•Family 1 glycosyltransferases comprise the greatest number of glycosyltransferases found in plants. The widespread occurrence and diversity of glycosides throughout the plant kingdom underscore the importance of these glycosyltransferases.
•Here, we describe the identification and characterization of a late-flowering Arabidopsis (Arabidopsis thaliana) mutant, in which a putative family 1 glycosyltransferase gene, UGT87A2, was disrupted. The role and possible mechanism of UGT87A2 in the regulation of flowering were analyzed by molecular, genetic and cellular approaches.
•The ugt87a2 mutant exhibited late flowering in both long and short days, and its flowering was promoted by vernalization and gibberellin. Furthermore, the mutant flowering phenotype was rescued by the wild-type UGT87A2 gene in complementation lines. Interestingly, the expression of the flowering repressor FLOWERING LOCUS C was increased substantially in the mutant, but decreased to the wild-type level in complementation lines, with corresponding changes in the expression levels of the floral integrators and floral meristem identity genes. The expression of UGT87A2 was developmentally regulated and its protein products were distributed in both cytoplasm and nucleus.
•Our findings imply that UGT87A2 regulates flowering time via the flowering repressor FLOWERING LOCUS C. These data highlight an important role for the family 1 glycosyltransferases in the regulation of plant flower development.
The photoperiod, one of the most important environmental factors affecting the floral transition, is mediated by the interactions between environmental light signals and intrinsic time-keeping mechanisms associated with the circadian clock (Doyle et al., 2002; Yanovsky & Kay, 2002; Hayama & Coupland, 2003). CRYPTOCHROME1 (CRY1), CRY2, PHYTOCHROME A (PHYA), PHYB and CO have been shown to be the key upstream genes in the photoperiod pathway regulating the expression of integrators (Lin, 2000; Doyle et al., 2002; Yanovsky & Kay, 2002; Hayama & Coupland, 2003). The genes in the autonomous pathway (e.g. FCA, FPA, FVE, FLOWERING LOCUS D (FLD), LUMINIDEPENDENS (LD) and FY) promote flowering by the suppression of FLC through chromatin or RNA modification. The vernalization response of flowering is governed by dominant alleles of two genes, FRIGIDA (FRI) and FLC (Michaels & Amasino, 2001). FRI acts as a scaffold protein to assemble a large protein complex (FRI-C) with other regulators, and to yield the FRI-C-mediated transcriptional activation of FLC (Choi et al., 2011). Vernalization genes (VRNs) (e.g. VRN1, VRN2 and VRN3) are another important type of gene involved in the vernalization pathway, and act as inhibitors of FLC expression through chromatin modification mechanisms (Levy et al., 2002; Bastow et al., 2004; He & Amasino, 2005). GA promotes flowering and is absolutely required in noninductive short days. Mutations in the GA biosynthesis pathway and signaling pathway (e.g. GA1, GAI (GA INSENSITIVE) and RGA (REPRESSOR OF ga1-3)) result in changes in flowering time (Mouradov et al., 2002; Olszewski et al., 2002). SPINDLY (SPY) is a negative regulator of GA responses. The spy mutation causes constitutively active GA signaling and can partially suppress the effects of disrupted GA biosynthesis (Jacobsen et al., 1996).
Glycosyltransferases (GTs), the enzymes responsible for the glycosylation of plant compounds, are believed to play a very important role in the production of a range of plant secondary compounds. At the time of writing, 94 distinct GT families in the biosphere have been identified and described in the CAZy (carbohydrate active enzyme) database (http://www.cazy.org/GlycosylTransferases.html). Family 1 contains the largest number of GTs found in plants, and most contain a carboxy-terminal consensus sequence, termed the ‘plant secondary product glycosyltransferase box’ (PSPG box). Family 1 GTs usually recognize substrates of low-molecular-weight secondary metabolites and use uridine 5’-diphospho sugars as the sugar donors, and are thus termed ‘UDP-sugar GTs’ (UGTs) (Jones & Vogt, 2001; Bowles et al., 2005; Wang & Hou, 2009).
The genome sequencing of Arabidopsis thaliana has revealed that GTs of secondary metabolites are encoded by an unexpectedly large multigene family of 120 members. It is now recognized that the glycosylation of secondary metabolites of plants, by adding a sugar moiety to the acceptors, changes their chemical properties, alters their bioactivity and enables access to membrane transporter systems; thus, GTs might play an important role in the maintenance of cell homeostasis and regulation of plant growth, development and defense responses (Jones & Vogt, 2001; Lim & Bowles, 2004). For instance, several UGTs have been shown to glycosylate phytohormones in vitro and to be involved in phytohormone homeostasis in vivo (Jackson et al., 2001, 2002; Hou et al., 2004; Lim et al., 2005; Poppenberger et al., 2005; Wang et al., 2011). A novel role of UGTs in plant–fungal interactions and defense responses has also been reported (Chong et al., 2002; Poppenberger et al., 2003). In addition, UGTs have been implicated in the detoxification of xenobiotics, including herbicides, pesticides and pollutants (Loutre et al., 2003; Messner et al., 2003). Family 1 GTs are a large multigene family and glycosides of small molecules are their products. The widespread occurrence, diversity and complexity of the glycosides throughout the plant kingdom indicate that family 1 GTs might have a broad functionality. Although some UGTs have been functionally characterized, very little is known about actual biological functions and molecular mechanisms for most family 1 GTs in planta, particularly during plant growth and development.
In the present study, we isolated a late-flowering Arabidopsis (Arabidopsis thaliana) mutant in which UGT87A2, encoding a putative family 1 GT, was disrupted by T-DNA insertion. The mutant late-flowering phenotype was rescued by transgenic expression of wild-type UGT87A2 in complementation lines. The mutant late flowering was not affected by long or short days, but positively responded to vernalization and GA treatments. Furthermore, the molecular mechanisms resulting in mutant late flowering were investigated by quantitative reverse transcription-polymerase chain reaction (qRT-PCR) analysis of flowering-related genes. The expression of the flowering repressor FLC was increased substantially in the mutants, but decreased to wild-type levels in complementation lines. Accordingly, the expression levels of the floral integrators (FT and SOC1) and floral meristem identity genes (AP1 and LFY) were also affected in an oppositely manner in mutants and complementation lines. Finally, we found that the expression of UGT87A2 was developmentally regulated and its protein products were distributed in both cytoplasm and nucleus. Our results suggest that UGT87A2, an Arabidopsis family 1 GT gene, regulates flowering time via the flowering repressor FLC.
Materials and Methods
Plant materials and growth conditions
Wild-type Arabidopsis (Arabidopsis thaliana) plants and all T-DNA insertion mutants used in this work, which were obtained from the Nottingham Arabidopsis Stock Centre (NASC), Nottingham, UK, were of the Col-0 accession. Mutant plants were confirmed by PCR-based genotyping.
Arabidopsis plants were grown on Nutrition Soil (Shangdao Biotech Co. Ltd., Shandong, China) with vermiculite (Nutrition Soil : vermiculite, 2 : 1) or Murashige and Skoog (MS) basal medium plates containing 3% sucrose and 0.7% agar at 22 ± 2°C with a light intensity of c. 100 μmol m−2 s−1. Long day (LD) is defined as 16 h of light and 8 h of darkness and short day (SD) as 8 h of light and 16 h of darkness.
Plasmid construction and plant transformation
The full-length cDNA and the full-length promoter of the UGT87A2 gene (At2g30140) were amplified and cloned into the pBluescript SK vector. After sequence confirmation, the UGT87A2 cDNA was cloned into the pBI121 vector to replace the glucuronidase (GUS) gene and yield the cauliflower mosaic virus (CaMV) 35S promoter-driven over-expression construct used for the generation of complementation lines. The promoter of UGT87A2 was also cloned into the pBI121 vector to replace the CaMV 35S promoter and yield the UGT87A2 promoter::GUS fusion construct used for the generation of GUS transgenic lines. In addition, the full-length UGT87A2 cDNA without stop codon was cloned into the p326-SGFP vector, in which the coding sequence of the UGT87A2 gene was fused to the 5’-terminus of the green fluorescent protein (GFP) gene in frame, driven by the CaMV 35S promoter. In all experiments, DH5α was used for subcloning, and the Agrobacterium strain GV3101 was used for Arabidopsis plant transformation through the floral dip method (Clough & Bent, 1998). The GFP fusion construct and the control vector with GFP alone were transformed into onion epidermis cells using a particle gun (PDS-1000/He, Bio-Rad, USA).
Flowering time and other phenotype measurements
Plants were grown in soil in LD or SD conditions. Flowering time was measured by counting the number of rosette leaves at flowering and the days from sowing to floral bud formation. The sizes of the leaves, flowers and seeds of mutants and wild-type plants were measured when the organs to be compared were at the same developmental stage. At least 30–50 plants were measured and averaged for each measurement and statistical treatment.
Vernalization and GA treatments
For vernalization treatments, plants were germinated and grown at 4°C for 6 wk in LD conditions and transferred to normal growth conditions (22 ± 2°C). To examine GA effects, plants were grown on soil in LD conditions, and were sprayed with a 20 μM GA solution twice a week until flowering.
Total RNA extraction and qRT-PCR
To study the expression level of flowering-related genes, 3- and 4-wk-old plants were harvested by pooling aboveground tissues of seedlings. Plants were only used once for tissue harvest to avoid gene expression alterations as a result of wound effects, and plant tissues were collected at 4 h after the growth chamber lights had been turned on. Total RNA was extracted using Trizol reagent (TaKaRa, Shiga, Japan) and treated with RNase-free DNase I (TaKaRa, Shiga, Japan). For first-strand cDNA synthesis, 500 ng of RNA was used as a template, employing the PrimeScript RT reagent kit (DRR037, TaKaRa, Shiga, Japan), and the PCR comprised 24–37 cycles of 30 s at 94°C, 30 s at 55°C and 50 s at 72°C. qRT-PCR was performed using the SYBR Premix Ex Taq II kit (DRR081, TaKaRa, Shiga, Japan) according to the manufacturer’s instructions and run on a LightCycler480 (Roche) real-time PCR system. Thermocycling conditions were 95°C for 30 s, followed by 40 cycles of 95°C for 5 s, 60°C for 20 s, with a melting curve detected at 95°C for 15 s and 65°C for 15 s. The relative transcript level was normalized with Tubulin2 gene (abbreviated as TUB in this article) according to the method (Livak & Schmittgen, 2001). All primers used in qRT-PCR and semi-quantitative RT-PCR are listed in Supporting Information Tables S1, S2.
For the complementation of the ugt87a2 mutant, UGT87A2 cDNA was amplified through RT-PCR with the primer pair 5′-GCGTCTAGAATGGATCCAAATGAATCTCC-3′ and 5′-GCGGAGCTCTTAATTTGTATTGGTAATAT-3′. The amplified fragment was cloned into the pBI121 vector and transferred into mutant plants as described in the ‘Plasmid construction and plant transformation’ section. The transgenic plant lines were selected on MS medium with 40 μg ml−1 of kanamycin. Kanamycin-resistant and PCR-positive transgenic plants were transferred to a glasshouse and maintained up to the T2 generation. At least two independent homozygous lines were then selected and used for further analysis.
Tissue-specific expression of UGT87A2
A GUS-coding sequence was transcriptionally fused to the UGT87A2 promoter. The fusion vector was transferred into Arabidopsis plants, and homozygous plants were subjected to GUS staining. Samples were collected in microcentrifuge tubes and placed in 90% acetone on ice for 15 min, and then washed with staining buffer (50 mM sodium phosphate, pH 7.2, 0.2 mM K3Fe(CN)6, 0.2 mM K4Fe(CN)6 and 0.2% Triton X-100) without X-Gluc (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) twice before adding staining buffer with X-Gluc (final concentration of 2 mM) and incubation overnight at 37°C. The samples were washed in 70% ethanol for 30 min before observation under a dissecting microscope (Olympus, Tokyo, Japan).
Analyses of the cellular location of GFP-fused proteins
The recombinant plasmid containing the UGT87A2–GFP fusion gene and the control plasmid with GFP alone were transformed into onion epidermal cells using a particle gun (PDS-1000/He, Bio-Rad). After 24 h of incubation on MS medium, GFP fluorescence in transformed onion cells was observed under a Nikon fluorescence microscope (Nikon, New York, USA). Photographs were taken with > 10 cells for each plant material in which GFP fluorescence was observed, and all cells showed a nearly identical pattern representing the localization of the fusion protein.
Experiments were repeated at least three times. P values were determined by Student’s t-test (quantification of flowering-related gene expression) or by one-way ANOVA using the protected least-significant difference (LSD) test (quantification of phenotypic differences). Data are expressed as mean values ± SE.
ugt87a2 was late flowering
With the aim of exploring the functions of GTs in the regulation of plant growth and development, we obtained many mutants of putative GTs from NASC. One mutant, ugt87a2, drew our attention because of its severe late-flowering phenotype as measured by the days to bolting and the rosette leaf number at flowering time. Under normal growth conditions (LD photoperiod), the wild-type plants bolted nearly 26 d after sowing, compared with 45 d for mutant plants. In order to confirm that the late-flowering phenotype was caused by a mutation in the UGT87A2 gene, we identified two independent UGT87A2 T-DNA insertion homozygous mutants (SALK_124038 and SALK_061574), one with the T-DNA insertion in the first exon and the other in the 5’-untranslated region (5’-UTR) (Fig. 1a). The suppression of UGT87A2 expression in these two mutants, named ugt87a2-1 and ugt87a2-2, respectively, in this study, was confirmed by RT-PCR assays (Fig. 1b). When grown under normal growth conditions, the two independent mutations exhibited a similar late-flowering phenotype to that mentioned above (Fig. 1c). Therefore, the gene UGT87A2 was probably related to flowering time control.
Other phenotype changes were also observed in ugt87a2 mutants. The leaves, flowers and flower organs of ugt87a2 mutants were much larger than those of the wild-type (Fig. 2a–d; Table S3), and the numbers of flower organs (sepals, petals, stamens and pistils) were the same as those of the wild-type. At the silique ripening stage, ugt87a2 mutants were taller than the wild-type plants (Fig. 2e). Consistent with the larger flower organs, the seed size and thousand-seed weight of ugt87a2 mutants were much greater than those of the wild-type (Fig. 2f; Table S3).
UGT87A2 encoded a putative family 1 GT
The UGT87A2 gene consists of two exons and is located in chromosome 2. It encodes a protein of 455 amino acid residues with a predicted molecular mass of 51.7 kDa. There was a consensus motif of 44 amino acids (PSPG motif) in the C-terminus of deduced sequences (Fig. 3a) – characteristic of plant secondary product GTs (Wang & Hou, 2009) – suggesting that UGT87A2 is probably a member of family 1 GTs. It is known that Arabidopsis possesses > 100 family 1 GTs. The phylogenetic tree generated from 107 members of Arabidopsis family 1 GTs (Ross et al., 2001) showed that UGT87A2 was placed within a small clade containing five other UGTs (Fig. 3b). However, the six UGTs in this clade belong to four different groups, i.e. group K (UGT86A1 and UGT86A2), group J (UGT87A1 and UGT87A2), group I (UGT83A1) and group N (UGT82A1). To date, no function or enzymatic activity has been identified for these six UGTs.
Flowering phenotype was restored in complementation lines of UGT87A2
To further verify that the mutant phenotype was indeed caused by the loss-of-function of UGT87A2, two independent complementation lines, ugt87a2-1/UGT87A2 and ugt87a2-2/UGT87A2, were generated by 35S promoter-driven UGT87A2 expression in the mutant background. RT-PCR analysis showed that the expression levels of UGT87A2 in the two complementation lines were greatly increased relative to that in the ugt87a2 mutant, but were comparable with wild-type expression (Fig. 4a). Correspondingly, the two complementation lines showed much earlier flowering than the ugt87a2 mutant, but a similar flowering time to the wild-type (Fig. 4b), suggesting that the UGT87A2 transgene restored wild-type flowering to the mutant. These results provide further evidence that the GT gene UGT87A2 is an important factor in the regulation of plant flowering time. In addition to the flowering time, the present study showed that other phenotypes of complementation lines, such as leaf number, plant height and flower size, were also restored to those of the wild-type. However, the seed size and thousand-seed weight were incompletely restored (Table S3).
ugt87a2 was late flowering in both LD and SD
To determine whether the late-flowering phenotype of mutants was related to the photoperiod pathway, the wild-type and mutant plants were grown under LD or SD conditions to determine their flowering time. Under LD, the wild-type plants started to produce inflorescences c. 26 d after sowing, compared with 45 d for ugt87a2 mutants. Under SD, ugt87a2 mutants flowered much later than did wild-type plants (Fig. 5a). It is well established that late-flowering plants form more leaves at flowering time (Koornneef et al., 1991) – in the present study, the ugt87a2 mutants had c. 24% more rosette leaves at flowering time compared with the wild-type, whether grown under LD or SD (Fig. 5a). Thus, the flowering of ugt87a2 mutants was delayed in SD and LD in terms of both the days to flowering and the total rosette numbers at flowering initiation. This observation suggests that UGT87A2 is not involved in the photoperiod pathway of flowering control, but is possibly involved in the autonomous flowering pathway.
ugt87a2 was responsive to vernalization and GA treatments
To examine the responsiveness of ugt87a2 to vernalization, the ugt87a2 plants were germinated and grown at 4°C for 6 wk and transferred to normal growth temperature. The vernalization-treated ugt87a2 plants flowered much earlier than did untreated plants, when the total leaf number was 13–14 (Fig. 5b), although this was still later than for wild-type plants. The untreated mutant plants flowered at a rosette leaf number of 16–18. The time to mutant flowering was also shortened by vernalization. The incomplete recovery of flowering time by vernalization may be caused by vernalization having two mechanisms to promote flowering: FLC-dependent and FLC-independent pathways (Michaels & Amasino, 2001).
To analyze the effects of GA on the flowering of mutants, a 20 μM GA solution was sprayed twice a week on growing plants until they flowered. GA greatly stimulated the flowering of ugt87a2 mutants (Fig. 5c). The ugt87a2 plants initiated flowering at a total leaf number of 11–12 under GA treatment, which was comparable with that of the wild-type. These observations indicate that ugt87a2 mutants positively respond to both vernalization and GA. Because the later flowering of autonomous pathway mutants can be reversed by vernalization (Martinez-Zapater & Somerville, 1990; Koornneef et al., 1991) or by GA (Mouradov et al., 2002; Moon et al., 2003), our observations are consistent with the possibility that UGT87A2 is involved in the autonomous pathway.
Expression of the flowering repressor FLC was up-regulated in ugt87a2 mutants
To investigate the molecular mechanism underlying the later flowering phenotype of ugt87a2 mutants, we compared the expression level of several key genes specifically involved in the photoperiod (CRY1, CRY2 and CO), autonomous (FVE, FCA and FPA), vernalization (VRN1 and VRN3) and GA (GAI, RGA and SPY) pathways, and those genes universally regulated by two (FLC) or more (SOC1, FT, AP1 and LFY) pathways between the mutants and wild-type plants. The expression of the genes specifically involved in the four pathways was not altered in ugt87a2 mutants (Figs 6, S1). However, the expression of those genes universally regulated by different pathways (i.e. FLC, FT, SOC1, AP1 and LFY) was changed substantially in the ugt87a2 mutants compared with wild-type plants. FLC is a strong flowering repressor (Michaels & Amasino, 1999). In ugt87a2 mutants, the FLC transcript level was up-regulated dramatically, which is in good agreement with the late-flowering phenotype. Accordingly, the two floral integrators, FT and SOC1, which are down-regulated by FLC, were suppressed significantly in the mutants; eventually, the expression of the floral meristem identity genes AP1 and LFY was also inhibited. Together, the data of the present study indicate that the loss of function of UGT87A2 in the mutants primarily results in increased expression of FLC and, consequently, the down-regulation of FT, SOC1, AP1 and LFY, and thus a late-flowering phenotype.
Expression of FLC was suppressed in the complementation lines
The expression levels of flowering-related genes were also tested for the ugt87a2 complementation lines. Our data showed that the expression of the floral repressor FLC was suppressed substantially in the two independent complementation lines when compared with the mutants, but was comparable with the expression level of wild-type plants (Figs 6, S2). By contrast, the expression of the floral integrators FT and SOC1, and the floral meristem identity genes AP1 and LFY, was increased in complementation lines to levels comparable with those in wild-type plants. These results indicate that the over-expression of a wild-type copy of UGT87A2 in the mutants can recover the flowering phenotype through the recovery of the expression of FLC and its downstream flowering-related genes, suggesting an important role of UGT87A2 in the regulation of the flowering time.
Expression of UGT87A2 was developmentally regulated
Histochemical staining for GUS activity of the transgenic lines showed strong expression of UGT87A2 in germinating seed, cotyledons and primary roots (Fig. 7a–c), but weak expression in the hypocotyls (Fig. 7b,c), of early seedlings. UGT87A2 was also expressed in roots, leaf apices, leaf margins and shoot apices of 1–2-wk-old seedlings (Fig. 7d,e). UGT87A2 expression was very low in 4-wk-old plants (data not shown). In the reproductive stage, UGT87A2 was highly expressed in sepals, anther filaments, stigma, pedicel–silique junction and capsule (Fig. 7f–i). All of these results indicate that UGT87A2 expression is spatiotemporally regulated.
Subcellular localization of UGT87A2
To determine the subcellular localization of UGT87A2 protein, the vector 35Spro::UGT87A2-GFP was transformed into onion epidermis. The distribution of green fluorescence signals of the fusion protein transiently expressed in cells indicated that UGT87A2 protein was typically located in both cytoplasm and nucleus (Fig. 8).
The late-flowering phenotype of mutant ugt87a2 and the phenotypic rescue by UGT87A2 expression clearly indicate that the GT UGT87A2 is a novel factor regulating the flowering time. It has been established that FLC is a strong flowering repressor, and both the autonomous and vernalization pathways act to repress FLC expression to induce flowering (Bäurle & Dean, 2006). FT and SOC1 encode for flowering activators and are positioned downstream of FLC in flowering signal transduction. Our data showed that FLC expression is elevated in the mutant, but that the expression of other upstream flowering genes is unaffected. Thus, it seems likely that the loss of UGT87A2 enzyme activity, in an as yet unknown way, contributes to the increased expression level of FLC, which then causes late flowering by decreased expression of FT and SOC1, and, eventually, the reduced expression of floral meristem identity genes AP1 and LFY in ugt87a2 plants.
Previous studies have revealed four major flowering pathways: photoperiod, autonomous, vernalization and GA pathways (Mouradov et al., 2002; Boss et al., 2004; Henderson & Dean, 2004). Mutations in genes that cause a late-flowering phenotype have been grouped into two classes based on details of their phenotype. One group of mutations results in day length-dependent delayed flowering, suggesting a compromise in the responsiveness to day length in these mutants. These mutations define genes in a photoperiod pathway. A second class of mutations results in day length-independent late flowering (Koornneef et al., 1991). In effect, these mutations are ‘blind’ to the inductive photoperiod and show late flowering in both LD and SD. Therefore, these mutations define genes in a pathway often designated as autonomous to indicate that it is distinct from the photoperiod pathway. The autonomous pathway mutants are highly sensitive to vernalization (Koornneef et al., 1991; Martínez-Zapater et al., 1995) and their delayed flowering can also be reversed by GA (Mouradov et al., 2002; Moon et al., 2003). ugt87a2 mutant plants flowered much later than did wild-type plants in both LD and SD photoperiods, and their flowering could be promoted by exposure to a prolonged period of cold or GA treatment. This suggests that UGT87A2 is a novel and important player in the autonomous pathway, and we propose a UGT87A2 working model (Fig. 9).
The genes in the autonomous pathway, for example, FLD and FVE, promote flowering by suppressing FLC. Mutations in FLD and FVE result in hyperacetylation of histones in FLC chromatin, up-regulation of FLC expression and delayed flowering, indicating that the autonomous pathway regulates flowering, in part, by histone modification (He et al., 2003; Ausin et al., 2004; Kim et al., 2004). The data of the present study indicate that the ugt87a2 mutants delayed flowering in a similar manner to the mutants in the autonomous pathway, in which the expression of FLC was increased substantially. Interestingly, the data showed that the ugt87a2 mutants did not influence the expression levels of autonomous pathway upstream genes, such as FVE, LD, FCA and FPA, which generally regulate FLC through histone or RNA modifications. These results may imply that UGT87A2 does not regulate FLC through these typical autonomous pathway genes. It is not known how UGT87A2 down-regulates the FLC transcript level. However, given that UGT87A2 encodes a putative family 1 GT, which would glycosylate certain plant compounds, we speculate that the glycosylation of certain molecule(s) may play a key role in the regulation of FLC transcription. Further analysis of the GT UGT87A2 will reveal the nature of glycosylation and the mechanisms of UGT87A2 affecting FLC transcription activity.
This research was supported by research grants from the National Natural Science Foundation of China (NSFC Grant Nos. 90917006, 30971543 and 30770214 to B.K.H.). We thank the Nottingham Arabidopsis Stock Centre for providing T-DNA insertion mutant seeds SALK_124038 and SALK_061574.