Control of floral homeotic gene expression and organ morphogenesis in Antirrhinum

The identity of floral organs depends on the expression of genes in specific domains of the floral meristem (Coen and Meyerowitz, 1991; Weigel and Meyerowitz, 1994). Although much progress has been made in characterising these genes, less is known about how their expression is coupled to organ formation. One possibility is that there is a class of genes that regulate both the identity and formation of organ primordia. Mutations in such regulatory genes might be expected to have pleiotropic effects, altering both organ identity gene expression and other aspects of floral development. The phenotype of these mutants would therefore display homeotic transformations similar to those due to loss or misexpression of organ identity genes but would also have additional defects. Here, we describe three genes of this type involved in the regulation of the organ identity gene plena (ple) in Antirrhinum.

Wild-type Antirrhinum flowers consist of four whorls of organs: five sepals in the first whorl, followed by five petals, four stamens and two carpels at the centre (Figs 1A, 2A). The petals are united along part of their length to form a corolla tube ending in five separate lobes. Three homeotic functions, a, b and c, have been proposed to specify organ identity in the combination, a, ab, bc and c in whorls one to four respectively (Carpenter and Coen, 1990; Schwarz-Sommer et al., 1990; Bowman et al., 1991). Mutants that lack the b function, such as deficiens (def) and globosa (glo), have sepals instead of petals and carpels instead of stamens (Carpenter and Coen, 1990; Schwarz-Sommer et al., 1990). In mutants lacking the c function, such as ple, petals replace stamens, and organs with sepaloid, carpeloid and/or petaloid features replace the carpels (Carpenter and Coen, 1990; Schwarz-Sommer et al., 1990). The ple gene is also required for floral meristem determinacy, as multiple whorls of petaloid organs proliferate internal to the fourth whorl of ple mutants. Organ identity genes are transcribed from early stages in specific domains of the floral meristem and expression is subsequently maintained in floral organs as they develop (Sommer et al., 1990; Schwarz-Sommer et al., 1992; Tröbner et al., 1992; Bradley et al., 1993). The domain of ple expression is expanded in mutants lacking the a function, in which carpels replace sepals and stamens replace petals (Bradley et al., 1993; Lönnig and Saedler, 1994). Ectopic expression of ple in these mutants is caused by the insertion of a transposable element in the second intron of ple (Bradley et al., 1993). The organ identity genes interact with another set of genes (cycloidea and dichotoma) that establish differences in organ type along the dorsoventral axis (Carpenter and Coen, 1990; Luo et al., 1996). Dorsoventral asymmetry is most notable in the second and third whorl, in which three types of organ can be clearly distinguished: dorsal (upper), lateral (side) and ventral (lower).

Several genes involved in the regulation of the organ identity genes have been described in Antirrhinum (Coen et al., 1990; Simon et al., 1994; Hantke et al., 1995; Ingram et al., 1997; Motte et al., 1998). One of these is the floral meristem identity gene, floricaula (flo), which is needed for the activation of ple and def early in development (Hantke et al., 1995). In addition, flo may play a role in the regulation of organ identity genes at later stages, as it is transiently expressed in sepals and petals and eventually in carpels as well (Coen et al., 1990). The fimbriata (fim) gene also plays a major role in the regulation of organ identity genes, particularly those involved in the b function (Simon et al., 1994; Ingram et al., 1997).

To understand further how organ identity gene expression is regulated, we have characterised several pleiotropic mutants in which c activity is altered in Antirrhinum. One of these mutants, polypetala (poly), has two effects on the central dome of the floral meristem. First, expression of ple is greatly reduced in poly mutants, resulting in a proliferation of petals within the flower. Secondly, the formation of organ primordia is altered, leading to a delay in the appearance of organ primordia, an increase in the number of organ primordia and loss of whorled phyllotaxy. This implies that poly normally regulates both ple expression and the formation of organ primordia in the central dome of the meristem. Whether the wild-type role of poly is to positively or negatively regulate these processes depends on whether the semi-dominant allele described represents a loss or gain of function. Another mutant, fistulata (fis), specifically affects the distal domain of petal primordia, which normally form the lobes. As was also recently shown by Motte et al. (1998), ple is ectopically expressed in this domain in fis mutants, resulting in the development of anther-like tissue. In addition, the size of the petal lobes is greatly reduced in fis mutants, even in the absence of ple activity. These results imply that fis interacts with a proximo-distal prepattern to inhibit ple expression and to promote lobe growth. The effect of fis on ple expression is greatly enhanced by a weak allele of flo, showing that flo also plays a role in negative regulation of ple in outer whorls. Altogether, the results show that genes involved in ple regulation in Antirrhinum have additional roles in organ formation, perhaps reflecting underlying mechanisms for coupling homeotic gene expression to morphogenesis.

Plant growth conditions were as previously described (Carpenter et al., 1987). The semi-dominant poly-704 mutant allele arose spontaneously in a cross between stock JI 69 and stock JI 75. The recessive fis mutant was obtained from the Gatersleben collection of mutants (Stubbe, 1966; Hammer et al., 1990). The same allele was used by Motte et al. (1998). A similar mutant, rhinanthoides, was described by Lotsy (1916) and Baur (1924) though it is not known if the two mutants are allelic. The flo-662 allele was first identified in a heterozygous condition with flo-613, as a mutant with split and distorted flowers amongst progeny arising from revertant flowers on flo-613 plants (Carpenter and Coen, 1995). The flo-613, squa-641, squa-701, def-621 and ple-625 alleles were all derived from stock JI 98 (which was itself derived from a cross between JI 69 and JI 75) during large scale transposon mutagenesis experiments (Carpenter and Coen, 1990). In all cases, mutants were compared to their nearest wild-type relatives: poly mutants were compared to their wild-type siblings; fis mutants were compared to the standard wild-type line, Sippe 50, from Gatersleben; ple mutants were compared to the progenitor line, JI 98.

DNA extractions and genomic DNA blots were carried out as described previously (Coen et al., 1986). To determine the molecular basis of the flo-662 allele, genomic DNA blots from the flo-662 mutant were probed with various fragments from the flo locus. A 15 kb genomic clone of the flo locus (λ-JAM98) was obtained by screening a partial Sau3A library cloned into the EMBL3 vector cut with BamHI and EcoRI. The probes used were a 4.8 kb SalI-EcoRI upstream fragment and the adjacent 5.8 kb EcoRI fragment, which includes the first two exons of flo, which were each subcloned into Bluescript SK+ to give plasmids pJAM 106 and pJAM 107 respectively. PCR and DNA sequencing was carried out as described by Hantke et al. (1995). The oligonucleotides used for identifying the upstream insertion in flo-662 were ACAGCTATATACAACGTATG for flo and AGGTAGATCGGAGTTCGCGGC for Tam5. The Tam5 transposon was isolated from a mutation of the incolorata gene and DNA sequencing showed that it belonged to the CACTA family (Nacken et al., 1991; S. Doyle, R. Carpenter and E. Coen, unpublished results). Sequencing revealed that flo-662 had a DNA footprint in the second exon due to imprecise Tam3 excision. The footprint was identical to that previously described for flo-639, and would be predicted to give a two amino acid substitution in the encoded protein (Coen et al., 1990). The similarity between the footprints in flo-662 and flo-639 reflects their common origin, as revertants from the same progenitor, flo-613, carrying Tam3.

The genotype at the flo locus in fis:flo-662 double mutants was confirmed by probing blots of EcoRV-digested genomic DNA from these plants with the 5.8 kb EcoRI insert from pJAM 107. This revealed a polymorphism in the flo gene between the genetic backgrounds of the fis and flo mutants, allowing the flo genotype in the double mutants to be determined. To ensure that the upstream transposon in flo-662 was still present in the double mutants (i.e. the transposon had not excised), EcoRI-digested DNA from the double mutants was probed with the 4.8 kb SalI-EcoRI insert from pJAM106. This revealed a larger fragment diagnostic of the transposon insertion, confirming that it had not excised.

Plants were grown in growth cabinets at 20°C and inflorescence apices moulded as described in Green and Linstead (1990). Casts of the moulds were then sputter coated with gold and viewed and photographed by SEM. In Fig. 7G the sample was fixed in 3% v/v glutaraldehyde, dehydrated through an acetone series, critical point dried (Polaron, UK) and viewed by SEM as above.

RNA in situ hybridisation was performed according to Coen et al. (1990) with modifications as described by Bradley et al. (1993). DIG-labelled antisense RNA probes of flo, def, fim and ple were transcribed as described elsewhere (Bradley et al., 1993; Simon et al., 1994).

Plants homozygous for poly produced flowers with a first whorl of sepals, followed by an indeterminate number of separate petals (Figs 1C, 2D). The petals were arranged as two outer whorls followed by a proliferation of narrow petals which usually could not be assigned to specific whorls. Staminoid, carpeloid or sepaloid tissue was occasionally seen on the innermost petals.

The poly mutation was not fully recessive. Heterozygous poly plants usually had five whorls of organs in the order sepal, petal, petal, stamen, carpel (Figs 1B, 2C). Adjacent petals were often united, both within and between whorls and stamen filaments often bore petaloid tissue (64% of stamens in 60 flowers). Various defects were seen in the fifth whorl: occasionally carpels were petaloid (25%) or the style was aborted (3.3%) and sometimes three carpels were present instead of two (29.4%). In some heterozygous poly plants, the flowers had a weaker phenotype and were distinguished from wild type only by the presence of additional narrow petals in dorsal positions between the second and third whorl.

Scanning electron microscopy was used to determine how the development of poly differed from wild type and ple. Floral meristems are formed sequentially on the periphery of the apical inflorescence meristem in the axils of bract primordia; the youngest meristems at the apex, and progressively older meristems below. Flower development has been divided into several morphological stages (Bradley et al., 1993; Carpenter et al., 1995; Zachgo et al., 1995).

At stage 4, when sepal primordia surround the central dome, all genotypes were similar (Fig. 3A-D). By stage 5, when primordia of the second and third whorl were visible, wild type and ple resembled each other (Fig. 3E,F). However, poly mutants had a distinctive appearance: the primordia in whorl 3 appeared to be less advanced and did not encroach on the central dome as much as in wild type (Fig. 3G,H). By stage 6, when carpel primordia of wild type were clearly visible, wild type and ple were still difficult to distinguish (Fig. 3I,J), even though the primordia in whorl 3 of ple were destined to form petals rather than stamens, and those in whorl 4 were destined to form organs with sepaloid, petaloid and carpeloid features rather than normal carpels. By contrast, poly mutants had a larger central dome surrounded by several additional primordia (indicated with asterisks in Fig. 3K) destined to form stamens or petals in heterozygous or homozygous poly respectively (Fig. 3K,L). At stage 7, the anther thecae were visible as swellings on the stamen primordia of wild type (arrowed in Fig. 3M). At this stage, the whorl 3 primordia of ple could be distinguished from those of wild type by the absence of thecae and a more petal-like morphology (Fig. 3N). Differences between heterozygous and homozygous poly also became apparent at stage 7. In heterozygous poly, several carpel primordia formed the fifth and final whorl of organs at the centre of the floral meristem (arrowed in Fig. 3O). However, in homozygous poly, petal primordia continued to be initiated with irregular phyllotaxy around the periphery of an enlarged central dome (Fig. 3P).

The replacement of reproductive organs by petaloid organs in poly mutants suggested that ple expression might be reduced. To test this, RNA in situ hybridisations were performed on young floral meristems of poly mutants using DIG-labelled antisense RNA of ple as a probe.

At stage 4, there was usually no detectable expression of ple in homozygous poly floral meristems (early stage 5 is shown in Fig. 4B), unlike wild type where ple was strongly expressed in the centre of the floral meriste (Fig. 4A; Bradley et al., 1993). The first detectable expression of ple in homozygous poly mutants was usually at late stage 5, when the primordia of the second and third whorls were visible (Fig. 4E). At this stage and later, ple RNA was more restricted than wild type, being detected in only one to three outer cell layers of part of the central dome (Fig. 4E,H). Transcripts of ple were also detected at the tips of occasional mosaic petaloid/staminoid organs (not shown). In heterozygous poly floral meristems, ple expression was restricted to a small patch of approximately 10 cells wide near the centre of the floral meristem at stage 4 (not shown). By late stage 5 and stage 6, ple was expressed at wild-type levels in stamen and carpel primordia as they arose but not in the primordia of whorl 3, which were destined to form petals. Consecutive sections were also probed with flo and def to confirm the identity of organs (Coen et al., 1990; Schwarz-Sommer et al., 1992; Bradley et al., 1993). Surprisingly, early expression of def was reduced in homozygous and heterozygous poly floral meristems. At stage 4, the expression of def in the central dome of the floral meristem was less extensive than wild type. Expression was patchy in homozygous poly (Fig. 5C) while in heterozygous poly, the def expression domain did not extend to the edges of the central dome (compare Fig. 5A,B). However, by stage 5, def was expressed in petal primordia and was absent from the central dome as seen in wild type (not shown). At later stages, def was expressed in whorls 2, 3 and 4 of heterozygous poly (Fig. 5E) and in all petal primordia in homozygous poly (Fig. 5F). Expression of flo was also observed in all petal primordia (not shown). Mosaic organs were seen in older floral buds of homozygous poly. Some of these expressed ple and flo (typical of wild-type carpels) on the adaxial side and def and flo (typical of wild-type petals) on the abaxial side, implying that they were destined to form carpeloid petals. Similarly, primordia with staminoid regions (expressing ple and def) and sepaloid regions (expressing flo only) were also occasionally seen. In heterozygous poly, abnormally shaped carpels were sometimes visible (Fig. 5E) which may have been a reflection of the tricarpellate gynoecium seen in some heterozygous poly flowers.

The early changes in def and ple activity in poly mutants did not appear to be mediated by alterations in fim or flo expression, as the temporal and spatial expression pattern of these genes was unchanged early on in development (not shown). Later in development, fim was ectopically expressed between all organs similar to the expression of fim in ple mutants (Simon et al., 1994).

Plants homozygous for the recessive fis mutation had much smaller petal lobes than wild type (Fig. 1D,E); the lobes were sometimes so small that the stamens and style protruded from within the petals (Fig. 1F). The petals were also partially staminoid: anther-like tissue was observed on the lateral and ventral petal lobes, although pollen was not normally produced (Figs 1F, 2E). Staminody was more pronounced on plants grown outside during the summer and on older plants.

The development of fis floral meristems was as wild type up to stage 6, by which time the primordia for all four whorls had appeared (Fig. 6A,B). However, by stage 7, when the petal lobes of wild type started to increase in size and cover the stamen primordia, the lobes of fis floral meristems exhibited a reduced growth rate and did not cover the stamens as extensively (compare Fig. 6C,D). This was even more apparent at later stages, when the petal lobes completely covered the stamens of wild type (Fig. 6E,G) while the lobes of fis were so reduced that the internal reproductive organs remained clearly visible (Fig.6F,H). Also at these stages, the lateral and ventral lobes of fis mutants began to assume a staminoid appearance, forming anther-like bulges at their margins (arrowed in Fig. 6F,H).

The staminoid appearance of petals in fis mutants suggested that ple might be ectopically expressed in the second whorl. To test this, RNA in situ hybridisations were performed on young floral meristems using ple as a probe.

Ectopic expression of ple was observed in fis floral meristems from early stage 5, when petal primordia first became visible (Fig. 4C). In wild type at this stage, ple was only expressed in the central dome of the meristem (Fig. 4A); whereas in fis meristems, additional expression of ple was observed in a few cells of the ventral second whorl primordia (arrowed in Fig. 4C). Ectopic expression of ple in the second whorl organs increased as they developed further (Fig. 4F,I). In all cases, expression was restricted to one or two cell layers along the length of the inner (adaxial) surface of ventral and lateral second whorl organs. Similar results were recently reported by Motte et al. (1998). These results show that a wild-type role of fis is to prevent ple expression in the second whorl of the flower.

As controls, consecutive sections were probed with flo and def. The expression of these genes was altered in ventral and lateral second whorl organs after stage 6, consistent with their partial staminoid identity: there were clearings of def expression, consistent with the absence of def from male sporogenic tissue, and flo expression was restricted to the outer (abaxial) surface or the base of the organs, consistent with the lack of flo expression in stamens (not shown).

If the staminody of fis mutants was the result of ectopic c function activity in the second whorl, then loss of b activity in a fis mutant background, should result in carpeloidy of second whorl organs. As predicted, the def:fis double mutant had carpeloid tissue in the second whorl (Fig. 7B). The lateral and ventral organs resembled carpels and produced ovules whereas the two dorsal organs resembled sepals with stylar and stigmatic tissue at the tip and occasional ovules on the abaxial surface (Fig. 7B,C). Surprisingly, one or two of the ventral sepals in the first whorl of the double mutant were also slightly carpeloid, showing stylar tissue at the tips (Fig. 7B). In agreement with this, RNA in situ hybridisation showed that ple was ectopically expressed in patches in the first whorl as well as in the second whorl of the def:fis double mutant (not shown).

To determine if all of the phenotypic effects in fis mutants were due to ectopic expression of ple, the ple:fis double mutant was constructed. If the only role of fis was to inhibit ple, then the phenotype of ple:fis double mutants would be expected to be indistinguishable from the ple single mutant. The ple:fis double mutants resembled ple single mutants in that they produced indeterminate whorls of petals in place of reproductive organs (compare Fig. 7E,D). However, unlike ple, all petals had reduced lobes similar to those of fis single mutants. Scanning electron microscopy showed that the petal lobes of ple:fis exhibited reduced growth at a similar time as fis mutants (not shown). This showed that the reduced lobe size in fis single mutants was not caused by ectopic ple activity and therefore, that fis played an additional role in petal lobe development.

The fact that the transformation of petals to stamens was only partial in fis mutants, suggested that other genes might interact with fis to negatively regulate ple. The floral meristem identity gene, flo, was a candidate for such a gene as it is expressed in the first and second whorl and it had previously been proposed to play a role in the regulation of organ identity genes (Coen et al., 1990). To examine how flo interacted with fis, we constructed a flo:fis double mutant. However, because flowers are not produced in the absence of flo activity, a double mutant between fis and a null flo allele was not expected to be informative. We therefore used a very weak allele of flo, termed flo-662, which produces almost wild-type flowers, for the double mutant analysis.

Before constructing the double mutant, the molecular basis of the flo-662 allele was determined. Probing of genomic DNA blots from the flo-662 allele with fragments from the flo locus (see Materials and Methods) revealed that flo-662 carried a 5 kb insertion about 7 kb upstream from the first exon, which was absent in wild type (data not shown). The 5′-end of the insertion could be amplified by PCR, using an oligonucleotide matching the flo locus together with one matching the end of Tam5, a previously identified CACTA-type transposon. Probing the DNA blots with a fragment of Tam5 confirmed that flo-662 carried an upstream Tam5-like insertion. In addition, DNA sequencing revealed that flo-662 carried a two amino acid substitution in the second exon, reflecting a DNA footprint left by imprecise excision of Tam3 from its progenitor, flo-613 (see Materials and Methods). The phenotypic consequences of these alterations in flo-662 could be observed in plants heterozygous for flo-662 and a null flo allele: these plants had flowers with split and distorted petals, indicating that flo activity was reduced (Simon et al., 1994). Plants homozygous for flo-662 produced almost wild-type flowers, showing that two doses of this allele were sufficient to confer near wild-type development.

The flo-662:fis double mutant had a much more extreme phenotype than fis alone (Figs 7F, 2F). Organs in the second whorl of the double mutant exhibited a complete homeotic transformation to stamens. This transformation preserved dorsoventral asymmetry: pollen-producing stamens were observed in lateral and ventral positions while small arrested stamens (staminodes) developed in dorsal positions, similar to the corresponding organs in the third whorl of wild type. Thus, seven stamens were found in mature flowers (Fig. 7F). The stamens in the second whorl were sometimes united with each other or with the stamens in the third whorl. Narrow strips of petal tissue were occasionally found between second whorl organs. In addition to these changes in the second whorl, the first whorl of flo-662:fis double mutants was also altered: one or two of the ventral sepals showed slight carpeloidy, with stylar and stigmatic tissue visible at their tips (arrowed in Fig. 7F). Light microscopy showed that stylar tissue extended from the tip down along the inner surface of the sepal. The organs in the third and fourth whorls were phenotypically normal except that sometimes one of the third whorl stamens was united with the carpel and this was associated with deformation of the carpel.

Scanning electron microscopy of young floral buds was used to investigate the phenotype of the flo-662:fis double mutant further (Fig. 7G). At stage 5, five primordia were visible in the second whorl of flo-662:fis meristems, similar to the situation in wild type (not shown). However, unlike wild type, these primordia developed with a stamen rather than petal identity: lateral and ventral primordia formed stamens, whereas the dorsal primordia formed vestigial organs, similar to the arrested dorsal stamen of wild type (arrowed in Figs 7G, 2F). In some cases, primordia at stage 5 were irregularly arranged and united with each other (not shown).

To determine if there was increased ectopic expression of ple in the flo-662:fis double mutant, in situ hybridisations were performed. As in fis floral meristems, ple was ectopically expressed in the second whorl of flo-662:fis floral meristems from early stage 5 (early stage 6 is shown in Fig. 8B). However, ectopic ple expression was more extensive than in fis single mutants, with expression being observed throughout the organs in whorl two instead of being restricted to their adaxial surface (compare Fig. 8B with Fig. 4F). In older floral buds, strong ple expression, similar to that of wild-type stamens, was seen in the ventral and lateral organs of the second whorl (not shown). Consistent with carpeloidy of the sepals, ple was also ectopically expressed in the first whorl organs of the flo-662:fis double mutant: patches of ple RNA were detected on the inner surface of the ventral sepal from stage 4 onwards (arrowed in Fig. 8B). This expression was restricted to the tip of the sepal and was only one or two cell layers deep. In addition, very weak expression of ple was sometimes detected in the epidermis of young bracts from stages 0 to 3.

As flo expression is normally absent in the third whorl stamens of wild type, we tested whether flo was also reduced in second whorl stamens of flo-662:fis. Control experiments showed that flo was expressed in the normal temporal and spatial pattern in flo-662 single mutants though subtle quantitative differences in expression would not have been detected (not shown). At stage 6, flo was very weakly expressed in the second whorl primordia of flo-662:fis double mutants, at a much lower level than in wild type or homozygous flo-662 (compare Fig. 8C,D). Expression was restricted to the outer (abaxial) surface or the base of second whorl primordia (Fig. 8D). In older buds, flo was not expressed in the second whorl stamens (not shown). Therefore, ectopic expression of ple in flo-662:fis correlated with reduced expression of flo in second whorl stamens. However, the possibility that fis rather than ple is responsible for the reduced levels of flo cannot be ruled out.

To determine whether another floral meristem identity gene, squa, interacted with fis, double mutants were constructed between fis and two alleles of squa (Carpenter et al., 1995; Huijser et al., 1992). The squa-641:fis and the squa-701:fis double mutant had an additive phenotype. The inflorescence structure was like the squa single mutants, with inflorescences in the axils of bracts instead of flowers. When flowers were produced, they had reduced petal lobes and staminody due to fis. Therefore, unlike flo, squa did not appear to interact with fis.

We have identified three genes, poly, fis and flo, involved in the regulation of ple in Antirrhinum. In addition, these genes play other roles in flower development: poly affects the formation of organ primordia, fis plays a role in petal lobe development and flo affects meristem identity.

The replacement of reproductive organs by indeterminate perianth organs in homozygous poly mutants can be accounted for by the observed reduction in ple expression in the central dome of the floral meristem. However, plants homozygous for poly have additional defects in the formation of organ primordia which are not seen in ple mutants: there is a delay in the development of internal primordia, the central dome of the meristem becomes larger after stage 5, and additional organ primordia arise with an irregular phyllotaxy. Moreover, plants heterozygous for poly have five whorls of organs and additional carpels, a phenotype quite distinct from that conferred by weak alleles of ple (which have four whorls of organs with petaloid stamens in the third whorl).

A further difference from ple mutants is that, at early stages, poly mutants have reduced def expression. Later, when petal primordia form, def expression recovers, probably due to its positive autoregulation in petals (Schwarz-Sommer et al.,1992; Trobner et al., 1992; Zachgo et al., 1995). Therefore, poly affects both the initial activation of def and ple expression as well as the pattern of organ initiation in the central dome of the floral meristem. These effects together imply that the normal role of poly may be to ensure that the expression of organ identity genes is coordinated with the formation of organ primordia. In this regard, poly may act in a similar way to superman and fim which are thought to couple the regulation of class b gene expression to the formation or maintenance of whorl boundaries (Sakai et al., 1995; Simon et al., 1994; Ingram et al., 1997).

Determining the wild-type role of poly is complicated by the fact that the poly mutation is semi-dominant. One possibility is that the poly mutation leads to a loss or reduction of function (either because two doses of Poly⁺are needed for wild-type activity or because the poly mutant allele actively interferes with Poly⁺). According to this view, Poly⁺would normally act at the centre of the floral meristem to promote organ formation and early ple and def expression. Alternatively, poly could be a gain-of-function mutation, perhaps resulting in ectopic Poly⁺ activity in the central dome. If this were the case, Poly⁺ might normally act outside the central dome to inhibit organ identity gene expression and primordium initiation. Regardless of whether this poly allele is caused by a loss- or gain-of-function mutation, the results suggest that poly acts upstream of both the regulation of gene expression and the formation of organ primordia.

Other mutants affecting the behaviour of the central dome of floral meristems leading to the production of extra floral organs include the clavata mutants in Arabidopsis (Okada et al., 1989; Leyser and Furner, 1992; Shannon and Meeks-Wagner, 1993; Crone and Lord 1993; Clark et al., 1993, 1995, 1997) and fasciated mutants in Arabidopsis, tomato and other species (Leyser and Furner, 1992; Szymkowiak and Sussex, 1992; White, 1948). The defects in clavata mutants have been traced to altered proliferation of undifferentiated cells in the central dome of the floral meristem (Clark et al., 1993, 1995). However, unlike poly, clv mutants have only minor effects on homeotic gene expression.

Semi-dominant mutations with extra whorls of petals have also been described in other plant species. Such mutants, often called double-flowered, have been described in Meconopsis (Welsh Poppy), Althaea (Hollyhock), Dianthus (Carnation) and Petunia (Saunders, 1910, 1913, 1917; de Vlaming et al., 1984; Natarella and Sink, 1971; Reynolds and Tampion, 1983; van der Krol and Chua, 1993). In the case of the do1 mutant of Petunia, sectioning of young floral buds reveals similarities to poly mutants, with irregular initiation of many primordia throughout the centre of the floral meristem (Natarella and Sink, 1971). However, do1 differs from poly in that many of the primordia initiated near the centre of the floral meristem are stamens and eventually carpels form, implying that there is a higher level of class c activity in do1 (Saunders, 1910; Natarella and Sink, 1971; Reynolds and Tampion, 1983; van der Krol and Chua, 1993).

As was also reported by Motte et al. (1998), part of the phenotype of fis mutants, weak staminody of the petals, can be accounted for by the observed ectopic expression of ple in petal primordia (Fig. 4). However, the phenotype of the ple:fis double mutant demonstrates that the fis mutation confers reduced petal lobes even in the absence of ple, suggesting that fis plays an additional role in petal lobe development.

The effects of fis may be explained by assuming that each primordium is subdivided early on in development into a proximal and distal region with distinct identities. In the case of a petal, these regions will grow to form tube (proximal) and lobe (distal). In the case of a stamen, the regions would form the filament (proximal) and anther (distal). Mutations in fis have two effects specific to the distal region of petals. First, ectopic expression of ple results in anther-like features, typical of the distal region of stamens. Secondly, growth is altered so that the petal lobes end up being much smaller. The effects on lobe growth are first detectable at about stage 6, presumably after the time that proximal/distal identities have been established. Therefore, fis acts in response to an underlying proximal/distal prepattern. In this respect the fis mutation is similar to the blind mutant of Petunia which has anther sacs in place of lobes (de Vlaming et al., 1984; Vallade et al., 1987; Tsuchimoto et al., 1993; van der Krol and Chua, 1993).

The level of ectopic ple expression in fis mutants differs between organs within a whorl, reflecting the dorsoventral asymmetry of Antirrhinum flowers. Only the ventral and lateral petals of fis mutants express ple, implying that additional genes inhibit ple expression in dorsal petals. This effect may also extend to the first whorl because only the ventral sepals are carpeloid in flo-662:fis and def:fis double mutants. Candidate genes for this dorsal-specific inhibition are cyc and dich, which are needed for dorsoventral asymmetry in Antirrhinum (Luo et al., 1996). These genes are strongly expressed in the dorsal region of the floral meristem where they may help to restrict ectopic ple expression. Enhanced staminody of ventral and lateral petals is also seen in flowers of the heptandra mutant of Digitalis purpurea, a species closely related to Antirrhinum which also exhibits dorsoventral asymmetry (Henslow, 1882; Saunders, 1911; Shull, 1912). By contrast, dorsal-specific inhibition of staminody is not seen in comparable mutants from species with more radially symmetrical flowers such as ap2, lug, clf and ap1 in Arabidopsis (Komaki et al., 1988; Kunst et al., 1989; Bowman et al., 1989, 1991; Liu and Meyerowitz, 1995; Goodrich et al., 1997; Irish and Sussex, 1990; Schultz and Haughn, 1993; Bowman et al., 1993), blind in Petunia (Vallade et al., 1987; Tsuchimoto et al., 1993; van der Krol and Chua, 1993) and lax-A in barley (Larsson, 1985; Laurie et al., 1997).

The fact that the transformation of petals to stamens is only partial in fis mutants, suggested that other genes might interact with fis to negatively regulate ple. Mutations in these genes might be expected to enhance the degree of staminody in fis mutants. One such mutation is flo-662, a weak allele of flo, that normally gives a near wild-type phenotype when homozygous. When introduced into a fis background, flo-662 gives flowers showing a complete conversion of petals to stamens, correlating with high levels of ectopic ple expression in whorl 2. This suggests that both flo and fis are normally involved in negatively regulating ple in petals. Moreover, the flo-662:fis double mutant has slightly carpeloid sepals with weak ectopic ple expression, showing that fis and flo also play a role in the negative regulation of ple in the first whorl. Thus, in addition to the early role flo plays in determining floral meristem identity, we show that flo plays a later role in flower development in negative regulation of ple.

The interaction between flo and ple changes during flower development. Early on, flo is expressed throughout the floral meristem and is needed to activate ple (Coen et al., 1990; Hantke et al., 1995). However, the analysis of flo-662:fis shows that, at later stages (stage 5), flo is involved in preventing ple expression in the outer whorls of the flower. Perhaps flo interacts with other transcription factors to either positively or negatively regulate ple at different times during development. As well as showing that flo negatively regulates ple, our results also imply that expression of ple inhibits flo expression in stamen primordia (Coen et al., 1990, Fig. 8). This mutual antagonism between flo and ple correlates with their complementary expression domains during stages 4 and 5 of wild type (Bradley et al., 1993). However, by stage 6, flo and ple are expressed in overlapping domains in the carpel implying that they are no longer mutually antagonistic in this context. Mutual antagonism of flo and ple may require combinatorial interactions with other genes which are not present in the fourth whorl.

A common feature to emerge from this study is that many of the genes involved in ple regulation have additional roles in flower development. A similar conclusion has been derived from studies on the regulation of the def and glo genes involved in the b function of Antirrhinum (Ingram et al., 1997). The general significance of these findings may be that they reflect underlying mechanisms for coupling organ identity gene expression to morphogenesis.

We thank Lucy Copsey for help in the construction and scoring of double mutants, Peter Walker for growing the plants, and Andrew Davies for photography. We thank Zsuzsanna Schwarz-Sommer and Elizabeth Schultz for useful discussions and Desmond Bradley and Oliver Ratcliffe for comments on the manuscript. The work was supported by grants from the Human Frontier Science Program Organisation and Biotechnology and Biological Sciences Research Council.