Gibberellin Acts through Jasmonate to Control the Expression of MYB21, MYB24, and MYB57 to Promote Stamen Filament Growth in Arabidopsis

Precise coordination between stamen and pistil development is essential to make a fertile flower. Mutations impairing stamen filament elongation, pollen maturation, or anther dehiscence will cause male sterility. Deficiency in plant hormone gibberellin (GA) causes male sterility due to accumulation of DELLA proteins, and GA triggers DELLA degradation to promote stamen development. Deficiency in plant hormone jasmonate (JA) also causes male sterility. However, little is known about the relationship between GA and JA in controlling stamen development. Here, we show that MYB21, MYB24, and MYB57 are GA-dependent stamen-enriched genes. Loss-of-function of two DELLAs RGA and RGL2 restores the expression of these three MYB genes together with restoration of stamen filament growth in GA-deficient plants. Genetic analysis showed that the myb21-t1 myb24-t1 myb57-t1 triple mutant confers a short stamen phenotype leading to male sterility. Further genetic and molecular studies demonstrate that GA suppresses DELLAs to mobilize the expression of the key JA biosynthesis gene DAD1, and this is consistent with the observation that the JA content in the young flower buds of the GA-deficient quadruple mutant ga1-3 gai-t6 rga-t2 rgl1-1 is much lower than that in the WT. We conclude that GA promotes JA biosynthesis to control the expression of MYB21, MYB24, and MYB57. Therefore, we have established a hierarchical relationship between GA and JA in that modulation of JA pathway by GA is one of the prerequisites for GA to regulate the normal stamen development in Arabidopsis.


Introduction
Arabidopsis flowers are organized into four concentric whorls of distinct organs (sepals, petals, stamens and pistils). Stamens, the male reproductive organs of flowering plants, form the third whorl. Processes of stamen filament elongation and anthesis are precisely controlled so that they coincide with the pistil development to determine the fertility [1]. Mutations that impair stamen development such as filament elongation, pollen maturation or anther dehiscence will result in male sterility [2,3]. Many genes have been found to control stamen development [4,5]. Stamen development is also subjected to hormonal control. For example, mutations affecting biosynthesis of two plant hormones gibberellin (GA) (e.g ga1-3 mutation) and jasmonate (JA) (e.g opr3 mutation) both confer male sterile phenotype due to failure of stamen filament elongation and of completion of anthesis and anther dehiscence [6,7].
The development of floral organs, especially petals and stamens, is impaired in the ga1-3 mutant. Detailed anatomical analysis showed that the male sterile phenotype of ga1-3 is due to the arrestment of stamen filament cell elongation and failure of completion of anthesis [8]. Application of exogenous GA can restore all the floral defects of ga1-3 [7]. Further studies revealed that the arrested floral development in ga1-3 is mediated by DELLA proteins [8,9]. DELLAs are a subfamily of the plant GRAS family of putative transcription regulators [10,11] and have been revealed to function as negative regulators of GA response in diverse plant species including Arabidopsis, barley, rice and wheat etc [12][13][14][15][16][17]. There are five DELLAs in Arabidopsis, namely GAI, RGA, RGL1, RGL2 and RGL3 [18,19]. Genetic studies have shown that RGA, RGL2 and RGL1 act synergistically in repressing petal and stamen development and GA triggers the degradation of these DELLAs to promote floral development [8,9,[20][21][22]. Severe JA deficient mutant opr3 and JA-signaling mutant coi1 also displayed retarded filament elongation, delayed anther dehiscence, and reduced pollen viability. As a consequence, the opr3 and coi1 mutants are male sterile [6,23]. Application of exogenous JA can fully restore the stamen development to opr3 [6].
It is intriguing to know whether GA-mediated and JA-mediated stamen development are via two parallel pathways or in a hierarchical way to control stamen development. In Arabidopsis, the known GA-response genes encoding transcription factors involved in stamen development are GAMYBs (MYB33 and MYB65), a subset of MYB genes [24]. GAMYB is the best characterized GAregulated transcription factor and was first identified in barley. GAMYB was found to bind to the GA-response elements (GARE) in the promoter of the a-amylase gene in cereals [25,26]. Genetic studies showed that Arabidopsis GAMYBs (MYB33 and MYB65) are essential to anther maturation but not for the elongation of stamen filament in Arabidopsis [24]. Previous studies have shown that GA regulates GAMYB through DELLA protein SLN1 and SLR1 in barley and rice, respectively [27,28]. However, several reports failed to identify MYB33 and MYB65 as GA-inducible genes in Arabidopsis and these two MYB genes are in fact regulated at the post-transcriptional level by miRNA159 [24,[29][30][31]. Two recent reports showed that three MYB genes (MYB21, MYB24 and MYB108) are responsive to JA treatment in opr3 mutant and lossof-function of MYB21 and MYB24 resulted in a short stamen phenotype [32] whereas MYB108 is involved in stamen and pollen maturation but not stamen filament elongation [33]. Interestingly, in an expression profiling study, we identified several MYBs including MYB21, MYB24, and MYB57 as DELLA-downregulated genes in ga1-3 flower buds [30]. This fact prompted us to investigate if there might be a cross-talk between GA signaling and JA signaling during stamen development. MYB21 and MYB24 have been shown to be expressed in all four whorls of the flower [32,34,35]. In this report, we showed that MYB21, MYB24, and MYB57 are down-regulated in the ga1-3 single mutant and the sterile quadruple mutant ga1-3 gai-t6 rga-t2 rgl1-1 (loss-of-function of GAI, RGA, RGL1 three DELLA genes but RGL2 is normal) but restored to wild type levels in the fertile penta mutant ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 (loss-of-function of GAI, RGA, RGL1 and RGL2 four DELLA genes). We also showed that absence of the four DELLAs (GAI, RGA, RGL1 and RGL2) cannot suppress the short stamen phenotype conferred by the lossof-function of MYB21 and MYB24. In addition, we observed that application of exogenous JA onto the ga1-3 gai-t6 rgl1-1 rgl2-1 quadruple mutant flower buds could restore the expression of MYB21, MYB24 and MYB57 whereas application of exogenous GA onto opr3 mutant flower buds failed to increase the expression of these three MYBs. Most importantly, we showed that GA upregulates JA-biosynthetic genes DAD1 and LOX1 and the JA content in the young flower buds of the GA-deficient quadruple mutant ga1-3 gai-t6 rga-t2 rgl1-1 is much lower than that in the WT and penta mutant. Therefore, we conclude that GA upregulates the DAD1 and LOX1 expression to promote JA production to promote the expression of the three MYBs necessary for stamen filament development.

Identification of DELLA-Repressed Stamen-Enriched Genes
The ga1-3 mutant is retarded in floral development, suggesting that the transcriptome for floral development in the ga1-3 mutant must be kept at a repressive state. Conversely, the fact that the ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 mutant (penta mutant) confers GA independent flowering suggests that the transcriptome responsible for floral development must have been constitutively activated in the penta mutant. We compared the expression profiles between ga1-3 and ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 and identified 360 DELLA-repressed and 273 DELLA-activated genes essential for floral development [30]. To identify DELLA-repressed stamenenriched genes, we examined expression of 43 DELLA-repressed genes in the sepal, petal, stamen and pistil via semi-quantitative RT-PCR. These 43 genes were chosen based on two criteria: 1) they are homologous to transcription factors known to regulate GA-response (e.g MYB gene family) and 2) genes whose expression showed drastic changes between the ga1-3 and penta mutant [30]. Only genes whose expression are either enriched in the stamen or highly expressed in the stamen and also in some other floral organs but not ubiquitously highly expressed in all four floral organs were classified as the stamen-enriched genes. A total of 34 genes, including two APG-like genes (At1g75880, At1g75900) and three genes (IRX1, IRX3, IRX5) encoding the cellulose synthase subunits which are known to be enriched in the stamen, were identified as DELLA-repressed stamen-enriched genes ( Figure 1; Table 1).

DELLAs Repress the Expression of MYB21, MYB24, and MYB57
Three MYB genes, namely MYB21, MYB24 and MYB57, were among the identified DELLA-repressed stamen-enriched genes ( Figure 1; Table 1). Based on the phylogenetic tree, MYB24 and MYB21 are classified into the subgroup 19 of R2R3-MYB family [36]. MYB57 shares high similarity with this subfamily and is a close member to this subfamily [37]. Overall, MYB21 shares 61.6% and 51.0% identity with MYB24 and MYB57 at the amino acid level, respectively ( Figure S1). The expression of these three MYBs in the young flower buds were reduced to a very low level in ga1-3 but restored to the wild type (WT) level in the ga1-3 gai-t6 rga-t1 rgl1-1 rgl2-1 penta mutant (Figure 2A). In order to find out which DELLA (RGL1, RGL2, RGA and GAI) is more effective in repressing the expression of MYB21, MYB24 and MYB57, transcript levels of each individual MYB gene were studied in four quadruple mutants in which only one of the four DELLA genes remains intact. All three MYB genes were almost undetectable in the Q1 (ga1-3 gai-t6 rgl1-1 rgl2-1, wild type for RGA) and barely detectable in the Q3 (ga1-3 gai-t6 rgl1-1 rga-t2,

Author Summary
Gibberellin and jasmonate are plant hormones that mediate diverse plant developmental processes and responses to the environment. Deficiency in either gibberellin or jasmonate causes male sterility, in part due to the short stamen filament conferred. In this report, we sought to study the interaction between gibberellin and jasmonate during stamen filament development. We focused on three MYB genes, namely MYB21, MYB24, and MYB57, which have been proven to be essential for stamen filament development in Arabidopsis. These three MYB genes are regulated by both gibberellin and jasmonate. We performed various molecular analyses and found that GA activates the expression of DAD1 and LOX1, two genes essential for jasmonate biosynthesis. The hypothesis of GA regulating JA biosynthesis is proved by the fact that the JA content in the young flower buds of the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant is much lower than that in the WT. This evidence demonstrates that GA promotes the production of jasmonate and high level of jasmonate will induce the expression of MYB21, MYB24, and MYB57 to promote stamen filament development. This is most likely the first molecular and genetic evidence to show how gibberellin and jasmonate interact to control stamen filament development.
wild type for RGL2) mutants but were detected at high levels in the Q2 (ga1-3 rga-t2 rgl1-1 rgl2-1, wild type for GAI) and Q4 (ga1-3 gai-t6 rga-t2 rgl2-1, wild type for RGL1) mutants ( Figure 2B), suggesting that RGA and RGL2, but not GAI nor RGL1, were the more effective DELLAs in repressing the expression of these three MYB genes. Interestingly, we showed previously that while Q1 and Q3 mutants, as the ga1-3 mutant, were retarded in floral development both Q2 and Q4 mutants produced normal fertile flowers ( Figure 2C) [8]. Therefore, it seems there is a nice correlation between normal floral development and the expression of MYB21, MYB24 and MYB57, suggesting that these three MYBs are probably necessary for normal floral development.

MYB21, MYB24, and MYB57 Function Redundantly in Controlling Stamen Filament Elongation
Expression analysis showed that MYB21 and MYB24 [32,34] as well as MYB57 are flower-specific genes ( Figure 2D). To determine if the spatial and temporal expression patterns of MYB21 and MYB24 correlate with their proposed role during stamen filament elongation, we examined MYB21 expression via in situ hybridization and generated pMYB24:GUS transgenic for examining MYB24 expression. Our in situ hybridization result showed that, starting from floral stage 12 [1,38], MYB21 is expressed in the anther vascular tissue and in cells at the junction between anther and stamen filament ( Figure S2A,B) where rapid filament elongation is hypothesized to occur starting from the floral stage 13 after a successful pollination [1]. MYB21 expression is also detected in the nectaries and ovules ( Figure S2A,B). Similarly, staining the young inflorescence of the pMYB24::GUS plants revealed that strong GUS activity was detected in the vascular tissue of stamen filament and sepals whereas only weak GUS activity was detectable in the petals starting from floral stage 12 ( Figure S3A-E). GUS activity was also detected in the upper part of the pistils ( Figure S3A-3E).
To investigate their roles in GA-mediated floral organ development, we identified T-DNA insertional mutant lines corresponding to these three MYB genes from the Salk Institute Genomic Arabidopsis Laboratory (SIGnAL) database. Mutant alleles were confirmed (data not shown) and designated as myb21-t1 (SALK_042711) for MYB21, myb24-t1 (SALK_017221) for MYB24, and myb57-t1 (SALK_065776) for MYB57 ( Figure 3A). myb24-t1 and myb57-t1 are both likely null alleles since MYB24 and MYB57 transcripts were undetectable in myb24-t1 and myb57-t1   Table 1) based on their relative more abundant expression in the stamen than in one or more of the rest of the floral organs. Primer pairs corresponding to these genes were listed in Table S1 mutant flower buds, respectively ( Figure 3B). On the other hand, MYB21 transcripts were still detectable in myb21-t1 although its level was greatly reduced in the mutant, suggesting that myb21-t1 is likely a leaky allele ( Figure 3B). After two rounds of backcross, we found that myb24-t1 and myb57-t1 mutant plants were phenotypically indistinguishable from the WT control plant ( Figure 3C; Table 2). However, in myb21-t1 the early developed flowers (,the first 10 flowers) bore short stamens ( Figure S4) with greatly reduced fertility and only the late developed flowers yielded proper seed settings ( Figure 3C; Table 2). A close look at the matured early flowers in myb21-t1 showed that the stamens did produce pollens ( Figure 3D, panel d). Cross-pollinating the pollens onto the myb21-t1 stigma yielded seeds that were homozygous for myb21-t1 and onto the WT stigma yielded myb21-t1 heterozygous seeds (data not shown), demonstrating that the short stamen is responsible for the partial sterile phenotype. Although myb21-t1 is likely a leaky allele, the short stamen phenotype conferred by the myb21-t1 mutation is identical to a MYB21 null allele we obtained later from Gabi-Kat stock (stock number N311167, data not shown). The WT-like phenotype displayed by myb24-t1 and myb57-t1 and mild floral phenotype displayed by myb21-t1 suggest that these MYB genes might function redundantly during stamen development. To prove this hypothesis, crosses were made among homozygous myb24-t1, myb21-t1 and myb57-t1 plants. Three RT-PCR analysis shows that the expression of the three MYB genes in the young flower buds are greatly reduced in ga1-3 but restored to the WT level in ga1-3 gai-t6 rgat2 rgl1-1 rgl2-1 (penta). (B-C) RT-PCR analysis shows that the repressed expression of the three MYB genes in ga1-3 was restored in ga1-3 gai-t6 rga-t2 rgl2-1 (Q2) and ga1-3 rga-t2 rgl1-1 rgl2-1 (Q4) two quadruple mutants but not in ga1-3 gai-t6 rgl1-1 rgl2-1 (Q1) and ga1-3 gai-t6 rga-t2 rgl1-1 (Q3) two quadruple mutants (B). This restoration of MYB expression nicely correlates with the recovery of fertility in Q2 and Q4 (C). Total RNA used in RT-PCR analysis was extracted from the young flower buds. (D) RT-PCR analysis shows that the three MYB genes are floral specific genes. IF, inflorescence; CL, cauline leaves; RL, rosette leaves; IT, internodes; RT, roots; SL, siliques. (E) Amino acid alignment of MYB21, MYB24 and MYB57 proteins. The conserved R2 and R3 domains and the NYWSV/ME/DDlWP/S motif are highlighted in red, blue and green, respectively. doi:10.1371/journal.pgen.1000440.g002 double mutants (myb21-t1 myb24-t1, myb21-t1 myb57-t1, and myb24-t1 myb57-t1) and one triple mutant (myb21-t1 myb24-t1 myb57-t1) were generated and used in our phenotypic analysis.
The flower development of myb24-t1 myb57-t1 double mutant at all stages was indistinguishable from the WT control ( Figure 3D; Table 2) [1]. Stamens in mature flowers of the myb21-t1 myb24-t1 double mutant were shorter than that of the myb21-t1 single mutant and shorter stamens were also observed in majority of the late developed mature flowers in the double mutant ( Figure S4). As a result, the myb21-t1 myb24-t1 double mutant is more severely sterile than myb21-t1 by having fewer siliques with seed settings ( Figure 3C and 3D; Table 2), an observation also reported by Mandaokar et al [32]. The myb21-t1 myb57-t1 double mutant had shorter stamens in early developed mature flowers and some of the later flowers ( Figure 3D; Figure S4) and its seed settings displayed an intermediate phenotype between myb21-t1 single and myb21-t1 myb24-t1 double mutants ( Figure 3C and 3D; Table 2). Interestingly, cross-pollination showed that the short stamens in both myb24-t1 myb21-t1 and myb21-t1 myb57-t1 double mutant plants produced viable pollens (data not shown), suggesting that the short stamen ( Figure S4) is responsible for the reduced fertility in these mutants.
The myb21-t1 myb24-t1 myb57-t1 triple mutant, as the myb21-t1 myb24-t1 double mutant, had short stamens but was even more severely sterile than myb21-t1 myb24-t1 ( Figure 3C and 3D; Figure  S4; Table 2). For myb21-t1 myb24-t1 double and myb21-t1 myb24-t1 myb57-t1 triple mutant plants, we occasionally observed that while, in the same inflorescence, most of the flowers did not set or set very few seeds, some were able to develop normal siliques filled up with seeds ( Figure 3C and 3D). Cross pollination showed the pollens produced by myb21-t1 myb24-t1 myb57-t1 triple mutant plants were partial viable (data not shown), suggesting the short stamens in the triple mutants are the main cause of the sterility. It is possible that environmental factors (e.g. temperature) may influence male fertility in these mutants, an observation also reported for MYB33 and MYB65 [24]. Therefore, MYB21, MYB24 and MYB57 function redundantly to control the stamen filament development in the late developed flowers.
Since MYB21 and MYB24 are also expressed in sepals and petals, we examined the sepal and petal development in the single, double and triple myb mutants. As shown in Figure 3D, sepal development appeared normal in all mutants whereas petal development varied in different mutants. Petals in the myb24-t1 and myb57-t1 two single mutants grew to a final length longer than the pistils, as that did the WT petals. Petals in the myb21-t1 single, myb24-t1 myb57-t1 and myb21-t1 myb57-t1 two double mutants grew to a final height parallel to the pistil ( Figure 3D). Petals in the myb21-t1 myb24-t1 double mutant grew just out of the sepals but ended at a lower level than the stigma ( Figure 3D). The growth of petals in the myb21-t1 myb24-t1 myb57-t1 triple mutant was arrested and the petals never grew out of the sepals ( Figure 3D). myb21-t1 myb24-t1 Is Epistatic to gai-t6 rga-t2 rgl1-1 rgl2-1 in Controlling Stamen Filament Elongation MYB21 and MYB24 were repressed in ga1-3 but their expressions were restored to the WT level in the ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 penta mutant, suggesting that GA regulates MYB21 and MYB24 through inactivating DELLA proteins. Application of  exogenous GA could not rescue the stamen development in myb21 myb24 mutant (data not shown), suggesting that MYB21 and MYB24 are needed in GA-mediated stamen development. To further confirm this hypothesis, we crossed myb21 myb24 with ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 to generate two hexa mutants (hexa1: ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 myb21-t1; hexa2: ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 myb24-t1) and one hepta mutant (ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 myb21-t1 myb24-t1). The two hexa mutants overall appeared similar to each other and had wildtype-like stamens and were largely fertile ( Figure 4A; Figure S5). The hepta mutant plant displayed no difference from the penta mutant plant in its vegetative growth. However, its mature flowers showed a short filament phenotype identical to that in the myb21-t1 myb24-t1 double mutant ( Figure 4A; Figure S5). This observation demonstrated that myb21-t1 myb24-t1 double mutations are epistatic to DELLA mutations. SEM analysis showed that the short stamen phenotype in the hepta mutant was due to reduced cell length ( Figure 4B and 4C), rather than to a reduction in cell number ( Figure 4D). Therefore, MYB21 and MYB24 act downstream of DELLAs in GA signaling pathway to control the stamen filament development.
GA Application Fails to Induce the Expression of MYB21, MYB24 and MYB57 in JA-Deficient Mutant We showed in the above that the expression of MYB21, MYB24 and MYB57 was repressed in the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant (wild type for RGL2) but restored to normal in the ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 penta mutant (Figure 2A and 2B). Mandaokar et al reported that the expression of MYB21 and MYB24 was downregulated in opr3 mutant and application of exogenous JA could restore their expression [32]. These results suggest that there might be a crosstalk between the GA and JA pathways in regulating the expression of MYB21, MYB24 and MYB57 during stamen development. Genetically, there are three possible ways of interaction between GA and JA. Firstly, GA might act through the JA pathway to regulate the expression of these MYB genes. In this case it is expected that JA application onto ga1-3 gai-t6 rga-t2 rgl1-1 would induce the expression of MYB21, MYB24 and MYB57 whilst GA application onto opr3 would have no effect on their expression. Conversely, JA may act upstream of the GA pathway to regulate the expression of these three MYB genes. In this case, GA application onto opr3 would induce whilst JA application onto ga1-3 gai-t6 rga-t2 rgl1-1 would have no effect on the expression of MYB21, MYB24 and MYB57. The third possibility is that GA and JA may not act in a hierarchical manner but rather via parallel pathways to regulate the expression of the three MYB genes. If this is the case, GA application onto opr3 and JA application onto ga1-3 gai-t6 rga-t2 rgl1-1 would probably both induce the expression of the three MYB genes. To find out which is the likely case, we first examined the effect of GA application on JA-deficient mutant opr3 and found that GA application failed to rescue the opr3 mutant phenotype and failed to induce the expression of MYB21, MYB24 and MYB57 in opr3 even at 96 hrs after GA treatment ( Figure 5A). Failure in induction of expression of MYB21, MYB24 and MYB57 in GA-treated opr3 mutants could be due to inactivation of GA signaling in JA-deficient background. GA3ox1 and GA20ox2 are two key genes that contribute to the biosynthesis of bioactive GA and these two genes are under negative feedback regulation by GA signaling pathway (GA-down) [30]. On the other hand, GA2ox1 is a GA-up gene responsible for GA catabolism [30]. Examination of the GA3ox1 and GA20ox2 and GA2ox1 expression in GA-treated opr3 mutants showed expected GA-response ( Figure 5B). Meanwhile, expression of GA3ox1 and GA2ox1 appeared normal in opr3 ( Figure 6A). These results suggest that JA-deficiency specifically blocks the GA-signaling leading to the induction of MYB21, MYB24 and MYB57 expression but not the negative feedback pathway for GA-biosynthesis.
GA Suppresses DELLA to Upregulate the JA-Biosynthesis Gene LOX1 and DAD1 Considering the fact that JA application was able to induce the expression of MYBs in the ga1-3 gai-t6 rga-t2 rgl1-1 mutant it is reasonable to argue that JA biosynthesis, instead of JA signaling pathway, is likely affected in the ga1-3 gai-t6 rga-t2 rgl1-1 mutant. To test this hypothesis, we examined the expression of known or putative JA biosynthesis genes including DAD1 (Defective in anther dehiscence 1), LOX1 (Lipoxygenase 1), LOX2 (Lipoxygenase 2), AOS (Allene oxide synthase), AOC1 (Allene oxide cyclase 1, At3g25760), AOC 2 (At3g25770), AOC 3 (At3g25780), AOC 4 (At1g13280) and OPR3 (OPDA reductase 3) in La-er WT, Ws WT, ga1-3 single, ga1-3 gai-t6 rga-t2 rgl1-1 quadruple, ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 penta, and opr3 mutants. We found that only MYB21, LOX2 and AOC1 showed reduced expression in the opr3 mutant whereas all the other genes, including GA-biosynthesis genes, expressed similarly in the opr3 mutant and Ws WT control ( Figure 6A and 6B), suggesting that JA-deficiency does not affect GA biosynthesis. Expression of these genes in GA-related mutants was more complicated. We found that the expression levels of AOS, AOC1, AOC3, AOC4, LOX2 and OPR3 did not show significant differences in all GA-related mutants when compared to the La-er WT control ( Figure 6A and 6B), suggesting these genes are probably regulated in a GA-independent fashion. The expression of AOC2 was obviously induced in ga1-3 and then reduced to the WT level in the quadruple and penta mutants ( Figure 6A), suggesting AOC2 is a GA-down gene. In contrast, the expression of LOX1 was significantly reduced in ga1-3 but was restored both in the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple and penta mutants ( Figure 6B), suggesting that although LOX1 is a GA-up gene and its expression is not repressed by RGL2. Interestingly, DAD1 expression was found to be downregulated to approximately 20% of the WT level in both ga1-3 and the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant whereas was restored to approximately 60% of the WT level in the penta mutant (approximately three folds increase in penta versus One expected consequence of downregulation of DAD1 expression by RGL2 is the reduction of JA levels in the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant (Q3). To text this hypothesis, we measured the JA contents in the young flower buds in WT, opr3, the Q3 quadruple mutant and the ga1-3 gai-t6 rga-t2 rgl1-1 rgl2-1 penta mutant (penta). The data obtained clearly showed that the JA content was greatly reduced in the young flower buds of the quadruple Q3 mutant whereas was restored in the penta mutant when compared to that in the WT and opr3 mutant ( Figure 6D; Table S2).
GA Application Induces DAD1 Expression Prior to the Induction of MYB21, MYB24, and MYB57 DAD1 is a stamen specific gene encoding chloroplastic phospholipase A1 protein that catalyzes the first step of JA biosynthesis. Mutation in DAD1 resulted in a typical JA-deficient phenotype in stamen development [40], a phenotype similar to that of myb21-t1 myb24-t1 double mutant. As mentioned earlier, JA likely acts downstream of GA to regulate the expression of MYB21, MYB24 and MYB57. To study whether there is a correlation between GA-regulated DAD1 expression and MYB21, MYB24 and MYB57 expression, we treated the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant with GA. We first confirmed the GAresponsiveness in the quadruple mutant plants by examining the expression of known GA-response genes GA3ox1, GA20ox2 and GA2ox1 ( Figure 7A). Then we examined the expression of DAD1 Figure 6. GA Regulates the Expression of JA Biosynthesis Genes DAD1 and LOX1. (A) RT-PCR analysis of GA-and JAbiosynthesis genes in the young flower buds of La-er WT control, ga1-3, ga1-3 gai-t6 rga-t2 rgl1-1 (Q3), ga1-3 gai-t6 rga-t2 rgl1-1rgl2-1 (penta), opr3 (in WS background) and Ws WT control. (B) Northern analysis of MYB21, GA20ox2, LOX1, LOX2 and OPR3 in the young flower buds of the La-er WT control, ga1-3, Q3, penta, opr3 and Ws WT control. (C) Semi-quantitative analysis of DAD1 expression in ga1-3, Q3 and penta relative to that in WT (Laer), respectively. Data were averaged from three independent batches of samples and ACTII was used as the normalization control. The expression level of WT is set as 1. (D) Comparison of JA contents among WT (La-er), opr3, ga1-3 gai-t6 rga-t2 rgl1-1 (Q3) and ga1-3 gai-t6 rga-t2 rgl2-1 rgl1-1 (penta) ( Table S2). For WT and the penta mutant JA contents were averaged from four repeats. For the Q3 mutant, JA was detected in three out of the four repeats. For opr3, JA was detected only in one out of the four repeats. FW, fresh weight. doi:10.1371/journal.pgen.1000440.g006 Surprisingly, compared to the induction of MYB21 and MYB24 expression by JA treatment which is detectable at 18 hrs post treatment ( Figure 5D), GA induction of the expression of these two MYB genes in ga1-3 gai-t6 rga-t2 rgl1-1 happens much later and only became detectable at 72 hrs ( Figure 7B). More interestingly, GA induction of the expression of DAD1 is obviously detectable at 48 hrs which is prior to GA-induced expression of MYB21 and MYB24 in the ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutant ( Figure 7B). Our data suggest that GA might first induce the expression of DAD1 to promoter JA production then via JA signaling to regulate the expression of MYB21 and MYB24.
Expression of MYB21, MYB24, and MYB57 Is Necessary But Insufficient for Normal Stamen Filament Elongation in ga1-3 gai-t6 rga-t2 rgl1-1 As shown in the above, MYB21, MYB24 and MYB57 act downstream of DELLAs in controlling stamen filament elongation. Expression of MYB21, MYB24 and MYB57 was repressed and floral development was arrested in the ga1-3 gai-t6 rga-t2 rgl1-1 Q3 quadruple mutant ( Figure 2B and 2C). Regarding the fact that JA content is reduced in the young flower buds of Q3 we questioned whether restoration of expression of these MYBs by exogenous application of JA could rescue the stamen development to the ga1-3 gai-t6 rga-t2 rgl1-1 Q3 plants. We analyzed the flowers of JAtreated ga1-3 gai-t6 rga-t2 rgl1-1 plants and found that repeated JA application was unable to rescue the stamen development ( Figure 8) though JA could restore the expression of the three MYB genes ( Figure 5D), indicating that expression of MYB21, MYB24 and MYB57 alone was insufficient for normal stamen development in the ga1-3 gai-t6 rga-t2 rgl1-1 mutant. Furthermore, we found that exogenous application of GA to the ga1-3 gai-t6 rga-t2 rgl1-1 plants was able to induce the expression of MYB21, MYB24 and MYB57 ( Figure 7B) and recover normal floral development (Figure 8). Taken together, our results demonstrate that besides these JA-inducible MYBs, other important GAregulated JA-independent factors are needed for normal stamen filament development in ga1-3 gai-t6 rga-t2 rgl1-1.

Overexpression of MYB21 Restored Stamen Filament Elongation and Fertility to opr3 Flowers
To test our hypothesis that GA acts through JA to control expression of the MYB genes to promote filament elongation, we fused MYB21 gene with the CaMV35S promoter (pCAMBIA1301 vector) and this construct was used to generate transgenic plants in the opr3 mutant background. Semi-quantitative RT-PCR showed that MYB21 was overexpressed in the transgenic plants in the opr3 background ( Figure 9A). We found that overexpression of MYB21 could restore the stamen filament growth ( Figure 9B) and restore  the fertility (Figure 9C and 9D) to the opr3 mutant partially. Together with the fact that loss of function of four DELLA (GAI, RGA, RGL1 and RGL2) could not restore the fertility and filament elongation to the coil1 mutant ( Figure S6), we have now provided strong evidence to show that GAs act through JA to control expression of the MYBs and promote stamen filament elongation.

Discussions
In this report, we first identified 34 DELLA-repressed stamenenriched genes by RT-PCR analysis of candidate genes selected from microarray data [30]. We then selected MYB21, MYB24 and MYB57 for detailed genetic analysis because GAMYBs are the best characterized transcription factors involved in GA-response. We showed that Arabidopsis MYB21, MYB24 and MYB57 are highly expressed in the stamen. The stamen-enriched expression pattern is consistent with the observation that the myb21-t1 myb24-t1 myb57-t1 triple mutant is impaired in the stamen development, especially in the stamen filament elongation. During the course of studying these three MYBs, Mandaokar et al reported that MYB21 and MYB24 are also JA-inducible [32] which immediately attracted our attention to study the hierarchical relationship between GA and JA in regulating the expression of these three MYBs.
We first tested the responses of GA-and JA-deficient mutants (i.e. ga1-3 gai-t6 rga-t2 rgl1-1 Q3 quadruple mutant and opr3 mutant, respectively) to GA and JA treatments and found that JAtreatment induced the expression of MYB21, MYB24 and MYB57 in the GA-deficient plants whereas GA-treatment failed to do so in the JA-deficient plant. This result suggests that JA likely acts downstream of GA pathway to control the expression of these three MYBs. It is possible that JA acts downstream by modulating the stability or activity of DELLA proteins to induce the expression of the three MYBs. If this is the case, we would expect that JAtreatment would lead to RGL2 degradation or would change the expression patterns of GA-response genes in ga1-3 gai-t6 rga-t2 rgl1-1. However, we found that neither the RGL2 protein level (data not shown) nor the expression patterns of three GA-response genes GA2ox1, GA3ox1 and GA20ox1 were obviously altered in the JAtreated ga1-3 gai-t6 rga-t2 rgl1-1 plants at 18 hrs post treatment although the three MYBs are highly expressed at this time point, suggesting that destabilization or inactivation of DELLA proteins is unlikely the cause for JA-induced expression of MYBs in ga1-3 gai-t6 rga-t2 rgl1-1. Alternatively, it is possible that GA suppresses DELLA to promote JA production or modulate JA-signaling to induce the expression of the three MYBs. The fact that JA application can restore the expression of the three MYBs in the GA-deficient background strongly suggests that, at least in part, JA biosynthesis is impaired in the ga1-3 gai-t6 rga-t2 rgl1-1 mutant. JA biosynthesis is accomplished by a sequential biochemical reactions mediated by JA-biosynthesis genes including DAD1, LOX1, 2, AOS, AOC1, 2, 3, 4 and OPR3 and is regulated by OPDA compartmentalization and a JA-mediated positive feedback loop [41]. Biotic and abiotic stresses also induce JA formation [42][43][44]. In our experiment, we found that JA biosynthesis gene DAD1 was greatly down-regulated in both ga1-3 single and ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutants but partially restored to a relatively high level in the penta mutant, suggesting that GA is required for the expression of DAD1 to control the production of JA via repression of DELLA proteins. In flowers of dad1 null mutant, the JA levels were only 22% of that of WT [40], demonstrating that limited initial substrate generation by DAD1 reaction acts as a control point for JA biosynthesis in flowers. Therefore, it is highly possible that reduced expression of DAD1 in ga1-3 gai-t6 rga-t2 rgl1-1 or ga1-3 mutant may result in relative low JA production. This hypothesis is strongly supported by the observation that JA content was greatly reduced in the young flower buds of the GA-deficient quadruple mutant ga1-3 gai-t6 rga-t2 rgl1-1 (Q3 mutant). Furthermore, the fact that the induction of DAD1 expression happens prior to the expression of MYBs by GA in the ga1-3 gai-t6 rga-t2 rgl1-1 mutant strongly support our hypothesis that GA may regulate the MYBs' expression via mobilization of the biosynthesis of JA. A recent report showed that DAD1 expression is directly controlled by AGAMOUS (AG) [45]. Interestingly, Yu et al reported that AG expression was downregulated in the GA-deficient mutant ga1-3 and exogenous GA application promoted the AG expression [20]. It will be interesting to study if there is a relationship among DELLAs, AG and DAD1 in the future. High level of JA would induce the expression of the three MYB genes essential for stamen development. In addition to DAD1, we also observed that expression of LOX1 was down-regulated in ga1-3 mutant and restored to the WT level in the penta mutant. On the other hand, another JA biosynthesis gene AOC2 was up-regulated in ga1-3 mutant. These observations suggested that GA may be one of the endogenous signal involved in the regulation of JA biosynthesis genes.
Genetic studies have shown that MYB21, MYB24 and MYB57 are indispensable for stamen development. The stamen phenotype of myb21-t1 myb24-t1 myb57-t1 triple mutant is similar to that of JAdeficient mutants including opr3 and dad1 mutants. Overexpression of MYB21 restored the stamen filament elongation and fertility to the opr3 flowers, strongly suggesting that JA-mediated stamen filament growth is mainly through the MYB pathway. Both ga1-3 single and ga1-3 gai-t6 rga-t2 rgl1-1 quadruple mutants showed a more severe flower phenotype than myb21-t1 myb24-t1 myb57-t1 triple mutant. The fact that expression of these MYBs in ga1-3 gai-t6 rga-t2 rgl1-1 plants was not enough to rescue the mutant flower phenotype indicates that these MYBs are necessary but not sufficient for GA-mediated floral development. These data also indicate that modulation of JA pathway may be only one of the branches of GA function in regulating stamen development.
Active cross-talk between different hormone signaling pathways have been revealed in many developmental processes [46]. For example, it was reported that auxin was necessary for GAmediated Arabidopsis root growth by promoting GA-dependent degradation of DELLA proteins [47]. In contrast, ethylene inhibits Arabidopsis root growth by delaying the GA-induced destabilization of DELLA [48]. Recently, the complexity of interactions between ethylene and GA signal transduction pathways were analyzed by using combinations of different ethylene and GA related mutants [49]. Hormone-hormone interaction also plays an important role in controlling flowering. For example, it was found that stress induced hormone ethylene control floral transition via DELLAdependent regulation of floral meristem-identity genes LEAFY and SUPPRESSOR OF OVEREXPRESSION OF CONSTANS 1 (SOC1) [50]. We have here established a linear relationship between GA and JA in that GA modulates the expression of DAD1 that in a likely scenario to promote JA biosynthesis and in return JA induces the expression of MYB21, MYB24 and MYB57 to control the normal stamen development in Arabidopsis.

RT-PCR and Northern Analysis
Different organs (sepal, petal, stamen, pistil and peduncle) of stage 11-12 flowers were dissected under microscope and pooled for RNA extraction. Flowers younger than stage 11 were pooled as young flower buds for RNA extraction. Total RNA was extracted from the young flower buds of respective genotypes treated with or without GA and JA using Tri Reagent (Molecular Research Center, Cincinnati, OH). The residue genomic DNA in the total RNA was removed via treatment with RNase-free DNase I (Roche, Germany) and the total RNA further purified with the RNeasy Mini kit (QIAgen, Valencia, CA, USA). First strand cDNA was synthesized using SuperScript TM II RNase H 2 Reverse Transcriptase (Invitrogen, USA). First strand cDNA was used as the template in PCR using gene specific primers. Primer pairs used in identification of DELLA-repressed stamen-enriched genes were listed in Table S1. Primer pairs for RT-PCR analysis of GA2ox1, GA3ox1, GA20ox2, DAD1, AOS, OPR3, LOX1, 2 and AOC1, 2, 3, 4 were listed in Table 4. For quantifying the gene expression levels, PCR products were stained with ethidium bromide and the intensity was quantified using software Molecular Analyst (Bio-Rad). The gene expression level was normalized to the expression level of ACTII and then displayed as a ratio of expression levels of GA (or JA) treated samples versus untreated control.
Quantification of JA 500 mg young flower buds harvested from different genotypes were frozen in liquid N 2 and ground to a fine powder with a mortar and pestle. Following addition of 600 mL methanol, homogenates were mixed and kept at 4uC overnight, then centrifuged at 4,800 g for 10 min. The supernatant was transferred to a new 5 mL glass tube and the residue was reextracted with 200 mL of methanol. 3000 mL ddH 2 O was added to the combined extracts and this solution was applied onto the Sep-pak C 18 cartridge. The cartridge was washed with 200 mL 20% methanol and 250 mL 30% methanol 300 mL, respectively. Finally, the cartridge was eluted with 300 mL 100% methanol and the eluted solution was collected and used as the samples. Preprepared JA solutions (three concentrations were used: 10 ng/mL; 100 ng/mL; 1000 ng/mL) were used as the internal normalization standard. Samples were analyzed by a Thermo TSQ Quantum Ultra LC-MS-MS system.

pMYB24::GUS Transgenic Plants
To make the pMYB24::GUS construct, a 3098 bp (68 bp upstream of MYB24 start codon ATG) genomic DNA fragment was PCR amplified from Col-0 genomic DNA using primers 18F (PstI, 59 TTCTAGGCTGCAGCTAAACGACTTC 39) and 2934R (59 GTAATAGAAAGGGAGAGTTGTGAAAG 39). PCR amplifications of promoter regions were performed using PfuTurbo DNA polymerase (Stratagene). The amplified DNA fragment was digested with PstI and then cloned into PstI/NcoIcleaved pCambia 1301 vector and their sequences were confirmed by sequencing. The pMYB24::GUS fusion construct was then introduced into Arabidopsis thaliana ecotype Col-0 plants using flower dip method [19]. More than three independent lines were examined at various stages of floral development in this study.