Regulation of APETALA3 floral homeotic gene expression by meristem identity genes

Arabidopsis plants undergo a transition from vegetative growth, in which the shoot apical meristem produces leaves and axillary buds, to reproductive growth, when the meristem begins to form flowers. The meristem identity genes LEAFY (LFY) and APETALA1 (AP1) are both necessary and sufficient for this transition to reproductive growth and the concomitant formation of flowers. Loss of LFY function leads to leaves and shoots in place of flowers, while constitutive expression of LFY results in precocious floral development (Weigel et al., 1992; Weigel and Nilsson, 1995). Similarly, constitutive AP1 expression also results in premature flowering, and loss of AP1 function results in a partial transformation of flowers to more inflorescence-like structures (Bowman et al., 1993; Irish and Sussex, 1990; Mandel and Yanofsky, 1995).

In turn, the development of floral structures depends on the action of three classes of floral homeotic genes, A, B and C. These ABC floral homeotic genes function in overlapping domains to specify different floral organ identities. AP1, in addition to its role as a meristem identity gene, has a second role as an A class gene and is required for the development of sepal and petal primordia (Bowman et al., 1993; Irish and Sussex, 1990; Mandel et al., 1992). The B class genes APETALA3 (AP3) and PISTILLATA (PI) specify petal and stamen identities (Bowman et al., 1989; Goto and Meyerowitz, 1994; Jack et al., 1992), while AGAMOUS (AG), the C class gene, is responsible for conferring stamen and carpel identities (Bowman et al., 1989; Yanofsky et al., 1990). To a large extent, the functions of these ABC genes in specifying different organ identities correspond to their domains of expression in the developing flower (Riechmann and Meyerowitz, 1997; Weigel and Meyerowitz, 1994).

LFY is required for the transcription of representatives of all three classes of ABC genes (Weigel and Meyerowitz, 1993). LFY encodes a nuclear-localized product that can bind to DNA and so could act directly to regulate transcription of the floral homeotic genes (Parcy et al., 1998). In fact, LFY protein has been demonstrated to bind to sequences in the AP1 and AG enhancer regions that are required for normal levels of expression from these genes (Busch et al., 1999; Parcy et al., 1998). Ectopic expression of LFY is sufficient to induce the expression of AP1 outside the flower (Parcy et al., 1998). Furthermore, the activation of early AP1 expression by LFY is not dependent on protein synthesis, demonstrating that LFY is a direct transcriptional activator of AP1 (Wagner et al., 1999). While ectopic expression of LFY is insufficient to ectopically activate AG, expression of a dominant, activated form of LFY, LFY:VP16, can induce AG expression in vegetative tissues (Parcy et al., 1998). These observations suggest that AG expression does not depend on LFY alone, but the requirement of other factors for AG activation can be bypassed by LFY:VP16 (Parcy et al., 1998). One such factor is the WUSCHEL (WUS) homeodomain protein, which cooperatively interacts with LFY to regulate AG expression (Lenhard et al., 2001; Lohmann et al., 2001).

The effects of LFY on AP3 expression are more complex. LFY is required for AP3 expression, since a loss-of-function lfy-6 mutant shows a dramatic reduction in the levels and domain of AP3 activation (Weigel and Meyerowitz, 1993). However, ectopic expression of either LFY or LFY:VP16 does not significantly affect AP3 expression (Parcy et al., 1998). These observations suggest that the regulation of AP3 is considerably different from that of AP1 or AG, and that LFY may not directly activate AP3 expression. Alternatively, LFY activation of AP3 may occur directly by binding to the AP3 promoter, but other cofactors may be required for transcriptional activation to ensue.

AP1 has also been implicated in the regulation of AP3 gene expression. AP1 encodes a MADS-domain containing protein that binds to sequences in the AP3 promoter that are required for normal AP3 expression (Hill et al., 1998; Tilly et al., 1998). While AP3 expression is almost normal in ap1 mutant flowers, plants containing both the strong lfy-6 and ap1-1 alleles show a complete abolition of AP3 expression, reflecting the synergistic action of both LFY and AP1 in activating AP3 expression (Weigel and Meyerowitz, 1993). Furthermore, plants containing an activated form of AP1, AP1:VP16, display a partial transformation of medial first whorl organs into petals that is dependent on AP3 function, supporting the idea that AP1 positively regulates AP3 (Ng and Yanofsky, 2001).

In order to begin to dissect the molecular mechanisms by which these meristem identity genes function, we have analyzed the role of LFY and AP1 in regulating AP3 transcription. Previously, we have shown that the AP3 promoter contains distinct cis-acting elements that are required for the different spatial and temporal aspects of AP3 expression (Hill et al., 1998; Tilly et al., 1998). Here we show that LFY protein can bind to sequences within the AP3 promoter that are required for early AP3 expression. Furthermore, using an inducible form of LFY, we show that LFY acts both directly and indirectly to regulate AP3 expression, and that the indirect pathway depends on the function of the AP1 floral homeotic gene. Mutations of the LFY binding site in the AP3 promoter fail to abrogate AP3 expression in planta, suggesting that the indirect pathway may be sufficient to induce AP3 expression in this context. Based on these observations, we propose a model for how these meristem identity genes act together to activate the expression of the AP3 floral homeotic gene.

Plants were grown under 16 hours light/8 hours dark conditions. To generate 35S::LFY-GR; 35S::UFO seedlings, 35S::UFO/+ pollen was used to fertilize 35S::LFY-GR/35S::LFY-GR plants and the resulting offspring used in bulk for experiments. Similarly, pollen from 35S::UFO/+ plants were was used to fertilize 35S::LFY/35S::LFY plants and pollen from 35S::UFO/+; ap1-1/ap1-1 plants was used to fertilize 35S::LFY; ap1-1/ap1-1; the appropriate offspring were chosen by phenotype in the next generation. Pollen from 35S::LFY-GR homozygous plants was used to fertilize lfy-6/lfy-6 and ap1-1/ap1-1 plants, the offspring allowed to self fertilize and the correct genotype identified by PCR in the next generation. Transgenic lines were generated using the floral dip method (Clough and Bent, 1998) and selected on medium containing 30 mg/l kanamycin. β-glucoronidase (GUS) enzymatic activity was detected as previously described (Hill et al., 1998). At least three independent transgenic lines containing a single insert (as assessed by Southern analysis) were crossed into the various genetic backgrounds and were evaluated for their patterns of GUS expression.

Proteins for EMSA were produced by in vitro transcription of the LFY cDNA which was translated in vitro using wheat germ extract (Promega, Madison, WI). Preparation of DNA probes, binding reactions, and gel conditions were as described previously (Hill et al., 1998). Fragments used for cold competitors in the binding reactions corresponded to the following sequences within the AP3 promoter and were generated by polymerase chain reaction (PCR): competitor 1, –727 to –554; competitor 2, –705 to –587; competitor 3, –662 to –554; competitor 4 –727 to –626; competitor 5, –618 to –554; competitor 6, –727 to –678. Competitors were cleaned by ammonium acetate precipitation, quantitated, then added to the binding reaction in amounts of 10,000-fold molar excess over labeled probe. The following oligos were annealed with their complimentary oligos and cloned into the EcoRV site of the pBluescript SKII+ vector (Stratagene, La Jolla, CA): AP3I (site I), 5′-CTT AAA CCC TAG GGG TAA TA-3′; AP3Im, 5′-CTT AAA CCC TAT TATTAA TA-3′; AP3II (site II), 5′-TTC TAT TTT CCA AGG ATC TTT AGT TAA AGG C-3′; AP3IIm, 5′-TTC TAT TTT CCA ATT ATC TTT AGT TAA AGG C-3′; AP3I-II, 5′-CTT AAA CCC TAG GGG TAA TAT TCT ATT TTC CAA GGA TCT TTA GTT AAA GGC-3′; AP3Im-II, 5′-CTT AAA CCC TAT TAT TAA TAT TCT ATT TTC CAA GGA TCT TTA GTT AAA GGC-3′; AP3I-IIm, 5′-CTT AAA CCC TAG GGG TAA TAT TCT ATT TTC CAA TTA TCT TTA GTT AAA GGC-3′; AP3Im-IIm, 5′-CTT AAA CCC TAT TAT TAA TAT TCT ATT TTC CAA TTA TCT TTA GTT AAA GGC-3′. Mutated sequences are shown in bold. These fragments were released from the vector using HindIII and EcoRI to produce gel shift probes.

An in frame GAL4AD:LFY fusion was created in the pGAD424 vector and transformed into yeast strain YM4271. The DEE promoter fragment was trimerized and cloned into the EcoRI site of the placZi vector resulting in trimers fused in both the + and – orientations with respect to the lacZ reporter gene. The DEE-lacZ fusions were then integrated into the yeast strain YM4271 containing the GAL4AD:LFY construct. Yeast transformation protocols, yeast vectors and yeast strains were obtained from Clontech (Palo Alto, CA). To determine lacZ expression, yeast colonies were grown on selection medium for 2 days, and then 1 ml of the culture was resuspended in 2 ml of YPD and grown at 30°C. After 3-4 hours, the OD₆₀₀ of the culture was measured, and 1 ml of the culture was spun down and resuspended in 800 μl of Z buffer (60 mM Na₂HPO₄, 40 mM NaH₂PO₄, 10 mM KCl, 1 mM MgSO₄, 50 mM 2-mercaptoethanol, pH 7). One drop of 0.1% SDS and 2 drops of chloroform were added to each tube to lyse the cells. The suspension was then equilibrated at 30°C for 15 minutes. Subsequently, 160 μl of 4 mg/ml ONPG (Sigma, St. Louis, MO) was added, and the reaction was allowed to take place for 2 hours. The reaction was stopped with 400 μl of 1 M sodium carbonate, and the OD₄₂₀ and OD₅₅₀ of the suspension was noted. The units of β-gal activity were calculated using the following formula: U=1000 × [(OD₄₂₀) – (1.75 × OD₅₅₀)]/[time (minutes) × OD₆₀₀]. Replicate assays were conducted using five colonies of each construct tested per assay, and similar results were obtained in all assays.

The mutated promoter constructs were generated using PCR from plasmids p5D3, pD3-36 and pD3-18 (Hill et al., 1998) using the primers AP3IM-II, AP3I-IIM or AP3IM-IIM and vector primers. The resulting products were cloned into the TOPO TA vector (Invitrogen, Carlsbad, CA) and sequenced. In frame translational fusions at the ATG of the mutated AP3 promoter constructs and the GUS reporter gene were created by cloning the various promoter constructs into the SalI and BamHI sites of pBI101 (Clontech, Palo Alto, CA). Expression constructs were transferred to Agrobacterium tumefaciens strain GV3101 by electroporation.

Dexamethasone (DEX; Sigma, St. Louis, MO) was dissolved in ethanol and used at a final concentration of 1 μM on seedlings. For inhibition of protein synthesis, 10 μM cyclohexamide (CHX; Sigma, St Louis, MO) was added simultaneously with the DEX treatment. For seedling treatments, wild-type or transgenic seedlings were grown for 5 days on growth medium then transferred to media containing either DEX or DEX/CHX for 16 hours. For inflorescence treatments, plants were grown on soil under long day conditions (16 hours light, 8 hours dark). The primary bolt was cut 1-2 days after the start of bolting and after 24-48 hours individual inflorescences were treated with 5 μM DEX, 10 μM CHX, or DEX/CHX as described previously (Wagner et al., 1999). RNA was extracted from seedlings and inflorescences using Trizol (GibcoBRL, Frederick, MD) according to the manufacturer’s instructions. cDNA was synthesized using Superscript II RNase-Reverse Transcriptase (Invitrogen, Carlsbad, CA) according to the manufacturer’s instructions. Real time PCR reactions were carried out using an ABI Prism 7000 Sequence Detector (Applied Biosystems, Foster City, CA) in MicroAmp Optical 96-well reaction plates with optical covers, according to manufacturer’s instructions. PCR reactions (final volume 50 μl) contained TaqManMGB gene-specific probe and primers and the passive reference dye ROX, in order to normalize fluorescence across the plate. In all experiments, controls without template were used and at least two replicates using at least two independent RNA samples were used. AP3-specific TaqManMGB probe was conjugated to the fluorescent dye JOE and AP1-specific TaqManMGB probe was conjugated to the fluorescent dye FAM. Reaction conditions were: 50°C for 2 minutes, 94°C for 10 minutes, followed by 40 cycles of 94°C for 15 seconds, 60°C for 1 minute. Primers and probes were designed using Primer Express software (Applied Biosystems, Foster City, CA) to flank introns so genomic DNA contamination would not amplify. AP3 primers: AP3F 5′ CCACCAGAACCATCACCACTATT; AP3R 5′ GTCAGAGGCAGAGGGTGCAT. AP3 TaqManMGB probe: 5′ CCCAACCATGGCCTT. AP1 primers: AP1F 5′ TGAGCTGGAACTAAGAGCTGAAGA; AP1R 5′ AACTGAGTCGTAATCTCCTCCATTG. AP1 TaqManMGB Probe: 5′ CCTCACTATGGACTACTAG. Relative quantification values and standard deviations were calculated using the standard curve method according to the manufacturer’s instructions (ABI Prism 7000 Sequence Detection System User Guide). Values were normalized to the mock treated sample and results analyzed with Microsoft Excel software.

Previously we have identified discrete cis-acting elements in the AP3 promoter which are required for different aspects of AP3 expression (Hill et al., 1998). We defined two elements, the proximal early element (PEE) and distal early element (DEE), which are required for AP3 expression during early stages of floral development, from stages 3 to 5 (Fig. 1). In addition, other promoter elements were identified that are required for petal or stamen-specific expression at later stages of development.

The 5D4 promoter fragment consists of 727 bp and contains all promoter elements required for AP3 expression (Hill et al., 1998). We crossed a 5D4::GUS reporter gene construct into a lfy-6 mutant background to test whether LFY acts via these AP3 promoter sequences. The strong lfy-6 mutation results in the transformation of flowers to a more inflorescence-like structure of spirally arranged organs that have leaf-like, sepaloid and/or carpeloid features (Weigel et al., 1992). Flowers homozygous for lfy-6 lack stage 3-5 expression driven by the 5D4::GUS construct (Fig. 2A,C). This demonstrates that LFY is required for initial activation of AP3 expression and this action is mediated by this 727 bp sequence within the AP3 promoter. Disruption of 5D4::GUS expression is also seen at later stages in a lfy-6 background, with GUS activity limited to just the base of the developing second and third whorl organs (Fig. 2B,D). This reduction in later AP3 expression presumably reflects the lack of AP3 autoregulation (Jack et al., 1992).

Since LFY has been shown to bind to cis-acting regulatory sequences in the AP1 and AG floral homeotic genes (Busch et al., 1999; Parcy et al., 1998; Wagner et al., 1999), it seemed likely that LFY protein could also bind to AP3 regulatory sequences. Using electrophoretic mobility shift assays (EMSA), we assayed the ability of LFY protein to bind to different sequences within the AP3 promoter. LFY protein was found to bind to a promoter fragment that contains the distal early element (DEE) (Fig. 3A). LFY did not bind to any other AP3 promoter fragments in EMSA, including the PEE that is also required for early AP3 expression (Fig. 1; data not shown).

Competition assays demonstrated that LFY protein binding mapped to sequences contained within the DEE (Fig. 3A). Cold competitors corresponding to different regions of the –727 to –554 region of the AP3 promoter were used to compete for LFY binding to the labeled –727 to –554 test fragment. The smallest region, as defined by various competitors, that is capable of competing for binding to LFY corresponded to the region from –705 to –626. Furthermore, the sequence from –727 to –678 did not interfere with LFY binding, delimiting the LFY binding site region to the 52 bp region from –678 to –626.

This 52 bp sequence could be divided into two regions, termed site I (corresponding to basepairs –678 to –658) and site II (–657 to –626), each of which showed some sequence similarity to the LFY binding sequences found in the regulatory regions of the AP1 and AG genes (Fig. 3B). These two sites each contain a palindromic sequence that overlaps the putative LFY binding sites. Mutated versions of site I and site II were generated that disrupted both the palindromic sequences and the presumptive LFY binding sites (Fig. 3B). We tested the ability of LFY to bind to these sites by assaying the wild-type and mutated versions of each site either individually or within the same 52 bp DNA fragment using several assay systems. First, we checked the ability of these sequences to bind to LFY in yeast one-hybrid assays (Fig. 3C). Trimerized versions of wild-type or mutated site I and/or site II sequences were fused to the lacZ coding sequence and introduced into the yeast genome; these strains were assayed for lacZ expression in the presence of a LFY-GAL4 activation domain (LFY-AD) fusion gene product. In addition, the wild-type and mutated oligonucleotide sequences were used in EMSAs to test their ability to bind to LFY (Fig. 3D). Both these assays gave similar results and indicated that site I is necessary and sufficient for LFY binding, while the site II sequence alone is not sufficient. However, LFY cannot bind when site II is mutated in the context of the entire fragment, indicating that intact site II is required for LFY binding in this context. This may reflect a requirement for a particular DNA conformation for LFY binding.

We assessed whether LFY acts directly or indirectly to activate AP3 transcription in vivo by utilizing a ubiquitously expressed inducible form of LFY, 35S::LFY-GR (Wagner et al., 1999) to induce AP3 expression in several contexts in the presence or absence of the protein synthesis inhibitor, cyclohexamide (CHX). The LFY-GR fusion protein is localized to the cytoplasm and thus is inactive, but can be induced to localize to the nucleus and function by treatment of plants with dexamethasone (DEX) (Wagner et al., 1999).

Because AP3 is not normally expressed in seedlings, we took advantage of the fact that constitutive expression of LFY in conjunction with UFO results in ectopic AP3 transcription in seedlings (Parcy et al., 1998). Inducing LFY expression at the seedling stage provides an ideal way to directly assess the effects of LFY expression on AP3 transcriptional activation. 35S::LFY-GR; 35S::UFO seedlings were grown without DEX for 5 days, then treated with 1 μM DEX. Upon DEX treatment, these seedlings arrest their normal development, but recover if DEX is removed. The 35S::LFY-GR; 35S::UFO seedlings were treated with DEX alone, or concomitantly with 10 μM CHX at day 5, and seedling tissue was harvested after 16 hours of treatment and examined for levels of AP3 expression. The relative levels of AP3 expression with the various treatments were assessed using a quantitative real time reverse transcription PCR (RT-PCR) approach. The treatment of 35S::LFY-GR; 35S::UFO seedlings with DEX posttranslationally activated LFY and resulted in induction of AP3 expression (Fig. 4A). CHX treatment somewhat reduced the levels of AP3 expression in these DEX-induced seedlings, but these levels are significantly above that of seedlings treated with CHX alone (Fig. 4A). These observations indicate that part of the LFY-dependent activation of AP3 expression in the seedling requires protein synthesis, and implies that the activation of AP3 by LFY in this context is indirect. In addition, the fact that we could observe significant levels of AP3 expression in these DEX/CHX-treated seedlings indicates that LFY can also function in a direct manner to activate AP3 expression.

We also examined whether AP1 expression can be directly induced in the presence of DEX and CHX in 35S::LFY-GR, 35S::UFO seedlings. AP1 is directly activated in response to LFY in floral tissue (Wagner et al., 1999). In contrast to what we observed for AP3, the levels of DEX-inducible AP1 expression in the presence of CHX was equivalent to that of treating with DEX alone (Fig. 4B). This implies that all AP1 expression in these seedlings is dependent on direct activation by the LFY protein.

Since LFY directly activates AP1 expression in seedlings, we examined whether LFY-dependent expression of AP3 required AP1 activity. 35S::LFY; 35S::UFO; ap1-1 seedlings were examined for the presence of AP3 transcripts using quantitative real time RT-PCR. An examination of the relative levels of AP3 ectopic activation in a 35S::LFY; 35S::UFO; ap1-1 background, as compared to a 35S::LFY; 35S::UFO background indicates that overall levels of AP3 expression are significantly reduced when AP1 function is absent (Fig. 4C). However, some AP3 expression is still detectable in the 35S::LFY; 35S::UFO; ap1-1 seedlings, supporting the idea that LFY can act independently of AP1, and presumably directly, to activate AP3 transcription.

We also examined AP3 expression induced by 35S::LFY-GR in floral tissue. DEX application to young 35S::LFY-GR flower buds did not induce AP3 expression above endogenous levels (data not shown). In order to eliminate AP3 expression dependent on endogenous LFY-dependent pathways, we examined the induction of AP3 expression in a 35S::LFY-GR; lfy-6/lfy-6 background. In these flower buds, DEX application induced AP3 expression, and simultaneous DEX/CHX application somewhat reduced AP3 expression, but did not abrogate this expression completely (Fig. 4D). This is similar to what we observed in 35S::LFY-GR, 35S::UFO seedlings (Fig. 4A) and implies that LFY can act in both a direct and an indirect manner to activate AP3 expression in flowers. As a control, we also examined the levels of AP1 expression induced in 35S::LFY-GR; lfy-6/lfy-6 flowers (Fig. 4E). Similar to what has been observed previously, LFY appears to act directly to activate AP1 transcription in these tissues (Wagner et al., 1999).

To determine whether LFY-dependent induction of AP3 depended on AP1 activity in flowers, we examined the levels of AP3 expression in 35S::LFY-GR; ap1-1/ap1-1 flower buds. DEX treatment of 35S::LFY-GR; ap1-1/ap1-1 plants (Fig. 4F) resulted in levels of AP3 expression that were higher than the level produced in the presence of DEX and CHX, indicating that LFY can act via an AP1-independent indirect pathway to activate AP3 expression in the flower.

Because our real time RT-PCR results indicated that LFY acts in both a direct and an indirect manner to regulate AP3 transcription, we chose to test the significance of the in vitro defined LFY binding sites in planta. We generated transgenic plants in which site I and/or site II sequences were mutated in the context of several diagnostic AP3 promoter constructs (Fig. 1). The same site I and/or site II mutations used in the EMSA and yeast-one hybrid assays were introduced into the 5D3 and the D3-18 (lacking the PEE but containing the DEE) AP3 promoter constructs (Fig. 1) and fused to the GUS reporter gene. These mutated reporter gene constructs were stably transformed into Arabidopsis plants, and at least three independent single insertion lines for each construct were analyzed for reporter gene expression in flowers. We also examined the expression of these reporter constructs in 35S::LFY; 35S::UFO seedlings. Mutation of sites I and II individually, or of both sites simultaneously, has no obvious effect on reporter gene expression in either the wild-type or 35S::LFY; 35S::UFO backgrounds (Fig. 5 and data not shown). This is quite surprising in the case of the site I and site II mutated D3-18 construct, since this construct is mutated for the LFY binding site in DEE as well as being deleted for the PEE, the other promoter region required for stage 3-5 AP3 expression (Hill et al., 1998). Based on these observations, these results suggest that LFY-dependent activation of AP3 does not require that the identified LFY binding site be intact.

LFY is required for the onset of normal AP3 expression, since mutations in lfy result in significantly reduced levels of AP3 transcripts during early stages of floral organogenesis (Weigel and Meyerowitz, 1993). Using both in vitro electrophoretic mobility shift assays and in vivo yeast one hybrid assays, we have demonstrated that LFY protein binds to specific regulatory sequences in the AP3 promoter. This binding site lies within the DEE, which is required for stage 3 to 5 expression from the AP3 promoter (Hill et al., 1998). This binding is sequence specific, since mutation of this region in the AP3 promoter abolishes binding in vitro and in yeast one hybrid assays. Comparisons of the LFY binding sites in the AG and AP1 regulatory regions defines a consensus that is only loosely matched by the binding site we have defined in the AP3 promoter (Fig. 3B). The binding of LFY to regulatory sequences is therefore presumably quite sensitive to sequence context, similar to the context dependence of many other transcription factors. LFY does not appear to bind to any other sequences within the AP3 promoter in vitro (data not shown).

Based on our results, we propose a model for LFY and AP1 activation of AP3 transcription (Fig. 6). We suggest that there are at least four separate pathways that regulate the onset of AP3 expression in the flower.

One pathway appears to be the direct activation of AP3 transcription by LFY. Two lines of evidence support the existence of this pathway. First, we can demonstrate, using 35S::LFY-GR to induce AP3 expression in the presence of cycloheximide, that LFY protein can function in the absence of protein synthesis to activate AP3 transcription. Second, the DNA binding studies indicate that LFY can bind directly to sequences in the AP3 promoter that are required for early expression. However, mutation of the defined LFY binding site in planta does not abrogate AP3 expression (Fig. 5), which would appear to contradict the idea that LFY acts directly by binding to AP3 promoter sequences. We can suggest two possibilities to reconcile these observations. First, we may not have identified the relevant binding site, although we have not been able to detect LFY binding to any other sequences in the AP3 promoter using EMSA (data not shown). Alternatively, LFY may act to directly activate AP3 transcription through the binding site we have defined, but this pathway may be redundant with the indirect LFY-dependent pathways leading to AP3 activation. Therefore, loss of LFY binding to the AP3 promoter may not be sufficient to disrupt LFY-dependent activation of AP3 in planta.

The second LFY-dependent pathway we have defined is indirect and depends on the activity of AP1 to activate AP3 transcription. Loss of AP1 function significantly reduces the level of AP3 transcription that can be induced by LFY action (Fig. 4C). AP1 has been previously shown to be a positive regulator of AP3 (Hill et al., 1998; Krizek and Meyerowitz, 1996; Ng and Yanofsky, 2001; Weigel and Meyerowitz, 1993). Since LFY directly activates AP1 (this work) (Wagner et al., 1999), and AP1 has been shown to bind to sequences within the PEE (Hill et al., 1998), this short regulatory cascade may activate transcription through the PEE sequences. LFY direct activation of AP1 expression appears to be limited to early stages of floral development, when the pattern of AP1 expression is largely coincident with that of LFY in the floral meristem (Parcy et al., 1998; Wagner et al., 1999). At later stages, spatially restricted AP1 expression depends on other factors, which may include AG (Gustafson-Brown et al., 1994; Parcy et al., 1998; Wagner et al., 1999).

The fact that mutation of AP1 does not completely abolish LFY-dependent expression of AP3 suggests that LFY also regulates AP3 via a third pathway that is independent of AP1 (Fig. 4F). This appears to be an indirect pathway of activation via an unknown factor ‘X’ (Fig. 6). This indirect pathway potentially could depend on the products of the CAULIFLOWER (CAL) and FRUITFULL (FUL) genes, which are both paralogs of AP1 and appear to have overlapping functions (Ferrandiz et al., 2000); and so may act in a partially redundant fashion to weakly activate AP3 in the absence of AP1.

A number of other identified genes are also candidates for being involved in this third, LFY-dependent indirect pathway of AP3 activation. One such gene is UFO, which has been shown to encode a region-specific factor that is required in conjunction with LFY to activate AP3 (Lee et al., 1997). UFO does not appear to have DNA binding activity, and so presumably does not act as part of the transcriptional machinery (data not shown). It is more likely that UFO acts as part of an SCF (SKP1-Cullin-F-box) complex and targets specific proteins for ubiquitin-dependent degradation, since UFO encodes an F-box containing protein that has been shown to interact with SKP1-like gene products (Bai et al., 1996; Ingram et al., 1995; Samach et al., 1999). This postulated role of UFO has led to a model whereby UFO acts to promote the degradation of a putative negative regulator of AP3 (Samach et al., 1999). One possibility to explain the role of UFO in the indirect pathway regulating AP3 expression would be that LFY activates the transcription of the putative negative regulator while UFO acts to target it for degradation.

At least one other candidate gene has been identified which may act in this LFY-dependent indirect pathway activating AP3 transcription. A myb-domain containing DNA binding protein has recently been identified that binds to AP3 promoter sequences and appears to act as a positive regulator of AP3 transcription in vivo (C. Juarez, E. Chae, Q. K.-G. T. and V. F. I., unpublished data).

Finally, a fourth pathway that is independent of LFY can be defined, which requires an as yet unidentified factor or factors (‘Y’, Fig. 6). Low levels of AP3 expression are detectable at the base of the second and third whorls in lfy-6 mutant plants, indicating that not all AP3 expression is dependent on LFY function (Fig. 2). This pathway also appears to be independent of UFO as well as ASK1, a putative subunit of a UFO-containing SCF ubiquitin ligase complex, since mutations in either UFO or ASK1 still result in AP3 expression at the base of the second and third whorls (Levin and Meyerowitz, 1995; Zhao et al., 2001).

LFY has now been implicated in activating the transcription of representatives of all three A, B, and C classes of floral homeotic genes (Busch et al., 1999; Wagner et al., 1999; Weigel and Meyerowitz, 1993) (this work). Despite this global control of floral homeotic gene expression by LFY, each of these organ identity genes is expressed in a different spatially limited domain, implying that LFY acts in conjunction with other factors to delimit ABC gene activation. Our results suggest that LFY acts to regulate expression of AP3 in a manner distinct from that of AP1 or AG. The multiple AP3 regulatory pathways we have defined could act as a failsafe mechanism to ensure appropriate expression of AP3 and may reflect a requirement for the strict temporal control of expression of this floral homeotic gene. In light of these results, it seems likely that there are multiple complex regulatory interactions that serve to reinforce the precise spatial and temporal control of floral homeotic gene expression, which in turn is critical for normal floral patterning.

We would like to thank Dr Detlef Weigel for a gift of the LFY cDNA and the 35S::LFY and 35S::UFO lines, and Dr Doris Wagner for a gift of the 35S::LFY-GR line used in this study. We thank Dr Savithramma Dinesh Kumar for the use of his ABI Prism 7000 Sequence Detector and Tessa Burch-Smith for technical help with real-time PCR. We thank members of the Irish lab for comments on the manuscript. This work was supported by a grant from the USDA (no. 2001-35304-09990) to V. F. I.