Dual roles of the bZIP transcription factor PERIANTHIA in the control of floral architecture and homeotic gene expression

Flowers have attracted the attention of humanity for millennia, not only because of their beauty, but also because they give rise to fruits and seeds. As flowers are the sites of sexual reproduction of flowering plants, their proper development is essential for survival of the species. Evolution has brought about a dazzling array of floral phenotypes as flowering plants have found their way into virtually every ecological niche on land. Still, the molecular machinery that controls the identity and stereotypic arrangement of floral organs has been well conserved(Theissen et al., 1996). Flowers of the reference plant Arabidopsis thaliana contain four major organ types: sepals, petals, stamens and carpels, which are arranged in four concentric rings or whorls. Organ identity is specified by the overlapping activities of three classes of homeotic genes, termed A, B and C,as predicted in the ABC model (Bowman et al., 1991; Coen and Meyerowitz, 1991). The activity of a floral homeotic gene is typically confined to two adjacent whorls, with the A class represented by APETALA1 (AP1) and APETALA2 (AP2) acting in whorls one and two, the B class genes APETALA3 (AP3) and PISTILATA (PI) in whorls two and three, and the C class gene AGAMOUS (AG) in whorls three and four (reviewed by Lohmann and Weigel, 2002). With the exception of AP2, all floral homeotic genes code for MADS domain transcription factors. Essential co-regulators in this process are the MADS domain transcription factors of the SEPALLATA class (SEP1, SEP2, SEP3 and SEP4), which form tetrameric complexes with different combinations of homeotic ABC factors (Ditta et al.,2004; Honma and Goto,2001; Melzer et al.,2009; Pelaz et al.,2000).

While the primary function of floral homeotic genes appears to be the specification of floral organ identity, several of them have additional roles. The most prominent example is the C class gene AG. In addition to specifying reproductive organs, stamens and carpels, it has a key role in limiting stem cell proliferation in the center of emerging flowers(Bowman et al., 1989; Bowman et al., 1991; Mizukami and Ma, 1995; Sieburth et al., 1995).

Although the molecular nature of AG was uncovered almost two decades ago (Yanofsky et al.,1990) our understanding of AG regulation is far from complete. The proper spatiotemporal expression of AG RNA is dependent on sequences located in the second intron, which is one of the longest found in the A. thaliana genome(Sieburth and Meyerowitz,1997). Functional characterization of this intragenic enhancer demonstrated that multiple redundant regulatory modules mediate the response to competing activating and repressing inputs(Bomblies et al., 1999; Busch et al., 1999; Deyholos and Sieburth, 2000). One of the most prominent direct regulators is the meristem identity factor LEAFY (LFY) (Busch et al.,1999; Parcy et al.,2002; Parcy et al.,1998), which acts in concert with the homeodomain transcription factor WUSCHEL (WUS) to activate AG in the center of developing flowers (Lohmann et al.,2001). Additional inputs are provided by SEP3, which can act as activator (Castillejo et al.,2005) or repressor of AG transcription(Sridhar et al., 2006),depending on the regulatory environment. Once AG transcription has been activated, an autoregulatory mechanism is in place to ensure stable expression throughout flower development(Gomez-Mena et al., 2005).

Important repressors of AG expression are LEUNIG (LUG) and SEUSS(SEU), which act as co-repressors in higher order complexes with DNA-binding transcription factors, such as AP1, SEP3 and AGAMOUS-LIKE 24(Franks et al., 2002; Gregis et al., 2006; Liu and Meyerowitz, 1995; Sridhar et al., 2006). In addition, AG expression is repressed in the outer whorls by the AP2 transcription factor (Drews et al.,1991), which in turn is under negative regulation by the microRNA miR172 (Aukerman and Sakai,2003; Chen,2004).

An additional layer of regulatory complexity is introduced by epigenetic silencing of the AG locus, which is mediated by trimethylation of lysine 27 on histone H3 proteins. It has been shown that these modifications are dependent on a complex containing CURLY LEAF (CLF) and EMBRYONIC FLOWER 2(EMF2), two members of the polycomb group protein family, as well as the plant-specific protein EMBRYONIC FLOWER 1 (EMF1)(Calonje et al., 2008). Consequently, mutations in the corresponding genes cause ectopic AGexpression (Goodrich et al.,1997). The activity of the repressor complex is antagonized by ARABIDOPSIS HOMOLOG OF THRITHORAX 1 (ATX1)(Alvarez-Venegas et al.,2003).

As the proper spatiotemporal expression of AG involves a plethora of regulators, which act through redundant modules in the large intragenic enhancer, direct identification of important cis-regulatory elements has been difficult. Candidates for cis-regulatory motifs have instead been identified using phylogenetic footprinting and shadowing(Hong et al., 2003), although many of the corresponding trans-factors have remained unknown. Recent studies have shown that regulatory elements have been a driving force for the evolution of floral diversity (Causier et al., 2009). Here, we identify a new activator of AG, the bZIP factor PERIANTHIA (PAN), which was previously known to affect floral organ number (Chuang et al.,1999; Running and Meyerowitz,1996), but not homeotic gene expression.

Descriptions of all plant lines used are given in Table 1. All experiments were carried out in the KB14 AG::GUS reporter in Col-0 background under long-day conditions, unless otherwise noted.

GUS staining was performed as described(Lohmann et al., 2001). In situ hybridizations were performed in accordance with standard protocols(Weigel and Glazebrook, 2002)with the addition of 10% PVA to the staining solution. RNA probes were prepared from full-length cDNAs. Real-time RT-PCR analyses were performed on three replicates of independently grown plant material. RNA from a pool of primary inflorescences was prepared using RNeasy Plant Mini Kit (Qiagen),followed by cDNA synthesis with RevertAid H Minus First Strand cDNA Synthesis Kit (Fermentas). PCR was carried out in the presence of SYBR Green using BETA-TUBULIN-2 (At5G62690) for normalization. Primer sequences can be obtained upon request.

A list of all constructs is given in Table 2.

For construction of RH146 the AAGAAT box was deleted from KB8(Busch et al., 1999) and the 5′AG enhancer was cloned in forward orientation into pDW294(Busch et al., 1999) by BamHI/HindIII restriction sites. pSST210 was obtained using the QuikChange Site-Directed Mutagenesis Kit (Stratagene) on KB8 following the manufacturer's protocol and subcloning of the 5′ enhancer into pDW294 using BamHI/HindIII. Primer sequences can be obtained upon request.

As bait for the yeast one-hybrid screen, the 33 bp AAGAAT box was trimerized using XhoI and SalI sites, inserted in the XhoI site of pEMBLYi22 (Baldari and Cesareni, 1985) to yield pSST90, linearized and integrated into the URA3 locus of the W303-A1a yeast strain (for detailed information see: http://wiki.yeastgenome.org/index.php/CommunityW303.html). Using standard protocols the screen was performed with a cDNA library constructed of RNA from young floral tissue in pEXPAD-502(Wigge et al., 2005). Double transformants were selected on CSM-TRP/-HIS + 15 mM 3AT plates. Plasmids of positive clones were rescued, re-transformed and confirmed by X-Gal filter-lift-assay before sequencing. For independent confirmation, full-length cDNAs of candidate factors were cloned into pGEM-T easy (Promega) and subcloned into pEXPAD-502 (TRP1, Invitrogen). All were transformed stably into the yeast strain EGY48(Golemis et al., 1996) that contained wild-type or mutated versions of the trimerized AAGAAT box upstream of the lacZ reporter in the vector KF1, a derivative of pLG718(Guarente and Mason, 1983). pSST093 contained the wild-type AAGAAT box sequence, for pSST197 two nucleotides of the putative binding sites for bZIP were mutated [using the QuikChange Site-Directed Mutagenesis Kit (Stratagene) following the manufacturer's protocol] and the resulting ΔbZIP AAGAAT box trimer was inserted using SpeI/XbaI sites into KF1. For pSST209 the same procedure was followed by mutating two nucleotides in each of the putative GARP-binding sites as described in Fig. 2A and inserting the fragment into KF1.

GARP and bZIPs transcription factor cDNAs were cloned in p423Gal1 and p424Gal1 (Mumberg et al.,1994), respectively, and co-transformed into yeast. Liquid cultures were assayed for β-galactosidase activity to quantify transcriptional activation of the AAGAAT box reporter gene. Color reactions were carried out using ONPG (o-nitrophenyl-beta-D-galactoside) as a substrate and an ELISA reader. Relative reporter gene activity was calculated according to Miller (Miller, 1972).

Phylogenetic footprinting and shadowing had previously been used to identify conserved motifs in the second AG intron, which is essential for proper transcriptional regulation of AG(Busch et al., 1999; Deyholos and Sieburth, 2000; Hong et al., 2003). One of the least variable sequences was the 33-bp-long AAGAAT box, which is not an obvious candidate for a motif bound by known regulators such as LFY, WUS or MADS domain factors (Fig. 1A). Recently, Causier et al. (Causier et al.,2009) have shown that the AAGAAT box has been conserved in sequence and relative position in monocots and dicots, suggesting that it arose before the split of the two lineages roughly 140 million years ago. The enhancer in the second AG intron is composed of two redundant regions(Bomblies et al., 1999; Busch et al., 1999; Deyholos and Sieburth, 2000),and the AAGAAT box is located in the 5′ fragment defined by the KB14 reporter (Fig. 1A), which drives GUS reporter expression in the AG domain (Fig. 1B)(Busch et al., 1999). The deletion of the AAGAAT box in this context (RH146) caused a severe reduction of reporter gene expression (Fig. 1B-D), indicating that this element is essential for activity of the 5′AG enhancer. The AG activator LFY acts through DNA motifs contained in KB14, which are distinct from the AAGAAT box. Thus, we crossed KB14 and RH146 to plants expressing an activated form of LFY, LFY:VP16(Busch et al., 1999; Parcy et al., 1998), to test the functionality of the RH146 reporter. We observed strong and ectopic GUS expression, confirming that the regulatory input of other factors binding to the 5′AG enhancer is not impaired in the RH146 lines (see Fig. S1 in the supplementary material).

To identify factors that act through the AAGAAT box, we performed a yeast one-hybrid screen, using a cDNA fusion library prepared from RNA of microscopically dissected inflorescence apices(Wigge et al., 2005) and an AAGAAT box trimer as bait (pSST90). Two independent screens of more than 2.5 million transformants identified 50 positive clones, which contained inserts coding for transcription factors of the GARP or bZIP families. We found four different bZIPs, At1g08320/ATBZIP21, At1g68640/PERIANTHIA (PAN),At3g12250/TGA6, and At5g06960/TGA5, and five GARP factors, At1g79430/ALTERED PHLOEM DEVELOPMENT (APL), At3g24120, At4g13640/UNFERTILIZED EMBRYO SAC 16(UNE16), At4g16110/ARABIDOPSIS RESPONSE REGULATOR 2 (ARR2), and At4g37180. Full-length cDNAs of all but ARR2 were fused with the GAL4 activation domain and interaction with the AAGAAT box was confirmed.

The AAGAAT box contains one consensus binding motif for bZIP factors(CACGTC) and two potential GARP-binding motifs (AATCT and AGATA)(Fig. 2A)(Foster et al., 1994; Hosoda et al., 2002). Within the AAGAAT box, these elements were almost invariant, even outside the Brassicacae (Hong et al.,2003; Causier et al.,2009), supporting the conserved functional importance of these sequences. To determine if the consensus motifs are indeed bona fide binding sites for the identified transcription factors, we selectively mutated them in the context of the yeast reporter construct. Mutations in the putative bZIP-binding site (ΔbZIP) (Fig. 2A,B) resulted specifically in the loss of response to the four bZIP factors. Conversely, mutations in the GARP motifs selectively interfered with reporter gene expression in response to the GARP transcription factors(ΔGARP) (Fig. 2A,B). These results demonstrated that the bZIP and GARP transcription factors identified in the one-hybrid screen bind in a sequence-specific manner to distinct motifs within the AAGAAT box.

If any of the identified transcription factors were to play a role in the activation of AG transcription, one would expect overlapping expression in early flowers, where AG mRNA is first detected. To address this issue, we queried the AtGenExpress database(Schmid et al., 2005). From the nine candidates, only the bZIP transcription factor gene PANshowed strong and specific expression in apices and flowers(Fig. 2C)(Chuang et al., 1999), while the other mRNAs are expressed more uniformly across the various tissues. pan mutants have prominent defects in floral organ number(Fig. 3G,H)(Chuang et al., 1999; Running and Meyerowitz, 1996),whereas T-DNA insertion mutants for the other factors identified in the one-hybrid screen did not have any floral phenotypes (see Table 1), suggesting that among the one-hybrid candidates only PAN has important roles in flower development, or that genetic redundancy is masking the function of the other regulators.

In situ hybridization confirmed a large overlap of PAN and AG expression domains in flowers(Fig. 3A,B). To investigate the importance of PAN for the activity of AG regulatory sequences, we introduced the KB14 reporter into a pan T-DNA insertion mutant, in which PAN RNA expression was reduced below the levels detectable by in situ hybridization(Fig. 3B,C). 5′AG::GUS reporter activity was drastically reduced in early pan flowers (Fig. 3D,E; see Fig. S4 in the supplementary material), with some residual GUS activity at later stages of flower development(Fig. 3E, white arrowheads). These results demonstrated that PAN is required for activity of AG regulatory elements and were in agreement with PAN not being expressed beyond floral stage 7 (Fig. 3B).

A similar effect on AG::GUS expression as in the panmutant background was observed when the bZIP-binding site was mutated in the context of the KB14 reporter (Fig. 3F; see Fig. S4 in the supplementary material). In 18 of 25 primary inflorescences from T1 plants, GUS activity could not be detected at all, while the remaining seven apices showed only weak staining in young flowers. This was consistent with results of transactivation assays in yeast,which showed that PAN can synergistically activate transcription from the AAGAAT box in concert with a GARP transcription factor, At4g37180 (see Fig. S2 in the supplementary material), which is expressed in an overlapping domain with PAN mRNA (see Fig. S3 in the supplementary material). Interestingly, the closely related TGA proteins showed repressive activity in the same assay (see Fig. S2 in the supplementary material). Taken together,these results confirmed the relevance of PAN and the bZIP-binding motif for transcriptional activation mediated by AG enhancer sequences.

Quantitative analyses by RT-PCR, however, did not indicate any reduction of endogenous AG RNA levels in pan mutants (data not shown),suggesting that other factors act redundantly with PAN. This is consistent both with pan mutants not having ag-like defects(Fig. 3G-I)(Running and Meyerowitz,1996), and with the 5′ enhancer fragment (KB14)from the second AG intron acting redundantly with the non-overlapping 3′ enhancer fragment (Fig. 1A) (Busch et al.,1999; Deyholos and Sieburth,2000).

In contrast to the pan mutant defects, which are restricted to flowers, PAN protein is much more widely expressed, indicating that PAN has redundant functions in several different regulatory networks(Chuang et al., 1999; Running and Meyerowitz, 1996). We speculated that genetic perturbations have so far not been able to fully expose PAN function, and we therefore tested whether variations in environmental conditions could be a means to elucidate the role of PAN in AG activation. Photoperiod has been shown before to affect the phenotypes of plants that are homozygous for a mutation in AG, or heterozygous for a mutation in the AG activator LFY (Okamuro et al.,1996).

Thus, we compared pan mutants and wild-type plants grown under short days (SD; 8 hours light), long days (LD; 16 hours light) and continuous light (CL), all at 23°C. Whereas the floral defects of panmutants grown under CL and LD were limited to the floral organ number defects known before, SD caused a loss of determinacy, reminiscent of agloss-of-function defects (Fig. 4A-F). The phenotypes ranged from partially unfused carpels, which developed into deformed and bulged siliques, to fully unfused central organs that had carpeloid features and on which new floral meristems arose. These meristems in turn produced mostly petaloid, stamenoid and carpeloid tissues(Fig. 4D,F; see Fig. S5 in the supplementary material). These strong phenotypes were most apparent among the first ten flowers of the primary inflorescence. Later-arising flowers showed less severe defects consisting mainly of bulged carpels(Fig. 4B; see Fig. S5A in the supplementary material), which after dissection showed growth of floral organs in an aberrant fifth whorl. Together, these defects were detectable in roughly half of all fruits on the primary inflorescence(Fig. 4G). Similar defects were found in plants carrying other pan mutant alleles grown in SD (not shown) and plants expressing dominant-negative alleles of PAN in LD(Das et al., 2009). Whereas the reduction of KB14 activity in pan mutant inflorescences (see Fig. S6 in the supplementary material) was similar to that observed in LD-grown plants, SD-grown pan mutants showed in addition a marked reduction in AG mRNA expression in early flowers(Fig. 5A,B; see Fig. S7 in the supplementary material), which was in agreement with the phenotypic defects.

At later stages of flower development in SD-grown pan mutants we detected ectopic expression of AG(Fig. 5C,D, white arrowheads)and activation of the KB14 reporter (see Fig. S8 in the supplementary material), consistent with the ectopic formation of reproductive floral organs(Fig. 4A-F; see Fig. S5C in the supplementary material). This could be regarded as a consequence of losing proper early AG activation in pan mutants and thus floral meristem determinacy. Consistent with such a scenario, we found tissue outgrowths that resembled floral meristems in shape and tissue layer organization (Fig. 5D, black arrowhead). To determine the identity of these structures more directly, we assayed expression of the stem-cell regulator WUS(Mayer et al., 1998). In wild-type flowers, WUS expression is terminated in an AG-dependent fashion around stage 6, when patterning of the flower is accomplished (Lenhard et al.,2001; Lohmann et al.,2001). In line with the observation of ectopic meristems initiating inside the developing gynoecium of SD-grown pan mutants,we observed WUS expression in these tissues(Fig. 5F, black arrowhead; see Fig. S9A in the supplementary material). Although during early stages WUS expression was rather diffuse in the emerging meristems, once these meristems had firmly established themselves, WUS mRNA was confined to the organizing center (see Fig. S9B in the supplementary material). We conclude that ectopic expression of AG and WUSat later stages of flower development reflects indirect effects of compromised PAN activity during earlier stages, and that redundant activators of AG are normally active throughout flower development.

Taken together, our results show that PAN plays an essential role in the activation of AG and that both genes are embedded in a regulatory network that is sensitive to environmental conditions.

The fact that PAN mRNA is not restricted to flowers but is also highly expressed in proliferating cells of the shoot apical meristem suggested more general functions of PAN in meristem maintenance. A similar case is represented by WUS, which has dual roles in stem-cell induction and floral patterning via AG activation. Consequently, WUSexpression and thus stem-cell maintenance is terminated during flower development to allow for tissue differentiation, and this process is dependent on AG activity (Lenhard et al.,2001; Lohmann et al.,2001). Thus, we tested whether a similar feedback interaction existed between PAN and AG. Consistent with such a scenario,we found that strong PAN RNA expression persisted much longer in ag mutant flowers than in wild type(Fig. 6A,B), demonstrating that PAN is not only an essential activator of AG during early floral stages, but that at the same time PAN expression is under control of AG at later stages. As the PAN expression domain in large parts overlaps with WUS RNA, and because PAN expression also persists in clavata mutants(Chuang et al., 1999), in which WUS is ectopically expressed, we investigated whether WUS activity might mediate the feedback between AG and PAN. To this end,we analyzed PAN RNA distribution in flowers ectopically expressing WUS from the CLAVATA3 (CLV3) promoter(Brand et al., 2002). Flowers of plants with intermediate phenotypes develop meristematic tissues from which stamenoid and carpeloid organs arise, due to the activation of AG in these cells. In agreement with the hypothesis that WUS mediates at least a good part of the regulatory interaction between AG and PAN, we observed ectopic PAN expression as well as accumulation of AG transcripts in WUS-positive cells(Fig. 6C,D).

In this study, we have used a short, conserved enhancer motif as a starting point for the identification of a new regulator of AG expression, the bZIP transcription factor PAN. We have shown that PAN not only has an important role in the control of perianth organ number specification(Chuang et al., 1999; Running and Meyerowitz, 1996),but also in the regulation of floral determinacy through the direct activation of AG. The identification of PAN, which is also expressed in shoot meristems (Running and Meyerowitz,1996), as an AG activator supports the model that flower-specific factors such as LFY interact with factors expressed in similar patterns in both shoot and floral meristems to control region- and flower-specific expression of floral homeotic genes(Lee et al., 1997; Lenhard et al., 2001; Lohmann et al., 2001; Parcy et al., 1998).

PAN belongs to the D-group of bZIP transcription factors, which share a high degree of sequence similarity in the bZIP region(Jakoby et al., 2002). Intriguingly, all bZIP transcription factors isolated in our screen fall into this group. Despite the fact that we have recovered multiple clones for all factors identified in the screen, we did not isolate all D-group members in our screen. As at least TGA2 and TGA3 are expressed at substantial levels in meristematic and floral tissue and therefore should have been represented in our library, it is tempting to speculate that these factors might have different DNA-binding specificities.

The findings that PAN can synergize in yeast with the GARP transcription factor At4g37180 and that both share overlapping expression domains in the SAM and early flowers suggest that At4G37180 and related GARP transcription factors are co-factors of PAN in AG regulation. Regulators of diverse molecular nature can act together in a redundant fashion during flower development and in particular during the activation of AG, as recently demonstrated by Prunet et al.(Prunet et al., 2008). However, because the family of GARP transcription factors is rather large(Riechmann et al., 2000), and because we have found members of both GARP subgroups (ARR-B and GARP-G2), it will be difficult to identify and functionally test the most promising candidates for roles in AG regulation.

Similar to what has been described for the stem-cell regulator WUS(Lenhard et al., 2001; Lohmann et al., 2001), PAN not only contributes to AG activation but in turn is under negative control by AG during later stages. The finding that WUS is able to ectopically activate PAN expression, along with the fact that PAN RNA also persists in clv mutants(Chuang et al., 1999),suggests that WUS could at least partially mediate this feedback regulation. In such a scenario, WUS at the same time contributes to the activation of PAN and AG in a feed-forward loop at early stages of flower development, while the repression of WUS by AG at later stages would also lead to a reduction of PAN expression(Fig. 7). The lack of WUS repression in ag mutants(Lenhard et al., 2001; Lohmann et al., 2001) would in turn cause ectopic activation of PAN. However, as only the regulatory interactions between WUS and AG and PAN and AG have been characterized mechanistically, the true nature of this regulatory module remains to be elucidated.

Finally, our work has revealed that other factors must act redundantly with PAN and that this network is sensitive to variation in environmental conditions. This finding shows that the combination of genetic perturbation with environmental variation can be a powerful tool to uncover redundant regulatory mechanisms that are normally not thought to be under environmental control.