Arabidopsis thaliana ecotype Columbia-0 was used as the wild-type in this study. w34-1 (SALK_133019), w2-1 (Salk_020399), vq20-1 (CS827030), myb97-1 (Salk_112329C), myb101-2 (Salk_149918C), and myb120-3 (SALK_063698C) were obtained from the Arabidopsis Biological Resource Center (ABRC) and confirmed by PCR using a T-DNA primer and gene-specific primers (Supplementary Table S2). Seeds were surface sterilized and plated on half-strength Murashige and Skoog medium containing 1% sucrose and 0.6% agar. After stratification at 4°C for 3 days, the plates were incubated in an artificial growth chamber at 22°C under a 14-h light/10-h dark photoperiod for 1 week. Arabidopsis seedlings were then transferred to nutrient soil and grown in an artificial growth chamber at 22°C under 14-h light/10-h dark or 16-h light/8-h dark photoperiods.

For the MYB97-overexpressing transgenic plants, the coding region of MYB97 was amplified from Arabidopsis ecotype Columbia-0 using primers that contained BamHI and SalI restriction enzyme sites. PCR fragments were cloned into the vector pOCA28 under the control of a pollen-specific expression gene (AT5G09500) promoter (Honys and Twell, 2004; Winter et al., 2007). Transgenic plants were selected using 50 mg/mL kanamycin. The primers used are listed in Supplementary Table S2.

Pollen isolation was performed as described previously with some modifications (Honys and Twell, 2003, 2004). To obtain mature pollen, inflorescences collected from over 500 plants were placed in a large plastic pipe. After adding 30 mL of ice-cold 0.3 M mannitol, the flask was shaken vigorously three times for 30 s. The pollen suspension was sequentially filtered through 100- and 53-μm nylon meshes. Pollen grains were concentrated in 1.5-μL Eppendorf tubes by repeated centrifugation steps at 700 × g for 5 min at 4°C. The final compact pollen pellet was stored at -80°C. For the three stages of immature male gametophytes, after collection and concentration, the spores were loaded onto the top of 25%/45%/80% Percoll step gradient in a 10-mL centrifuge tube and centrifuged 450 × g for 5 min at 4°C. Three subfractions of immature pollen were obtained: microspores; microspores and bicellular pollen mixture; and bicellular pollen. Spores in each fraction were concentrated in Eppendorf tubes by centrifugation at 700 × g for 5 min at 4°C and stored at -80°C.

Total RNA was extracted from Arabidopsis pollen grains using TRIzol reagent (Invitrogen) and used for oligo (dT)₁₈-primed cDNA synthesis according to the Fermentas’ reverse transcription protocol (Lei et al., 2017). qRT-PCR was then performed using a SYBR Premix Ex Taq kit (Takara) on a Light-Cycler 480 real-time PCR machine (Roche) (Jiang et al., 2014). Three biological replicates of all qRT-PCR amplifications were performed with different pollen grains, and the ACTIN2 gene was used as a control. Student’s t-test was used for the analysis of statistical significance. All primers are listed in Supplementary Table S2.

The yeast one-hybrid assay was performed as described previously with some modifications (Vidal and Legrain, 1999; Ding et al., 2013). Three fragments (PF1, PF2, and PF3) enriched with two adjacent W-boxes and three corresponding mutant fragments (MPF1, MPF2, and MPF3) with mutant W-boxes of the promotor of MYB97 were independently cloned into the pAbAi vector (Clontech). pAbAi constructs were then separately integrated into the genome of the Y1HGold strain by homologous recombination to generate six bait reporter strains. WRKY2- and WRKY34- coding regions were fused to the pGADT7 prey vector (Clontech). For the yeast transformation, pGADT7-WRKY2, pGADT7-WRKY34, and pGADT7 were independently transformed into six bait reporter strains using yeast transformation system (Clontech). Transformations were plated onto SD media (-Leu) to select for transformed cells and incubated at 28°C for 4 days. Large healthy colonies were selected from the transformed and control cells. Each colony was resuspended in 0.9% NaCl, and the OD₆₀₀ adjusted to ∼0.002. Then, the suspensions were diluted in a 10×series and 100 μL of the dilution series was spread on plates containing SD (-Leu) and SD (-Leu) supplemented with 150 ng/mL Aureobasidin A (AbA) Clontech. The plates were then incubated for 3 days at 28°C.

In preparation for the assay, the cDNA sequences of WRKY2 and WRKY34 were amplified and independently cloned into pGreenII 62-SK under the control of the 35S promoter to create the effectors (Hellens et al., 2005). At the same time, the cDNA sequence of VQ20 was amplified and cloned into pGreenII62-SK. A 3,062 bp MYB97 promoter and a 3,062 bp mutant MYB97 promoter (MPF1 and MPF3) were amplified and fused respectively into pGreenII 0800-LUC vector to generate reporters (Hellens et al., 2005). The preparation of Arabidopsis mesophyll protoplasts from wild-type (Col-0) leaves and subsequent transfections were performed as described by Yoo et al. (2007). The effectors and corresponding reporters were cotransformed into prepared Arabidopsis mesophyll protoplasts to measure firefly luciferase (LUC) and renilla luciferase (REN) activities using a dual-luciferase reporter assay system (Promega). The relative REN activity was used as an internal control, and LUC/REN ratios were calculated. All of the primers are listed in Supplementary Table S2.

Siliques and flowers were examined and photographed with a Leica stereoscopic microscope. Pollen viability was examined using fluorescein diacetate (FDA) staining as described in a previous study (Heslop-Harrison and Heslop-Harrison, 1970). Images of Alexander-stained tissues were obtained under bright light, and GFP-tagged tissues were obtained using a fluorescence microscope (Leica, DM2500) (Lei et al., 2017).

Pollen germination was assayed in vitro by placing pollen grains on plates containing germination buffer solidified with low-melting-point agarose as described previously (Fan et al., 2005). The pollen grains spread on the agar plates were cultured immediately at 23°C and 100% relative humidity. After 6 h, the pollen grains were observed and photographed with a Leica microscope. At least 200 pollen grains and 60 pollen tubes were examined per culture and used to calculate the average germination rate and average pollen tube length, respectively, with Image J software.

To assay in vivo pollen germination, emasculated pistils were pollinated as described previously (Ishiguro et al., 2001). The pollinated pistils were collected 8 h after pollination and fixed in ethanol:acetic acid (3:1) for 2 h at room temperature. The fixed pistils were washed three times with distilled water and then maintained in a softening solution of 8 M NaOH overnight. The pistil tissues were subsequently washed in distilled water and stained in aniline blue solution (0.1% aniline blue in 0.1 M K₂HPO₄-KOH buffer, pH 11) for 4 h in the dark. The stained pistils were observed and photographed with a fluorescence microscope (Leica).

Our previous results found that theVQ20 interacts with WRKY2 and WKRY34, and the WRKY2–VQ20 and WRKY34–VQ20 interactions affect the transcriptional functions of WRKYs and are required for pollen development (Lei et al., 2017). However, downstream target genes and mechanisms underlying the WRKY2/34–VQ20 mediated during male gametogenesis in plants are unknown. Therefore, we compared our previous wrky2-1 wrky34-1 vq20-1 (w2-1 w34-1 vq20-1) mature pollen microarray data with previously published pollen microarray data (Liang et al., 2013; Lei et al., 2017). We found that six significantly altered genes in the myb97-1 myb101-2 myb120-3 triple mutant (designated as myb97/101/120) showed opposite changes in the w2-1 w34-1 vq20-1 triple mutant (Supplementary Table S1). We then examined the expression patterns of WRKY2, WRKY34, VQ20, MYB97, MYB101, and MYB120 during male gametogenesis using publicly availabe microarray data, and the data indicated that the WRKY34 and VQ20 expression patterns had opposite trends as those of the three MYBs during this process (Supplementary Figure S1). We further confirmed this result using quantitative RT-PCR (qRT-PCR), transcripts of WRKY34 and VQ20 decreased gradually during male gametogenesis and declined markedly in mature pollen, while the expresssion levels of the three MYBs increased gradually through the developmental process and peaked in mature pollen (Figures 1A,B). These contrasting expression patterns prompted us to hypothesize that MYB97, MYB101, and MYB120 are the potential target genes of WRKY34 and WRKY2 during male gametogenesis, and the overlapping expression of WRKY34, WKRY2, VQ20 and the MYBs suggests that WRKY34/2 (VQ20) may regulate MYB transcription during male gametogenesis.

To test this hypothesis, we collected mature pollen grains from wild-type, wrky2-1 (w2-1), w34-1, vq20-1, w2-1 w34-1 double mutant, w2-1 w34-1 vq20-1 triple mutant and myb91/101/120 triple mutant plants and examined the expression levels of MYB97, MYB101, and MYB120 using qRT-PCR. qRT-PCR results showed that the single, double, and the triple mutants conferred very different effects on the expression levels of MYB97, MYB101, and MYB120 in mature pollens. The expression levels of MYB97 and MYB120 were up-regulated in the w2-1 w34-1 double mutant and were more significantly increased in the w2-1 w34-1 vq20-1 triple mutant plants (Figure 1C). There was only a slight rise in MYB101 expression in the w2-1 w34-1 double mutant plants and the w2-1 w34-1 vq20-1 triple mutant plants, and no remarkable changes in its expression were observed in the single mutants (Figure 1C). Thus, WRKY34, WKRY2, and VQ20 suppress the expression of MYB97, MYB101, and MYB120. To further detect the influence of WRKY34/2 and VQ20 on MYBs, we examined the temporal expression patterns of the MYBs, revealing that MYB97’s expression level is higher in the w2-1 w34-1 vq20-1 triple mutant plants than in wild-type throughout male gametogenesis (Figure 1D). The expression of MYB120 was significantly higher in the triple mutant than wild-type during the bicellular pollen stage (BCP), tricellular pollen stage (TCP), and in mature pollen grains (MPG), while MYB101 expression was slightly increased in TCP and MPG (Figure 1D). To confirm the regulatory effects of WRKY34/2–VQ20 on the three MYBs during male gametogenesis, we also detected the expression level of eight putative downstream target genes of MYB97/101/120 in the WRKY/VQ single mutants, w2-1 w34-1 double mutant, w2-1 w34-1 vq20-1 triple mutant and myb97/101/120 triple mutant plants. Six out of the eight selected genes were expressed highest in w2-1 w34-1 vq20-1 triple mutant plants, followed by w2-1 w34-1 double mutant plants when compared with the wild-type, whereas they were significantly reduced in myb97/101/120 triple mutant plants (Figure 2). The results verified the changes in the transcript levels of MYB97, MYB10, and MYB120 in the w2-1 w34-1 and w2-1 w34-1 vq20-1 triple mutant plants. Thus, WRKY2, WRKY34, and VQ20 appear to negatively regulate the three MYBs during male gametogenesis.

WRKY proteins regulate the expression levels of downstream target genes by binding directly to W-boxes in their promoters (Eulgem et al., 2000; Zhang and Wang, 2005; Rushton et al., 2010; Chen et al., 2012). Because WRKY2 and WRKY34 repress the expression of MYB97, MYB101, and MYB120 during male gametogenesis, we investigated whether these WRKY transcription factors might directly bind to W-boxes in the MYB promoters. There are 13 W-boxes with TGAC core sequences in the 2.7 kb MYB97 promoter, which harbors three W-boxes groups containing two adjacent W-box. We conducted a yeast one-hybrid assays to test whether the W-box-enriched region of the promotor (containing two adjacent W-boxes) was the binding site for WRKY34 and/or WRKY2. First, we cloned this enriched region into a pBait-AbAi plasmid and generated a bait reporter strain by introducing the construct into the genome of theY1H Gold strain using homologous recombination. We also constructed negative control bait reporter strains that contained, three groups of mutant W-boxes, using the same method (Figure 3A). The associations between WRKY2 (WRKY34) and these three promoter fragments were confirmed by cotransformation on selective media lacking Leu supplemented with, concentrations of up to 150 ng/mL of AbA, which is toxic to yeast at low concentrations was added to the media to suppress background activation (Figure 3B). The bait reporter strains with the natural promoter regions (PF1 and PF3) grew well in the selective media when transformed with WRKY2, but the strains containing the corresponding mutant promoter fragments (MPF1 and MPF3) did not. WRKY34 was able to bind to the PF2 of the MYB97 promoter but not to the corresponding mutant promoter (Figure 3B). The results indicated that WRKY2 and WRKY34 were able to bind these regions in the MYB97 promoter.

We then used a dual luciferase reporter in Arabidopsis mesophyll protoplasts to confirm whether the binding of WRKY2 and WRKY34 to the MYB97 promoter depended on W-boxes. In this assay, WRKY2 and WKRY34 under the cauliflower mosaic virus (CaMV) 35S promoter were used as effectors. The reporter consisted of 35S promoter-driven REN, as an internal control, and LUC driven by either a normal MYB97 promoter (ProMYB97)or a MYB97 promoter with mutated W-boxes (MProMYB97; Figure 4A) (Hellens et al., 2005; Lei et al., 2017). As shown in Figures 4B,C, the transformation of WRKY2 and WRKY34 inhibited the reporter activity driven by the MYB97 promoter, but not the mutant MYB97 promoter, which had mutations in W-boxes s1-s13. The coexpression of VQ20 and WRKY2 (or WRKY34) dramatically attenuated ProMYB97-LUC activity but did not change the ratio of MproMYB97-LUC/REN activity. Thus the region of the MYB97 promoter containing the W-boxes appears to be critical for its interaction with WRKY34 or WKRY2, and that VQ20, as a partner of WRKY2 and WRKY34, assists in WKRY functioning.

The expression of MYB97 was up-regulated during each phase of male gametogenesis in w2-1 w34-1 vq20-1 compared with the wild-type. To assess the influence of this elevated MYB97 expression on pollen development, we generated 17 MYB97 overexpression lines (namely, MYB97OE) and chose 2 lines (lines #1 and #3) as the representatives. These two MYB97OE lines exhibited reduced fertility and shorter siliques, but most of the vegetative parts appeared normal compared with those of wild-type (Figures 5A–C). To investigate whether MYB97 overexpression affects male or female functions, MYB97-overexpression plants were used as male or female parents in crosses with the wild-type plants (Table 1). When homozygous MYB97-overexpression plants were used as the female parent in a cross with a wild-type plant, the number of seeds per silique was similar to that of the wild-type. In a reciprocal cross, however, only a few ovules in the silique were fertilized and developed into mature seeds (Table 1). To determine the sex-related transmission efficiency of heterozygous MYB97- overexpression lines, a genetic analysis was performed. When the wild-type plants were used as pollen donors in crosses with heterozygous MYB97-overexpression lines, approximately 50% of the resulting progeny having kanamycin resistance (Table 2). In contrast, the number of resulting progeny harboring contained kanamycin resistance was significantly reduced when the wild-type plants were used as recipients in crosses with the heterozygous MYB97-overexpression lines pollen grains (Table 2).

To accurately assess pollen development, we performed FDA staining to determine the pollen viability of MYB97-overexpression lines. Many fewer pollen grains of the MYB97-overexpression lines showed FDA fluorescence compared with those of the wild-type (Figure 6B). We further investigated the pollen germination and tube growth of these MYB97-overexpression lines. We hypothesized that the reduced pollen viability may lead to a reduction in pollen germination. As expected, the mean germination ratio and pollen tube length of pollen grains in plants of MYB97-overexpression lines were dramatically lower than those of wild-type plants (Figures 6A,B). Like the pollen phonotype of the w2-1 w34-1 vq20-1 triple mutant plants, MYB97 overexpression lines plants showed male sterility, with defects in pollen germination and pollen tube length. Thus MYB97 overexpression affected male gametophytic functions.