Identification of plant hormones and candidate hub genes regulating flag leaf senescence in wheat response to water deficit stress at the grain‐filling stage

Abstract

In order to clarify the transcriptional regulatory network and physiological mechanisms governing leaf senescence response to drought stress in wheat, experiments were performed using two wheat varieties with contrasting drought tolerance: Fu287 (F287, a drought‐sensitive genotype) and Shannong20 (SN20, a drought‐resistant genotype). The latter has higher SPAD values, salicylic acid (SA), jasmonic acid (JA), zeatin (Z), zeatin riboside (ZR), and gibberellin (GA ₃) content as well as higher expression levels of Cu/Zn‐SOD, Mn‐SOD, Fe‐SOD,POD,CAT, and APX under various water deficit conditions. Conjoint analysis of physiological and biochemical indicators and transcriptome data by weighted gene co‐expression network analysis (WGCNA) in the present study provides a useful genomic and molecular resource for studying drought adaptation in wheat. The flag leaf senescence process was changed by altering the concentration of phytohormones. SA, JA, abscisic acid (ABA), Z, ZR, and GA ₃ coordinate with each other to control leaf senescence and plant adaptation under drought stress. Further, the leaf senescence process was divided into two phases: the persistence phase and the rapid loss phase. Shorter Chl_total (duration of the flag leaf being photosynthetically active), shorter Chl_per (persistence phase), reduced M (inflection point cumulative temperature when senescence rate is the maximum), decreased r _max (the maximum senescence rate), larger r ₀ (the initial senescence rate), and increased r _aver (the average senescence rate) were slightly associated with low grain mass. We speculated that extending the period of the persistence phase by cultivation or chemical control measures could further increase the drought survivability and productivity of wheat.

1. INTRODUCTION

Wheat (Triticum aestivum L.) is one of the world's major crops. Water deficit stress is its main environmental limiting factor and severely affects global wheat production (Anjum et al., 2011; Rivero et al., 2007). In recent years, with the intensification of global climate change, precipitation in the winter wheat area of northern China has decreased. Drought stress occurs frequently during the grain‐filling stage of wheat, which accelerates the leaf senescence process in wheat cultivars. As a result, the grains remain incompletely filled, which ultimately leads to agricultural yield losses worldwide (Abid et al., 2017; Gregersen, Culetic, Boschian, & Krupinska, 2013; Yang & Zhang, 2006).

Previous studies indicate that the effect of water deficit stress on the growth and development of crops is a complicated process. Correspondingly, plants have a series of mechanisms to mitigate the adverse effects of water deficit stress. The response of crops to water deficit stress has morphological, physiological, and biochemical characteristics as well as molecular mechanisms including changes in growth and development, morphological adaptations, photosynthesis, antioxidant enzyme activity, osmotic adjustment substances, endogenous hormones, and drought‐induced genes (Chaves, Flexas, & Pinheiro, 2009; Dasgupta, Das, & Sen, 2015; Zhu, 2016).

Researchers have discovered that water deficit stress is unfavorable to the growth and development of crops owing to decreased cell division and cell elongation (Lawlor & Cornic, 2002), which ultimately reduces the biomass and harvest index of crops (Kumar, Nayak, Pani, & Das, 2017). Moreover, the balance between the production and scavenging of ROS (particularly O₂ ⁻ and H₂O₂ in chloroplasts, mitochondria, and peroxisomes) in crop cells is disturbed by drought stress, which causes membrane injuries, protein degradation, and enzyme inactivation and thus induces oxidative stress (Zlatev & Lidon, 2012). Antioxidant enzymes such as SOD, POD, CAT, and APX form a defense system and play critical roles in removing the excessive ROS induced by drought stress (Ahanger & Agarwal, 2017).

Drought stress accelerates leaf senescence, while delayed leaf senescence induces extreme drought tolerance in flowering plants (Rivero et al., 2007). This complicated process evidently coordinates the onset, process, and completion phases of leaf senescence which are closely related to an extensive reprogramming of gene expression (Jan, Abbas, Ashraf, & Ahmad, 2018; Koyama, 2014). Flag leaves are major photosynthetic organs during the grain‐filling period in wheat, making a major (41%–43%) contribution to crop yield (Blake, Lanning, Martin, Sherman, & Talbert, 2007; Sharma, Sain, & Sharma, 2003). Therefore, a comprehensive understanding of the physiological and molecular mechanisms regulating the flag leaf senescence response to water deficit stress is required in order to maintain agricultural productivity under drought stress and facilitate the development of new drought‐resistant varieties.

Compared with Arabidopsis and rice, lack of a complete reference genome has limited the progress of wheat sequencing. Transcriptome analysis in model plants has provided some molecular insights into the complex signaling pathways and regulatory networks involved in drought resistance. For instance, Urano et al. (2017) has reported that plant hormones such as ABA, jasmonic acid isoleucine (JA‐Ile), SA, cytokinin (trans‐zeatin, tZ), auxin (indole‐acetic acid, IAA), gibberellin (GA₄), and their related genes play important roles in metabolism and signaling during moderate water deficit in Arabidopsis. Based on candidate gene association analysis, further research has indicated that OsLG3 plays a positive role in water deficit tolerance in rice by ROS (Xiong et al., 2018). RNA sequencing analysis of Arabidopsis has also revealed that transcription factors WRKY46, WRKY54, and WRKY70 play important roles in promoting BR‐mediated gene expression and inhibiting drought‐responsive genes (Chen et al., 2017). Additionally, the flowering repressor SHORT VEGETATIVE PHASE (SVP) is induced by drought stress and is associated with promoter regions of ABA catabolic pathway genes CYP707A1/3 and AtBG1. The regulatory pathway involving SVP CYP707A1/3 and AtBG1 plays a critical role in plant drought resistance (Wang, Wang, et al., 2018). Other researchers have found that overexpression of DWA1 significantly increases cuticular wax deposition in rice, thereby improving the drought resistance of crop varieties (Zhu & Xiong, 2013). Recently, researchers discovered the novel molecular mechanisms of drought resistance in rice. Overexpressed OsTF1L enhances drought tolerance by regulating drought‐inducible, stomatal movement, and lignin biosynthetic genes (poxN/PRX38, Nodulin protein, DHHC4, CASPL5B1, and AAA‐type ATPase) (Bang et al., 2019). Studies also show that overexpression of PtoMYB170 increases lignification and thickens secondary walls in xylem and enhances the drought resistance of Arabidopsis (Xu et al., 2017). Researchers found a significant upregulated NAC transcription factor under drought stress via transcriptome analysis. This transcription factor is translocated into the nucleus, balances oxidation–reduction by regulating genes involved in oxidation–reduction, and finally enhances drought resistance (Duan et al., 2017). Shi et al. (2017) also found that the ARGOS8 variants modified by CRISPR‐Cas9 could enhance the grain yield of maize under drought stress. Moreover, the deletion of allele‐338, which includes the deletion of an endoplasmic reticulum stress response element in the 5′‐UTR region of ZmPP2C‐A10, improves the drought tolerance of maize (Xiang, Sun, Gao, Qin, & Dai, 2017).

The above studies indicate that many genes play critical roles in the drought resistance of plants such as Arabidopsis and rice. Here, we performed a combined analysis of steady‐state transcript abundance changes and physiological and biochemical indicator changes in the flag leaves of two wheat varieties (Fu287 and Shannong20) under water deficit stress at the grain‐filling stage. A comprehensive view of the physiological and molecular mechanisms regulating the leaf senescence response to drought paves the way for manipulating wheat by seeking suitable cultivation and chemical control measures to enhance its drought resistance.

2. MATERIALS AND METHODS

2.1. Plant materials and culture conditions

The experiments were conducted at the experimental station of Shandong Agricultural University Farm, Tai'an, China (36°9′N, 117°9′E; altitude 128 m) during two wheat‐growing seasons (from October 2016 to June 2017 and from October 2017 to June 2018). Two wheat varieties, Fu287 (F287, a drought‐sensitive genotype) and Shannong20 (SN20, a drought‐resistant genotype), were grown in 300 cm × 250 cm × 200 cm cement pools (length, width, and height, respectively). The pools were separated from each other by a cement ridge (30 cm wide). All the pools were filled with sandy loam, and the 0‐ to 20‐cm soil layer contained 12.41 g/kg total organic matter, 0.95 g/kg total nitrogen (N), 80.23 mg/kg available N, 20.23 mg/kg available P₂O₅, and 100.20 mg/kg available K₂O. The sowing dates were October 11, 2016, and October 10, 2017. The planting density was adjusted to 270 plants/m² at the three‐leaf stage (GS13) (Zadoks, Chang, & Konzak, 1974). Initially, 120 kg/ha N, 75 kg/ha P₂O₅, and 100 kg/ha K₂O were applied as the basal fertilizer before planting. In addition, 120 kg/ha N was applied at the jointing stage (GS31) as topdressing. Pests, diseases, and weeds were controlled by appropriate chemical applications during the growing period.

2.2. Experimental design and soil drought treatments

The experiment had a 2 × 4 (two wheat genotypes and four levels of soil moisture treatment) factorial design. Each treatment had three plots as repetitions in a randomized complete block design, and a total of 24 plots were used in the experiment. The soil moisture treatments were set according to previous studies (Valluru, Davies, Reynolds, & Dodd, 2016; Zhang, Chen, Wang, & Yang, 2017): well‐watered (WW), where the target relative soil moisture content in the 0‐ to 30‐cm soil layer was 75%–80% of field capacity (FC); mild water deficit (MiWD), where the target relative soil moisture content in the 0‐ to 30‐cm soil layer was 65%–75% of FC; moderate water deficit (MoWD), where the target relative soil moisture content in the 0‐ to 30‐cm soil layer was 50%–60% of FC; and severe water deficit (SWD), where the target relative soil moisture content in the 0‐ to 30‐cm soil layer was 30%–40% of FC. All plants were well watered before anthesis. Water stress treatments were initiated at anthesis (GS65), and plots were sheltered from rain using a rain protection shed. ECH2O soil moisture monitoring systems (Decagon, USA) equipped with EM50 data collectors and EC‐5 sensors were used to monitor soil volumetric water content at a soil depth of 25 to 30 cm accurately and in real time by inserting a 5‐cm‐long sensor. ECH2O readings were recorded every minute. When readings dropped to the desired value, the plots were irrigated with a water meter. The amount of supplemental irrigation (SI) was calculated according to the following equations (Wang, Shi, Guo, Zhang, & Yu, 2015):

$$\text{SI}=10\times {\gamma }_{\mathrm{bd}}\times D_{\mathrm{h}}\times ({\theta }_{\mathrm{t}}-{\theta }_{\mathrm{n}})$$

where SI (mm) is the amount of supplemental irrigation; γ _bd (g/cm³) is the soil bulk density; D _h (cm) is the depth of the soil layer (it is 30 cm in the present paper); θ _t (%) is the target soil water content (SWC) on a weight basis after SI; θ _n (%) is the soil water content (SWC) on a weight basis before SI. The value for θ _t was calculated as follows:

$${\theta }_{\mathrm{t}}={\theta }_{max}\times {\theta }_{\mathrm{tr}}$$

where θ _max (%) is the field capacity (FC), and θ _tr (%) is the target relative SWC for each tested soil as the experimental design.

Soil samples were collected from the 0‐ to 30‐cm soil layers several days before anthesis using 200 cm³ rings. The 0‐ to 30‐cm soil layer was divided into three layers (0–10 cm, 10–20 cm, and 20–30 cm) with three repetitions per plot. Soil water content (SWC), volumetric water content (VWC), soil relative water content (SRWC), and the bulk density (γ _bd) were calculated using the following equations (Guo, Yu, Wang, Shi, & Zhang, 2014):

$$\text{SWC}=\frac{\mathrm{FW}-\mathrm{DW}}{\mathrm{DW}}\times 100$$

$$\text{FC}=\frac{\mathrm{SSW}-\mathrm{DW}}{\mathrm{DW}}\times 100$$

$$\text{SRWC}=\frac{\mathrm{SWC}}{\mathrm{FC}}\times 100$$

$${\gamma }_{\mathrm{bd}}=\frac{\mathrm{DW}}{\mathrm{V}}$$

$$\text{SWC}=\frac{\mathrm{VWC}}{{\gamma }_{\mathrm{bd}}}$$

$$\text{VWC}=\frac{\mathrm{SRWC}\times \mathrm{FC}\times {\gamma }_{\mathrm{bd}}}{100}$$

where FW (g) is the soil sample fresh weight, DW (g) is the soil sample dry weight, FC (%) is the field capacity, SSW (g) is the saturated soil weight when the soil moisture reached 100% FC, and V (cm³) is the volume of soil in rings (200 cm³ in the present paper).

2.3. RNA isolation, cDNA library construction, Illumina sequencing, read mapping, and differential gene expression analyses

Flag leaves of wheat varieties were collected at 5 (early grain‐filling stages), 15 (mid‐grain‐filling stages), and 23 days (late grain‐filling stages) after anthesis from October 2016 to June 2017, a total of 24 samples (two varieties, four levels of soil water deficit, and three stages) in three biological replicates for both varieties. At least 20–30 leaves were tagged for each biological replicate. The tissue samples were collected, plunged directly into liquid nitrogen, and then stored at −80°C until analysis.

Total RNA extraction was performed using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. The quality of the total extracted RNA was tested using the RNA Nano 6000 Assay Kit of the Bioanalyzer 2100 system (Agilent Technologies). All of the 72 libraries (24 samples in three biological replicates) were sequenced using the NEBNext^® UltraTM RNA Library Prep Kit for Illumina^® (NEB) following the manufacturer's recommendations to generate 150‐nucleotide‐long paired‐end sequence reads. The raw reads were evaluated for various quality parameters, and high‐quality reads were filtered using the NGS QC Toolkit (v2.3). Clean reads were obtained by removing low‐quality reads as well as those containing adapter and poly‐N from raw reads. The following analysis was based on clean data with high quality. At the same time, the Q20, Q30, and GC content of the clean data were calculated. All the downstream analyses were based on the clean data with high quality. The filtered high‐quality reads were mapped onto the wheat genome (IWGSC, v1.0). An index of the reference genome was built using Hisat2 v2.0.5, and paired‐end clean reads were aligned to the reference genome using Hisat2 v2.0.5. The read numbers mapped to each gene were counted using featureCounts v1.5.0‐p3. The FPKM of each gene was then calculated based on the length of the gene.

Differential expression analysis was performed using the DESeq2 R package (1.16.1). The resulting p‐values were adjusted using Benjamini and Hochberg's approach to control the false discovery rate. Genes with an adjusted p < .05 found by DESeq2 were assigned as differentially expressed. A Gene Ontology (GO) enrichment analysis of differentially expressed genes was implemented using the clusterProfiler R package, where gene length bias was corrected. GO terms with a corrected p < .05 were considered significantly enriched by differential expressed genes. We used the clusterProfiler R package to test the statistical enrichment of differentially expressed genes in KEGG pathways.

2.4. Co‐expression network analysis for construction of modules

The weighted correlation network analysis (WGCNA) (Langfelder & Horvath, 2008) package was used for co‐expression network analysis. To determine the association of modules with water deficit and stage‐specific expression for the two wheat varieties, we determined the correlation between each module eigengene (ME) with the binary indicator as described (Downs et al., 2013). A positive correlation indicates that genes in a module have a higher/preferential expression in a particular stage relative to all other samples. Further, we followed a cross‐tabulation approach to create a contingency table that reports the number of genes that fall into modules of F287 versus modules of SN20.

2.5. Extraction and determination of endogenous abscisic acid (ABA), salicylic acid (SA), jasmonic acid (JA), zeatin (ZT), and zeatin riboside (ZR) contents

Flag leaves were sampled at 5, 10, 15, 20, 25, and 30 days after anthesis (DAA), frozen in liquid nitrogen, and then stored at −80°C until extraction.

Abscisic acid, SA, and JA were extracted according to a previously published method (Engelberth et al., 2003) with some modifications. Three replicates of each frozen leaf sample (80~100 mg for each replicate) were grinding in liquid nitrogen. 1.8 ml extraction solvent consisting of acetone/50 mM citric acid (70:30, vol/vol) was added, and then, internal standards (20 ng) including DHJA, D6‐SA, and D6‐ABA were added in the centrifuge tube, followed by shaking for 3 hr at 4°C on a constant temperature oscillation box. Open the cover of the centrifuge tube, place it in a fume hood, and left it overnight for about 14 hr for the acetone to evaporate. The remaining aqueous phase was extracted with diethyl ether (2 × 700 μl) by vortexing and phase separation. The combined supernatant was evaporated in vacuum at room temperature and re‐solubilized in MeOH (100 μl) and filtered through a 0.22 μm membrane filter. Then, 5 μl sample solution was injected for analysis of SA, JA, and ABA by using Acquity UPLCۛ system coupled to a Waters XEVO TQ‐S triple‐quadrupole mass spectrometer with an ESI source (Waters).

Extraction and determination of zeatin (Z) and zeatin riboside (ZR) were performed as described by Liu, Li, Xiao, and Wang (2012). Leaf samples (80–100 mg) in three replicates were weighed and ground to a fine powder with liquid nitrogen. 750 μl precooling extraction solvent (methanol:water:acetic acid, 80:19:1, v/v/v) supplemented with internal standards (20 ng, D5‐ZR and D5‐ZT) was added in the samples and then shaken on a shaking bed for 16 hr at 4°C in dark. The supernatant was carefully transferred to a new centrifuge tube, and the sample residue was remixed with 400 μl extraction buffer, shaken for 4 hr at 4°C. The combined supernatants were dried by the vacuum centrifugal concentrator and then redissolved in 100 μl methanol. The samples were filtered through a 0.22 μm membrane filter prior to LC‐MS/MS analysis.

Gibberellin (GA₃) was extracted according to Dave et al. (2011), with modifications. Leaves were ground and extracted overnight at 4°C with 1.5 ml solvent (isopropanol:acetic acid, 99:1). Then, the centrifugal supernatant was removed to a new tube. The residues were re‐extracted with 0.5 ml extraction solvent for 2 hr. The supernatants were dried and then redissolved with 100 μl methanol. The analysis of GA₃ was also performed by UPLC‐MS/MS.

2.6. Quantitative PCR analysis

Total RNA was extracted using TRIzol reagent (Invitrogen, CA, USA) and treated with DNase I to remove any contaminant genomic DNA. Then, 2 μg of the total RNA was reverse‐transcribed in a 20 μl reaction using All‐in‐One First‐Strand cDNA Synthesis SuperMix with oligo‐dT primers for qPCR (TRANS). Quantitative real‐time PCR was conducted using the SYBR Green Master Mix kit (for the genes Cu/Zn‐SOD, Fe‐SOD, TaSAG5, TaSAG7, and TaSAG9) and the TaqMan probe PCR kit (for the genes TaSAG1, TaSAG2, TaSAG3, TaSAG4, TaSAG6, TaSAG8, Mn‐SOD, CAT, and APX) on an Applied Biosystems 7500 real‐time PCR System (Applied Biosystems). The expression of the TaACTIN gene was used as an endogenous control to normalize the gene expression. The qRT‐PCR protocol was performed following the manufacturer's instructions. The annealing temperature of all the genes was determined by specificity. The relative expression levels were calculated with the equation 2^−∆∆Ct. The primers used in the present study are listed in Table S2.

2.7. Flag leaf senescence process model

The size (leaf length, width, and thickness) and shape of flag leaves in two wheat genotypes are similar. The leaf length, width, and thickness at maturity in F287 were about 23.34 cm, 1.52 cm, and 0.26 mm, while the length, width, and thickness in SN20 were about 22.81 cm, 1.46 cm, and 0.29 mm, respectively. The SPAD values of flag leaves for all treatments were determined nondestructively with a handheld chlorophyll meter (SPAD 502, Minolta Camera Co). For each plot, fifteen healthy flag leaves per treatment were tagged at anthesis (GS61) (Zadoks et al., 1974) in 2016–2017 and 2017–2018. Measurements of SPAD values were taken on the middle portion of flag leaves (avoiding the midrib and major veins), which were performed at anthesis, 7, 14, 21, and 28 DAA in 15 repetitions. The Gompertz growth equation (Equation ) and leaf senescence‐related parameters were used to describe leaf senescence process according to previous research (Xie, Mayes, & Sparkes, 2016).

$$\mathrm{G}=\mathrm{a}{\mathrm{e}}^{-{\mathrm{be}}^{(+\mathrm{rt})}}$$ (1)

where G is the chlorophyll SPAD value at accumulated temperature t; t is the accumulated temperature after anthesis; a is the maximum SPAD value predicted by the model during the period of measurement; b is an adjustable parameter that indirectly reflects the early senescence rate of the flag leaf. The initial senescence rate $r_0=b$; the maximum senescence rate $r_{\max }=\frac{a\times r}{e}$; the average senescence rate $r_{\mathrm{aver}}=\frac{r_0+r_{max}}{2}$; inflection point cumulative temperature (M) is the cumulative thermal time when senescence rate is the maximum, $M=\frac{\ln b}{r}$; duration of the flag leaf being photosynthetically active (Chl_total is defined as the period from anthesis to 90% senescence, ${\text{Chl}}_{\mathrm{total}}=\frac{\ln [\ln (0.1A/a)/b]}{r}$, A is the SPAD value at anthesis; Chl_total consists of two components: persistence phase (Chlper), from anthesis to t_onset (the onset of senescence, 10% senescence), ${\text{Chl}}_{\mathrm{per}}=\frac{\ln \left[\ln (0.9A/a)/(-b)\right]}{r}$, and rapid loss phase Ch_loss, ${\text{Chl}}_{\mathrm{loss}}={\text{Chl}}_{\mathrm{total}}-{\text{Chl}}_{\mathrm{per}}$.

2.8. Ultramicrotomy of leaf and structure observations

At 20 DAA, three representative flag leaves were selected from each treatment. Square leaves (4 mm × 4 mm) were taken near the center vein on the middle portion of each flag leaf and then immediately fixed with 3.5% glutaraldehyde. A 0.2 mol/L phosphate buffer (pH 7.0) was used to wash the leaves. The samples were post‐fixed with 1% osmic acid and then dehydrated with gradient concentrations of ethanol. Dehydrated samples were infiltrated, embedded, and polymerized with Epon812 for 2 hr. The samples were sliced into ultrathin sections with an LKB‐V ultramicrotome (LKB Company, Bromma, Sweden) and counterstained with uranyl acetate and Pb citrate. The sections were observed and photographed with a JEM‐1400Plus transmission electron microscope.

2.9. Kernel mass

At maturity, about 30 uniform spikes of every treatment were sampled, and 10 spikes were sampled as a repetition. The grains on the spikes were divided into superior grains and inferior grains. The first and second basal grains on the spikelets were defined as superior grains, whereas the third and fourth grains on the same spikelets were defined as inferior grains. Then, the grains were dried at 70°C to a constant weight. The 1,000‐grain weight of superior grains and inferior grains was calculated.

2.10. Statistical analysis

Analysis of variance was performed with DPS 7.05 software (Zhejiang University, Hangzhou, China). Means were tested by least significant difference at p < .05 (LSD0.05). DPS was used to fit the Gompertz growth equation. Graphs and Gompertz growth curves were drawn with SigmaPlot 10.0 (Systat Software, Inc., San Jose, CA, USA). Regulatory network graphs were plotted by Gephi software (v0.8.2).

2.11. Accession numbers

The RNA‐Seq raw reads have been deposited in the Sequence Read Archive of the National Center for Biotechnology Information (accession number SUB5706797; https://submit.ncbi.nlm.nih.gov/subs/sra/SUB5706797/overview ).

3. RESULTS

3.1. Kernel mass

Figure 1 indicates that the weight of superior, inferior, and total grains in the drought‐resistant wheat SN20 were all higher than those of the drought‐sensitive wheat F287. The data in 2017–2018 showed that compared with that under WW treatment, the superior grain weight of F287 under MiWD, MoWD, and SWD treatments was decreased by 5.68%, 8.07%, and 13.86%, respectively, while the inferior grain weight of F287 was decreased by 5.84%, 11.88%, and 17.79%, respectively. The total grain weight of F287 was reduced by 4.00%, 12.45%, and 21.69%, respectively. Moreover, the superior grain weight of SN20 under MiWD, MoWD, and SWD treatments was decreased by 5.97%, 7.08%, and 13.74%, respectively, and the inferior grain weight of SN20 was decreased by 5.12%, 9.61%, and 17.10%, respectively. The total grain weight of SN20 was decreased by 2.34%, 12.19%, and 17.59%, respectively. The above results indicate that the weights of the total grains and the inferior grains in F287 were much more sensitive to water deficit than those in SN20. Furthermore, the weight of the inferior grains was decreased more than that of the superior grains in the two wheat genotypes under water deficit stress.

Figure 1.

Thousand‐grain weight of superior grains, inferior grains, and total grains in two wheat varieties under different water deficit treatments in 2016–2017 and 2017–2018. S represents superior grains, I represents inferior grains, and Grain represents total grains. F287, Fu287; SN20, Shannong20. WW, MiWD, MoWD, and SWD represent the treatments of well‐watered, mild water deficit, moderate water deficit, and severe water deficit. (a) and (c) denote the superior grain weight, inferior grain weight, and total grain weight of F287 in 2016–2017 and 2017–2018, respectively. (b) and (d) denote the superior, inferior grain weight, and total grain weight of SN20 in 2016–2017 and 2017–2018, respectively

3.2. Ultrastructural features of wheat flag leaves under different water deficit treatments

Figure 2 indicates that the chloroplasts adhered closely to the cell wall, and mitochondria were arranged among the chloroplasts. With increasing water stress severity, the shape and internal structure of chloroplasts in mesophyll cells underwent significant changes. The shape of chloroplasts became rounder. The osmiophilic globules accumulated more in the chloroplasts, and their color was darker. Moreover, the granal lamellae became loose and disordered under water deficit treatments, and the gaps between granal lamellae were increased. Another obvious change was in the membrane system of the chloroplasts. The damage to the chloroplast membrane system was increased instead of being repaired under water deficit conditions. As shown in Figure 2 (d,h), the outer membranes of the chloroplasts of F287 flag leaves were completely destroyed under SWD treatment. The chloroplast structures of F287 were also injured much more seriously than those of SN20.

Figure 2.

Mesophyll cell and chloroplast ultrastructure changes in flag leaves of two wheat varieties at 20 days after anthesis under different water deficit treatments. (a‐d) represent the ultrastructure (×50 K) of mesophyll cells for F287 of well‐watered (WW), mild water deficit (MiWD), moderate water deficit (MoWD), and severe water deficit (SWD) treatments, respectively. (e‐h) represent the ultrastructure (×50 K) of mesophyll cells for SN20 under WW, MiWD, MoWD, and SWD treatments, respectively. Chl, chloroplast; OG, osmium granules; CW, cell wall; GL, grana lamella; Mi, mitochondria; CM, chloroplast membrane

3.3. Effects of water deficit treatments on the flag leaf senescence process

The trend of variation in SPAD values was used to describe the senescence process. The SPAD values in F287 were higher than those in SN20 (Figure 3). The SPAD values of flag leaves were also decreased with increasing water deficit severity. From 2016 to 2017, compared with WW treatments, the SPAD values of F287 were decreased by 6.10%, 11.09%, and 16.94% (average values at different stages) under MiWD, MoWD, and SWD treatments, respectively, while the SPAD values of SN20 were reduced by 4.87%, 10.27%, and 15.41%, respectively. From 2017 to 2018, the SPAD values of F287 dropped by 6.03%, 11.62%, and 17.65% under MiWD, MoWD, and SWD conditions, while those of SN20 dropped by 5.49%, 10.65%, and 16.54%, respectively. The above results indicate that the extent of reduction of SPAD values in F287 was higher than that in SN20 under water deficit stress, implying that F287 was more sensitive to water deficit as compared with SN20. Therefore, F287 was identified as a drought‐sensitive wheat variety and SN20 as a drought‐tolerant wheat variety.

Figure 3.

SPAD values of flag leaves in two wheat varieties under different water deficit treatments. F287, Fu287; SN20, Shannong20. WW, MiWD, MoWD, and SWD represent the treatments of well‐watered, mild water deficit, moderate water deficit, and severe water deficit. (a) and (b) denote the SPAD values of F287 and SN20 in 2016–2017, respectively. (c) and (d) denote the SPAD values of F287 and SN20 in 2017–2018, respectively

Figure 3 also shows that the SPAD values declined as the growth stages moved forward, and they decreased slowly at the early stage, whereas they declined rapidly in the later stage. Therefore, the duration of flag leaf being photosynthetically active was divided into two phases by the Gompertz model: persistence and rapid loss. Additionally, the SPAD values response to water deficit treatments varied at different stages after anthesis. At the early grain‐filling stage, the extent of reduction of SPAD values in two wheat varieties was relatively low, whereas the extent of reduction in SPAD values was high at the late grain‐filling stage and reached a maximum at 25 DAA. The above results show that leaves at the rapid senescence stage were more sensitive to water deficit. The results for the two wheat varieties in two growing seasons were consistent.

The flag leaf senescence process was assessed using a Gompertz curve by estimating the chlorophyll content of two wheat varieties under different water deficit treatments. Larger r ₀ and r _max mean that flag leaf senescence process was accelerated, M is the cumulative thermal time when senescence rate is the maximum, Chl_total is the duration of the flag leaf being photosynthetically active, and larger Chlper means that flag leaf could maintain longer highly active photosynthetic duration. The data showed that compared with F287, SN20 had longer Chl_total and larger r _aver and M (Table 1). The total photosynthetic functional duration of SN20 was longer than that of F287, and fast leaf senescence of SN20 was initiated later. Moreover, the data also showed that r ₀ and r _aver were enhanced whereas M, Chl_per, and Chl_total were decreased with increasing water stress. It means that the senescence of flag leaves was accelerated and duration of flag leaf being photosynthetically active could not be maintained longer as degree of soil water deficit stress aggravated. The above results suggest that water deficit treatments could initiate rapid leaf senescence prematurely, accelerate the senescence process, and eventually lead to premature leaf senescence and shorten the active photosynthetic duration.

Table 1.

The flag leaf senescence model and related parameters of two wheat varieties under different water deficit treatments

Variety	Treatment	r 0	r max	r aver	Chltotal	Chlper	M
F287	WW	.0474	.0945	.0710	849.53	244.10	664.16
	MiWD	.0557	.0968	.0763	785.21	213.51	605.81
	MoWD	.0582	.0934	.0758	784.48	200.25	603.49
	SWD	.0774	.0893	.0834	748.45	186.37	559.06
SN20	WW	.0516	.0937	.0727	905.73	261.96	706.82
	MiWD	.0665	.0943	.0804	838.47	201.07	641.21
	MoWD	.0767	.0890	.0829	820.05	180.85	618.98
	SWD	.0835	.0852	.0844	812.06	116.90	607.84

Abbreviations: Chl_per, duration of chlorophyll persistence, from anthesis to the onset of senescence; Chl_total, duration of the flag leaf being photosynthetically active, the period from anthesis to the time of 90% senescence; M, the cumulative thermal time when senescence rate is at its maximum; r ₀, the initial senescence rate; r _aver, the average senescence rate; r _max, the maximum senescence rate.

3.4. Analysis of the contents of endogenous hormones under water deficit treatments

Figure 4(a) shows the changes in a variety of phytohormones under different water treatments. With increasing of water stress severity, the contents of SA, JA, and ABA and the ratio of GA₃ to ZR were enhanced. Compared with WW treatments, SA content was enhanced by 27.42%, 72.57%, and 144.86% under MiWD, MoWD, and SWD treatment, respectively. Moreover, JA content was increased by 73.26%, 146.11%, and 303.01%, while ABA content was enhanced by 34.51%, 79.32% and 104.71%, respectively. The results also show that the contents of endogenous Z, ZR, and GA₃ and the ratios of SA to JA and ABA to JA were decreased under water deficit stress. Compared with WW treatment, Z content was decreased by 14.40%, 36.08%, and 53.64% and ZR content was reduced by 46.26%, 71.16%, and 83.89%, respectively. In addition, the content of GA₃ was also reduced by 25.00%, 45.92%, and 67.01%, respectively. The above results indicate that JA was the most sensitive to drought stress, and Z was the least sensitive.

Figure 4.

(a) Experimental endogenous hormones under well‐watered treatment (WW, n = 36), mild water deficit (MiWD, n = 36), moderate water deficit (MoWD, n = 36), and severe water deficit (SWD, n = 36) in two wheat genotypes. (a) The content of salicylic acid (SA); (b) the content of jasmonic acid (JA); (c) the content of abscisic acid (ABA); (d) the content of zeatin (Z); (e) the content of zeatin riboside (ZR); (f) the content of gibberellin (GA ₃); (g) the ratio of SA to JA (SA/JA); (h) the ratio of ABA to JA (ABA/JA); (i) the ratio of GA ₃ to ZR (GA ₃/ZR). Solid and dashed lines indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, whisker caps indicate 90th and 10th percentiles, and circles indicate outliers. The data were obtained from two wheat genotypes. (b) Experimental endogenous hormones at 5 days after anthesis (5 DAA, n = 24), 10 days after anthesis (10 DAA, n = 24), 15 days after anthesis (15 DAA, n = 24), 20 days after anthesis (20 DAA, n = 24), 25 days after anthesis (25 DAA, n = 24), and 30 days after anthesis (30 DAA, n = 24) in two wheat genotypes. (a) The content of salicylic acid (SA); (b) the content of jasmonic acid (JA); (c) the content of abscisic acid (ABA); (d) the content of zeatin (Z); (e) the content of zeatin riboside (ZR); (f) the content of gibberellin (GA ₃). Solid and dashed lines indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, whisker caps indicate 90th and 10th percentiles, and circles indicate outliers. The data were obtained from two wheat genotypes. (c) Endogenous hormones of flag leaves in two wheat varieties with different drought resistance. (a) The content of salicylic acid (SA); (b) the content of jasmonic acid (JA); (c) the content of abscisic acid (ABA); (d) the content of zeatin (Z); (e) the content of zeatin riboside (ZR); (f) the content of gibberellin (GA ₃). Each box plot contains 72 data. Solid and dashed lines in box plot indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, and circles indicate outliers

Analyses of the changes in endogenous phytohormones in different periods after anthesis were performed. The endogenous hormones content first increased and then gradually decreased with the growth process (Figure 4(b)). The contents of all the endogenous hormones were lowest at 30 DAA. The contents of SA, JA, ABA, and Z were higher at 15 to 20 DAA, while the contents of ZR and GA₃ were higher at 10 to 15 DAA. The above results suggest that the responses of different hormones to the senescence process of wheat leaves differed.

As shown in Figure 4(c), the contents of SA, JA, Z, ZR, and GA₃ in the drought‐resistant genotype SN20 were higher than those in the drought‐sensitive wheat genotype F287, while the content of ABA in SN20 was lower. The contents of endogenous hormones differed in wheat varieties with varying drought‐resistant responses to water deficit.

The results in Table 2 show that the grain mass of the two wheat varieties was significantly negatively related to r ₀ and r _aver (r = −.72*, −.68*); however, it was significantly positively related to r _max, Chl_total, Chl_per, and M (r = .68*, .80**, .72*, .86**). Furthermore, the superior grain weight was significantly positively correlated with r _max, Chl_total, Chl_per, and M (r = .75*, .76*, .84** .85**), whereas it was markedly negatively related to r ₀ and r _aver (r = −.83**, −.81**). The inferior grain weight was significantly negatively related to r ₀ (r = −.68*); however, it was significantly positively related to Chl_total, Chl_per, and M (r = .87**, .72*, .91**). The above results imply that the flag leaf senescence process is closely associated with the weight of the superior, inferior, and total grains. In addition, the concentrations of SA, JA, and ABA had clearly positive correlations with r ₀ and r _aver, but significantly negative associations with r _max, Chl_total, Chl_per, and M. By contrast, the concentrations of Z, ZR, and GA₃ showed the opposite relationship with the senescence parameters, indicating that different endogenous hormones play diverse roles in leaf senescence. SA, JA, and ABA were not conducive to the enhancement of grain mass, whereas Z, ZR, and GA₃ were beneficial to the enhancement of grain mass.

Table 2.

Correlation coefficients of senescence parameters with S (superior grain weight), I (inferior grains), grain mass, and concentrations of SA, JA, ABA, ZR, Z, and GA₃ in two wheat varieties

3.5. The biosynthesis, transportation, and signaling of other phytohormones in the regulation of drought resistance

To further understand how the molecular mechanisms of phytohormones, such as auxin, cytokinin (CKs), ABA, ethylene (ETH), GA, JA, and SA, coordinate flag leaf senescence response to drought stress, differential genes from the transcriptome analysis were analyzed. Some of the genes were directly involved in the biosynthesis, transportation, and signaling of plant hormones. Based on the classified modules in Figure 9 and the GO enrichment analysis in Table S3, we obtained the genes participating in plant hormone signal transduction, shown in Figure 5. SAUR and AUXIAA, which encode auxin‐responsive protein, were negative regulators of auxin biosynthesis and transport (Kant et al., 2009; Teale, Paponov, & Palme, 2006). The results indicate that with increasing water stress, the expression levels of SAUR and AUXIAA were enhanced at 5 and 15 DAA, suggesting that the auxin signal transduction was sensitive to drought stress at the early grain‐filling stage. Additionally, the expression levels of SAUR and AUXIAA were low at 23 DAA. We speculated that auxin metabolism and signal transduction were not active when the leaf was entering the late aging period. Further, the expression levels of SAUR and AUXIAA were higher in SN20 than in F287. PYL, which encodes the abscisic acid receptor, was downregulated at 23 DAA, which was in accordance with the decreased endogenous ABA level. The expression of PYL in SN20 was higher than that in F287 under water deficit stress treatment, which was beneficial for promoting stomatal closure of guard cells to prevent water loss. The zeatin biosynthesis‐related gene CRE1 was downregulated with increasing water deficit stress at 15 and 23 DAA. In addition, CRE1 was first upregulated and then downregulated, indicating that the content of zeatin was highest at about 15 DAA. Furthermore, the transcript abundance of ERF1/2 (5DAA), ETR, EIN3 (5DAA), JAR1 (5DAA), PR‐1, GID1 (5DAA), and CRE1 was all higher in SN20 than in F287. Additionally, the expression of several other genes involved in the regulation of hormonal signaling of ethylene, GA, and JA was also altered in response to water deficit stress.

Figure 5.

Hierarchical clustering of the differential expressed genes related to plant hormone signal transduction. The genes involved in the biosynthesis or signal transduction of auxin, ABA, ethylene, CK, JA, SA, or GA were identified. The values of the heatmaps were FPKM values calculated by Log₂ ^FPKM+1

3.6. Analysis of the relative expression of ROS‐scavenging‐related genes under water deficit treatment

The expressions of antioxidant enzyme‐encoding genes were analyzed by RT‐qPCR. Cu/Zn‐SOD encodes the chloroplastic copper/zinc superoxide dismutase, Mn‐SOD encodes the mitochondrial manganese superoxide dismutase, Fe‐SOD (iron superoxide dismutase) encodes the chloroplastic iron superoxide dismutase, CAT encodes catalase, and APX encodes the H₂O₂‐scavenging enzyme ascorbate peroxidase. As shown in Figure 6(c), the expression levels of Cu/Zn‐SOD, Mn‐SOD, Fe‐SOD, POD, CAT, and APX in the drought‐resistant genotype SN20 were higher than those in F287 under drought stress conditions, implying that the scavenging ability of ROS in SN20 was greater than that in F287. The results also show that the expression of Cu/Zn‐SOD was downregulated with increasing water stress severity. By contrast, the expressions of the other genes (Mn‐SOD, Fe‐SOD, POD, CAT, and APX) were upregulated by the water deficit (Figure 6a). In addition, as the growth stages progressed, the expression of antioxidant enzyme‐encoding genes first increased and then decreased. The relative expression of Cu/Zn‐SOD, Mn‐SOD, and Fe‐SOD was high at 20 DAA to 30 DAA (Figure 6b). Additionally, the relative expression of POD was highest at 20 DAA, while that of CAT was high at 10 to 20 DAA and that of APX was high at 20 to 25 DAA, suggesting that at different stages of grain filling, the enzymes that play major roles in scavenging ROS in wheat flag leaves are different.

Figure 6.

(a) Experimental antioxidant competence under well‐watered treatment (WW, n = 36), mild water deficit (MiWD, n = 36), moderate water deficit (MoWD, n = 36), and severe water deficit (SWD, n = 36) in two wheat genotypes. (a) The relative expression of Cu/Zn‐SOD; (b) the relative expression of Mn‐SOD; (c) the relative expression of Fe‐SOD; (d) the relative expression of POD; (e) the relative expression of CAT; (f) the relative expression of APX. Solid and dashed lines indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, whisker caps indicate 90th and 10th percentiles, and circles indicate outliers. The data were obtained from two wheat genotypes. (b) Experimental antioxidant competence at 5 days after anthesis (5 DAA, n = 24), 10 days after anthesis (10 DAA, n = 24), 15 days after anthesis (15 DAA, n = 24), 20 days after anthesis (20 DAA, n = 24), 25 days after anthesis (25 DAA, n = 24), and 30 days after anthesis (30 DAA, n = 24) in two wheat genotypes. (a) The relative expression of Cu/Zn‐SOD; (b) the relative expression of Mn‐SOD; (c) the relative expression of Fe‐SOD; (d) the relative expression of POD; (e) the relative expression of CAT; (f) the relative expression of APX. Solid and dashed lines indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, whisker caps indicate 90th and 10th percentiles, and circles indicate outliers. The data was obtained from two wheat genotypes. (c) Expression of genes encoding antioxidant enzymes in wheat varieties with different drought resistance. (a) The relative expression of Cu/Zn‐SOD; (b) the relative expression of Mn‐SOD; (c) the relative expression of Fe‐SOD; (d) the relative expression of POD; (e) the relative expression of CAT; (f) the relative expression of APX. Each box plot contains 72 data. Solid and dashed lines in box plot indicate medians and means, respectively. Box boundaries indicate upper and lower quartiles, and circles indicate outliers

3.7. Global transcriptome analysis of flag leaves under different water deficit treatments in two wheat varieties

To explore the molecular mechanisms of flag leaf senescence influenced by stages and water deficit stress, RNA‐Seq analyses were conducted to generate global transcriptome profiles. We chose two wheat varieties, F287 (a drought‐sensitive genotype) and SN20 (a drought‐resistant genotype). A total of 72 libraries were constructed and analyzed (each sample had three biological replicates). Approximately 70–106 million total reads were generated for each sample (Table S1) and mapped to the wheat genome (IWGSC, v1.0) using Hisat2 v2.0.5. The mapped files were processed via featureCounts, which generated a total of 269,584 gene loci. The uniquely mapped reads for each replication totaled 62–94 million, and there were 62–99 million total mapped reads for each sample (Table S1). These were processed using feature Counts to determine the normalized expression level as fragments per kilobase of transcript length per million mapped reads (FPKM) of each transcript. There were 35–54 million high‐quality reads for each sample.

The number of genes identified in each sample is shown in Figure S1. The results indicate that 103,076, 104,961, 105,783, 106,538, 102,891, 109,208, 104,783, 106,471, 102,159, 101,397, 103,268, and 102,138 genes were detected in F287 under different water deficit treatments at 5, 15, and 23 DAA, respectively. Correspondingly, the numbers of genes expressed in SN20 were 99,157, 106,981, 102,599, 105,711, 104,801, 106,850, 104,834, 108,063, 103,487, 103,873, 97,513, and 102,333. The numbers of transcripts at different expression levels (distributed by FPKM values) are shown in Figure S1. The percentage of detected genes in the 0–0.1 FPKM range, 0.1–1 FPKM range, 1–10 FPKM range, 10–50 FPKM range, and >50 FPKM range were about 23%, 27%, 34%, 12%, and 4%, respectively (Figure S1).

We validated the gene expression patterns of 15 genes (including six antioxidative‐related genes and nine senescence‐associated genes) in the two wheat varieties via RT‐qPCR (Figure 7). Genes with high correlation (r > .70) between RNA‐Seq data and data validated by RT‐qPCR were analyzed. The expression patterns of antioxidative‐related genes are analyzed in Figure 6 in detail.

Figure 7.

Correlation between relative expression of the selected genes (Cu/Zn‐SOD, Mn‐SOD, and so on) obtained from RNA‐Seq and RT‐qPCR analysis in two varieties. Heatmaps represent expression profiles of selected genes (labeled on right side) obtained from RNA‐Seq (left) and RT‐qPCR (right) analysis. The color scale at the bottom represents Z‐score. The values of the left heatmaps were FPKM values calculated by Log₂ ^FPKM+1, and the values of the right heatmaps were the relative expression value (compared with FWW5). The values between the two heatmaps represent correlation value between the expression profiles obtained from RNA‐Seq and RT‐qPCR analysis for each gene. The correlation values above .60 are highlighted in bold

A large number of genes that are upregulated during senescence are senescence‐associated genes (SAGs). The expression of TaSAG3, TaSAG4, TaSAG5, TaSAG7, and TaSAG8 increased with the growth process. The expression levels of TaSAG3, TaSAG4, TaSAG5, TaSAG7, and TaSAG8 were enhanced with increasing of water deficit stress. In addition, the expression levels of TaSAG4, TaSAG5, and TaSAG8 in SN20 were lower than those in F287. By contrast, the expression of TaSAG3 at 5 DAA in SN20 was lower than that in F287, while it was higher than that in F287 at 15 and 23 DAA. The expression of TaSAG6 in SN20 was higher than that in F287 at 5 and 15 DAA, whereas it was lower than that in SN20 at 23 DAA. The expression of TaSAG7 in SN20 was higher than that in F287, while it was lower at 15 and 23 DAA. The above results also indicated that TaSAG4, TaSAG5, and TaSAG8 could be used as molecular markers to evaluate flag leaf senescence in wheat.

3.8. Construction of gene co‐expression networks

To obtain a comprehensive understanding of the gene regulatory network and identify specific gene responses to water deficit during the grain‐filling stages of wheat, we performed WGCNA (Langfelder & Horvath, 2008). Genes with low expression (FPKM < 1) were filtered out, and 65,535 genes were considered for the downstream analysis. Previous research has indicated that co‐expression networks are established based on the pairwise correlations of gene expression across all samples. Modules are defined as clusters of highly interconnected genes. In the present study, a total of 33 distinct modules (labeled with different colors) were identified for the two wheat varieties (Figure 8).

Figure 8.

WGCNA of genes in two wheat varieties under different treatments. Hierarchical cluster tree showing co‐expression modules identified by WGCNA. Each “leaf” represents individual gene. The major tree branches constitute 33 modules, labeled with different colors

We evaluated the correlation between the modules and our traits of interest (the traits related to drought resistance, such as plant hormones, antioxidant enzymes, and osmotic adjustment substances). For example, the pink 4 module with 1416 genes was highly related to soluble sugar, sucrose synthetase, sucrose phosphate synthase, and proline. The correlation coefficients and p‐values are shown in Figure 9. Based on the correlation coefficients, we chose 14 modules that had a high relationship with plant hormones, antioxidant enzymes, or osmotic adjustment substances. A functional analysis of the 14 modules is shown in Table S3 (the statistical significance of enrichment analysis was p < .05). Subsequently, we needed to further associate the functions and gene modules so as to screen out the key gene modules. Heatmaps characterizing the expression level of co‐expression genes of the above 14 modules were plotted to screen gene modules once more. After filtering out the modules, seven modules were selected as the key gene modules for the subsequent analysis as follows.

Figure 9.

Module‐trait associations. Each row corresponds to a module eigengene, column to a trait (physiological and biochemical traits). Each frame contains the corresponding correlation and p‐value. The table is color‐coded by correlation according to the color legend

The lightpink 2 module‐specific genes (261) and thistle 4 module‐specific genes (267) were overrepresented in the drought‐resistant wheat variety SN20 during the grain‐filling stage (Figure 10). The blue 2 module‐specific genes (236) were highly expressed in the drought‐sensitive variety F287 across all the grain‐filling stages. Genes in the skyblue 4 module were upregulated in SN20 at 5 and 15 DAA, but they were highly expressed in F287 at 23 DAA. The expression patterns of genes in the plum 3 module were contrary to those in the skyblue 4 module. The expression levels of genes in the lightslateblue module were enhanced as the growth stages moved forward. The darkolivegreen module‐specific identified 1,921 genes which were highly expressed at 5 DAA. These were determined to be the key modules for downstream analysis.

Figure 10.

Expression profile and co‐expression regulatory network analysis of specific modules (showing opposite expression patterns in F287 and SN20, or representing different expression profile at different stages). Heatmaps show the expression profile of all the coexpressed genes (number given on the top) in the modules (labeled on top). Bar graphs (below the heat maps) show the expression pattern of coexpressed genes in each module. (a), lightpink 2 module and thistle 4 module and (b), blue 2 module showing SN20‐up/F287‐down and SN20‐down/F287‐up expression, respectively. (c), skyblue 4 module and (d), plum 3 module showing SN20‐up/F287‐down and SN20‐down/F287‐up at 5, and 15 days after anthesis (DAA), respectively. (e), lightslateblue module and (f), darkolivegreen module showing SN20‐up & F287‐up at 23 DAA and SN20‐up & F287‐up at 5 DAA, respectively. The top 6 genes of the correlation networks for the above modules were exhibited (graphs given on the below). The large nodes in the graphs were identified as candidate hub genes

Next, we constructed gene networks, where each node represents a gene, and the interconnected lines between genes represent co‐expression interrelations. Hub genes are those that show the most interactions with other genes in the network (Du et al., 2017). The top six hub genes in the network of the above key modules are shown in Figure 10, and their KEGG enrichment analysis is shown in Table 3.

Table 3.

Candidate hub genes in lightpink 2, thistle 4, blue 2, skyblue 4, plum 3, darkolivegreen, and lightslateblue modules

Gene name	Description
SN20‐specific lightpink 2 module
TraesCS1B01G107600	Sucrose‐6‐phosphatase
TraesCS1A01G203700	Chitinase 8
TraesCS1A01G450800LC	Serine/threonine‐protein phosphatase 7
TraesCS1B01G115900	Peroxidase
TraesCS1B01G175500	Uncharacterized LOC4344889
TraesCS1B01G048900	Histone H2A
SN20‐specific thistle 4 module
TraesCS1B01G008200	Charged multivesicular body protein 5
TraesCS1B01G038100	PGR5‐like protein 1A, chloroplastic
TraesCS1B01G038800	Protein kinase APK1A, chloroplastic
TraesCS1B01G080300	Low temperature‐induced protein lt101.2
TraesCS1A01G315400	E3 ubiquitin‐protein ligase
TraesCS1B01G103700	DIBOA‐glucoside dioxygenase BX6
F287‐specific blue 2 module
TraesCS1A01G104500	Replication factor A1
TraesCS1A01G050000LC	Actin‐depolymerizing factor 7
TraesCS1A01G445500	Eukaryotic translation initiation factor 2C
TraesCS1B01G119400	Glycine‐rich cell wall structural protein 1.0
TraesCS1B01G335700LC	Syntaxin‐binding protein 1
TraesCS1B01G446300	7‐dehydrocholesterol reductase
SN20‐specific at 5, 15 DAA skyblue 4 module
TraesCS1A01G539700LC	Uncharacterized protein
TraesCS1A01G542500LC	Uncharacterized protein ycf70
TraesCS1A01G065500	Uncharacterized LOC4348457
TraesCS1A01G264500	Amino acid permease 3
TraesCS1A01G379200LC	Uncharacterized protein
TraesCS1B01G454800	Hexose carrier protein HEX6
F287‐specific at 5, 15 DAA plum 3 module
TraesCS1A01G087100	Uncharacterized LOC4329709
TraesCS1A01G250300	Ubiquinol‐cytochrome c reductase subunit 9
TraesCS1A01G352800	F‐type H+‐transporting ATPase subunit G
TraesCS1A01G386300	Large subunit ribosomal protein L37Ae
TraesCS1A01G120900	Hypoxanthine‐guanine phosphoribosyltransferase
TraesCS1B01G413700	Large subunit ribosomal protein L37Ae
Specific at 5 DAA darkolivegreen module
TraesCS1A01G030000	Uncharacterized LOC4337587
TraesCS1A01G057200	Glutamyl‐tRNA reductase
TraesCS1A01G315300	UDP‐glucose 4,6‐dehydratase
TraesCS1A01G392000	Photosystem I reaction center subunit VI
TraesCS1A01G540400LC	Uncharacterized protein
TraesCS1A01G056200	Uncharacterized protein
Specific at 23 DAA lightslateblue module
TraesCS1A01G182700	Plant UBX domain‐containing protein 10
TraesCS1A01G173500	F‐box/LRR‐repeat protein
TraesCS1A01G072200	Acetylornithine aminotransferase
TraesCS1A01G049700	Serine/arginine‐rich splicing factor RS2Z3
TraesCS1A01G221800	Sorting nexin 2A
TraesCS1A01G183000	Alpha, alpha‐trehalase

4. DISCUSSION

Understanding the physical, biochemical, and molecular mechanisms regulating drought resistance of wheat is of great importance for maintaining yield under water deficit stress. Previous studies have reported that drought leads to low grain mass, and moderate water deficit stress (55%, field water capacity) decreases 1,000‐grain weight by 9.20% (Ahmed, Zhang, Ma, & Dell, 2018; Fang et al., 2017). In the present study, water deficit stress at grain‐filling stages led to low weights of superior and inferior grains, especially inferior grains. Therefore, the total grain weight was decreased (Figure 1). Additionally, the total grains and inferior grain weights of F287 were much more sensitive to water deficit than those of SN20 (Figure 1), indicating that F287 was the drought‐sensitive variety.

Flag leaves of wheat are the primary organs attacked by water deficit stress at the grain‐filling stages, causing various physiological and molecular disorders (Jan et al., 2018). Preserving the chloroplast structure plays critical roles not only in photosynthesis but also in transduction of stress‐associated signals (Sheng et al., 2014). We found that under water deficit treatments, the shape of chloroplasts became much rounder, the osmiophilic globules accumulated more, and their color was darker (Figure 2). Moreover, the granal lamellae became loose and disordered and the gaps between granal lamellae were increased. Further analysis suggested that the chloroplast structure of F287 was injured much more seriously than that of SN20. We concluded that the chloroplasts were seriously damaged instead of being repaired in response to water deficit stress. This was because drought stress led to the destruction of endogenous hormone homeostasis as well as an imbalance in the scavenging and production of ROS (Figures 4 and 6). As a result, leaf senescence was induced. Leaf senescence consists of three major phases: the initiation, reorganization, and terminal phases (Holland, Koller, Lukas, & Brüggemann, 2015). In the present study, we evaluated the flag leaf process using a Gompertz curve, which was divided into two phases: the stable persistence phase of chlorophyll and the rapid loss phase (Xie et al., 2016). We found that the Chl_total of SN20 was larger than that of F287. The senescence‐associated parameters r ₀ and r _aver were increased while M, Chl_per, and Chl_total were decreased with increasing water deficit stress (Table 1). The results demonstrate that water deficit stress could provoke the advance of the flag leaf rapid senescence process, speed up the senescence rate, and finally shorten the active photosynthetic duration. Table 2 also indicates that the leaf process was closely associated with 1,000‐grain weight. The same results were also shown in the superior grain weight and the inferior grain weight (Table 2). This is why the grain mass was decreased under water deficit stress.

It is well established that the leaf senescence process is regulated by plant endogenous hormones (Luo et al., 2018). For instance, the reduction in cytokinins and enhanced levels of ABA regulate the reorganization phase of leaf senescence (Sarwat, Naqvi, Ahmad, Ashraf, & Akram, 2013). Ethylene, JA, ABA, and SA act as enhancers of leaf senescence, while cytokinins, GA, and auxins delay senescence (Ha, Vankoa, Yamaguchi‐Shinozaki, Shinozaki, & Tran, 2012; Jibran, Hunter, & Dijkwel, 2013). Cytokinins and ethylene are the most significant phytohormones regulating the timing of leaf senescence (Jan et al., 2018). The hormone cross talk, especially the interaction between cytokinins and ABA, is believed to be involved in controlling leaf senescence response to water deficit stress (Munné‐Bosch & Alegre, 2004). In our study, the varied endogenous JA content response was the most sensitive to water deficit stress. By contrast, the content of endogenous Z was the most insensitive to water deficit (Figure 4). Table 2 also shows that the concentration of SA, JA, and ABA had positive correlations with r ₀ and r _aver, but had significantly negative association with r _max, Chl_total, Chl_per, and M. On the contrary, the concentration of Z, ZR, and GA₃ show the opposite relationship with the senescence parameters, indicating that different endogenous hormones play diverse roles in leaf senescence response to drought stress.

To construct the gene regulatory network and screen hub genes involved in conferring drought tolerance and delaying leaf senescence during grain‐filling stages in these wheat varieties, we performed WGCNA (Langfelder & Horvath, 2008). The WGCNA analysis used in our study is based on the premise that genes in the same modules, with similar expression patterns, tend to be functionally relevant (Eisen, Spellman, Brown, & Botstein, 1998; Takehisa & Sato, 2019). Conjoint analysis of physiological and biochemical indicators as well as large‐scale transcriptome data was used to identify leaf senescence‐stage‐specific or drought‐resistance‐specific gene modules (Figure 10) according to previous studies (Du et al., 2017; Garg, Singh, Rajkumar, Kumar, & Jain, 2017). Figure 10 shows that the lightpink 2 module‐specific genes (261) and thistle 4 module‐specific genes (267) were overrepresented in the drought‐resistant wheat variety SN20 at different grain‐filling stages. Moreover, the lightpink 2 module was highly correlated with hormones, while the thistle 4 module was strongly relevant to antioxidant enzymes (Figure 9). We speculated that upregulation of the genes in these modules would enhance drought resistance by influencing phytohormones and the ROS scavenging system. Hub genes are those that show the most connections in the network. After constructing the network of genes, we identified the top six genes as candidate hub genes (Figure 10). Table 3 shows the KEGG pathway enrichment analyses of the hub genes. The gene encoding E3 ubiquitin‐protein ligase was one of the hub genes. We speculated that overexpressed E3 ubiquitin‐protein ligase would lead to enhanced drought resistance of wheat. Whether or not the E3 ubiquitin‐protein ligase plays a positive role in response to drought stress remains controversial. On the one hand, it has been reported that RZFP34/CHYR1, a RING‐type ubiquitin E3 ligase of Arabidopsis, enhances drought resistance (Ding, Zhang, & Qin, 2015). Others have also reported that the E3 ubiquitin ligase gene ZmAIRP4 plays an active role in enhancing drought stress tolerance (Yang et al., 2018). Further, overexpression of TaSAP5 increases the drought tolerance of Arabidopsis and wheat seedlings (Zhang, Yin, et al., 2017). Joo, Lim, and Lee (2019) reported that the RING‐type E3 ligase CaASRF1 positively enhances the drought resistance of pepper by modulating ABA signaling. However, there have been reports indicating that the E3 ligase (DHS) causes a negative regulation of water deficit stress by decreasing cuticular wax biosynthesis (Wang, Tian, et al., 2018). Cui, Min, Byun, Oh, and Kim (2018) also reported that a putative RING E3 Ligase gene, OsDIRP1, plays a negative role in drought and salt stress, but a positive role in cold stress response. Chitinase 8, found in the lightpink 2 module, is related to studies that previously demonstrated that a class III chitinase, DIP3, induced by drought, might play critical roles in the regulation of the rice stress response (Guo et al., 2013). Chitinase may be involved in the protection of cucumber against water stress (Chen, Li, Cheng, & Chen, 2007). Peroxidase is involved in scavenging ROS and positively regulates the plant against oxidative damage induced by water deficits. Previous studies have demonstrated that other selected candidate hub genes also play critical roles in leaf senescence and plant response to abiotic stress (Chen et al., 2005; Henty‐Ridilla, Li, Day, & Staiger, 2014; Huang, Huang, Hong, Lur, & Chang, 2012; Lee, Wang, Huang, & Chen, 2001; Li et al., 2017; Nishikawa et al., 2012; Pourcher et al., 2010; Rao, Naresh, Reddy, Reddy, & Mallikarjuna, 2017; Yamada, Takatsu, & Manabe, 2003; Yang et al., 2016). The above results demonstrate that WGCNA is particularly useful and reliable for identifying candidate hub genes. Of course, the uncharacterized proteins may be novel genes influencing the drought tolerance of wheat. Additionally, the blue 2 module‐specific genes were highly expressed in the drought‐sensitive variety F287 across all the grain‐filling stages. The blue 2 module was highly related to osmotic adjustment substances (Figure 9). We hypothesized that downregulated hub genes in the blue 2 module could enhance drought resistance by regulating the osmotic regulation system. In addition, the genes of the skyblue 4 module were upregulated in SN20 at 5 and 15 DAA, but were highly expressed in F287 at 23 DAA. By contrast, the plum 3 module‐specific genes (224) were highly upregulated in F287 at 5 and 15 DAA, but were overexpressed in SN20 at 23 DAA. The skyblue 4 module had a high correlation with antioxidant enzymes, and the plum 3 module was highly relevant to osmotic adjustment substances (Figure 9). We speculated that the hub genes of the modules, which played roles in drought response, had a phase difference. The candidate hub genes might be valuable targets for the engineering of drought‐tolerant wheat varieties. We also found stage‐specific genes in the lightslateblue module and the darkolivegreen module. The expression levels of these genes in the lightslateblue module were increased with the growth process. However, the 1,921 genes in the darkolivegreen module were highly expressed at 5 DAA. This indicated that the lightslateblue module was tightly correlated with osmotic adjustment substances, whereas the darkolivegreen module was closely related to chlorophyll content (Figure 9). We hypothesized that the hub genes in the lightslateblue module were upregulated with the leaf senescence process by regulating the osmotic adjustment system. These are defined as senescence‐associated genes (SAGs), while the hub genes in the darkolivegreen module were identified as senescence‐downregulated genes (SDGs), which directly affect chlorophyll content.

5. CONCLUSION

Wheat plants responded to water deficit stress during the grain‐filling stage by changing the chloroplasts’ ultrastructure, establishing ROS scavenging mechanisms, and altering the leaf senescence process: r ₀ and r _aver were enhanced by 0.02°C and 0.009°C, whereas Chl_total, Chl_per, and M were decreased by 79.51°C, 69.87°C, and 79.43°C, respectively. Shorter Chl_total, shorter Chl_per, reduced r _max, decreased M, larger r ₀, and increased r _aver were closely associated with low grain mass under drought stress in the two wheat varieties. Moreover, the content of SA, JA, and ABA in flag leaves was enhanced; however, those of Z, ZR, and GA₃ were decreased in response to water deficit stress. The content of endogenous JA was the most sensitive to water deficit treatment, while endogenous Z was the most insensitive. Plant endogenous hormones (SA, JA, ABA, Z, ZR, and GA₃) coordinate with each other to regulate leaf senescence‐related parameters. Our conjoint analysis of physiological and biochemical indicators as well as transcriptome data provide a useful genomic resource and molecular insights into the co‐expression network conferring wheat varieties with enhanced drought tolerance related to endogenous hormones, osmotic adjustment substances, and antioxidant enzymes.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

AUTHOR CONTRIBUTIONS

Zhenlin Wang and Yong Li designed the research; Yongli Luo conducted experiments and wrote the paper. All other authors discussed the data and made comments on the manuscript.

Supporting information

 

 

 

 

 

 

 

 

 

 

 

Luo Y, Pang D, Jin M, et al. Identification of plant hormones and candidate hub genes regulating flag leaf senescence in wheat response to water deficit stress at the grain‐filling stage. Plant Direct. 2019;3:1–23. 10.1002/pld3.152