Wheat AGAMOUS LIKE 6 transcription factors function in stamen development by regulating the expression of Ta APETALA3

Flower development is the basis for seed development in angiosperms. On the basis of analyses of flower mutants in the model dicot species Arabidopsis thaliana and Antirrhinum majus, the ABCDE model was proposed to interpret the molecular mechanism underlying flower development (Coen and Meyerowitz, 1991; Pelaz et al., 2000; Theißen, 2001; Ditta et al., 2004).

In rice (Oryza sativa), the functions of many floral genes have been elucidated. These genes included OsMADS14, OsMADS15 and OsMADS18 (Kobayashi et al., 2012; Wu et al., 2017), B class gene OsMADS16/SPW1 (Nagasawa et al., 2003), C class genes OsMADS3 and OsMADS58 (Yamaguchi et al., 2006; Dreni et al., 2011; Hu et al., 2011), D class gene OsMADS13 (Dreni et al., 2007; Li et al., 2011b), E class genes OsMADS7 and OsMADS8 (Cui et al., 2010), OsMADS1 (Jeon et al., 2000), OsMADS34 (Gao et al., 2010; Kobayashi et al., 2010; Lin et al., 2014), and carpel identity gene DROOPING LEAF (DL) (Yamaguchi et al., 2004; Li et al., 2011b).

In the MADS-box gene family, AGAMOUS LIKE 6 (AGL6) genes constitute an ancient sub-family of MADS-box genes and can be found extensively in the plant kingdom (Becker and Theissen, 2003; Reinheimer and Kellogg, 2009). Two AGL6 genes have been identified in Arabidopsis: AGL6 and AGL13. Because of probable functional redundancy, null mutant plants of agl6-2 showed no obvious phenotypes in flowering time and floral development (Koo et al., 2010). Analyses of transgenic Arabidopsis overexpressing AGL6 showed its function in flowering time (Koo et al., 2010; Yoo et al., 2011b). Additionally, AGL6 functions in leaf movement and is involved in the formation of the axillary bud (Yoo et al., 2011a; Huang et al., 2012). AGL13 regulates the development of male and female gametophytes (Hsu et al., 2014).

In Petunia, pMADS4/PhAGL6 showed redundant functions with E class genes FBP2 and FBP5 (Rijpkema et al., 2009). In grasses, the maize (Zea mays) AGL6 gene ZAG3 regulates the development of floral organs and meristems (Thompson et al., 2009). Rice OsMADS6 affects the development of paleas, lodicules, stamens and ovules, as well as the floral meristems (FMs) and seeds (Ohmori et al., 2009; Li et al., 2010; Zhang et al., 2010; Duan et al., 2012). Analyses of genetic interactions between OsMADS6 and other floral genes indicated that OsMADS6 is a master floral regulator (Li et al., 2011a).

The bread wheat (Triticum aestivum) is a hexaploid plant (Shewry, 2009). Although an ABCDE model has been proposed for wheat (Murai, 2013), the biological functions of most wheat floral genes have not yet been elucidated.

Previous studies have shown that TaMADS12 and TaAGL37 are AGL6 homologues in wheat (Paolacci et al., 2007; Reinheimer and Kellogg, 2009). In this study, we have investigated the functions of wheat AGL6 homoeologous genes. Our results show that TaAGL6 transcription factors play an essential role in stamen development through the transcriptional regulation of TaAP3.

For the sake of cloning the full cDNAs of TaAGL6, 3′ and 5′ rapid amplification of cDNA ends (RACE) experiments were performed (Table S1). About 1000 bp and 400 bp PCR products in which 205 bp overlapped were obtained (Fig. S1A) and cloned into T vectors. Sequencing results showed that both 3′ and 5′ RACE products included three conserved but distinctive sequences (Fig. S1B), indicating that there are three homoeologous TaAGL6 genes, which is consistent with the wheat genome (Appels et al., 2018). According to chromosome locations, they were named as TaAGL6-A, TaAGL6-B and TaAGL6-D. TaAGL6-A corresponds to reported TaMADS37 and TaAGL6-B corresponds to TaMADS12. Every cDNA includes an open reading frame and encodes one AGL6 protein (Fig. S2). In addition to MADS-box and K domains, these proteins have AGL6 motifs I and II (Fig. S2). They were classified into the grass AGL6-I-ZAG subgroup (Dreni and Zhang, 2016). Although some SNPs resulted in some differences in the protein sequence, they did not significantly affect the functions, according to the prediction with the software Protein Variation Effect Analyzer (Fig. S2B). These results indicate that they have similar functions.

First, we analyzed TaAGL6 gene expression pattern using quantitative RT-PCR with primers simultaneously applicable for TaAGL6-A, TaAGL6-B and TaAGL6-D. Consistent with a previous report (Feng et al., 2017), results showed that TaAGL6 genes are strongly expressed in the inflorescences from stage 3-8.5 (Fig. 1A,B), according to previous classification (Waddington et al., 1983). The transcripts mainly accumulate in the paleas, lodicules and pistils at stage 6-7. Relatively, the expression in the lemmas and stamens is very low (Fig. 1C). In addition, qRT-PCR with specific primers showed that these three TaAGL6 genes have similar expression patterns in the floral organs (Fig. S3).

We further performed in situ hybridization. TaAGL6 mRNA was not detected in the inflorescences at stage 2 (Fig. 1D), and the transcripts first emerged in the FMs at stage 3 (Fig. 1E) and then in the palea and lodicule primordial at stage 3.5-4 (Fig. 1F,G). TaAGL6 mRNA was then continually detected in the developing lodicules and paleas at stage 4-7.5 (Fig. 1H-P). Meanwhile, TaAGL6 transcripts were detected in the carpels and ovules at stage 4.5-7.5 (Fig. 1I-P), while signal in the lemmas and stamens was undetectable (Fig. 1I,K-P). When hybridized with sense probes, no signal was found (Fig. 1Q).

TaAGL6 genes were overexpressed in Arabidopsis. Generally, the transgenic plants harboring different TaAGL6 genes showed similar phenotypes that could be classified into two types: early flowering and dwarf (Fig. S4A-E); and with an increased number of stems or branches (Fig. S4F-J). Relatively, the expression level of the target gene was higher in the early flowering lines than in the multi-branch lines (Fig. S5A). These observations further indicate that three TaALG6 genes have similar functions.

FLOWERING LOCUS T (FT) in Arabidopsis is a master flowering regulator (Gu et al., 2013). We analyzed the expression of FT in our transgenic plants. The expression level of FT is higher in the transgenic lines than the control during the vegetative stage, consistent with the early flowering phenotype (Fig. S4K).

We also generated transgenic wheat overexpressing TaAGL6-B and obtained 27 positive lines. The expression level of the target gene was increased in most lines, especially in lines 18 and 30 (Fig. S5B). The flowering time of 12 lines was 10-20 days earlier than that of the control (Fig. S6A–E), and the expression of wheat FT homologs was upregulated in the transgenic wheat (Fig. S6F); however, no floral phenotype was observed.

Because all three TaAGL6 genes are expressed in the flowers and showed similar functions in transgenic Arabidopsis, we speculated that their functions are redundant. Therefore, we knocked down these TaAGL6 genes simultaneously using RNA interference (RNAi). A total of 21 transgenic wheat lines were obtained. qRT-PCR results showed that the expression of TaAGL6 genes in lines 17-2, 28-3 and 3-2 was decreased drastically (Fig. S5C), so they were selected for further analyses. At the vegetative stage, none of them showed obvious phenotypes. At the reproductive stage, abnormal floral phenotypes were observed.

In about 30% of RNAi florets, stamens display abnormality in the number or/and the morphology. One normal floret generates three stamens (Fig. 2A), whereas some transgenic florets bear only two stamens (Fig. 2B). The morphology of some RNAi stamens was also affected (Fig. 2C,D). As shown in Fig. 2E, a normal stamen has four anther cavities; however, in some transgenic stamens, two anther cavities fuse together or one stamen has only two anther cavities (Fig. 2F,G). The vigor of pollen grains was affected in all RNAi stamens, irrespective of morphology (Fig. 2H-J). As a result, the seed-setting rate is lower (35.81%) than the control (96.34%) (Fig. 2K). Occasionally, some florets displayed strong phenotypes by generating three or four carpels without stamens or with anther-stigma mosaic organs (Fig. 2L-N).

Scanning electron microscope (SEM) observation showed the developmental defects of stamens at the early stage, which is consistent with these phenotypes. In normal cases, three stamen primordia are formed; however, in some transgenic florets, only two normal stamen primordia were observed (Fig. 2O,P). In addition, all stamens in the wild type are quadrangular; however, in some transgenic florets, only one stamen was quadrangular and the morphology of the other two stamens was not normal (Fig. 2Q,R). These phenotypes were observed in not only T₀ plants but also T₁ and T₂ plants.

Because only a single AGL6 gene was found in the wheat A, B and D sub-genome, and these homoeologous genes are mainly expressed in the paleas and lodicules, we predicted that wheat AGL6 genes function as rice OsMADS6 genes, and also play roles in paleas and other organs.

Surprisingly, no obvious phenotype was observed in RNAi paleas and lodicules. Reinheimer and Kellogg (2009) proposed that the expression of grass AGL6 genes in the ovule is conserved, whereas the expression domain of the palea is acquired during the evolution process. TaAGL6 genes might acquire the expression domain in the paleas, but have no functions in palea development. During the evolution process, the functions of AGL6 become divergent, and OsMADS6 gains the function of specifying palea identity. Another possibility is that the low expression of TaAGL6 genes in RNAi lines is enough to maintain normal palea and lodicule development.

Several AGL6 transcription factors interact with floral regulators. ZAG3 forms a heterodimer with C-function protein ZAG1 (Thompson et al., 2009); OsMADS6 interacts with D-function protein OsMADS13 (Li et al., 2011a); and OsMADS6 and OsMADS17 form protein complexes with B-function proteins (Seok et al., 2010). Similar to the E-function protein FBP2, petunia AGL6 interacts with C- and D-function proteins (Rijpkema et al., 2009). These and other results indicate that AGL6 transcription factors have partial E-function proteins.

To clarify whether TaAGL6 proteins interact with other floral regulators, yeast two-hybrid (Y2H) and bimolecular fluorescence complementation (BiFC) assays were performed. The results showed that TaAGL6-A interact with TaAP3, TaAG and TaMADS13 not only in yeast cells, but also in tobacco (Nicotiana benthamiana) leaf cells (Fig. 3A,B). We further verified these interactions in planta by performing co-immunoprecipitation (Fig. 3C). In addition, TaAGL6-A displayed transcription activity, whereas TaAP3, TaAG and TaMADS13 did not (Fig. S7). Meanwhile, TaAGL6-B and TaAGL6-D interact with TaAP3, TaAG and TaMADS13 in yeast cells and tobacco leaf cells (Fig. S8). These results showed that TaAGL6 proteins act as E-class transcription factors, and further implied redundant functions.

Previously, we investigated the expression patterns of TaAP3, TaAG, TaDL, TaMADS13, TaSEP and TaLHS1 in floral organs (Li et al., 2016). We analyzed the expression of related genes in TaAGL6 RNAi stamens. qRT-PCR results showed that the expression of TaAP3 and TaMGH3 is obviously decreased in RNAi stamens. The ectopic expression of TaDL, TaMADS13 and TaLHS1 was detected, whereas the expression of TaAG and TaMADS58 is slightly increased in the RNAi stamens (Fig. S9). These results further verified the developmental defects in TaAGL6 RNAi stamens at the molecular level.

OsMGH3 is an auxin-responsive gene and is regulated by OsMADS6 indirectly; it also functions in flower development (Zhang et al., 2010; Yadav et al., 2011). We screened the promoters of three TaMGH3 genes but no CArG motif was found. Therefore, they are indirect downstream genes of TaAGL6. In wheat, two TaAP3 homoeologous genes exist: TaAP3-B and TaAP3-D. In the promoters of TaAP3-B and TaAP3-D, two CArG motifs were found, respectively (Fig. 4A). Taking the transcription activity of TaAGL6 proteins into account, we wondered whether TaAGL6 transcription factors regulate the expression of TaAP3 directly. Therefore, we expressed and purified the TaAGL6-B-GST fusion protein (Fig. 4B,C) for electrophoresis mobility shift assays (EMSAs). The EMSA results showed that the fusion protein could bind to motif 1 (Fig. 4D), but it could not bind to the other three motifs (Fig. S10). Results of yeast one hybrid (Y1H) further verified the interactions between three TaAGL6 transcription factors and M1 (Fig. 4E,F). Transient expression assay further showed the positive regulation of TaAGL6 to TaAP3-B in planta (Fig. 4G,H).

As transcription factors, MADS-box proteins function through transcriptional regulation of target genes. In addition to regulating the expression of OsMGH3 indirectly (Zhang et al., 2010), OsMADS6 regulates the expression of OsMADS58 and OsFDML1 directly (Li et al., 2011b; Tao et al., 2018). However, we failed to identify the ortholog of OsFDML1 in the wheat genome (data not shown). These results indicate that the detailed mechanism is different.

Mutations of AP3 resulted in the homeotic transformation of stamens to carpels. Similar phenotypes were observed in Antirrhinum majus defa, maize Silky1 and rice spw1 mutant plants (Schwarz-Sommer et al., 1990; Ambrose et al., 2000; Nagasawa et al., 2003; Whipple et al., 2004), suggesting important and conserved functions for AP3 genes during stamen development.

The disrupted expression of AP3 in TaAGL6 RNAi stamens may be the cause of abnormality in stamen development (Fig. 2). In addition, SPW1 represses the expression of DL (Nagasawa et al., 2003). This mechanism might be conserved in wheat (Murai, 2013). Consistent with this prediction, TaDL is ectopically expressed in TaAGL6 RNAi stamens, accompanied with deceased expression of TaAP3 (Fig. S9), indicating the potential homeotic transformation from stamens to carpels. The ectopic expression of TaLHS1 and TaMADS13 (Fig. S9) and the severe phenotypes (Fig. 2) further support this prediction.

In general, our results extended understanding of AGL6 genes and provided further evidence that AGL6 proteins have E-functions. Moreover, our findings indicate the distinctive functions and mechanisms of action of TaAGL6 genes in stamen development. We propose a model to illustrate these findings (Fig. S11) but further studies will help us to reveal the mechanism.

The wheat varieties Chinese Spring (CS) and Kenong199 (KN199) were used in this study. They were planted in a glasshouse under a photoperiod of 12 h of light at 22°C and 12 h of darkness at 16°C, or in an isolated experimental field under natural conditions at the Northwest A&F University (108°4′E, 34°15′N). The Arabidopsis ecotype Columbia-0 (Col-0) was used to generate the transgenic lines in growth chambers with a light period of 16 h and dark period of 8 h at 23°C. Phenotypes were observed in the T2 and T3 plants. Tobacco was planted under the same growth conditions as those of Arabidopsis.

Total RNA was extracted from CS flowers, and the SMART RACE kit (TaKaRa) was used to synthesize the cDNA and generate the 5′-RACE and 3′-RACE products, according to the protocol. The specific primers for 5′-RACE (5GSP in Table S1) and 3′-RACE (3GSP in Table S1) were designed by Primer Premier 5.0. Specific PCR products were cloned into the pMD-19T vector (TaKaRa) and different positive clones were sequenced.

The roots, stems, leaves and flowers of wheat were collected at the heading stage. Inflorescences at different stages were collected and judged according to a previous study (Waddington et al., 1983). Glumes and different floral organs were collected from florets at stages 6-7. Total RNAs were extracted from different organs using TRIzol reagent (Sangon Biotech), according to the manufacturer's instructions. cDNAs were synthesized with AMV reverse transcriptase (Roche). qRT-PCR was performed using TB Green Premix Ex TaqII (TaKaRa) and CFX96 real-time PCR detection system (Bio-Rad). Three technical and three biological replicates were used. The data were analyzed using the 2^−ΔΔCT method (Livak and Schmittgen, 2001). The common primers were designed according to the consensus sequences, while the specific primers (TaAGL6AF/TaAGL6AR, TaAGL6BF/TaAGL6BR and TaAGL6DF/TaAGL6DR) for analyzing the expression of every TaAGL6 gene were designed according to the differences between the full cDNA sequences, specifically a 5 bp insertion into the 5′ UTR of TaAGL6-A, and a 9 bp deletion in the 3′UTR of TaAGL6-D. The specificity of each pair of primers was verified by sequencing the RT-PCR products (data not shown). The primers are listed in Table S1.

CS inflorescences at different stages were fixed in FAA (50% ethanol, 5% acetic acid and 3.7% formaldehyde) overnight at 4°C. The materials were then treated and the experiments were performed as described previously (Li et al., 2011b; Tao et al., 2018). The probes labeled with digoxigenin were prepared using the in vitro transcription kit (Roche), according to the protocol and previous studies (Li et al., 2011b; Tao et al., 2018). The probes were amplified with primers TaAGL6RNAiPF/TaAGL6T7-R (henceforth referred to as the ‘antisense probe’) and TaAGL6T7-F/TaAGL6RNAiPR (henceforth referred to as the ‘sense probe’) (Table S1).

To construct the overexpression vectors, one pair of primers TaAGL6OF/TaAGL6OR (Table S1) was designed to amplify the coding region, because of the identical sequence after the ATG and before the TGA in three genes (Fig. S1). The PCR products were then cloned into pMD18-T vectors, three different cDNAs were selected by sequencing different clones and the encoding TaAGL6 cDNAs were cloned into the pCAMBIA1301 vectors (TaKaRa) with NcoI and BglII sites under the control of the 35S promoter. The dsRNAi vector was constructed according to a previous study (Li et al., 2010). Wheat was transformed via gene gun bombardment, according to a previous study (Zhang et al., 2015). Transformation of Arabidopsis was mediated by Agrobacterium tumefaciens GV3101.

The materials were fixed in FAA at 4°C overnight, dehydrated in graded ethanol then xylene. They were then embedded in Paraplast Plus (Sigma-Aldrich) and cut into 8 μm sections, which were stained with 0.2% Toluidine Blue and photographed using a Nikon E600 microscope and a Nikon DXM1200 digital camera. Scanning electron microscopy (SEM) images were obtained using a JSM-6360LV (JEOL) scanning electron microscope, as described previously (Liu et al., 2016).

Y2H assays were performed using the GAL4-based two-hybrid system (Clontech). Self-activation assays of three TaAGL6, TaAP3, TaAG and TaMADS13 were performed according to the protocol. The coding sequence (CDS) of three TaAGL6 genes was cloned in the frame of pGADT7-Rec to generate pGAD-Preys. The coding regions of TaAP3, TaAG and TaMADS13 were cloned in the frame of pGBKT7 to generate pGBK-Baits. The different combinations of pGAD-Preys and pGBK-Baits were co-transformed into Y2HGold yeast cells (Clontech). Positive clones on SD/-Trp/-Leu solid medium were transferred to SD/-Trp-Leu-His or SD/-Trp-Leu-His-Ade liquid medium to detect the interactions. Clones harboring pGADT7-TaAGL6 and pGBKT7 were used as the negative control. The primers are listed in Table S1.

BiFC vectors pSPYNE and pSPYCE were used to construct nYFP-TaAGL6-A/B/D, TaAP3-cYFP, TaAG-cYFP and TaMADS13-Cyfp, with the one-step cloning kit (Yeasen Biotech). Agrobacterium GV3101 cells harboring different nYFP-TaAGL6 and different cYFP constructs were infiltrated into N. benthamiana leaves. YFP fluorescence was observed under a confocal laser scanning microscope (Olympus FV10 ASW). Three independent N. benthamiana leaves were observed for the analysis of every interaction.

The partial cDNA (1-507 bp) of TaAGL6-B was amplified with the primers TaAGL6-BGSTF and TaAGL6-BGSTR (Table S1), and inserted into EcoRI and XhoI sites of the pGEX-6p-1 vector to express the fusion protein glutathione-S-transferase (GST)-TaAGL6-B. The construct was then introduced into BL21 (DE3) pLysS E. coli cells. Positive clones were cultured in Luria-Bertani (LB) medium containing 50 mg/ml ampicillin at 37°C to OD₆₀₀=0.6. Expression of the fusion protein was induced by adding 1 mM isopropyl β-d-1-thiogalactopyranoside. The fusion proteins were purified using glutathione-agarose beads (Sangon Biotech), according to the manufacturer's instructions.

The purified proteins were separated on 12% SDS-PAGE gels and transferred to a polyvinylidene difluoride (0.45 μm) membrane (Millipore, cat: IPVH00010F). After blocking for 3 h in Tris-buffered saline Tween (TBST) with 5% nonfat milk, the membrane was washed with TBST three times (5 min each time) and incubated with the GST antibody at a dilution of 1:1000 (D190101, Sangon Biotech) at 4°C overnight. Secondary goat anti-mouse IgG conjugated with horseradish peroxidase (D110087, Sangon Biotech) was then incubated with the membrane for 2 h at a dilution of 1:5000. The target protein bands were visualized using a chemiluminescence reagent (Beyotime Biotechnology) (Song et al., 2018).

The fragments including the CArG cis-element in the TaAP3 promoter were amplified using PCR with biotin-labeled or non-labeled primers (Table S1). The biotin-labeled DNA probes were incubated with the TaAGL6-B-GST protein, and competing experiments were performed by adding excess non-labeled probes. Biotin-labeled probes incubated with GST protein served as the negative control. EMSA assays were performed using the LightShift Chemiluminescent EMSA Kit (20148, Thermo), according to the manufacturer's instructions. Briefly, the reaction mixture [in 20 μl: 10 μg purified fused protein or GST, 100 fmol biotin end-labeled probes, 2 μl 10× binding buffer, 1 μl of 1 µg/μl poly (dI·dC), 1 μl 50% glycerol, 1 μl 1% NP-40, 1 μl of 1 M KCl, 1 μl of 100 mM MgCl₂, 0.5 μl of 200 mM EDTA and double-distilled water] was incubated for 30 min at room temperature for the binding reaction and electrophoresed on a 10% native polyacrylamide gel. The proteins were then transferred to a nylon membrane (S4056, Millipore) in 0.5× TBE buffer at 380 mA for 45-60 min, and the binding between the proteins and biotin-labeled probes was detected by chemiluminescence (Feng et al., 2014).

For the Y1H assay, three copies of the CArG element were cloned into the pAbAi vector to create the bait construct. The CDS of three TaAGL6 genes were fused to the GAL4 AD in the pGADT7 vector to generate the prey constructs AD-TaAGL6-A, AD-TaAGL6-B and AD-TaAGL6-D. The prey vector and the empty vector (AD), serving as the negative control, were then transformed separately into yeast cells containing bait constructs, with the primers TaAP3F/TaAP3R and Mut TaAP3F/Mut TaAP3R (Table S1). The transformed yeast cells were diluted with a 10× dilution series and dotted on the SD plates lacking Leu and Ura with 200 ng/ml (optimized according to the protocol) Aureobasidin A. The binding between prey protein and bait sequence is judged by the growth of cells harboring bait construct and prey construct.

To generate reporter construct, a 1746 bp region upstream of TaAP3-B start codon was amplified and cloned into a pGreenII 0800-LUC vector (Hellens et al., 2005). To create an effecter construct, TaAGL6-B CDS were cloned into pGreenII 62-SK vector (Hellens et al., 2005). The recombinant effecter and reporter plasmids were transfected into A. tumefaciens strain GV3101 (pGreenII series holding psoup plasmid) separately and co-infected into tobacco leaves. A dual-luciferase reporter assay system (Promega) was used to measure firefly LUC and renilla luciferase (REN) activities. The REN gene under the control of the CaMV 35S promoter and the LUC gene were in the pGreenII 0800-LUC vector (Hellens et al., 2005). Relative REN activity was used as an internal control, and LUC/REN ratios were calculated. At least three assay measurements were performed for each assay.

The HA-tag vector Gold and GFP-tag vector pCAMBIA1302 were used for co-immunoprecipitation analysis. Full-length CDSs of TaAGL6-B, TaAP3, TaAG and TaMADS13 were amplified and inserted into the tag vectors. Agrobacterium strains carrying TaAP3-GFP, TaAG-GFP or TaMADS13-GFP were co-infiltrated into tobacco leaves with an Agrobacterium strain carrying TaAGL6-B-HA and P19 silencer, respectively. After 3 days, leaves were harvested, frozen in liquid nitrogen and homogenized with protein extraction buffer. Homogenates were centrifuged (14,000 g) for 15 min at 4°C. The expressed HA-TaAGL6 and TaAP3-GFP, TaAG-GFP, TaMADS13-GFP or GFP proteins, were immunoprecipitated with GFP-Trap MA beads (ChromoTek) at 4°C for 2 h. The immunoprecipitated proteins were washed four times with the lysis buffer and then eluted by boiling for 5 min with 2× loading buffer. Immunoblots were detected by performing western blotting with an anti-GFP antibody (D190750-0100; Sangon; 1:1000), an anti-HA antibody (AH158; Beyotime; 1:1000) and the secondary goat anti-mouse IgG conjugated with horseradish peroxidase (D110087, Sangon Biotech; 1:5000). Total proteins (50 mg) extracted from tobacco leaves co-expressing HA-TaAGL6 and TaAP3-GFP, TaAG-GFP, TaMADS13-GFP or GFP were used as input control (Li et al., 2019). Western blotting was carried out as described above.

All statistical analyses were performed using Student's t-test in the Statistical Product and Service Solutions (SPSS) software. The differences were examined using Student's t-test and the significance level was set at 0.05 (P<0.05).

TaAGL6-A, TaAGL6-B and TaAGL6-D are in the GRAMENE database (www.gramene.org) under TraesCS6A02G259000, TraesCS6B02G286400 and TraesCS6D02G240200, respectively.

We appreciate the kind help from Dr Dejun Han and Dr Yu Wang at Northwest A&F University (Shaanxi, China).