PETAL LOSS, a trihelix transcription factor gene, regulates perianth architecture in the Arabidopsis flower

Flowers share a common underlying architecture. In general, reproductive organs occur centrally, with the female carpels internal to the male stamens. Surrounding these are the perianth organs, which in many species are differentiated into inner, showy petals and outer, protective sepals. The number of floral organs in a flower, their spatial relationships and their degree of fusion are relatively conserved properties, presumably under genetic control.

A range of genes that regulate floral architecture has been identified in Arabidopsis thaliana. The PERIANTHIA (PAN) bZIP transcription factor gene ensures that flowers arise with the appropriate numbers of organs in the outer three whorls (four sepals, four petals and six stamens) (Chuang et al., 1999). In pan mutants there are usually five in each case. By contrast, the architecture of the second and third whorls is supported by the UNUSUAL FLORAL ORGANS (UFO) F-box gene(Durfee et al., 2003; Laufs et al., 2003). Another gene, PRESSED FLOWER (PRS), encoding a homeodomain protein,has roles in defining the lateral regions of the flower primordium rather than the radial regions (Matsumoto and Okada,2001). In prs mutants, lateral sepals are frequently absent.

Other genes function to define boundaries. For example, the SUPERMAN (SUP) zinc finger gene acts to control cell proliferation in the boundary between the stamens and carpels(Sakai et al., 1995). In addition, boundaries between individual organs are controlled by CUP-SHAPED COTYLEDON (CUC) genes encoding NAC transcription factors. These act to keep the primordia of adjacent floral organs, especially the sepals, separate(Aida et al., 1997).

Genes required for the development of specific organ types have also been described. For example, another function of UFO is to promote petal outgrowth, perhaps by targeting an inhibitor of this process for degradation(Durfee et al., 2003; Laufs et al., 2003). Another gene that specifically promotes petal growth, RABBIT EARS(RBE), encodes a zinc finger protein(Takeda et al., 2004). In null rbe mutant plants, petals are mostly filamentous or absent.

The PETAL LOSS (PTL) gene of Arabidopsis plays a unique role in controlling perianth development(Griffith et al., 1999). In mutant plants, sepals are mis-shapen and sometimes fused with an adjacent sepal. Petals are often absent, and their mean number per flower falls progressively so that later-formed flowers usually have none. Those petals that do arise are often smaller than normal and are sometimes trumpet-shaped. Petal primordia occupy the same regions of the mutant flower primordium as in the wild type (internal to each of the inter-sepal zones), although the four regions are somewhat enlarged. Also, their initiation may be delayed. The number of petals per flower is influenced by the presence of a dominant allele of the PETAL LOSS MODIFIER (PMD) gene, with more petals arising per flower in pmd-1d background.

Griffith et al. proposed that PTL normally functions to support the action of a petal initiation signal (Griffith et al., 1999). This was proposed to act in four regions of the developing flower, internal to the inter-sepal zones. These regions might be enlarged in ptl mutants such that response to the signal is weakened and the threshold is only occasionally reached. It was also proposed that sensitivity to the signal is boosted in the pmd-1d background.

The orientation of petals within the flower is also disrupted in ptl mutants. Some face sideways, and others are reversed. Griffith et al. suggested that the response to another signal, acting with defined polarity within the flower primordium, was being disrupted in ptlmutants, such that petal primordia sometimes adopted an inappropriate, or even default, orientation (Griffith et al.,1999).

In this study, we report the identification of PTL as a transcription factor gene of the plant-specific trihelix family. PTLis the first member of this family known to control morphogenesis –others known to date are associated with the regulation of light-responsive genes. PTL is expressed in four zones between newly arising sepals,where it may help to maintain their separation. Consistent with this idea,ectopic expression of PTL results in growth inhibition. In addition,some fusion of adjacent sepals occurs in ptl mutant plants. Surprisingly, PTL is apparently not expressed in developing petal primordia. This suggests that the disruptions to petal initiation and orientation in ptl mutants are caused indirectly, perhaps as a consequence of overgrowth in the nearby inter-sepal zones.

ptl-1 (in Columbia) was induced by ethyl methane sulphonate (EMS),and ptl-2 (in C24) arose following transformation(Griffith et al., 1999). Three new mutants were provided by Stuart Baum (ptl-3) and Yuval Eshed(ptl-4 and ptl-5), both from the University of California Davis, following EMS mutagenesis of Landsberg erecta (Ler)plants (carrying the dominant modifier allele pmd-1d), and were shown not to complement ptl-1. The intermediate pinoid-2 mutant allele in Ler was from laboratory stocks(Bennett et al., 1995), and cuc1-1 and cuc2-1 mutants (in Ler) were provided by Masao Tasaka (Aida et al.,1997).

Plants were grown at 20-25°C in natural daylight supplemented with continuous Cool White fluorescent light. Stages of flower development follow Smyth et al. (Smyth et al.,1990).

Clones of Yeast Artificial Chromosomes (YACs) and TAMU Bacterial Artificial Chromosomes (BACs) for generating contigs in the region overlapping PTL were obtained from The Arabidopsis Biological Resource Center. PTL was localized to a 10.7 kb XbaI genomic fragment present in the right end of BAC T24M12 (Columbia ecotype). This was cloned into pBluescript SK(+) (Stratagene) for sequencing, or into the binary vector pBIN19 for complementation tests. In the latter, the construct (D289) was transferred to Agrobacterium tumefaciens strain AGL1, and then into ptl-1 mutant plants by bacterial infiltration of flower buds.

A cDNA library, made from Ler inflorescence tissue with buds up to stage 12 using λZAPII (Stratagene) (provided by Detlef Weigel, Salk Institute), was probed with BAC T14K22, and the positive clones probed in turn with the right end XbaI subclone of BAC T24M12. One positive clone(D171) was obtained (GenBank Accession number AY555728).

PTL coding sequences in cDNA clone D171 were translationally fused to the C terminus of the GFP coding sequence from pBIN mGFP5-ER (with ER deleted). This was inserted into plasmid pART7(Gleave, 1992) downstream of the 35S CaMV promoter sequence. Transient expression of 35S::GFP-PTL in onion epidermal cells, and GFP fluorescence, followed the protocol of Weigel and Glazebrook (Weigel and Glazebrook,2002).

RT-PCR followed the OneStep procedure (Qiagen) starting with total RNA extracted using the RNeasy procedure (Qiagen). Primers for transcripts of PTL and the two control genes APETALA3 (AP3) and ACTIN2 (ACT2) were designed to flank or overlap introns,generating cDNA products of 738 bp (PTL), 356 bp (AP3), or 1,153 bp (ACT2). Each reaction started with approximately 0.5 μg total RNA, and the PTL and AP3 reactions were identical(55°C annealing temperature with 20 seconds extension for 30 cycles)except that `Q solution' (Qiagen) was added to the PTL reaction. The ACT2 reaction conditions were 52°C, 30 seconds and 28 cycles.

In situ hybridization, using digoxigenin-labelled sense and antisense probes made using the cDNA plasmid D171, essentially followed the protocol of Heisler et al. (Heisler et al.,2001).

The PTL regulatory region was translationally fused with the uidA gene of E. coli encoding β-glucuronidase (GUS). The BAC clone T14K22 (Columbia genomic DNA) supplied the sequences used,either restriction fragments or PCR products. These were ligated into the shuttle vector pRITA (Eshed et al.,1999; Eshed et al.,2001) using an NcoI site present at the first methionine codon of GUS. After sequencing to confirm appropriate cloning, the pPTL::GUS casettes were cloned into the NotI site of the binary vector pART27(Gleave, 1992) (conferring kanamycin resistance in plants), or a Basta resistance derivative of this,pMLBART. These were then transferred into Columbia plants as before.

Five PTL reporter constructs were made, three with the first exon and intron (p8.0i::GUS, p2.0i::GUS and p1.3i::GUS), and two without (p8.0::GUS and p2.0::GUS). For the first three constructs, the fusion involved either the first methionine (p2.0i) or the first tyrosine (p8.0i and p1.3i) of the second exon. For the last two (p8.0 and p2.0), the fusion was at the second methionine of the first exon. Where PCR had been used for cloning, full sequencing revealed one change in both p8.0 and p8.0i (one less T in a string of 15 located 4,957 bp upstream of the first methionine), and 9 base substitutions in p2.0i. The latter did not change the translated sequence, and as the expression pattern for p2.0i was closely similar to that of the error free p1.3i, it was assumed the changes had had no functional consequences.

GUS staining of transformed lines was carried out by briefly fixing material in 90% acetone, staining overnight in 2 mM X-Gluc in 50 mM phosphate buffer (pH 7.2) at 37°C, and then removing chlorophyll using ethanol. Most staining solution also included 6 mM potassium ferri- and ferro-cyanide to reduce bleeding. Tissues were examined as whole mounts in 70% ethanol, or were embedded in Paraplast Plus, sectioned at 8 μm, and viewed using dark-field optics. The precipitated product of the β-glucuronidase reaction appears blue in bright field, but pink in dark field.

In this system (Moore et al.,1998), a driver line carries a promoter of choice driving a hybrid transcription factor (the lac repressor translationally fused with the GAL4 activation domain, LhG4). The target line carries the lac operator (pOp) in tandem copies upstream of a gene of interest. Expression of the target gene is induced in progeny of crosses between the two lines, here indicated as pDRIVER>>TARGET(Eshed et al., 1999; Eshed et al., 2001).

Two PTL driver constructs were made containing either 1.3 kb upstream of the first methionine codon, the first exon and the intron [pPTL(1.3i)], or 2.0 kb of the upstream region [pPTL(2.0)]. These PTL regulatory sequences were each inserted into the modified shuttle vector pBJ36-LhG4(Eshed et al., 2001), and then into the NotI site of pMLBART. ptl-1 plants were transformed, and 11 independent insertion lines carrying pPTL(1.3i), and seven of pPTL(2.0) were recovered. Their expression was assessed by intercrossing with a pOp::GUS target tester line (provided by Yuval Eshed, University of California Davis), and the patterns closely matched those of comparable pPTL::GUS translational fusion lines. Three other driver lines, carrying either 1.7 kb of the APETALA1 upstream promoter region (pAP1)(Emery et al., 2003), or 0.42 kb of the APETALA3 promoter region (pAP3), or the CaMV 35S control region present in 1.35 kb of pART7 (p35S)(Gleave, 1992), were also provided by Yuval Eshed.

A PTL target construct was made by inserting the PTL coding sequence from cDNA clone D171 downstream of pOp present in plasmid p12OpBJ36(Eshed et al., 2001). The insert was excised with NotI and inserted into pART27. Four independent target insertion lines were obtained in ptl-1 background,and 12 in Columbia background.

Scanning electron microscopy (SEM) was carried out as described previously(Griffith et al., 1999).

Previously it was briefly reported that neighbouring sepals are sometimes fused basally along their lateral edge in ptl-1 and ptl-2mutant plants (Griffith et al.,1999) (Fig. 1A,B). The three new ptl mutants were similarly affected. Intersepal fusions also occur in single mutants of either of the organ boundary genes cuc1 and cuc2 (Aida et al., 1997). We tested whether this fusion represented disruption of the same process as in ptl by scoring fusions in ptl cuc1and ptl cuc2 double-mutant flowers. In each case there was a dramatic increase in the number and extent of fusions(Table 1; Fig. 1C-F). Levels now approached the near-complete fusion seen in cuc1 cuc2 double-mutant plants (Aida et al., 1997). Thus the three genes PTL, CUC1 and CUC2 each apparently contribute to a process that results in sepal separation.

The PETAL LOSS gene was cloned based on its map position [0.5 map units above the TERMINAL FLOWER1 (TFL1) locus on chromosome 5 (Griffith et al., 1999)]. The ptl-1 mutant in Columbia background was crossed with the tfl1-2 mutant in Landsberg erecta. Using F2 plants, the PTL locus was mapped distal to TFL1 (four recombinants out of 854 chromosomes) and proximal to the DNA marker UBQ6121 (one recombinant out of 220 chromosomes). A YAC contig was generated in the proximal direction from UBQ6121, and PTL was found to lie between the right ends of YACs CIC2B9 and CIC10H2. Using these ends, a BAC contig was assembled across this region, and PTL was localized to BAC T14K22. A HaeIII polymorphism was located within this BAC using the right end of another BAC,T24M12, as a probe. Fortuitously, the ptl-1 mutant was uniquely associated with loss of one of the HaeIII sites identified by this probe.

A cDNA covering this site was isolated from an inflorescence library, and the candidate gene shown to correspond to PTL, in that a 10.7 kb genomic clone encompassing the gene (Fig. 2B) fully complemented the ptl-1 mutant phenotype. Also,all five mutant alleles (of similar phenotype) were associated with sequence changes that are likely to result in loss of function of the deduced polypeptide (Fig. 2A).

The PTL gene (At5g03680) encodes a transcription factor of the plant-specific trihelix family. Proteins of this family were first identified through their specific binding to conserved GT boxes within the promoters of light-regulated genes (hence they are also known as GT factors)(Zhou, 1999). The trihelix region (helix-loop-helix-loop-helix), duplicated in PTL, is a conserved DNA-binding domain distantly related to that of MYB transcription factors(Nagano, 2000).

The deduced PTL protein (Fig. 2A) contains a long helical region between the two trihelices that is conserved in family members closely related to PTL(Nagano et al., 2001). The center of this is predicted to form a coiled-coil(Fig. 2A), possibly associated with multimer formation. In addition, there is a serine-threonine rich region and a glycine rich region, each commonly found in transcription factors. There are also several putative nuclear localization signals(Dehesh et al., 1995), and we have shown that the PTL protein accumulates in the nucleus(Fig. 2C,D).

We were unable to detect PTL transcripts by northern hybridization, so we screened tissues using RT-PCR(Fig. 3). In comparison with AP3, a gene strongly and specifically transcribed in floral tissues(Jack et al., 1992), PTL is weakly expressed. The highest level was found in inflorescences (including flowers up to stage 12). Relatively high expression was also seen in 7-day-old seedlings, although not in seedlings one day younger. AP3 transcripts also increased markedly in seedlings between 6 and 7 days, suggesting that inflorescences had commenced development. Lower levels of PTL transcription were detected in the other tissues examined (Fig. 3).

Detection of PTL transcripts by in situ hybridization was difficult. A weak PTL signal was observed on occasion in the flower primordium from stages 4 to 6, occurring between the developing sepal primordia and in sepal margins (Fig. 4A).

To assess expression patterns in detail, a reporter gene construct was made that included all of the 8 kb region upstream of the PTL coding sequence (up to the neighbouring gene), the first exon and the single large intron translationally fused with GUS (p8.0i::GUS)(Fig. 2B). Twelve out of 14 independent insertion lines showed a consistent pattern of GUS staining(Table 2). Closely similar patterns were seen in transformants of two similar constructs including less of the 5′ upstream region (p2.0i::GUS and p1.3i::GUS)(Fig. 2B); the results will be considered together (Table 2).

Expression occurred in a range of tissues, but the strongest signal was seen in developing flower primordia (Fig. 4B,D-H). Discrete expression zones were present on the lateral flanks of each flower primordium from the time they arose (stage 1)(Fig. 4D-H). Expression became localized to the lateral indentations where flower primordia separate from the inflorescence meristem at stage 2. As this expression faded, two dumbbell-shaped expression zones arose alongside and internal to the lateral sepals that initiated at stage 3 (Fig. 4D-H). This expression then resolved into four discrete,cone-shaped zones lying between each of the developing sepal primordia at stage 4. This expression was relatively strong at first and matched the in situ hybridization pattern observed during stages 4-6(Fig. 4A), but the level fell progressively until it was undetectable by the end of stage 7.

Expression of PTL was also seen fully around the edges of developing sepals (Fig. 4A,D-H)from stage 5 when they enclose the developing flower meristem. It extended several cells inward from the epidermis, particularly in adaxial regions at the sepal base, persisting until around stage 9.

Petal primordia arise between the sepals and internal to them at stage 5. It was not possible to determine whether the few petal-initiating cells carried GUS product at this stage, but, surprisingly, it was not observed in the small developing primordia at stages 6-7(Fig. 4I,J). GUS product was first detected at around stage 8, just before petal primordia commence rapid growth (Fig. 4K). Even then, it was present only basally in the flanks, and this faded by stage 13 when petals were fully grown.

GUS product was also present in lateral regions of stamen primordia,specifically at the top of the filament and the base of the anther, from the time they differentiate from each other at stage 7(Fig. 4L) until stage 9.

In seedlings, PTL expression was first seen in developing leaves(Fig. 4M,N), commencing only after their primordia had extended over the shoot apical meristem. It was first present in a continuous band around their perimeter, extending through all three cell layers (Fig. 4N). Later, expression became limited to the basal region,including the petiole (Fig. 4M). Strong GUS staining was also seen in stipules throughout their life (Fig. 4N).

Weaker PTL expression was seen in the axils of cauline leaves and floral pedicels, with the intensity of GUS staining increasing as individual axils aged (Table 2). Staining intensity was somewhat lower on average in the shorter promoter lines,p2.0i::GUS and p1.3i::GUS, than in p8.0i::GUS lines. Some other vegetative regions were also stained, although relatively weakly(Table 2). These included the basal node of the rosette, the floral receptacle, and the vasculature of the root.

Overall, localization of the GUS reporter gene product is consistent with RT-PCR results (Fig. 3). Furthermore, PTL expression in ptl mutant plants, and in pmd-1d (Ler) plants, was not noticeably different from the wild-type pattern in either case.

A striking difference in staining pattern was seen in reporter lines when the first exon and intron were absent (p8.0::GUS and p2.0::GUS)(Table 2). Staining in the flanks of flower primordia, and subsequently in the four inter-sepal zones,could not be detected (Fig. 4C). Staining in basal regions of developing petals was also absent. Thus, the intron is required to drive expression in these regions. (An involvement of exon 1 sequences is possible although regulatory sequences rarely occur in translated regions.) By contrast, staining around the margins of sepals and leaves was similar with or without the exon and intron(Fig. 4C; Table 2).

Expression in developing stamens was present without the intron, but only in the 8.0 kb construct (Table 2). It is apparently controlled by redundant elements present both in the intron and upstream of –2.0 kb. The intron also contains elements that dampen PTL expression in other vegetative regions(Table 2).

In ptl mutants, visible defects were limited to sepals and petals(e.g. Fig. 5A). To test which components of the PTL expression pattern are associated with these defects, we carried out complementation tests using expression lines with or without the intron (i.e. with or without expression in the early flower primordium and inter-sepal zone). To do this, we used the two-component system of Moore et al. (Moore et al.,1998) (see Materials and methods). Two sets of driver lines containing PTL regulatory sequences (one set with the intron[pPTL(1.3i)], and one without [pPTL(2.0)], were each crossed with four replicated target lines containing the PTL coding sequence, all in ptl-1 mutant background. [Although the latter driver carries an additional 0.7 kb of the upstream region (–1.3 to –2.0 kb), no controlling elements were identified in this region(Table 2).]

All 11 lines containing the PTL promoter region and intron[pPTL(1.3i)] could fully restore the wild-type sepal and petal phenotype in combination with at least one of the target lines(Fig. 5B), whereas complementation was not seen among progeny of any of the seven driver lines that lacked the intron [pPTL(2.0); data not shown]. Thus the loss of PTL function specifically in the early flower primordium and inter-sepal zone is likely to be responsible for mutant disruptions to sepal and petal development. Sepal margin expression, conferred by both drivers, is insufficient to restore these disruptions to normal.

Interestingly, in six of the 11 combinations in which the intron was present in the driver [pPTL(1.3i)>>PTL], a new defective flower phenotype was seen (Fig. 5C). The severity of this phenotype was correlated with the strength of expression of each driver line (previously assessed by crossing them with a GUS target line). The phenotype was characterized by narrow sepals and sometimes loss of those in lateral positions (Fig. 5E, compared with 5D). The number of petals was usually restored to four, although they were often narrow and sometimes fused to each other at the base in lateral positions within the flower. Such fusion may be the consequence of petal primordia arising much closer than normal,associated with the absence of a lateral sepal (see stage 4 bud in Fig. 5E). The two lateral stamens were also often absent. This new floral phenotype was apparently controlled by regulatory sequences within the intron because similar floral defects were not seen in any of the pPTL(2.0)>>PTL combinations. One explanation is that overexpression of PTL in the flanks of the floral meristem and the four inter-sepal zones inhibits growth of these regions. To test this, we overexpressed PTL in other defined regions of the flower meristem.

The APETALA1 (AP1) gene is expressed predominantly in developing flower primordia from stage 1, becoming limited to sepal and petal primordia as they arise at stages 3 and 5(Mandel et al., 1992). When an AP1 driver line was crossed with another 12 independent PTL target lines, the pAP1>>PTL progeny revealed a spectrum of floral defects(Fig. 5F). For the most severely affected target line (#10), plants developed normally until the production of the inflorescence meristem. Flower primordia arose continuously,but their development was arrested at around stage 2(Fig. 5F). In less severely affected target lines (e.g. #7), flower primordia were also arrested but some filamentous floral organs, mostly carpelloid, eventually arose.

A second floral gene was also tested. AP3 drives expression specifically in the petal and stamen sectors of flower primordia from stage 3(Jack et al., 1992). An AP3 driver line was crossed with the same 12 PTL target lines. Strikingly, in the resulting pAP3>>PTL plants, petal and stamen growth was often abolished(Fig. 5G,H), or was reduced to thin, filamentous outgrowths. In addition, the lateral sepals were often narrow and stunted (AP3 is expressed in lateral sepal margins). The severity of the defects across the 12 target lines was strongly correlated with those seen using the AP1 driver (e.g. strongest in each case using line#10, intermediate with line #7). Interestingly, the strongest phenotype closely matched that seen when the diphtheria toxin gene DTA was expressed by the AP3 promoter, ablating the tissues involved(Day et al., 1995). The weaker effect was similar to that seen in Brassica napus flowers when cell division was inhibited by AP3 driven expression of ICK1, an inhibitor of cyclin-dependant kinase (Zhou et al., 2002).

The consequences of mis-expression of PTL more generally through the plant were tested using the CaMV 35S promoter. When the strongly responsive target line #10 was crossed with a 35S driver line, the p35S>>PTL progeny never developed beyond the production of a very short root (Fig. 5I). Using a less responsive target line (#7), progeny were arrested soon after the cotyledons emerged.

Thus it seems that strong expression of the PTL gene generally results in inhibition of growth of any tissue in which the expression occurs.

Sepal development is severely disrupted in mutants of the PINOID(PID) gene (Bennett et al.,1995). This gene encodes a protein kinase associated with auxin signalling (Christensen et al.,2000). In pid mutant flowers, sepals can be fused laterally with each other, are irregularly spaced, and are variable in number and size (Fig. 6A). The first whorl arises initially as a ring of tissue without inter-sepal zones, and sepal primordia develop irregularly from its rim much later than normal(Bennett et al., 1995).

We were interested to see whether the apparent early loss of the inter-sepal zone in pid mutant flowers was associated with the loss of PTL expression. Surprisingly, this was not the case. Instead,ectopic PTL expression occurred, firstly throughout the newly arising outer ring of tissue in the flower primordia(Fig. 6B,C), and later continuously around its internal side (Fig. 6D). Thus it seems that early expression of PTL is negatively regulated by PINOID function. We further tested whether this ectopic PTL expression in pid mutants is associated with the delayed outgrowth of sepal primordia by looking to see if such disruptions were ameliorated in pid-2 ptl-5 double-mutant plants. This is apparently not the case as sepal development was not significantly different between the two genotypes [mean number of sepals per flower (±s.e.m.)was 3.48±0.16 for 40 pid-2 ptl-5 flowers compared with 3.73±0.14 for 45 pid-2 single mutant flowers).

PTL apparently acts in the developing flower to dampen growth in four small regions lying between the sepals from the time they arise, resulting in their separate outgrowth. Evidence for this is that: (1) PTL expression is strongest in inter-sepal zones during early flower development(Fig. 6); (2) adjacent sepals are sometimes fused in ptl mutants, a likely consequence of ectopic growth occurring between them; (3) extensive sepal fusion occurs in double mutants of ptl with either of the organ boundary genes cuc1or cuc2, as would be expected if all three genes function in keeping sepals separate; and (4) strong ectopic expression of PTLconsistently represses growth, suggesting that growth suppression is its normal role. Concerning the latter observation, we realise that the consequences of ectopic PTL expression might represent a gain of function if, for example, abnormally large amounts of PTL protein sequester accessory factors required for the activity of other growth-promoting transcription factors. Alternatively, excess PTL protein might inappropriately activate growth-suppressing genes (Le Gourrierec et al., 1999). These possibilities can be tested by overexpressing PTL with its DNA-binding domains inactivated.

PTL does not itself define the sites of origin of the four sepals, as four still arise in appropriate positions in the absence of PTL function. Other genes are apparently involved. One of these may be PINOID that encodes a serine/threonine kinase implicated in auxin signalling(Christensen et al., 2000),possibly through its positive regulation of polar auxin transport(Benjamins et al., 2001). PID is expressed in sepal primordia as they arise and grow, but not between them (Christensen et al.,2000). Recently it has been shown that auxin is involved in defining the sites of initiation of organ primordia(Reinhardt et al., 2003). Thus, sepal initiation may be promoted by the localized concentration of auxin acting through PID. We have shown that PID excludes PTL expression from sepal initiation regions. Furthermore, the absence of PIDexpression in inter-sepal zones would allow PTL to be expressed there, hence resulting in suppression of growth in these zones.

Although loss of PTL function severely affects petal development, PTL is not expressed in petal primordia until they start to expand at stage 8, and even then it occurs only in their basal margins(Fig. 7). Consistent with this,when PTL is ectopically expressed strongly in the second floral whorl using the AP3 promoter, petal initiation is frequently abolished, but overexpression of PTL controlled by its own regulatory sequences does not disrupt petal initiation. Thus it seems likely that PTL influences petal development indirectly (assuming it acts cell autonomously). It is true that PTL function is required for expression of the RBE petal development gene (Takeda et al.,2004), but this positive regulation may occur earlier (stages 3-4)in the inter-sepal zone, where their expression patterns overlap.

If PTL normally functions to inhibit growth, additional growth may be expected to occur in ptl mutants. Thus, the number of petals per flower might be predicted to increase if more space were available for them to arise. However, the number is observed to decrease. It may be that overgrowth in inter-sepal zones diffuses and weakens a petal initiation signal so that they arise less often, later on average, and over a wider area than usual(Griffith et al., 1999).

In ptl mutants, disruption of petal orientation within the flower is also seen (Griffith et al.,1999). Petal primordia normally initiate when cells in the L2 layer divide outwards (periclinally) rather than laterally (anticlinally)(Hill and Lord, 1989). Sector analysis has suggested that two adjacent cells are initially involved(Bossinger and Smyth, 1996). Presumably, these cells usually lie side by side on a circumference of the flower primordium, defining the ultimate orientation of the developing petal. In ptl mutants, however, petals arise in any orientation(Griffith et al., 1999). It may be that additional growth within the petal-initiating zone in ptlmutants has relieved the constraint that only cells lying adjacent on a circumference commence petal development.

In addition to adopting an orientation within the flower primordium, the two faces of developing petals acquire adaxial-abaxial polarity. Petal orientation, once established, may automatically define the later developing polarity, or they may be independent processes(Griffith et al., 1999). It has been proposed that polarity of lateral organs is defined by a signal emanating from the center of the shoot or flower meristem(Eshed et al., 2001; McConnell et al., 2001; Emery et al., 2003). This could ensure that adaxializing genes of the class III HD-ZIP family are expressed in adaxial regions of newly arising organ primordia. In ptlmutants, disruption to such a signal is also likely to be indirect because PTL is not expressed in petal primordia, the targets of the signal.

Strikingly, PTL expression occurs in the margins of most developing lateral organs (Fig. 7). Of floral organs, only carpels lack such expression. One appealing hypothesis is that the PTL protein dampens the growth in these regions of lateral organs, helping sculpt their final shape. Thus, PTL may moderate extension around the edges of leaves and sepals, ensuring that they keep pace with expansion of more centrally located tissue. Also, constriction in the region of the petiole of leaves, the claw of petals, and the filament/anther boundary of stamens may be the consequence of growth suppression. Flower primordia, too, may be constrained from lateral expansion by early PTL expression in their flanks.

Even so, loss of PTL function does not seem to be associated with any obvious phenotypic consequences in these regions. Other genes with the same growth suppression function may still be active (see below). The homeodomain-encoding gene PRESSED FLOWER is expressed specifically in lateral regions of primordia of all of these organ types, and in newly arising flower primordia (Matsumoto and Okada,2001). However, it apparently promotes growth in lateral regions,and it will be interesting to assess whether it competes or otherwise interacts with PTL.

Twenty-eight trihelix genes have been identified in the Arabidopsis genome (Riechmann et al., 2000), but only three others have been characterized in detail to date. All were identified as encoding proteins that bind specifically to promoter elements required for light responsiveness. The GT-2 protein of rice binds to GT boxes in the promoter of the phytochrome A gene and is associated with its dark induction(Dehesh et al., 1990). A closely related protein, DF1 from pea, also binds to regulatory elements necessary for dark induction (Nagano et al., 2001). Contrastingly, GT-1a from tobacco binds promoter elements associated with induction by light(Gilmartin et al., 1992; Perisic and Lam, 1992).

In addition to PTL, six of the 28 trihelix genes in Arabidopsis have (or probably once had) duplicated trihelix domains. These include AtGT-2 (Kuhn et al.,1993) and AtDF1 (Nagano et al., 2001), but not AtGT-1a(Hiratsuka et al., 1994). Even so, neither AtGT-2 nor AtDF1 is closely similar to PTL (60-65% identical in amino acid sequence in the N-terminal trihelix, and 42-45% in the C-terminal trihelix). This divergence is apparently reflected by divergence in the targets of regulation, as PTL does not play any obvious role in light response.

Expression of PTL in early flower primordia and the inter-sepal zone is controlled intragenically, probably by sequences within the intron. Other genes are known that are regulated by intronic elements (e.g. Sieburth and Meyerowitz,1997). Interestingly, an intron occurs at this position only in those 7 members of the trihelix family that arose following duplication of the trihelix. The closest relative of PTL, At3g10000 (94% identical in the N-terminal trihelix), shares some tracts of similar sequence within its intron. It will be of interest to see whether it shares functions with PTL in defining the inter-sepal zone. Also, it may be that this and other very close relatives of PTL act redundantly in lateral organ margins where loss of PTL function is apparently without effect.

In conclusion, this study has revealed that the PETAL LOSS gene helps to control morphogenesis of the perianth. It may repress growth between sepal primordia keeping them separate, and later allow petal developmental signals to be perceived appropriately. This represents a new developmental function for a family of transcription factors previously known only for their role in regulating light response.

We are grateful to Yuval Eshed for advice and discussion, and for generously supplying components of the LhG4/pOp transactivation system, and the ptl-4 and ptl-5 alleles. We also appreciate discussions with Stan Alvarez, John Bowman, Marcus Heisler and other laboratory colleagues, and the excellent input and assistance from John Arvanitakis,Gunta Jaudzems and Georgia Stamaratis. We thank Stuart Baum for providing the ptl-3 allele, and The Arabidopsis Biological Resource Center, Ohio State University, for the YAC and BAC clones. This work was supported by the Australian Research Council (Grants A19801319 and A00103617).