Dehydrins (DHNs) are hydrophilic proteins with conserved K-segment and belong to Group II of the late embryogenesis abundant (LEA) protein family. K-segment is a lysine-rich amino acid sequence (EKKGIMDKIKEKLPG and similar sequences), which is highly conserved in most DHNs. In addition, Y-segment [(V/T)D(E/Q)YGNP] and S-segment (5 to 7 Ser residues) are often found in DHNs. DHNs can be classified into five architecture types, such as K_n, SK_n, K_nS, Y_nK_n, and Y_nSK_n, according to the presence and order of K-, S- and Y-segments within protein sequence (Close 1996; Graether and Boddington 2014).

DHNs are found in various subcellular locations such as cytoplasm, nucleus, mitochondria, chloroplast, vacuole and in the vicinity of the plasma membrane (Graether and Boddington 2014). DHNs are considered as molecular chaperones that protect various cellular components under stress conditions, such as proteins, enzymes, nucleic acids and membranes (Hara 2010; Hanin et al. 2011). Furthermore, the expressions of many DHN genes have shown a positive correlation with plant stress tolerance under abiotic stress conditions (Bravo et al. 2003; Hara et al. 2003; Houde et al. 2004; Brini et al. 2007; Graether and Boddington 2014; Kumar et al. 2014).

Until now, a number of DHN genes have been identified in many plant species. (Zolotarov and Strömvik 2015). Among those, many were also identified to encode RAB (responsive to abscisic acid) and COR (cold-regulated) proteins (Kusano et al. 1992; Welin et al. 1995; Danyluk et al. 1998; Nylander et al. 2001). For example, 10 and eight DHN genes were identified in Arabidopsis thaliana and Oryza sativa ssp. japonica, respectively, and their stress responsiveness was studied (Hundertmark and Hincha 2008; Verma et al. 2017). Although a number of studies have identified DHN genes and tried to reveal their biological roles, the precise function of the proteins still remains unclear.

The Cucurbitaceae family is composed of about 120 genera and more than 800 species, including many economically important crops such as cucumber, melon and watermelon (Lu and Jeffrey 2011). So far, nuclear genomes of seven crop species in the Cucurbitaceae family have been sequenced. Among those, Cucumis sativus var. sativus (cucumber) is an economically important crop and the genomes of three cucumber cultivars have been sequenced (Huang et al. 2009; Cavagnaro et al. 2010; Wóycicki et al. 2011). In addition, a wild cucumber species, Cucumis sativus var. hardwickii, was also analyzed for its genome sequencing (Qi et al. 2013). Similarly, genomes of some other important crops of the Cucurbitaceae family such as Cucumis melo (melon), Citrullus lanatus (watermelon) and Momordica charantia (bitter gourd) have also been sequenced (Garcia-Mas et al. 2012; Guo et al. 2013; Urasaki et al. 2016). Recently, genome sequences of bottle gourd (Lagenaria siceraria) and pumpkin species (Cucurbita maxima and Cucurbita moschata) became available (Wu et al. 2017; http://cucurbitgenomics.org/).

Like other crops, the growth of Cucurbitaceae crops is also adversely affected by abiotic stresses, which results in severe yield loss. In case of cucumber, its growth is inhibited at a temperature higher than 35°C and lower than 12°C as well as under stress conditions such as low humidity and high salt concentration (RDA 2013; Ali et al. 2014a, 2014b). Many studies have been carried out to understand the physiological and molecular mechanisms of stress response and to enhance abiotic stress-tolerance in Cucurbitaceae crops. However, the studies are still in their early phases for molecular breeding in the crops.

Although several DHN genes were identified as a part of LEA protein study in cucumber, melon and watermelon (Celik Altunoglu et al. 2016; Celik Altunoglu et al. 2017), no comprehensive study on the DHN gene has been carried out in Cucurbitaceae species. On this account, this study identified and compared DHN genes through genome-wide searches in five Cucurbitaceae species. The DHN genes studied here can be used in understanding the stress response mechanism and developing abiotic stress-tolerant crops in the Cucurbitaceae family.

Identification of DHN genes

Eight gene sets of cucumber cultivars, wild cucumber, melon and watermelon cultivars were retrieved from genome database of respective plant. Gene set of bitter gourd was kindly provided by Prof. Hideo Matsumura in Shinshu University, Japan. Deduced protein sequences of 10 genomes including Arabidopsis and rice genomes (as references) were used to investigate DHN genes (Table 1). InterProScan software (ver. 5.16–55.0 or 5.18–57.0, https://www.ebi.ac.uk/interpro/, Jones et al. 2014) was employed to identify DHN protein sequences, with default parameters. Protein sequences with K-segment (InterPro IDs IPR000167 and IPR030513) were selected as DHNs and used for further study.

Investigation of DHN architectures and subcellular location

Protein architectures of selected DHN genes were further divided on the basis of presence and order of conserved motifs such as K-, Y-, and S-segments (Close 1996; Graether and Boddington 2014). To find amino acid sequence conserved in DHNs, in-house Python script was used with conserved amino acid sequence information of three segments, such as “[KR][LIMV][KMR][DE][KQ] [LFI]⋯.” for K-segment, “S[4,10]” for S-segment, and “DE.GNP” for Y-segment and then the retrieved conserved sequences were manually confirmed. Subcellular location of DHNs was predicted using three different web tools, PSORT (https://wolfpsort.hgc.jp/), TargetP 1.1 (http://www.cbs.dtu.dk/services/TargetP/), and ProtComp 9.0 (http://www.softberry.com/berry.phtml?topic=protcomppl&group=programs&subgroup=proloc) with default parameters.

Phylogenetic analysis of DHN genes

The deduced protein sequences of DHN genes were multiple aligned using MUSCLE included in MEGA 7 software (http://www.megasoftware.net/, Kumar et al. 2016) and then the phylogenetic tree was generated based on the alignments using the Maximum-Likelihood (ML) method by MEGA 7 software with the following parameters: Poisson model, pairwise gap deletion and 1,000 bootstraps. The deduced protein sequences of 10 and eight DHN genes reported in A. thaliana (Hundertmark and Hincha 2008) and rice (Verma et al. 2017), respectively, were also included in the phylogenetic tree for comparison.

In silico expression profiling of DHN genes in cucumber and melon

The expression of DHN genes in cucumber and melon was profiled using public RNA-Seq data [Bioproject accession no. PRJNA80169 (http://www.ncbi.nlm.nih.gov/bioproject/PRJNA80169, Li et al. 2011) for cucumber cv. Chinese long and PRJNA300582 (https://www.ncbi.nlm.nih.gov/bioproject/PRJNA300582, Kim et al. 2016) for melon (C. melo var. makuwa, oriental melon, Korean landrace KM)] deposited in GenBank SRA database. First, the RNA-Seq data were downloaded and trimmed using NGS OC Tool kit (v2.3.1). Trimmed RNA-Seq reads of high quality (Phred score > 20) were mapped on protein-coding sequences (CDSs) predicted in the whole genome sequences, based on which Fragment Per Kilobase of transcript per Million mapped reads (FPKM) values were calculated using RSEM program with default parameters. Afterward, FPKM values for DHN genes were chosen and used to draw heatmap using MeV s/w (Saeed et al. 2003) with default parameters.

In silico analysis of DHN promoter sequences

Putative promoter sequences of 2-kb upstream from start codon of DHN genes were retrieved from genome databases of cucumber, melon, and watermelon. Putative cis-acting elements in the sequences were investigated using PlantPAN 2.0 (http://plantpan2.itps.ncku.edu.tw/promoter.php, Chow et al. 2016) and then elements related to abiotic stress response, such as dehydration response element (DRE, 5′-CCGAC-3′ or 5′-CCGAA-3′) and abscisic response element (ABRE, 5′-ACGCG-3′ or 5′-ACGTG-3′), were selected. Afterward the elements with similar score of 1.0 (i.e. 100% identical to DRE and ABRE sequences) were further selected and used for generating a scheme showing distribution of cis-acting elements.

DHN genes and their features in five Cucurbitaceae species

Genome-wide searches identified a total of 32 genes encoding DHNs in eight genome sequences of five Cucurbitaceae species (Table 2, Supplementary Table S1). Among those, four, three, and four genes were identified in cucumber cv. Chinese long, cv. Gy14, and cv. Borszczagowski, respectively, and four in wild cucumber. In addition, four, five, four, and four genes were identified in melon, watermelon cv. 97103, cv. Charleston Gray, and bitter gourd, respectively (Table 2).

Among reported five protein architecture types of DHNs (Close 1996; Graether and Boddington 2014), three types, K_n, SK_n and Y_nSK_n, were found in the deduced protein sequences of Cucurbitaceae DHN genes whereas remaining two types, K_nS and Y_nK_n, could not be found (Table 2). The K_n, SK_n and Y_nSK_n-type DHNs were encoded by 6, 15 and 11 Cucurbitaceae genes, respectively.

Subcellular location was predicted using three different tools such as PSORT, TargetP 1.1 and ProtComp 9.0. However, predicted subcellular locations of Cucurbitaceae DHNs were not consistent among the tools. All Cucurbitaceae DHNs were predicted to be located in nucleus by PSORT whereas some were predicted to be in cytoplasm or extracellular regions by ProtComp 9.0 (Supplementary Table S1). However, the predicted subcellular locations of Cucurbitaceae DHNs were similar to those of DHNs in A. thaliana and rice (Supplementary Table S2, S3).

Total 50 DHN genes, including 32 Cucurbitaceae (in this study), 10 Arabidopsis (Hundertmark and Hincha 2008, Supplementary Table S2) and eight rice DHN genes (Verma et al. 2017, Supplementary Table S3) could be divided into five groups in phylogenetic tree based on similarity of amino acid sequences (Fig. 1). Although some DHN architecture types were grouped closely together with the same types, the whole grouping pattern seemed to be irrelevant to protein architecture of the used DHNs. Group I included all SK₃-type DHNs in Cucurbitaceae, three SK₂-type DHNs and a SK₃-type DHN in A. thaliana, and a SK₃-type DHN in rice. Group II included K₆ and KS-type DHNs in A. thaliana and a KS-type DHN in rice. Group III included Y_nSK_n-type and YK-type DHNs. Group IV included mainly SK₂-, Y_nSK_n- and YSK_n-type DHNs. Group V included mainly SK₂- and K-type DHNs. Group II and Group IV did not include DHNs identified in Cucurbitaceae, whereas Group V did not contain those in A. thaliana and rice. Among the five groups, in particular, Group I was comprised of DHNs involved in abiotic stress tolerance in Arabidopsis and rice, such as COR47, ERD10, and OsDHN1 (Hanin et al. 2011; Kumar et al. 2014), and also included eight DHNs (seven SK₃ type and one SK₂ type) identified in Cucurbitaceae species.

The Cucurbitaceae DHN genes showed 2% to 100% similarity to each other and 2% to 42% similarity to Arabidopsis and rice DHN genes at amino acid level (Supplementary Table S4). Furthermore, the comparison of protein sequences revealed that K-segment was highly conserved in all DHNs in Cucurbitaceae species, A. thaliana, and rice. In addition, well conserved S- and Y-segments were also found in some DHNs (data not shown). For example, DHNs in Group I had highly conserved K- and S-segments (Fig. 2). In addition, whole protein sequences of DHNs in Cucurbitaceae species were more similar to each other than to those in A. thaliana and rice, as expected.

Expression profiles of DHN genes in cucumber and melon

Public cucumber and melon RNA-Seq data were used for in silico expression profiling of DHN genes in cucumber and melon, as mentioned in Materials and Methods. Among four cucumber DHN genes, only SK₃-type gene (Csa6M358710.1) showed strong expression in the ten examined tissues (leaf to ovary), whereas other three genes [Csa2M249810.1 (Y₃SK₂), Csa4M045040.1 (SK₂), Csa4M045050.1 (K)] showed very low or no expression in the tissues (Fig. 3A, Supplementary Table S5). Similarly, only melon SK₃-type DHN gene (MELO3C026550T1) was strongly expressed in the five examined tissues among four melon DHN genes (Fig. 3B, Supplementary Table S6). Melon SK₂-type DHN gene (MELO3C016402T1) showed high expression only in fruit, which was slightly different from the expression profiles of cucumber DHN genes.

Cis-acting elements related to abiotic stress response in DHN promoter sequences

Through in silico searching of cis-acting elements on putative promoter sequences of DHN genes in cucumber (cv. Chinese long), melon (DHL92), and watermelon (97103), examined all 13 sequences were identified to contain DRE-like and/or ABRE-like sequences (Fig. 4). DRE and ABRE are well known as cis-acting elements for stress-induction of many stress responsive genes (Baker et al. 1994; Yamaguchi-Shinozaki and Shinozaki 1994; Jiang et al. 1996; Yamaguchi-Shinozaki and Shinozaki 2005). In particular, promoter sequences of SK₃-type DHN genes, Csa6M358710.1, MELO3C026550T1 and Cla021949, showed similar distribution of DRE- and ABRE-like sequences among cucumber, melon and watermelon.

Understanding the molecular mechanisms of abiotic stress responses in Cucurbitaceae crops is as important as in other crops to breed new cultivars with stress tolerance. On this account, DHN genes encoding molecular chaperons were identified in Cucurbitaceae species through genome-wide search. To the best of our knowledge, this study is the first to present the results of comprehensive analysis of DHN genes in Cucurbitaceae species.

In this study, a total of 32 DHN genes were identified from eight genome sequences of five Cucurbitaceae species (Table 2). Each genome of the five species had three to five DHN genes, which were almost identical to number of those genes reported previously in cucumber, melon and watermelon (Celik Altunoglu et al. 2016; Celik Altunoglu et al. 2017). On the other hand, DHN gene number identified in each Cucurbitaceae genome was smaller than those in Arabidopsis and rice genomes. In addition, K_nS and Y_nK_n-type DHN genes could not be found in Cucurbitaceae genomes, unlike Arabidopsis and rice. Although conserved domain searches using reliable program and manual inspection were employed in this study, still it cannot be excluded that another DHN gene with less conserved K-segment might be present in Cucurbitaceae genomes. In fact, DHNs lacking complete K-segments were identified in Pinus pinaster (Perdiguero et al. 2014).

Although the detailed function of DHNs has not been known yet, several studies have revealed the function of DHNs (Rorat 2006). For example, some YSK_n-type DHNs bind to lipid vesicles, which are thought to contribute to maintaining membrane structure. K_nS-type DHNs have the activity of reactive oxygen species (ROS) scavenging. SK_n and K_n-type DHNs are closely related to cold acclimation (Rorat 2006). Therefore, considering the association of SK₃-type DHN genes to cold tolerance (Danyluk et al. 1994; Houde et al. 2004; Rorat 2006; Yin et al. 2006; Xing et al. 2011), Cucurbitaceae SK₃-type DHNs in Group I might be important for cold tolerance (Fig. 1). This possibility can be supported by the Group I members, COR47 and OsDHN1, that are downstream target genes regulated by CBF/DREBs, key transcription factors for cold tolerance (Seki et al. 2001; Thomashow et al. 2001; Lee et al. 2005; Lee et al. 2013). In addition, Yin et al. (2006)’s study showed that overexpression of a Solanum sogarandinum SK₃-type DHN gene increase cold tolerance in transgenic cucumber seedling, which is also in accordance with our findings.

Among DHN genes whose expressions were investigated, only SK₃-type DHN genes were strongly expressed in cucumber and melon (Fig. 3), although further analysis about expression change under stress condition is needed. In addition, DRE- and ABRE-like sequences were well conserved among promoter sequences of SK₃-type DHN genes in cucumber, melon, and watermelon (Fig. 4), which supports the role of the SK₃-type DHNs in stress tolerance in Cucurbitaceae crops.

In conclusion, the present research characterized DHN genes and their expression pattern in five Cucurbitaceae species through genome-wide searches and in silico expression profiling. The DHN genes identified in this study will be valuable genetic resource for understanding stress response mechanism and assisting molecular breeding in Cucurbitaceae crops.

We thank Prof. Hideo Matsumura in Shinshu University, Japan for kindly providing gene set of bitter gourd. This work was supported by Korea Institute of Planning and Evaluation for Technology in Food, Agriculture, Forestry and Fisheries (IPET) through Agri-Bio industry Technology Development Program, funded by Ministry of Agriculture, Food and Rural Affairs (MAFRA) (grant numbers 116076-03-2-HD090 and 116076-03-2-HD0b0).