In vitro shoot cultures of rootstock M26 from M. domestica were sub-cultured at 4-week intervals on an Murashige and Skoog (MS) agar medium containing 0.5 mg L⁻¹ indole butyric acid (IBA) and 1.0 mg L⁻¹ 6-benzylaminopurine (6-BA). Initial growing conditions were 25°C, 100 μmol photons m⁻² s⁻¹, and a 16-h light/8-h dark photoperiod. For root induction, 4-week-old shoots were shifted to MS medium supplemented with 0.3 mg L⁻¹ indoleacetic acid (IAA).

Seeds of Arabidopsis ecotype “Columbia” (Col-0) were first incubated in dark at 4°C for 3 days, then surface-sterilized according to the protocol described by Weigel and Glazebrook (2002), and sown on medium containing ½-strength MS salts, sucrose (3%, w/v), and agar (0.75%, w/v). Seedlings were transferred to soil (rich soil:vermiculite, 2:1, v/v) after germination for 7–10 days and grown in a climate chamber at 22°C, 70% relative humidity, 100 μmol photons m⁻² s⁻¹, and 16-h light/8-h dark photoperiod.

For subcellular localization, the coding sequence of MpCYS4, without the termination codon, was cloned into the XhoI/SacI site of pBI221-GFP vector (Clontech, Palo Alto, CA, USA). The specific primers are shown in Table S1. For transient expression in onion epidermal cells, pBI221-MpCYS4-GFP plasmids or blank vector pBI221 were inserted into the cells using a PDS-1000 biolistic helium gun device (Bio-Rad, Hercules, CA, USA), as the protocol described by Diaz et al. (2005). Then, the cells were cultured on MS medium in dark for 24 h before observation of the fluorescent proteins with a LSM510 confocal laser-scanning microscope (Carl Zeiss, Thornwood, NY, USA).

The coding region of MpCYS4 was cloned into the EcoRV/XhoI site of pET-32a expression vector (Novagen, Gibbstown, NJ, USA), introduced into Escherichia coli BL21 strain. The specific primers are shown in Table S1. Bacterial cells expressing the fusion protein were harvested after induction with isopropyl β-D-thiogalactopyranoside (IPTG). The recombinant protein was purified using a nickel column chromatography procedure (Novagen) and quantified by method of Bradford (1976). Testing the capacity of the recombinant MpCYS4 protein to inhibit cysteine proteases has been described previously (Gaddour et al., 2001), using papain (EC3.4.22.2; Sigma, St. Louis, MO, USA) and benzoyl-L-arginine-p-nitroanilide (BANA) as substrate. Control assays used the equivalent molar amount of purified recombinant protein isolated from E. coli that harbored the empty pET-32a vectors. Absorbance of the enzyme reaction was recorded at 405 nm with a benchmark microplate reader (Bio-Rad).

The coding region of MpCYS4 was cloned into the pBI121 vector (Clontech) under the CaMV 35S promoter. The specific primers are shown in Table S1. Subsequently, the resultant vectors were genetically transformed into Arabidopsis ecotype “Col-0.” Transformation was accomplished via the floral dip method using Agrobacterium tumefaciens strain EHA105 harboring the recombinant plasmid, as described by Clough and Bent (1998). Putative transgenic Arabidopsis plants were selected on MS medium containing 50 mg L⁻¹ kanamycin. After further selecting at a 3:1 segregation ratio, T3 homozygous transgenic lines were derived for the phenotypic analysis.

For germination assays, Arabidopsis seeds after surface-sterilized, were placed on the germination media containing various concentrations of mannitol (an osmotic agent; Wang et al., 2011) or ABA (A1049; Sigma). Germination was considered complete when radicles had emerged by 1 mm. The experiments were repeated for three times, and approximately 50 seeds per line were sown for each experiment.

During drought period, net photosynthesis (Pn) was monitored using a Li-6400 portable photosynthesis system (LiCor, Huntington Beach, CA, USA), at a constant airflow rate of 500 μmol photons m⁻² s⁻¹, with vapor pressure deficit of 2.0–3.4 kPa and cuvette CO₂ concentration of 400 μmol CO₂ mol⁻¹ air. Data were obtained from light exposed, fully-expanded leaves on each of five plants. Values for relative water content (RWC) were determined according to method described by Gaxiola et al. (2001). Electrolyte leakage (EL) was determined from leaves as described by Dionisio-Sese and Tobita (1998), with an electrical conductivity meter (DSS-307; SPSIC, Shanghai, China). Chlorophyll was extracted with 80% acetone, and the concentration was determined spectrophotometrically according to the method of Lichtenthaler and Wellburn (1983). ABA was extracted and measured as described by Li et al. (2015), with a high-performance liquid chromatograph (HPLC) (LC2010A; Shimadzu, Kyoto, Japan).

The ABA-induced stomatal closure was measured from leaves of 4-week-old plants that had been detached and floated abaxial side down on a stomatal opening solution containing 10 mM MES-KOH (pH 6.15), 20 mM KCl, and 1 mM CaCl₂. After they were incubated for 2 h under lights, they were then treated for 2 h with various concentrations of ABA added to the solution. Obtained from epidermal peels, the stomata were photographed and counted with a JSM-6360LV scanning electronic microscope (JEOL Ltd., Tokyo, Japan) as previously described (Tan et al., 2014b). Their sizes were measured using Image J software, using at least 50 apertures from each treatment.

The rate of transpiration (water loss) was evaluated by placing detached, fresh Arabidopsis leaves on open Petri dishes with abaxial side up and weighing them at regular intervals. The treatment was conducted at room temperature. Third to fourth true rosette leaves were collected from 3-week-old soil-grown plants under standard growing conditions as described above. In each experiment, approximately 50 leaves per line were used and the entire test was repeated for three times.

To assess drought resistance of the soil-grown plants, 1-week-old seedlings were transplanted in soil for 3 weeks under standard growing conditions before being subjected to water deficit by withholding water for 14 days (severe stress). The leaf RWC and stomatal apertures were measured from tissues sampled after 7 days. Digital images and plant survival rates were recorded on Day 14. Experimental variations were minimized by growing the same number of plants on each tray. The entire test was repeated at least three times.

For ABA treatment, 100 μM (±)-ABA (Sigma; A1049) was sprayed onto 5-day-old seedlings grown on MS plates to ensure total foliar coverage. In parallel experiments, water-only was sprayed as the control. Samples were collected after 0, 3, 6, 12, or 24 h of treatment, then frozen in liquid nitrogen and stored at −80°C for RNA extractions.

The MpCYS4 overexpression vector (pBI121-MpCYS4) was transferred into apple by Agrobacterium-mediated transformation, using leaf discs of apple rootstock M26, as described by Kotoda et al. (2010). After cultivation in dark for 2 weeks, the explants were moved under lights. Their shoots were excised and then propagated on multiplication medium containing 50 mg L⁻¹ kanamycin. The rooted plantlets were transferred to plastic pots (4 × 5 × 5 cm) filled with a mixture of forest soil, sand, perlite, and vermiculite. Growth conditions were 24 ± 2°C, 70 ± 5% relative humidity, 100 μmol photons m⁻² s⁻¹, and a 16-h light/8-h dark photoperiod.

After 2 months, the plants were transplanted in plastic culture pots (30 × 26 × 22 cm) contained a mixture of forest soil:organic substrate:sand (5:1:1, v:v:v), and placed in the greenhouse under ambient light, with temperature ranging from 20 to 35°C and a relative humidity of 65–70% at Northwest A&F University, Yangling, China (34°20′N, 108°24′E). After 3 months of growth under this condition, healthy and uniformly sized plants (30 per line, for a total of 60) were subjected to a water deficit. Another 30 plants per line (Control) continued to receive irrigation daily to maintain the saturated soil water content. Soil water content (v/v) was determined with an HH2 Moisture Meter (Delta-T Devices, Cambridge, UK). Drought was induced by withholding water for 8 days. At 9:00 a.m. on alternate days, Pn was recorded on the ninth to twelfth leaves from the stem base of five plants with similar soil water content. Then, the leaves were rapidly frozen in liquid nitrogen and stored at −80°C for RNA extractions. At each time point, another set of samples was collected to determine leaf RWC. After 3 days of treatment, additional leaves were sampled for stomatal observation, ABA measurement, and RNA-Seq analysis. Leaf chlorophyll concentrations and EL values were measured at the final time point.

Total RNA was isolated using E.Z.N.A.® Plant RNA Kit (Omega Biotek, Norcross, GA, USA) according to the manufacturer instructions. RNA integrity was checked on a 1.2% agarose gel and RNA quantity was assessed with a NanoDrop 1000 Spectrophotometer (Thermo Scientific, Wilmington, DE, USA). First-strand cDNAs were synthesized from 2 μg of total RNA with SYBR Prime Script RT-PCR Kit II (TaKaRa, Tokyo, Japan).

Semi-quantitative RT-PCRs were performed to examine the transcript levels of MpCYS4 in apple and Arabidopsis. Apple Actin and Arabidopsis Actin were used as respective loading controls. Quantitative real-time PCR (qRT-PCR) was conducted on an iQ5.0 detection instrument (Bio-Rad) using SYBR Green qPCR kits (TaKaRa). Reactions were performed in triplicate with a volume of 20 μL, and apple EF-1α or Arabidopsis Actin was amplified as the appropriate internal control. The relative quantity of target gene transcript was determined by applying the 2^−ΔΔCT method (Livak and Schmittgen, 2001). Primers are listed in Tables S1, S2. Each value was expressed as mean and standard deviation (SD) calculated from the result of three independent replicates.

Transcriptional profiling of WT apple and transgenic line #4 was conducted by RNA-Seq. Three biological replicates were used for each genotype, with samples collected from plants grown under normal conditions as well as after drought induced by withholding irrigation for 3 days. Each biological replicate comprised samples from three to four plants that were mixed after collection to generate one independent pool. Total RNA (2 μg per sample) was prepared and used for both cDNA library construction (Zhong et al., 2011) and RNA-Seq by the Gene Denovo Biotechnology Corporation (Guangzhou, China), with an Illumina HiSeq™ 2500 (Illumina, San Diego, CA, USA). The files of raw fastq were checked by FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc). After the adaptor sequences were removed using fastqmcf (https://code.google.com/p/ea-utils/wiki/FastqMcf), the reads were aligned to the database for M. domestica genome (Velasco et al., 2010; http://www.nature.com/ng/journal/v42/n10/full/ng.654.html) using TopHat2 (Kim et al., 2013). All usable reads were then normalized into fragments per kilobase of transcript per million mapped reads (FPKM) (Mortazavi et al., 2008). The transcript abundance differences among samples were computed with the ratio of FPKM values. The significance of those differences was evaluated with a false discovery rate (FDR) as the threshold of P-value for multiple tests (Yekutieli and Benjamini, 2001). Genes with FDR <0.05 and fold-changes >2 were used in later analyses. Our methods for assessing the significance of differentially expressed genes (DEGs) followed those described by Anders and Huber (2010).

All statistical analyses were conducted using SigmaPlot software (Systat Software). Data were analyzed via one-way ANOVA using the SPSS-11. Statistical differences were compared based on Student's t-tests. ^*, ^**, and ^*** indicates significance level at P < 0.05, P < 0.01, and P < 0.001, respectively.

MpCYS4 encoded a phytocystatin protein containing 248 amino acids with a predicted molecular weight of 27.72 kDa. Characteristic sequences and motifs of MpCYS4 are highlighted in Figure 1A, including the reactive site QVVAG in the central part, two glycine residues (GG) near the N-terminal, PW residues near the C-terminal, and a SNSL motif in the C-terminal extension. The typical sequence of PhyCys, LARFAV, was also localized at the N-terminus of the protein.

To investigate the subcellular localization of MpCYS4 protein, we fused it with GFP at the N-terminal and transiently expressed in onion epidermal cells. After 24 h, we detected the accumulation of MpCYS4-GFP fusion protein in the cell membrane, cytoplasm, and nucleus of those cells, which was the same as the subcellular localization of the GFP protein (Figure 1B). This result was consistent with previous reports for other cystatins, such as BvM14-cystatin from Beta vulgaris “M14” (Wang et al., 2012) and GsCPI14 from Glycine soja (Sun et al., 2014).

His-tagged fusion protein of MpCYS4 was expressed in E. coli cells and affinity-purified (Figure 1C). To confirm that this gene is a functional cysteine protease inhibitor, we used an in vitro assay with papain as substrate. Testing the behavior of the recombinant MpCYS4 protein against cysteine protease revealed that papain activity was blocked in a concentration-dependent manner (Figure 1D). Meanwhile, when the purified and induced protein of empty pET32a was added into the reaction as a negative control, this activity was not inhibited.

We previously demonstrated that MpCYS4 is up-regulated by drought, heat, exogenous ABA, or MV treatments in leaves of M. prunifolia (Tan et al., 2014a). We further investigated its potential functioning during abiotic stress responses by analyzing the phenotypes of MpCYS4-overexpressing plants. The coding sequence of MpCYS4 was inserted into a transgenic vector and constitutively expressed in transgenic Arabidopsis plants. In all, we obtained 16 independent T3 homozygous lines and verified them via PCR analysis (data not shown). Two representative lines with the highest expression—overexpressing (OE)-4 and OE-13—were selected for further examination (Figures S1A,B). On a MS growth medium, seed germination and elongation of the primary roots from P_CaMV35S:MpCYS4 was retarded when compared with the wild-type (WT) (Figures S1C–E), while no obvious difference was observed for soil-grown plants in their aerial parts.

Because MpCYS4 expression was strongly induced by water deficit in M. prunifolia (Tan et al., 2014a), our study focused on the drought tolerance of 35S:MpCYS4 transgenic Arabidopsis. Seeds from the 35S:MpCYS4 transgenic and WT plants were germinated and seedlings were then grown on a MS medium containing 400 mM mannitol. Both germination and post-germinative growth of all genotypes were repressed, albeit to varying degrees. For example, the seeds of transgenic lines (OE-4 and OE-13) germinated much faster compared to the WT. After 14 days, only 19% of the latter had germinated compared with 83% for OE-4 and 73% for OE-13 (Figure S1C). Seedlings of the transgenic lines also grew faster and had longer primary roots and higher fresh weights (FWs) than the WT (Figures S1D,E). The results suggested that MpCYS4 overexpression alleviates the adverse effects induced by osmotic stress on Arabidopsis growth.

The rate of water loss was slower for 35S:MpCYS4 plants than the WT, based on FW measurements from detached rosette leaves. As shown in Figure 2A, after 3 h on open Petri dishes, the FWs of detached rosette leaves were reduced to 69% of the original value for OE-4, 66.7% for OE-13, and 60.1% for the WT. In addition, the survival rates for 35S:MpCYS4 plants deprived of water for 14 days were higher (77.8% for OE-4, 73.9% for OE-13) than for similarly treated WT plants (10.7%) (Figure 2B). These results confirmed that overexpression of MpCYS4 could greatly improve drought resistance in transgenic Arabidopsis.

The function of MpCYS4 in controlling water losses prompted our investigation into the operation of stomatal behavior, a major factor affecting water-holding capacity in plant leaves. The stomatal density did not show significant difference between 35S:MpCYS4 Arabidopsis and WT plants before and after 7 days of drought treatment (Figure S2), which suggested that the higher drought resistance of 35S:MpCYS4 Arabidopsis is unrelated to stomatal density. Whereas, compared with the WT, stomatal apertures were smaller on 35S:MpCYS4 leaves when subjected to water-deficit conditions (Figure 2C). As shown in Figure 2D, no obvious difference was detected among 35S:MpCYS4 and wild-type plants under normal conditions. However, drought treatment led to a higher occurrence of closure in the transgenic lines, with stomatal dimensions for the WT being reduced to approximately 74% of the size measured from untreated plants vs. reductions of 51.5 and 51.8% of the original, unstressed sizes for transgenic lines OE-4 and OE-13, respectively. Consistent with these results, the RWC of leaves in the WT was 49.6%, as opposed to 78.4% for OE-4 and 70.9% for OE-13 after 7 days of drought treatment (Figure 2E). These findings indicated that, during periods of dehydration, the stomata of the transgenic lines responded to water deficits better than did the WT plants.

Stomatal closure is one of the most important ABA-induced environmental responses. Then, we examined 35S:MpCYS4 plants to determine whether MpCYS4 overexpression affects ABA sensitivity. When germinated on an MS medium supplemented with ABA, both the germination and post-germinative developmental stages of the 35S:MpCYS4 proved hyposensitive to exogenous ABA compared with the WT (Figures 3A,B). In the presence of 0.25 μM ABA, 35S:MpCYS4 seedlings were dramatically restrained while the WT seedlings showed expanded, green cotyledons (Figure 3A). To determine whether MpCYS4 is involved in ABA-induced stomatal movement, we treated WT and 35S:MpCYS4 leaves with ABA and found that such treatment resulted in a higher degree of closure in the transgenic plants (Figure 3C). Thus, MpCYS4 seems to have an important role in ABA-mediated control of the guard cells.

Our germination and post-germination assays revealed an increased sensitivity by 35S:MpCYS4 plants under ABA exposure, while endogenous ABA levels did not show significant difference between 35S:MpCYS4 Arabidopsis and WT seedlings before and after subjected to 100 μM ABA for 24 h (Figure S3). The specific and dramatic induction of expression for stress genes represented a prominent response to ABA at the molecular level. Therefore, we used 35S:MpCYS4 and WT plants treated with 100 μM ABA to assess the expression patterns of several ABA-responsive genes by qRT-PCR.

Upon ABA treatment, three A-type protein phosphatase 2C (PP2C) family members—ABA-INSENSITIVE1 (ABI1), ABI2, and AtPP2C—were induced in the WT and even more strongly in 35S:MpCYS4 seedlings (Figure 4). Other genes in the molecular network of the ABA signaling pathway were also tested, e.g., OPEN STOMATA1 (OST1); the NADPH oxidase F (AtrbohF); the anion channel SLOW ANION CHANNEL-ASSOCIATED 1 (SLAC1); ABA-responsive promoter element (ABRE) binding transcription factors ABI5 and ABF3; and other downstream targets such as LOW TEMPERATURE-INDUCED 65-kDa PROTEIN (RD29B), RESPONSIVE TO ABA 18 (RAB18), RESPONSIVE TO DEHYDRATION 22 (RD22), and COLD-REGULATED PROTEIN COR6.6 (KIN2), which serve as marker genes for monitoring ABA and stress responses in plants. In our experiments, those genes were highly induced upon ABA treatment in the WT plants and even significantly more in the 35S:MpCYS4 plants. Although expression was detected for all of them, the extent and kinetics of this induction differed among markers. These findings suggested that MpCYS4 could interact with ABA-mediated signaling process and affect multiple ABA-responsive genes.

To examine whether MpCYS4 confers tolerance to drought in apple, MpCYS4 transgenic apples were obtained. The three MpCYS4-overexpressing lines—#1, #3, and #4—which were detected with higher MpCYS4 transcript level were chosen for further investigation (Figures S4A,B). Under normal growing conditions, the aerial portions of those transgenic plants showed apparent dwarfing and slower development when compared with the traits of the WT (Figures S4C–F). However, the root systems did not show obviously different among the transgenic lines and the WT (data not shown). Drought stress was induced by withholding water from these plants for 8 days. No obvious difference was observed among the tested lines on their growth potential at the onset of the experiment. While, the transgenic lines grew much better than the WT as the stress period was prolonged. The rate of net photosynthesis decreased rapidly after drought treatment, with rates being significantly lower for the WT than for 35S:MpCYS4 plants (Figure 5A). In addition, the RWC calculated for leaves from each line was significantly reduced under drought conditions. However, overexpression of MpCYS4 alleviated this response obviously over time (Figure 5B). Consistently, this phenotype was also confirmed by other measurements, with the transgenic lines having much less EL and higher chlorophyll concentrations than the WT at the end of the treatment period (Figures 5C,D).

Meanwhile, the stomata characters of the 35S:MpCYS4 plants and WT plants were checked using scanning electron microscopy after water was withheld for 3 days. Under normal conditions, both the stomatal density and aperture sizes did not differ significantly between the two genotypes (Figure S5; Figures 6A,B). However, the transgenic lines showed much greater stomatal closure than the WT when drought stress was induced (Figure 6B). Stomatal apertures were also measured from overexpressing and WT plants in response to ABA treatment. In the absence of ABA, sizes were not obviously different among genotypes. However, after 2 h of exposure to 50 or 100 μM ABA, the degree of closure was greater in the 35S:MpCYS4 plants (Figure 6C), thereby indicating that they had a drought-insensitive phenotype associated with ABA-mediated stomatal closure under the water deficit. Then we examined 35S:MpCYS4 apple plants to determine whether MpCYS4 overexpression affects endogenous ABA levels. As shown in Figure S6, no obvious difference was detected among 35S:MpCYS4 and WT plants under normal conditions. Drought stress increased the levels of ABA in both WT and MpCYS4 transgenic lines, but no significant differences appeared among them. These data indicated that MpCYS4 could function in plants ABA sensitivity but not in the biosynthesis of ABA under drought stress conditions.

To obtain further insight into the molecular mechanisms by which MpCYS4 mediates tolerance to drought stress in apple, we performed a large-scale RNA sequencing (RNA-Seq) with RNA extracts from leaves of WT and #4 transgenic plants either grown under normal conditions or subjected to drought by withholding irrigation for 3 days. A total of 12 samples were used with three biological replicates for each genotype. Clean reads were obtained after removing rRNA contaminated reads, with mapping to the genome sequence made up approximately 70% (Table S3). To further evaluate the robustness of the RNA-Seq data, we calculated the correlation coefficients of the transcriptome profiles among the 12 samples. The value could reach 0.99 between each set of biological replicates (Table S4). Based on a two-fold criterion, 82 and 69 genes were up and down-regulated, respectively, in the transgenic line when compared with the WT in the absence of drought stress (Figure S7A). Under water deficit, even more genes exhibited differential expression in line #4 than in WT, with 1556 and 1327 genes being up and down-regulated, respectively (Figure S7B). Among the DEGs, 15 and 8 were commonly up or down-regulated, respectively, in #4 under control and after water was withheld for 3 days (Figure S7C).

It is worth mentioning that many DEGs involved in ABA signaling transduction and stress response were identified under water-deficit conditions (Table 1), including PYL4 (MDP0000228470) and PYL9 (MDP0000284624), which were down-regulated in stressed transgenic line. By contrast, genes encoding ABI1 (MDP0000437033), ABI2 (MDP0000231674), HAB1 (MDP0000265371 and MDP0000178692), OST1 (MDP0000224969), and the ABRE binding transcription factor ABF3 (MDP0000701734 and MDP0000248567) were up-regulated in those transgenic plants. Furthermore, transcript levels of several other regulatory proteins associated with stress signal transduction, such as mitogen-activated protein kinase, calcium-dependent protein kinase and transcription factors, including NAC, WRKY, MYB, MYC, MADS, DREB, basic helix-loop-helix (bHLH) transcription factor, heat stress transcription factor (HSF), and APETALA2/ethylene-responsive factor (AP2/ERF) were activated in the overexpression line. In addition, the genes encoding major stress-responsive functional proteins that have been demonstrated to play a direct role in stress response, such as ascorbate peroxidase (APX), catalase (CAT), glutathione S-transferase (GST), late embryogenesis abundant protein (LEA), RD29B, RD22, and KIN2 were also enriched in drought-treated 35S:MpCYS4 plants. To validate the results from this RNA-Seq analysis, we used qRT-PCR to evaluate the expression of 12 of those genes (Figure 7). It was evident that the expression patterns of those selected genes were largely consistent with what we found from our RNA-Seq data despite the difference in absolute fold-changes between the two methods. This demonstrated that our RNA-Seq results were reliable.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer HS and handling Editor declared their shared affiliation, and the handling Editor states that the process nevertheless met the standards of a fair and objective review.