Brassica napus “ZS11” and a near-isogenic line “S45AB” were sown at the experimental station of Huazhong Agricultural University (Wuhan, China) under natural conditions. Arabidopsis thaliana Col-0 was grown in plastic pots containing soil mixture (nutrient soil:roseite = 3:1) in a greenhouse at 22°C and under 16-h/8-h light/dark photoperiod.

Leaves, stems, whole inflorescence, flower buds (1.5, 2.5, 3.5, 4.5, 5.5, and 6.5 mm), opening anthers, and siliques (2, 11, and 29 dap) were collected from ZS11 plants; only buds without stamens were collected from S45A plants. Total RNA was extracted using RNeasy Plant Mini Kit (Qiagen, United States). The first-strand cDNA was then synthesized by reverse transcription and Bna.SP6 transcript was amplified using Bna.SP6-F and Bna.SP6-R primers (Supplementary Table S1) with 32 cycles. Ubiquitin-associated (UBA) protein gene BnUBA (Yang et al., 2014) was amplified as a control (Supplementary Table S1).

The genomic DNA of BnSP6_C08 (611 bp) and BnSP6_A08 (475 bp), the pBnaC.SP6 (1167 bp), and the full-length CDS of Bna.SP6 (246 bp) was amplified with specific primers (Supplementary Table S1) and sequenced. The Bna.SP6 CDS sequences were translated to their respective peptide sequences using Primer Premier 5 software. The potential cis-acting elements in pBnaC.SP6 were predicted using the PlantCARE program (Lescot et al., 2002). The presence and location of signal peptide cleavage sites in the amino acid sequences were predicted by SignalP 4.1 server (Petersen et al., 2011).

All the plasmids used for GUS assay were constructed in the backbone of pCAM2300-H2BYFP-gusplus-Nost vector. The H2BYFP-GUSplus-Nost fragment was restricted from pG2NHL-H2BYFP-GUSplus-Nost plasmid by KpnI and EcoRI and ligated into pCAM2300.

A series of fragments with 5′-deletion in the promoter were amplified using different forward primers (p647F, p447F, p375F, p306F, p210F, and p135F) and a single reverse primer, Pro-1R. The full-length promoter and six 5′-deleted derivatives were cloned at SalI and SmaI sites in the pCAM2300-H2BYFP-GUSplus-Nost vector and the resulting constructs were designated as pBnaC.SP6, p647, p447, p375, p306, p210, and p135, respectively. In addition, two 3′-deletions containing the regions from -306 to -52 and -210 to -52 were amplified using two forward primers (p306F and p210F) from the upstream region and a single reverse primer (Pro-d52R) from the downstream region. They were cloned into the pCAM2300-H2BYFP-GUSplus-Nost vector and the resulting constructs were designated as p254 and p158. The primers used are shown in Supplementary Table S1.

Col-0 plants were transformed with recombinant Agrobacterium tumefaciens GV3101 strain, harboring the promoter:H2BYFP-GUSplus construct, by floral dip method (Clough and Bent, 1998). Seeds from the wild-type were germinated on agar plates containing half-strength Murashige and Skoog’s medium (12 MS), with 1% (w/v) sucrose, 0.7% (w/v) agar at pH 5.8, and supplemented with kanamycin (50 mg l^-1) and timentin (75 mg l^-1). Thereafter, we randomly selected 100–160 seeds of each T₁ line to perform the segregation analysis. The kanamycin-resistant lines showing 3:1 segregation pattern were further carried onto the next generation to obtain homozygous lines. Three such homozygous lines were used for quantitative GUS activity assay. Chi-square values were calculated by the corrected formula χC2 =Σ(O−E−1/2)2/E, where O and E are the observed and expected values, respectively. The probability was calculated with two-degrees-of-freedom based on Chi-square distribution table.

The histochemical GUS and 4′,6-diamidino-2-phenylindole (DAPI) staining assays were performed as described by Luo et al. (2012) and Liu et al. (2017), respectively. Different tissues from heterozygous transgenic lines were stained overnight in X-Gluc solution. Thereafter, the chlorophyll was cleared out of the samples using 70% ethanol. Images of various tissues were taken using an Olympus DP72 Digital Microscope Camera.

GUS-stained flower buds were embedded in Technovit 7100 resin, as described previously by Zhu et al. (2010). Subsequently, transverse sections of the anthers (approximately 8 μm thick) were cut from the embedded blocks using a Leica Ultracut R ultramicrotome. The GUS-stained anthers were incubated at 60°C for 1 h in a DAPI staining solution containing 20% methanol and 1.0 μg ml^-1 of DAPI. The images were taken using a Nikon ECLIPSE 80i microscope.

Arabidopsis flower buds (flower development stage 12–15) of homozygous transgenic lines were collected for quantitative measurement of GUS activity according to the protocol described by Jefferson et al. (1987). Total proteins were extracted from the buds and the protein concentration was determined by the Bradford method (Bradford, 1976). The GUS assay buffer containing the substrate, 4-methylumbelliferyl-β-d-glucuronide (MUG), was added to the protein samples and the reaction was incubated for 30 min at 37°C. The resulting fluorescence was recorded using Tecan Infinite M200 PRO (Tecan Group Ltd., Switzerland). The relative GUS activities were calculated and expressed as nmole 4-MU generated per min per milligram of the total protein.

A Y1H assay was performed as mentioned in the manual of Matchmaker Gold Yeast One-Hybrid Library Screening System (Clontech). As the baits, pAbAi-p158, pAbAi-mp158, and pAbAi-p99 (part of p158, containing only the ABRE motif) were transformed into the Y1HGold yeast strain. Then the pGADT7-BnaA.bZIP1 fusion plasmid was introduced into three bait strains, respectively. The co-transformed yeast cells were cultured on SD/-Leu agar plates with or without AbA and incubated at 30°C for 3 days. p53 was used as a positive control.

pGBKT7-BnaA.bZIP1 and pGBKT7-atbZIP1 fusion plasmids were introduced into yeast reporter strain AH109 (Clontech). The transformants were transferred onto a filter paper and incubated at 30°C for 3–5 h in the presence of X-Gal to check the β-galactosidase activity by monitoring the generation of blue color. The primers used are shown in Supplementary Table S1.

The pET-32a-BnaA.bZIP1 recombinant plasmid was transformed into Escherichia coli BL21 cells. The recombinant fusion proteins were purified using Ni-NTA His•Bind^® Resin (Novagen). The complementary oligonucleotides (p59 and mp59) containing the consensus DNA-binding site C-box (CACGTC) and mC-box (Ctatga) were, respectively, annealed and used as DNA probes. The DNA–protein binding reactions were performed in a total volume of 10 μl, containing 5 μl BnaA.bZIP1-His or pET-32a-His protein, 2 μl 5× EMSA/Gel-shift-Binding Buffer (Beyotime Biotechnology, China), 30 nM Cy5-labeled probe, and 200- to 800-fold molar excess of unlabeled competitor. The reaction mixture was incubated for 30 min at 25°C. The electrophoresis was performed with 6% non-denaturing polyacrylamide gel and carried out in 0.5× TBE (45 mM Tris base, 45 mM boric acid, 0.5 mM EDTA, pH 8.3) at 4°C in a vertical electrophoresis system. The reaction mixture was loaded and electrophoresis was performed at 10 mA until the dye front migrated through 50% of the length of the gel. The gels were scanned to detect the fluorescent DNA using Fujifilm FLA-9000 plus DAGE (FujiFilm, Japan). The primers used are shown in Supplementary Table S1.

The dual luciferase reporter (DLR) assay was performed as described by Hao et al. (2010). The full-length BnaA.bZIP1 (139 a.a.), two C-terminal deletions (BnaA.bZIP1ΔC1, 114 a.a. and BnaA.bZIP1ΔC2, 89 a.a.) of BnaA.bZIP1, and AtbZIP1 (145 a.a.) restricted with XbaI and BamHI were inserted into pBDGAL4 vector as effectors. Luciferase (LUC) driven by the CaMV35S promoter was used as a reporter. The Renilla LUC gene driven by the Arabidopsis UBIQUITIN3 (AtUBI3) promoter was used as an internal control.

In addition, the pGreenII 62-SK and pGreen II 0800-LUC transient expression system was employed as described by Hellens et al. (2005). BnaA.bZIP1 restricted with BamHI and XhoI was inserted into the pGreenII 62-SK vector as the effector. The fragments p158, mp158, p103 (part of p158, containing only the C-box), and mp103 restricted with SalI and SmaI were inserted into the pGreen II 0800-LUC vector as reporters. The effector and reporter constructs were co-transformed into Arabidopsis protoplasts by polyethylene glycol (PEG)/calcium-mediated transformation (Yoo et al., 2007). After transfection, the luminescence from firefly LUC and Renilla LUC were recorded on a Tecan Infinite M200 PRO. The primers used are shown in Supplementary Table S1.

The BnaA.bZIP1 transcript was detected by quantitative real-time PCR (qRT-PCR) on a CFX96 Real-Time System (Bio-Rad, United States). The expression data were calculated using the 2^-ΔΔC_t method. Each sample was assayed in triplicate. The 1317-bp promoter region (pBnaA.bZIP1) of BnaA.bZIP1 restricted with SalI and BamHI was inserted into the pCAM2300-H2BYFP-GUSplus-Nost vector.

The pM999-CFP-GHD7 plasmid was used as a nuclear marker. The full-length CDS of BnaA.bZIP1, without the termination codon, restricted with XbaI, was inserted into the pM999-GFP vector. The preparation of Arabidopsis mesophyll protoplasts and their subsequent transfection was done as described above. The fluorescence signals were measured using a FV1200 Laser Scanning Microscope. The primers used are shown in Supplementary Table S1.

The fusion protein of full-length BnaA.bZIP1 with a 3XFlag-tag at the N-terminal was cloned into pTA7002 vector (Aoyama and Chua, 1997). The vector construct was transformed into p158 transgenic line. T₂ plants harboring BnaA.bZIP1 and p158 fragments were grown on soil and treated with DEX (Sigma–Aldrich, United States) at the flowering stage. Water (10 ml) containing 15 nM DEX was applied daily for 5 days to the soil. The day of first DEX treatment was recorded as 1 days. Buds (with p158 activity) were collected from T₂ plants on 0, 4, and 6 days after the DEX treatment. RNA extraction and qRT-PCR were performed as described above and Actin7 was used as the reference. The primers used are shown in Supplementary Table S1.

AT1G19500 belongs to small secreted protein gene family and named ATSP6 (Nakamura et al., 2012). Two ATSP6 homologs, BnSP6_C08 and BnSP6_A08, were cloned from B. napus “ZS11.” BnSP6_C08 was observed to have a significantly longer intron than that present in its paralog, BnSP6_A08 (Figure 1A). Their coding sequences (CDS) were also cloned from the opening anthers of “ZS11” and designated as BnaC.SP6 and BnaA.SP6. Sequence alignment showed that BnaC.SP6 and BnaA.SP6 only had five mismatches in their CDS sequences. They encoded 81-a.a. peptides containing putative signal cleavage sites between residues 25 and 26, and had not any discernable motif (Figure 1B). Thus, we used the cloning primers to analyze the full-length transcripts of Bna.SP6 (BnaC.SP6 and BnaA.SP6) by RT-PCR in different organs. The result revealed that Bna.SP6 existed high expression levels in whole inflorescence, opening anthers, and 29 dap (dap, day after artificial pollination) siliques. The expression levels were low in 5.5 and 6.5 mm buds, and 11 dap siliques; however, the expression was rarely detected in 1.5, 2.5, 3.5, and 4.5 mm buds, 2 dap siliques, buds without stamens, leaves, and stems (Figure 1C).

In the core promoter region, pBnaC.SP6 shared 95.7% nucleic acid identity with pBnaA.SP6. So pBnaC.SP6 (1167 bp) was selected for further analysis. pBnaC.SP6 (1167 bp) was fused to GUS reporter gene, and was introduced into wild-type Arabidopsis (Col-0) plants. Whole plant (14-days-old), cauline leaves, whole inflorescence, flower buds, and siliques (5 and 10 mm) of transgenic lines were evaluated by GUS staining. pBnaC.SP6 showed no GUS expression in the vegetative organs, such as root and rosette leaf of 14-day-old plant. During the reproductive phase, GUS expression was observed in the anthers at flower developmental stages 11–15. Also, GUS expression was observed in the silique septum of 5 mm siliques but not in the seeds and valves (Figure 2A). This could be the reason as to why Bna.SP6 transcript was rarely detected in 2 dap siliques using RT-PCR (Figure 1C). These results indicated that BnaC.SP6 encoded a small peptide and expressed in mature anthers and silique septum.

Motif search using PlantCARE revealed that the pollen-specific activation-related element POLLEN1LELAT52, LAT enhancer element, and GTGA MOTIF are located in pBnaC.SP6, which might help drive the expression of BnaC.SP6, predominantly in pollen of mature anthers (Supplementary Figure S1). Some known cis-elements, such as P1BS, C-box, ABRE motif, Skn_1-motif, and DOFCOREZM, were also detected. In addition, a potential TATA-box sequence was located at the position -136 (Supplementary Figure S1). Some of these elements, such as DOFCOREZM and C-box, were located around the basic promoter element, the TATA-box, and are likely to be required for the high levels of expression of BnaC.SP6 in the developing reproductive organs.

To identify functional regions in pBnaC.SP6 (p1167), a series of promoter deletions (p643, p447, p375, p306, p254, p210, p158, and p135) cassettes were introduced into Col-0 (Supplementary Figure S2A). With 5′- and 3′-deletion constructs, GUS expression could still be detected in mature anthers (Figures 2B–H); however, p135 failed to show GUS activity in any of the organs tested (data not shown). p158, containing the region from -210 to -52, showed pollen-specific GUS activity, indicating that the -210 to -135 region was crucial for the expression of BnaC.SP6 in pollen whereas the -52 to -1 region was not necessary (Figure 2H). Interestingly, only p643 showed weak GUS expression in the cotyledon (Figure 2B). When the -447 to -375 region was missing from p447, no GUS expression was observed in the silique septum (Figures 2C,D). This suggested that the -447 to -375 region was important for the silique septum activity of pBnaC.SP6.

The pollen-specific activity of p158 in developing pollen was determined using semi-thin anther sections and DAPI staining. Obvious GUS staining was first detected in the early bicellular pollen at the anther development stage 12 (Figures 3A,B,I,J). At stage 13, the GUS expression reached its maximal level in the tricellular pollen, and declined gradually in the pollen grains until mature pollen grains were released (Figures 3C,D,K,L). Similar results were obtained by staining the anthers with DAPI. The blue-fluorescing DAPI-stained nucleic acids were clearly detectable under ultraviolet illumination in the early bicellular, tricellular, and mature pollen (Figures 3E–H). Therefore, p158 could specifically drive the GUS expression from early bicellular pollen to mature pollen stage.

To determine the GUS activity driven by different promoter deletions of BnaC.SP6, quantitative GUS assays were performed after segregation analysis (Supplementary Table S2). As expect, p158 showed the minimum GUS activity in comparison to those obtained with other deleted BnaC.SP6 promoters (Supplementary Figure S2B). The p158 region was also amplified and sequenced from B. rapa, B. oleracea, B. juncea, and B. carinata. They shared 100% nucleic acid identity with the B. napus p158, suggesting that the pollen-specific activity of p158 was highly conserved in these species (Supplementary Figure S3). These results indicated that the minimal promoter p158 was required for late pollen-specific expression of BnaC.SP6 and highly conserved in Brassica species.

Quantitative real-time PCR analysis showed that BnaA.bZIP1 was predominantly expressed in flower buds and at low levels in stems, leaves, and 2, 11, and 29 dap siliques (Figure 4A). To obtain further insights into the expression pattern of BnaA.bZIP1 in developing flowers, the pBnaA.bZIP1 reporter construct was transformed into Col-0 plants and histochemical GUS assay was performed. During flower development, the initial GUS activity was specifically detected in the anthers at flower stages from 9 to 12. From stages 13 to 15, GUS expression was also observed in the stamen filaments and perianths (Figure 4B). These suggested that BnaA.bZIP1 was dynamically expressed in floral organs and coexpressed with BnaC.SP6 in anthers. Therefore, the full-length CDS of BnaA.bZIP1 were cloned from opening anthers. It contained a CDS of 420 bp, which encoded a 139 a.a. protein with a basic leucine zipper (bZIP) DNA-binding and dimerization domains localized between residues 17 and 68. The overall amino acid sequence identity between BnaA.bZIP1 and AtbZIP1 was 78.77% (Supplementary Figure S4). To determine the subcellular localization of BnaA.bZIP1, a vector pM999-BnaA.bZIP1-GFP was constructed by fusing GFP to the C-terminus of BnaA.bZIP1. We observed that the GFP signal coincided with the CFP signal of a previously characterized nuclear marker CCT domain protein (OsGHD7), indicating that BnaA.bZIP1 is a nuclear-localized protein (Figure 4C).

It had been suggested that AtbZIP1 functions as a transcriptional activator in yeast and its C-terminus is important for the transactivation activity (Sun et al., 2012). To investigate whether BnaA.bZIP1 has the similar activity as that of AtbZIP1, pGBKT7-BnaA.bZIP1, pGBKT7-atbZIP1, and pGBKT7, plasmids were separately transformed into the yeast strain AH109. All the transformants could grow well on the SD/-Trp medium, but on SD/-Trp-His medium only the yeast cells containing pGBKT7-BnaA.bZIP1 and pGBKT7-AtbZIP1 grew well. Consequently, the pGBKT7-BnaA.bZIP1 transformants showed blue color as did the colonies transformed with pGBKT7-atbZIP1 in the X-Gal staining for β-galactosidase activity (Figure 4D). Furthermore, the DLR assay system was used to confirm the transcriptional activation abilities of BnaA.bZIP1. The full-length BnaA.bZIP1 could activate the reporter gene but BnaA.bZIP1ΔC1 and BnaA.bZIP1ΔC2 failed to activate it, compared to the negative control pGAL4BD (Figure 4E). Furthermore, the positive control AtbZIP1 could also activate the reporter gene, which had almost 1.4-fold LUC activity compared to the full-length BnaA.bZIP1. These results indicated that BnaA.bZIP1 exhibited similar transcriptional activity and the C terminus of BnaA.bZIP1 was necessary for its transcription activation activity.

BnaA.bZIP1 belongs to bZIP TFs, which has been demonstrated that it can specifically recognize and bind to ACGT-containing elements in their target genes. Sequence analysis revealed that pBnaC.SP6 contained a C-box (CACGTC) and an ABRE motif (TACGTG) (Supplementary Figure S1). Moreover, BnaA.bZIP1 was coexpressed with BnaC.SP6 in anthers. These evidences imply that BnaA.bZIP1 might interact to the pBnaC.SP6. Thus, we performed a yeast one-hybrid assay with p158, mutated p158, and p99 promoter regions to examine whether BnaA.bZIP1 could bind to the C-box or ABRE motif. All the co-transformed yeast cells grew well on SD/-Leu medium without Aureobasidin A (AbA) and the positive control p53 also grew well on SD-Leu medium with AbA (50 ng ml^-1). Yeast cells co-transformed with pGADT7-BnaA.bZIP1 and pAbAi-p158 grew normally, but the yeast cells co-transformed with pGADT7-BnaA.bZIP1 and pAbAi-P99 or pAbAi-mp158 (with C-box sequence in the p158 changed from CACGTC to Ctatga) failed to survive on agar plates containing 50 ng ml^-1 AbA (Figure 5A). This indicated that BnaA.bZIP1 specifically binds to the C-box but not to the ABRE motif in p158 promoter region in yeast.

When 100 ng ml^-1 AbA was supplemented to the medium, only the p53 grew well on SD/-Leu medium (Figure 5A); thus, to further confirm the interaction of BnaA.bZIP1 with C-box, an electrophoretic mobility shift assay (EMSA) was performed with Cy5-labeled probes, p59 and mp59 (Figure 5B). The recombinant fusion protein, BnaA.bZIP1-His, was able to bind to the p59 probe, but failed to bind to the C-box-mutated probe, mp59. Furthermore, this specific binding could be reduced by competition with unlabeled probe at 200×, 400×, and 800× concentrations (Figure 5C). The negative control, pET-32a-His, failed to bind to the p59 probe. These results strongly provide further support that BnaA.bZIP1 can directly bind to the C-box in the p158 region. The results described above revealed that BnaA.bZIP1 has transcriptional activity; however, it was not clear whether it was an activator or a repressor. To determine this, a DLR assay using Arabidopsis protoplasts was performed to investigate how BnaA.bZIP1 interacts with the p158 region (Figure 5D). The presence of the effector 62-SK-BnaA.bZIP1 and the reporter 0800-p158-LUC resulted in a 50% reduction in LUC activity compared to that in the control, whereas such a reduction was not observed with the 0800-mp158-LUC reporter. Similar level of reduction in LUC activity was obtained for the combination of 62-SK-BnaA.bZIP1 and 0800-p103-LUC, but for 0800-mp103-LUC, the LUC activity was close to that in the control (Figure 5E). Thus, it was confirmed that BnaA.bZIP1 could specifically bind to the C-box sequence and suppress the p158 activity.

We also generated the stable transgenic plants expressing BnaA.bZIP1 with a dexamethasone (DEX)-inducible system in the background of p158 promoter transgenic line. qRT-PCR was performed to measure the suppression activity of BnaA.bZIP1 using total RNA from late buds of the DEX-inducible plants. Firstly, the expression level of BnaA.bZIP1 was induced to 37- and 154-fold higher than that in the control (Figure 6A), and the expression of GUS was decreased by 1.5-fold as expected (Figure 6B), which is consistent with the transient assay, indicating that BnaA.bZIP1 suppressed the activity of p158 (Figure 5E). Secondly, the expression of BnaC.SP6 homologous gene, ATSP6, was also decreased in the inducible plants (Figure 6C). Additionally, five genes (LEA27, GGP1, ATPS2, CML24, and RIN4) which were previously characterized to be bound and repressed by AtbZIP1, were also detected (Para et al., 2014). qRT-PCR results showed that the expression levels of these genes were consistently decreased in the inducible plants except for CML24 and RIN4 (Figures 6D–H). These results strongly supported that BnaA.bZIP1 acts as a transcriptional repressor of BnaC.SP6 in pollen activity.