnRK1 around 58 kDa and SnRK2 around 40 kDa, with more variability for SnRK3 [1]. The HvSnRKs separated clearly into these three subgroups, and clustered with the corresponding subgroups of rice and Arabidopsis in phylogenetic analyses, indicating that the three SnRK subgroups were established before the divergence of dicot and monocot plants. This was also inferred by Wang et al. [61] in the identi cation of BdSnRKs in Brachypodium distachyon. Motif analysis of the encoded proteins also distinguished the three subfamilies, and the presence/absence of different motifs could be used to assign SnRKs to the correct subfamily in species that are less well characterized and phylogenetic analysis is more di cult.
SnRK2s and SnRK3s are unique to plants, and have diverged further from SnRK1 than SnRK1 has from its fungal and animal counterparts, SNF1 and AMPK. Presumably the SnRK2 and SnRK3 genes rst arose from duplications of the SnRK1 gene, then evolved and diversi ed as they took on new roles, resulting in the burgeoning of the gene family as a whole. Our analysis revealed gene duplication (including segmental and tandem duplication) events in the evolution of the gene family, as well as differential regulation in response to ABA treatment. In addition, the Ka/Ks analysis indicated that the duplicated HvSnRK genes evolved slowly or were highly conserved [67].
SnRK1 genes of cereals have been subdivided into SnRK1a and SnRK1b according to their expression patterns, with SnRK1a expressed in multiple tissues and SnRK1b expressed predominantly in the endosperm [1,18,19]. Curtis et al. [16] then identi ed a subclass similar to SnRK1b that was expressed in the embryo as well as the endosperm and named it SnRK1b*. These three subclasses of HvSnRK1s could be readily distinguished in the phylogenetic analysis in this study.
SnRK2s were also divided into two subclasses (a and b) by Halford and Hardie [1], then further into three subclasses (I, II and III) by Kobayashi et al. [33], with II/III corresponding to SnRK2a and I to SnRK2b. Subclass II has been reported to be strongly activated by ABA, Class II weakly activated and Class I not at all [42][43][44][45]. The regulation of the barley HvSnRK2s by ABA appears to be more complicated. Phylogenetic analysis of the barley gene family readily distinguished the HvSnRK2a Class II/III and HvSnRK1b Class I types; however, all but two of the HvSnRK2s (HvSnRK2.8 and HvSnRK2.10) had ABA response elements in their promoters, and those that were shown to be signi cantly regulated by ABA treatment included members of all three subclasses.
Synteny analysis showed that one-to-one correspondence between the SnRKs of barley and rice was better than that between barley and Arabidopsis, consistent with the hypothesis of Wang et al. [61] that the monocot and dicot gene families have differentiated. The study also identi ed some HvSnRKs without corresponding genes in either rice or Arabidopsis, indicating continued divergence and differentiation of the barley gene family since its evolutionary line split from that of rice.

Conclusions
The barley HvSnRK gene family comprises 50 HvSnRK genes, six of which are of the HvSnRK1 subfamily, 10 HvSnRK2 and 34 HvSnRK3. The presence/absence of speci c motifs in the encoded proteins distinguishes between the subfamilies. Multiple putative regulatory elements are present in the gene promoters, with light-regulated elements (LRE), ABA response elements (ABRE) and methyl jasmonate response elements (MeJa) the most common. Many of the genes respond to ABA, some positively, some negatively and some with complex time-dependent responses.

Identi cation and phylogenetic analysis of the SnRK gene family in barley
The non-redundant amino acid sequences of Arabidopsis (Arabidopsis thaliana) and rice (Oryza sativa) SnRKs were collected from TAIR v10 (http://www.arabidopsis.org/) and RGAP v7 databases (http://rice.plantbiology.msu.edu/), respectively. Seventy-three SnRK sequences from Arabidopsis, 54 from rice and one (BKIN12) from barley were obtained directly from published papers [18,36,62,63,68] (Additional le 1).  [69] and SMART web tools (http://smart.embl.de/) [70]. The conserved protein kinase (Pkinase) domain regions of the SnRK family from Arabidopsis, rice and barley were then selected to perform multiple alignments using MAFFT v7.427 [71]; this region was selected and the C-terminal region omitted because of the high variation in the C-terminal sequences of the SnRK proteins. The amino acid substitution model was calculated by the ProTest-3.4.2 and the optimal model of "JTT + I + G + F" was selected [72]. The maximum-likelihood (ML) tree was constructed using the RAxML v8.2.9 program with bootstrap values for 1000 replicates [73], and the rooted tree was drawn by FigTree.

Protein Properties And Sequence Analyses
The molecular weights and isoelectric points of putative HvSnRK proteins were calculated by the ExPASy proteomics server (http://expasy.org/). MEGA 7.0 was used for constructing an unrooted phylogenetic tree for the HvSnRK family proteins. In brief, all of the barley SnRK protein sequences were aligned using ClustalW with the default parameters, and the phylogenetic tree was built by MEGA 7.0 using the Neighbor-Joining (NJ) method with 1000 bootstrap replications. The motifs were identi ed using the MEME program (http://meme.sdsc.edu/meme/intro.html), and information on the HvSnRK genes was obtained from the Hordeum vulgare genome database (Ensembl Plants). The gene structures and protein motifs were drawn by TB tools [74]. To analyze the cis-elements in the promoter regions, the 1.5 kb upstream region of the coding sequence region of each HvSnRK gene (Additional le 1) was obtained from the Hordeum vulgare Genome Database and analyzed with the PlantCARE databases [75] (http://bioinformatics.psb.ugent.be/webtools/plantcare/html/), and the cis-elements were drawn by TB tools .
Chromosomal Location, Genome Synteny And Gene Duplication Analyses The chromosomal location of the HvSnRK genes was downloaded from the Hordeum vulgare Genome Database (Ensembl Plants), and the distribution of the genes on chromosomes was drawn using the MG2C v.2 program (http://mg2c.iask.in/mg2c_v2.0/). Gene duplication, including segmental and tandem duplication, was analyzed using the MCScanX program [76]. Genes were considered to be segmentally duplicated if they occurred in collinear segments containing at least ve collinear gene pairs, whereas they were considered to be tandemly duplicated if they were located close to each other with no more than two interval genes. The KaKs ratio (the ratio of nonsynonymous (Ka) to synonymous (Ks) nucleotide substitution rates) was also calculated for the duplicated gene pairs using the KaKs Calculator 2.0 [77]. instructions, and mRNA was puri ed from total RNA using poly-Toligo-attached magnetic beads.
Nucleotide sequence analysis of the libraries was then carried out on an Illumina Hiseq Xten platform (Illumina Inc., San Diego, CA), and 150 bp paired-end reads were generated. The clean nucleotide sequence data ranged from 7.95 to 15.06 Gb (all > 6 Gb), and the Q30 percentages were all > 80% (Table  S1). These results suggested that the data were su cient and reliable enough for further analysis.
Spearman correlation analysis also showed that the three biological replicates of each sample met the requirements (all over 0.95) ( Figure S1).

Expression analysis of HvSnRKs by RNA-seq
Reference genome and gene model annotation les were downloaded directly from the Ensembl Plants database (http://plants.ensembl.org/Hordeum_vulgare/Info/Index, IBSC v2). An index of the reference genome was built and clean reads aligned to the reference genome using Hisat2 v2.0.5. HTSeq v0.6.1 was used to count the read numbers that mapped to each gene. The number of reads per kilobase of transcript, per million mapped reads (RPKM) was calculated based on the length of the gene and the number of reads that mapped to the gene. Differential expression analysis of the two treatments (with or without ABA) was performed using the DESeq2 R package (1.16.1). Genes with a p value < 0.05 were assigned as signi cantly differentially-expressed.