The seeds of two barley (Hordeum vulgare L.) cultivars, Amulet (Amu; a spring barley variety, originating from the Czech Republic) and Tadmor (Tad; a two-row barley selected from the Syrian landrace Arabi Aswad), which are differentially sensitive to drought, were obtained from Mendel University in Brno (CZE). The seeds were germinated for 2 days at 24°C in the dark; then the seedlings were grown in pots (bottom diameter 13 cm, upper diameter 21 cm, height 20 cm) filled with soil (a mixture of Alfisol with manure and sand, 6: 2: 1) at 25°C/20°C (light/dark, 14 h/10 h, irradiation intensity 350 μmol⋅m^-2⋅s^-1 provided by a high-pressure sodium lamp + incandescent bulb) in a growth chamber (Tyler, type T-16/4, Budapest, Hungary) for the next 9 days. The soil was maintained at 70% of maximum water capacity (MWC), with watering of the pots each day to maintain a constant weight (5,500 g). Next, one-half of the pots were kept under the same conditions until the plants reached the stage of a fully developed second leaf (well-watered, WW). The remaining half of the pots were withheld from water for the next 8 days (water-stress, WS) until the plants had reached the same growth stage as in the WW treatment (Figure 1). When the plants reached the stage of a fully developed leaf 2, the five crowns and five leaf 2 were sampled from five plants grown in five different pots with the same treatment (WW or WS) for determination of water-relationship parameters in three repetitions. The same number of samples was taken for abscisic acid (ABA) determination as well as dehydrin protein analysis. The experiment was two times repeated. The plants (the second leaf and crown separately) were sampled from both conditions (WW and WS), in order to measure their water-relationship parameters. The plant tissues used for determination of ABA level, content of dehydrins, and transcription activities were snap-frozen in liquid nitrogen and stored at -80°C.

The plant tissue was snap-frozen in liquid nitrogen and stored at -80°C, before being used for RNA extraction based on the TRIZOL reagent (Invitrogen, Carlsbad, CA, USA). The RNA was purified by passing through an RNeasy column in the presence of DNase (Qiagen, Hilden, Germany). RNA quality was assessed by both agarose gel electrophoresis and analysis in an Agilent 2100 Bioanalyzer (Agilent Technologies, USA). Each biological sample was represented by three independent replicates, each of which consisted of a bulk of four seedlings. Each RNA sample was hybridized to the Affymetrix 22 K Barley1 GeneChip Genome Array (Close et al., 2004). GeneChip^® hybridization quality was ensured by using standard controls supplied by manufacturer, and B2 oligonucleotides were added to each hybridization cocktail. PolyA controls (lys, phe, thr, dap) and hybridization controls (BioB, BioC, BioD and Cre) were used to monitor labeling and hybridization. Freely available software R (R Core Team, 2015) and associated library packages were employed for statistical computing. Raw data from microarrays were subjected to preprocessing analysis incorporating functions from Affy package library (Gautier et al., 2004), with emphasis on boxplots, density plots and Bland-Altman plot modification (MVA plot). Subsequently, Robust Multi-array Average method (RMA; Irizarry et al., 2003) from the same library was implemented for normalization. With the aid of this algorithm, background noise and processing artifacts were eliminated. Alongside, an iterative median polishing procedure summarized the data and generated a single expression value for each probe set. After normalization, repeated data verification was made by tools from initial analysis. Arrays exclusively meeting the criteria as recommended by manufacturer were submitted to further evaluation.

Differentially expressed genes (DEGs, |log₂-fold|≥ 2) were identified via linear model for microarray analysis from Limma library package (Smyth, 2005). Using this model, pairwise analysis of individual samples against all others was accomplished, considering P-value < 0.05 as threshold of expression differences significance. Duplication discarding procedure provided reduction of redundant probe sets to unique records, which figured in repeated RMA as a restriction criterion. To reach comprehensive classification of mutual interactions across observed samples, expression values matching sorted probe sets were tested using principal component analysis (PCA) from a map library package (Lucas, 2014). Enhanced heat map (gplots library; Warnes et al., 2015) combined with hierarchical clustering was involved in manifesting disparity within RNA samples originated from drought-treated plants. Log2-transformed differences in transcription activity of described samples over parallel controls were color-scaled based on expression strength and clustered as measurement of similarity. In probe sets arising of aforesaid comparison, searching was done for numbers of genes specific or common to distinct stressed samples. Differentially expressed genes specific/common to singly treated samples was visualized using quad-set Venn diagrams (VennDiagram; Chen, 2013).

The linear model combined with moderate F statistics (Smyth, 2005) was used to identify genes with variety-specific response to drought, i.e., genes with a strong response to drought in one variety and low response in the second variety. The resulting variety-specific response was plotted as a scatter plot. Genes that do not respond to stress or respond similarly in both varieties are located in the origin, whereas variety-specific responding genes are located outside the origin.

Annotation of differentially expressed genes was mediated by Plexdb annotation tool. Consensus sequences of selected probes were blasted against Uniref90 (last release) using blastx with maximum e-value of 1e-4.

The water content (WC, % of water weight to fresh weight) was measured in the crowns (under-ground bases (nodes) of plants), and the second leaves for both treatments. Moreover, the water saturation deficit (WSD) and osmotic potential (OP) were measured in leaf 2. OP was determined psychrometrically from the leaf sap, using a Wescor HR-33 microvoltmeter (USA) with C52 sample chambers. Results were expressed in MPa, determined by reference to calibration solutions of NaCl. WSD was determined in cut leaf segments (1 cm length) which were rapidly weighted (FW), then water-saturated in a wet sheet of foam, and weighted after 3 h (W3) and 6 h (W6), and finally after drying at 90°C (DW). WSD was calculated by Slavík (1974): WSD (%) = [(2W3 – W6 – FW)]/[(2W3 – W6 – DW)] × 100.

The samples for (±) ABA analysis were homogenized and extracted into distilled water (0.1 g FW/1 ml H₂O), shaken for 16 h under cold (4–5°C) and dark conditions, and processed by indirect ELISA according to Asch (2000). For each sample, we used three replicates on a microtitre plate. An extinction photometer SUNRISE Remote (Tecan, Germany) was used to measure color intensity of the final product at 405 nm, and the ABA concentration was calculated.

The dehydrin extraction as well as the 1D SDS-PAGE and immunoblots analyses were performed according to Vítámvás et al. (2007). Proteins were separated by SDS-PAGE on 10% gels (Laemmli, 1970; Multigel-Long, 24 well comb, Whatman Biometra). About 3 μg of soluble proteins (upon boiling) extracted from crowns and leaves were loaded on gels. To compare intensities of different samples on different membranes, calibration samples with different load (cca 1:5 of crown tissue) were added on gels as well. Precision Plus Protein All Blue Standards, mixture of ten blue-stained recombinant proteins (10, 15, 20, 25, 37, 50, 75, 100, 150, 250 kDa) were also added to gels. Gels were stained by Bio-Safe Coomassie (Bio-Rad) to obtain sample load controls. The relative accumulation of dehydrins was determined densitometrically using Quantity One software (Bio-Rad, v. 4.6.2., Munich, Germany).

Results were expressed as averages calculated from six repetitions, and compared using a T-test, ANOVA analysis, and Duncan’s multiple range test at the 0.05 level (Statistica v. 10 software, StatSoft, Tulsa, OK, USA).

Water content significantly decreased in all parts of plants under WS (water stress) conditions (Figure 2A). Reduction of WC was greater in the crowns than in the leaves. However, Tad retained significantly higher WC in the leaves and crowns than Amu did in WS plants. WSD increased and OP decreased in the second leaf in both varieties after WS conditions. WSD was higher and OP was lower in Amu than in Tad (Figures 2B,C).

These results showed that both varieties reached water scarcity under WS conditions. However, Tad lost less water and suffered lower water deficit than Amu.

The content of ABA significantly increased in both varieties under WS conditions. The level of ABA increased much more in Tad than Amu, although its lower content in Tad control plants (Figure 2D) indicating the important signaling role of ABA in the adaptation to drought stress.

Among DEGs, two were presented that encode key enzymes participating in ABA biosynthesis, 9-cis-epoxycarotenoid dioxygenases NCED9 (at1g78390.1) and NCED4 (at4g19170.1). These two genes, however, show different trend of expression (see Table 1). NCED9 significantly increased its expression in leaves of both genotypes with Amu having the higher level, while NCED4 lowered its expression in leaves of both genotypes and such decline was stronger in Amu. Although in both cold and drought stress ABA seems to play a central role in the basal stress response, the mechanisms which ABA is regulated might differ. For example the NCED4 gene was up-regulated in cold stress treatment in our previous study (Janská et al., 2014).

Immunoblot analysis of leaf and crown samples revealed five detectable dehydrin bands of different molecular weights: 82, 28, 25, 20, and 18 kDa (Figure 3A). In all samples, the most abundant dehydrin protein was 80 kDa, corresponding to dehydrin 5 (Kosová et al., 2010). The composition of individual dehydrins was different in the two varieties. For example, the 25 kDa dehydrin was characterized in Tad, whereas the 28 kDa dehydrin was found in Amu. The relative content of dehydrins was higher in the crowns than in leaves in both varieties, and the density of all dehydrins was higher in the crowns of Amu than those of Tad in WS conditions (Figure 3B).

In treated barley plants we identified 16 genes encoding dehydrins which compared to the non-treated plants were significantly expressed (see Table 2). Based on the expression changes, we can conclude that in the crowns Amu exhibited higher expression of dehydrin-encoding genes than Tad did. However, significant difference between the cultivars was recorded only for genes encoding dehydrins 5 (at3g50970.1) and 10 (at3g50970.1; at2g21490.1). Tad exhibited higher expression of certain dehydrin-encoding genes in leaves compared to Amu.

To investigate mutual interactions within dataset, pairwise analysis of every sample combination was performed followed by PCA of RMA subjected expression values matching DEGs (|log₂-fold|≥ 2, P ≤ 0.05). As is evident from PCA (Supplementary Figure 1) the distinction of biological replicates within individual samples represents a smaller component of total variance than variability across observed samples, thus adding further confirmation of data quality.

Susceptible variety Amu exhibited higher number of DEGs as a response to drought stress (Figure 4). Moreover, Amu has responded to stress more strongly in crown tissue (727) than in the leaf tissue (651), while Tad has been characterized by an inverse relationship (leaf – 489, crown – 360). Eighty DEGs were common to all treated samples (Figure 5).

Among crown-specific transcripts whose expression is reduced in Tad relative to Amu [39 DEGs (13 unknown), Figure 6B, Supplementary Table 3], there were some genes encoding HSPs such as HSP17 (at5g59720.1), HSP18 (at5g59720.1), HSP70 (at3g12580.1) and dehydrins [DHN 5 (at3g50970.1), DHN10 (at3g50970.1, at2g21490.1)] as well as genes encoding enzymes involved in carbohydrate metabolism [WSI76 (at1g09350.1), sucrose synthase 2 (SuSy2; at3g43190.1)]. Another gene reduced in Tad relative to Amu encodes cytosolic aldehyde dehydrogenase RF2C (RF2C; at3g24503.1). This cytosolic aldehyde dehydrogenase is involved in a biosynthesis of ferulic acid, which has been documented to respond positively to drought stimuli and to be more accumulated in drought tolerant genotypes (Hura et al., 2007, 2008, 2009).

Lower transcription activity of genes encoding some proteins with transcription regulation activities, such as MADS box protein 5 (at1g69120.1) and methyl binding protein MBD108 (at5g35330.3) were also detected in Tad crowns (Figure 6B, Supplementary Table 3). Transcriptional regulators are an important component of plant response to abiotic stresses including drought. Nevertheless, wide scale study of genes of MADS-box family showed that some representatives are up-regulated, while others are down-regulated upon drought (Arora et al., 2007).

There are some Tadmor specific transcriptional responses to drought stress that could be the important components of the complex response of the high drought stress tolerant cultivars. Tadmor leaves remarkably enhanced expression of various classes of heat shock proteins after drought treatment. Results of some papers suggest that some HSPs could represent a possible source of drought tolerance. For example, Grigorova et al. (2011a) monitored an effect of combined drought-heat stress on the expression of selected HSPs in two contrasting wheat cultivars and detected a higher accumulation of smHSPs in drought tolerant genotypes. In addition, Guo et al. (2009) in large scale expression profiling of three variously drought-resistant cultivars, registered HSP17.8 as a specific response of tolerant genotypes. On the contrary, authors recorded HSP17.9 in both tolerant and susceptible cultivar. Similarly, HSP70 was reported to confer drought tolerance in plants (Cho and Choi, 2009). However, HSP70s are a variable group and not all representatives are drought-responsive (Sarkar et al., 2013). In barley, the gene encoding this protein was induced in both tolerant and susceptible genotype (Guo et al., 2009). HSP80 was described to create complexes with HSP70 in Neurospora crassa (Ouimet and Kapoor, 1998) and, as such, they could be relevant in drought defense. Presence of HSP40s (DnaJ) could be credited to their co-activity with HSP70s and their capability to drive functional specificity of HSP70s (Cyr et al., 1994). However, some papers also suggest HSP40s significance for photosynthetic reactions (Chen et al., 2010). Thus these proteins could significantly contribute to drought tolerance. We also detected enhanced expression of gene encoding TPR-repeat protein, which is an important component of protein–protein interactions and coordinates HSP70 and HSP90 co-chaperone activity (Song and Masison, 2005). Increased TPR activity upon drought-stress is well documented (Rosado et al., 2006; Zhu et al., 2013).

Our results support also the statement of Chan et al. (2013) that certain steps in sulfur metabolism are of great importance in drought stress signaling and response. Based on the above, we deduced that one of the defense strategies of tolerant cultivar facing water scarcity is an improved utilization of sulfur.

It was proved that changes in lipid composition could help to maintain membrane integrity and preserve cell compartmentation under water shortage (Blum and Ebercon, 1981; Quartacci et al., 1995; Aziz and Larher, 1998; Toumi et al., 2008) as well as a positive relation between cutin content and drought tolerance (Yang et al., 2011). Because Tad also achieved higher expression level of genes encoding proteins related to lipid biosynthesis or transport as well as to cutin formation, we suggest that increased activity of GDSL as well as LTPs in response to drought could be another mechanism of tolerant cultivar to cope with water scarcity.

The altered lipid metabolism is probably important drought stress tolerance mechanism not only in leaves, but also in the crowns. Besides gene encoding LTPs, two transcripts encoding caffeic acid O-methyltransferase (COMT) were presented. COMT was recently described as modulator of lignin content and composition (Guo et al., 2001). There are also some evidences about COMT contribution to drought tolerance (Hu et al., 2009). Shikimate kinase, which expression was also enhanced in Tad crowns after drought stress, plays a pivotal role in the formation of aromatic secondary compounds in plants and finally leads to the formation of, for example, stilbenes, flavonoids, and lignins. Considering the above, our assumption is that tolerant genotype efficiently utilizes and modulates lignin biosynthesis. Such effort may result from an endeavor to preserve tissue itself. Second alternative is that crown alters lignin in order to strengthen roots endurance. Observed roots lignification upon drought in several thesis supports such hypothesis (Cruz et al., 1992; Vartanian et al., 1994).

Tad crowns enhanced expression of genes connected to jasmonate signaling such as small JIPs (JIP 32.6, JIP 32.7, JIP 23), OPR and thionins. JIP23 is the most abundant barley JA induced protein and was reported to be localized specifically in highly osmotically stressed cells (Hause et al., 1996). These findings together with our results suggests its role in drought-stress response of plants. OPR is an enzyme directly involved in jasmonate biosynthesis and was recorded to enhance osmotic and salt stress tolerance in Arabidopsis plants during seed germination (Gu et al., 2008). Thionin represents one of the main jasmonate-induced proteins usually documented as a response to biotic stresses (Bohlmann et al., 1988; Carmona et al., 1993; Molina et al., 1993; Epple et al., 1997; Muramoto et al., 2012). Nevertheless, thionin up-regulation after drought exposure was described as well (Ozturk et al., 2002).

In the crowns of Tad, there we found out enhanced expression of genes encoding proteins associated with nucleosome remodeling (histone H3 and H4). The altered expression of such genes was described to be mechanism of barley freezing and cold tolerance. Since the gene expression of such genes was altered exclusively in the crown (Janská et al., 2011, 2014) and because freezing similar to drought causes cell dehydration, we conclude that differential expression of histones encoding genes might be also one of the mechanisms that enables Tad to cope with water stress.

Our results provide support to the proposal made by Fowler and Thomashow (2002) that the selective repression of genes is likely to represent a major component of the acclimation response. We suggest this implication also for drought stress. This is probably true for groups of genes promoting growth and genes encoding proteins involved in programmed cell death or senescence in the face of drought stress. Such genes could help the tolerant genotypes to redirected the energy toward stress adjustment mechanisms and prevent leaf senescence or death. So, Tad negatively regulated genes encoding enzymes involved in biosynthesis of gibberellin and cytokinin, phytohormones inducing cell division and growth. That suggests expression of the genes involved in active growth and development in Amu during drought stress. On the other hand, Tad facing drought lowered transcription of such genes in order to survive. Endeavor of Amu for active growth facing drought is supported by higher activity of pseudo-response regulator (PRR; at5g60100.2) encoding gene. The PRR is a part of regulatory pathway, which (via CONSTANS and VRN3/FT1 genes) can induce transition from the vegetative to the reproductive phase under long day.

Also decreased transcription activity of genes for cellulase, cellulase synthase and expansin indicates reduced cell wall elongation in Tad contrary to Amu and therefore cessation of plant growth in Tad. Some negative effect on plant growth could also have a gene encoding NAC6 (at1g01720.1; Figure 6C, Supplementary Table 4). This transcriptional factor was described to induce drought tolerance (Nakashima et al., 2007, 2012). However, such improvement was redeemed by a retarded growth and a lower productivity of plants.

Drought leads to a repression of many genes related to photosynthesis at transcriptional level (Chaves et al., 2009). This is true also for our results. Repression of such genes may also resulted from photosynthetic accommodation strategies of Tad, such as reduced chlorophyll content (Tardy et al., 1998; Havaux and Tardy, 1999).

Higher transcription of genes encoding JIP60 along with genes for bowman-birk type trypsin inhibitor (at2g40070.2), subtilase (at5g11940.1) and subtilisin-like proteinase (at5g11940.1), in Tad relative to Amu (Figure 6C, Supplementary Table 4) indicates leaf senescence or cell death processes in susceptible genotype or their absence, possibly delaying in the tolerant one. Delaying of leaf senescence was described as a mechanism of drought tolerance acquisition (Rivero et al., 2007; Muchero et al., 2013). Leaf senescence is also related to the changes in peroxidase activities (Kar and Mishra, 1976). So, a diminution of peroxidase (at5g05340.1) expression in Tad relative to Amu (Figure 6C, Supplementary Table 4) could be explained by an ability of resistant genotype to avoid or postpone such drought consequences. On the other hand, peroxidase repression in Tad can be a part of an oxidative burst blocking or wilting precluding initiative (Lagrimini et al., 1990).

Co-inhibition of phenylalanine ammonia-lyase, chalcone synthase and bZIP protein HY5 by Tad lead us to a suggestion about anthocyanins accumulation blocking in Tad or its stimulation in Amu. Anthocyanins are well documented to accumulate upon drought and to confer desiccation tolerance (Chalker-Scott, 1999; Deeba et al., 2012; Nakabayashi et al., 2014). In affected plants, anthocyanins probably function as a barrier against photoinhibition. Considering such facts, one would expect higher activity in Tad. Nevertheless, results of some papers suggest that Tad is equipped by different mechanisms which enable it to prevent photoinhibition (Tardy et al., 1998; Havaux and Tardy, 1999). That could be a reason why Tad does not accumulate transcripts which code for anthocyanin-related proteins.

Recent studies reported an inhibition of root growth by aldehyde dehydrogenase RF2C in rice plants (Chi et al., 2013), which expression was reduced in Tad crowns relative to Amu in our study. Therefore we suggest that Tad might either inhibits or does not accumulate RF2C in order to sustain root growth.