INCOMPOSITA: a MADS-box gene controlling prophyll development and floral meristem identity in Antirrhinum

Plants can generate new organs post-embryonically throughout their life cycle, owing to the activity of meristems located at the tip of growing shoots(shoot apical meristem or SAM) and roots. Meristems are dynamic structures, in which mitosis continuously replenishes cells that are depleted by differentiation (Traas and Vernoux,2002). The SAM undergoes several distinct changes of identity during development of flowering plants(Pidkowich et al., 1999). In the vegetative phase the SAM generates lateral leaves and axillary meristems. In the reproductive phase the SAM becomes an inflorescence meristem (IM) that will give rise to a number of lateral leaves and floral meristems (FM) in their axils, which initiate floral organs.

Vegetative and reproductive organs are arranged in a species-specific phyllotaxis (Reinhardt and Kuhlemeier,2002). The switch from vegetative to reproductive growth is accompanied by a change of phyllotaxis in Antirrhinum majus(Carpenter et al., 1995). During the vegetative phase, the plants display decussate phyllotaxis, where the two leaves formed per node are positioned at opposite sides of the stem. After the transition to the reproductive phase, Antirrhinum plants switch to a spiral phyllotaxis, producing small leaf-like organs (bracts) at each node along the main (inflorescence) stem. Flowers arise in the axils of bracts and consist of four types of organs arranged in a whorled phyllotaxis. Five sepals in the calyx constitute the outer (first) whorl, followed by five petals (second whorl), four stamens and a stamenoid (third whorl), and two fused carpels in the inner (fourth) whorl. Mutations in two Antirrhinum genes, FLORICAULA (FLO) and SQUAMOSA (SQUA), transform flowers into indeterminate axillary inflorescences with bracts arranged in spiral phyllotaxis(Coen et al., 1990; Huijser et al., 1992). The phenotype of flo and squa mutants indicates that both genes play a crucial role in the specification of the floral meristem. The transcript level of SQUA and FLO in flo and squa mutants, respectively, is unchanged, but after independent transcriptional induction, the SQUA and FLO functions converge in the control of flower development. This is revealed by the enhanced squa or flo mutant phenotypes when the respective FLO or SQUA functions are reduced(Carpenter et al., 1995). In this report, we provide genetic evidence that INCOMPOSITA(INCO) is an additional factor required, in cooperation with FLO and SQUA, for the control of floral meristem identity.

INCO, like SQUA, is a MADS-box transcription factor(Schwarz-Sommer et al., 1990). MADS-box genes constitute a large family, which, throughout plant evolution,have been recruited as transcriptional regulators controlling the development of various plant structures and organs (Ng and Yanofsky, 2001). inco flowers display two extra organs, named prophylls or bracteoles(Bell, 1991; Weberling, 1989), which develop very close to the lateral sepals. We propose that INCOrepresses prophyll development in Antirrhinum, which is a novel function for a MADS-box transcription factor, and show that absence of this control results in impaired floral architecture.

Antirrhinum plants were grown in the greenhouse at 18-25°C with additional light during winter. The wild-type lines JI98 (the progenitor of line 165E), flo-640 (Carpenter et al., 1995) and flo-662(McSteen et al., 1998; Simon et al., 1994) were kindly provided by Rosemary Carpenter (John Innes Centre, Norwich, UK). The wild-type line Sippe 50 and the mutants, def-gli(Sommer et al., 1990), inco-pannosa and inco-deformis (referred to as def,inco-1 and inco-2, respectively) were obtained from the collection at the IPK, Gatersleben (Accession numbers MAM88, MAM162 and MAM161, respectively). inco-3 and inco-4 arose in a mutagenesis program, where the inco-1 allele has been targeted for transposon insertion (E. de Andrade Silva and Z.S.-S., unpublished). The squa-347 mutant has been previously described(Huijser et al., 1992). Because all inco mutant lines displayed identical phenotypes and lacked INCO expression (see Results), double mutants with squa-347 and flo-662 were constructed using the inco-1 allele. The presence of mutant alleles was confirmed by PCR and the phenotype of double mutants was corroborated by growing F3 populations of about 100 plants with defined genotypes.

Col0 Arabidopsis thaliana plants were transformed according to a dipping protocol (Clough and Bent,1998). 35S::INCO was constructed by inserting the full-size cDNA into the BamHI site of the pPCV072 vector(Koncz et al., 1990), and for 35S::SVP the XbaI site of pBAR-35S [modified from Becker et al.(Becker et al., 1992)] was used. The T2 progeny of several transgenic lines was grown in climate chambers at 20°C and 16 hours of light.

Scanning electron microscopy (SEM) was carried out with replicas of flowers and developing inflorescences as previously described(Green and Linstead,1990).

The vascular skeleton was viewed under bright field after processing the tissues according to Candela et al.(Candela et al., 1999).

In situ hybridisation with digoxigenin-labelled antisense RNA was performed as previously described (Bradley et al.,1993; Davies et al.,1996; Huijser et al.,1992). The INCO probe did not contain the MADS-box.

For northern blot analyses, 1 μg of mRNA was loaded per lane,transferred to nylon membranes and processed as previously reported(Sommer et al., 1990).

Leaf samples were ground in liquid nitrogen and suspended in extraction buffer (250 mM TrisHCl, pH 7.5; 1% SDS; 25 mM EDTA, 250 mM NaCl). After phenol/chloroform extraction, the DNA was precipitated and the pellet was resuspended in TE buffer containing 5 μg/ml RNAseA. The screening procedure followed the protocol described by Keck et al.(Keck et al., 2003). Detailed information on PCR primers and PCR conditions for these and all other experiments performed in this report are available upon request.

For yeast two-hybrid experiments the INCO or SQUA coding sequences were inserted into the EcoRI/SalI site of the pGAD424 and pGBT9(Clontech) vectors. INCOΔC (amino acids 1 to 206) was constructed by PCR amplification of the respective region of the INCO cDNA. This C-terminal deletion eliminates the transcriptional activator domain and prevents auto-activation in yeast. Ternary complex formation was tested with INCOΔMIK1/2 (amino acids 104 to 229 in the pGAD424 vector) using the full-size PLE cDNA (inserted into pGBT9) and the full-size DEFH200 cDNA(cloned into the EcoRI/SalI site of the pTFT1 vector(Egea-Cortines et al., 1999). Yeast libraries, screening protocols and all other constructs are described elsewhere (Davies et al., 1996; Egea-Cortines et al.,1999).

Semi-quantitative assays for comparing the strength of INCO homodimerisation and heterodimerisation with several partners were performed by liquid lacZ assays (Kippert,1995) using the SFY527 yeast strain. Activity in Miller units was calculated according the formula (1000 ≥A₄₂₀≥V_r)/(t ≥V_c ≥A₆₀₀), where V_r=final reaction volume in ml; V_r=volume of culture assayed in ml; t=time in minutes. Average and standard deviation of four independent assays are shown in Table 1.

DEFH70 is a MADS-box transcription factor first isolated during screening of an Antirrhinum genomic library for MADS-box genes and was also detected in yeast two-hybrid screens as a putative interacting partner of PLENA and DEFH200 (Davies et al.,1996). A reverse genetic approach(Keck et al., 2003) using sequence information of DEFH70 identified two classical allelic mutants, inco-1 and inco-2(Stubbe, 1966). Two additional inco alleles, inco-3 and inco-4, were obtained during a transposon mutagenesis program. All inco alleles carried CACTA-type transposons inserted in the gene(Fig. 1A). Among 162 inco-4 plants, a wild-type revertant was isolated, in which the Tam7-like element had been excised leaving a footprint behind (not shown). The occurrence of four alleles, the genetic instability of inco-4 and the absence of DEFH70 transcript in the mutants(Fig. 1B) provide the evidence that DEFH70 corresponds to the INCO gene.

Searches in databases showed that INCO shares 73.8% and 69.9% amino acid identity and 82.7% and 78.2% similarity with JOINTLESS (J)(Mao et al., 2000) and SHORT VEGETATIVE PHASE (SVP) (Hartmann et al.,2000), respectively (Fig. 2A,B). These genes belong to the StMADS11 subfamily(Becker and Theißen,2003), and perform very diverse functions in plant development;SVP is a floral repressor in Arabidopsis(Hartmann et al., 2000) and JOINTLESS organises the pedicel abscission zone in tomato(Szymkowiak and Irish,1999).

Like JOINTLESS, the INCO gene consists of eight exons and seven introns (Fig. 1A), while SVP has nine exons. The INCO MADS-box is encoded by the first exon, a typical feature of MIKC type MADS-box genes.

Northern blot and RT-PCR analyses indicate that INCO mRNA is present in all organs during the vegetative and reproductive phases (not shown). INCO mRNA is detectable in situ in two opposite domains of the SAM (Fig. 3A),corresponding to the position of incipient leaf primordia [P0(Waites et al., 1998)]. Similarly, INCO transcript is present in initiating bract primordia in the IM (Fig. 3B). Meristematic cells within the apical dome of the SAM and IM, however, do not express INCO. INCO is transcribed in emerging axillary meristems during the vegetative phase and in floral meristems until early stage 3(Fig. 3C-E). During stage 3, INCO mRNA disappears from the emerging sepal primordia and becomes more restricted to deeper layers of the floral meristem. Later, INCOexpression is only detectable in developing and mature anthers as revealed by northern blot analysis and in situ hybridisation experiments (not shown).

In summary, INCO expression during vegetative and reproductive development displays striking similarities, although, as shown below, mutation of INCO phenotypically affects only reproductive development.

The most conspicuous feature of inco mutants are two narrow, or filamentous structures beneath the two lateral sepals, outside the calyx,which are absent in the wild-type flower(Fig. 4)(Wilkinson et al., 2000). Occasionally, these organs reach the size of sepals, but in contrast to true sepals, and similar to bracts (Keck et al., 2003), they develop a glandular structure at their tip(Fig. 4D,E). In contrast to leaves and bracts with branched venation, however, the venation of prophylls resembles the parallel pattern observed in sepals(Fig. 4E,F). Mature flowers frequently display twisted or distorted petals, petaloid lateral sepals and petals that are partly fused to sepals(Fig. 4B,C). The lateral sepals are often smaller than in the wild type and are sometimes reduced to tiny narrow organs. The inco phenotype is variable, displaying nearly wild type to severely distorted petals and extra organs that are free-standing or fused to the adjacent lateral sepals. The phenotypes of different inco mutants are very similar, although in inco-4 flowers the additional two organs are sometimes positioned lower on the pedicel or very close to the subtending bract.

Comparison of developing wild-type and inco flowers by SEM revealed that the primordia of the additional two organs arise at stage 2,before the genuine sepals, very close to the position normally occupied by the two lateral sepal primordia (Fig. 5A,B). This resembles the position of prophylls [`the first leaf on the shoot' (Weberling,1989)] in species in which their development is not suppressed. As in wild-type flowers, the two ventral (abaxial) sepal primordia and the dorsal organ initiate almost simultaneously in inco flowers. In species with suppressed prophyll development the order of sepal initiation and the position of the organs is the same as if prophylls were present(Weberling, 1989). The lateral positioning of the additional organs in the inco mutant and the initiation pattern of sepals in the wild type (as well as in the mutant)suggest that the two additional inco floral organs are prophylls, the development of which is suppressed by the INCO gene product in the wild type.

However, subsequent initiation of the lateral sepals is delayed in inco, and the organs are displaced towards the centre of the flower primordium (Fig. 5C,D). Frequent petaloidy of lateral sepals and fusions between sepals and petals are most likely the consequence of this displacement. The mechanical nature of these distortions is corroborated by the reduced frequency of lateral fusions between organs in whorls 1 and 2 and the lack of organ size reduction in whorl 1 of def inco double mutants, where whorl 2 is occupied by small sepaloid organs instead of the larger petals, owing to the homeotic defect caused by mutation in the DEF gene(Sommer et al., 1990)(Fig. 6A,B and Fig. 5G,H).

Development of prophylls perhaps is due to some delay or incompleteness in determination of floral meristem identity. To test this possibility, double mutants between inco and mutants controlling Antirrhinumfloral meristem identity such as flo and squa were constructed.

Severe flo-640 mutants (Fig. 6C) display bracts arranged in a spiral phyllotaxis that carry,instead of flowers, axillary inflorescences composed of bracts in their axils(Coen et al., 1990). This severe phenotype is not affected when combined with inco (not shown). However, in combination with the weak flo-662 allele, which displays wild-type-like inflorescences with flowers(McSteen et al., 1998), inco flo-662 double mutants exhibit inflorescences(Fig. 6D,E). Expression of FLO is not altered in the inco mutant(Fig. 3H), indicating that the enhanced flo mutant phenotype is not due to impaired transcriptional regulation of flo-662 in the double mutant. The synergistic effect of mutations in inco and flo thus suggests that the INCO and FLO functions converge in establishing the floral meristem. It is possible, that the role of INCO in preventing prophyll development is related at least in part to its function in the control of floral meristem identity. If so, then prolonged delay in floral determination should result in more leaf-like organs before sepal initiation. Frequent development of an additional filamentous leaf-like organ (`third prophyll') in the inco flo-662 double mutant (not shown) supports this assumption.

squa mutants are affected in the transition from inflorescence to floral meristem, similar to flo mutants, but squa deletion mutants occasionally produce flowers, which are frequently misshapen and are subtended by prophyll-like organs (Huijser et al., 1992). Intriguingly, inco squa plants produced more flowers and flowered earlier than squa(Fig. 6F,G), and squaplants heterozygous for INCO had a phenotype intermediate between squa and the inco squa double mutant, indicating a dose effect of INCO.

The partial epistasis of inco to squa might suggest that INCO in the squa mutant background prevents flower formation and that SQUA counteracts this negative influence in the wild type. Enhanced flowering tendency of inco squa double mutants, if interpreted as indication for improved floral determination (see above),should be accompanied by improved flower morphology. Yet, the morphology of inco squa double mutant flowers(Fig. 6H) remains similar to squa flowers, and they even contain a `third prophyll' sometimes. This suggests that promotion of flowering, floral determination and prophyll initiation are not fully linked.

In the squa mutant, INCO transcript was not detectable in P0 bract primordia (Fig. 3G),in accordance with the presence of prophylls on the long pedicel of occasionally forming squa flowers(Huijser et al., 1992). SQUA is thus an activator of INCO expression. The level of INCO expression was slightly reduced in secondary inflorescence meristems (Fig. 3G), suggesting that SQUA is not absolutely necessary to establish and maintain INCO transcription in reproductive axillary meristems.

In summary, the functional relations of INCO, FLO and SQUA in the control of floral meristem identity, promotion of flowering and prophyll formation are complex, and the role of INCO in these processes cannot be simply reduced to the control of floral meristem identity.

Given the influence of INCO on floral meristem identity, we asked whether INCO interacts with MADS-box transcription factors involved in the same process. Yeast two-hybrid screens showed that INCO heterodimerises with several other MADS-box proteins such as SQUA and the so-called identity mediating (Im) proteins DEFH200, DEFH84 and DEFH72(Table 1), the orthologues of the Arabidopsis AP1 and SEPALLATA (SEP) proteins, respectively(Egea Gutierrez-Cortines and Davies,2000). Interestingly, according to semi-quantitative assays,heterodimer formation between INCO and SQUA, as well as with several other MADS-box proteins, is favoured compared with INCO homodimer formation(Table 1). It is likely therefore, that INCO homodimerisation is prevented in vivo by interactions with other proteins. As expression of all potential INCO heterodimerisation partners tested is controlled by SQUA(Davies et al., 1996), INCO homodimerisation appears to be favoured in the squa mutant background.

SVP is the closest Arabidopsis relative of the INCO protein(Fig. 2) and, similar to INCO,SVP interacts with AP1, SEPALLATA1 (SEP1) and SEP2, as well as with the respective Antirrhinum proteins(Table 1). The similarity in protein interactions is in line with the observed common features of Arabidopsis plants overexpressing SVP or INCO (see below). In contrast to INCO, however, SVP cannot homodimerise and cannot activate transcription in yeast on its own. In fact, the knockout phenotypes of inco or svp mutants differ, in part perhaps as a consequence of these differences.

INCO interacts with the floral organ identity MADS-box protein PLENA (PLE),and we also observed higher order complexes (e.g. ternary) between INCO, PLE and DEFH200 (Table 1). This could suggest that INCO is involved, together with PLE, in a developmental control function in stamens, in agreement with the expression of INCOmRNA in mature anthers. inco mutant stamens are not visibly affected in their development under our growth conditions, suggesting that other factors can mask here the role of INCO.

The observation that heterodimer formation is favoured over INCO homodimerisation prompted us to determine whether an excess of the INCO gene product resulting in over-production of INCO in the plant has some developmental consequences. Arabidopsis plants expressing the 35S::INCO transgene were generated to address this issue.

35S::INCO transgenic plants showed dramatic delay in the floral transition compared with wild-type plants (Fig. 7A). Their flowers displayed leaf-like features, such as branched trichomes on sepals, petals and carpels(Fig. 7B-D). Early arising flowers were more severely affected than later ones, and initiated inflorescences within the gynoecium (not shown). All changes observed point to incomplete floral transition in the presence of the 35S::INCO transgene. This is in line with the observed enhancement, compared with squa, of flowering in inco squa double mutants and supports the idea that, in the squa mutant, flowering is prevented by an excess of INCO, e.g. by formation of the INCO homodimer. In fact, Arabidopsis svp mutants flower earlier than wild type, suggesting that SVP prevents floral transition(Hartmann et al., 2000). In agreement with this function, and also with the yeast experiments, the behaviour of 35S::SVP and 35S::INCO in transgenic plants is similar(Fig. 7A,D), suggesting some common molecular and functional properties of the proteins. However, the svp and inco knockout phenotypes differ, indicating that this common potential is exploited in different ways in Arabidopsisand Antirrhinum. Thus, as previously noticed, the bona fide function of MADS-box proteins cannot be fully deduced from overexpression in transgenic plants in the absence of observations with loss-of-function mutants(Davies et al., 1999).

MADS-box genes, whose protein products belong to a particular subfamily,frequently reveal similar spatial and temporal expression patterns and perform similar control functions (Theißen et al., 1996). This `rule' does not hold true for the StMADS11 group, the members of which participate in diverse functions and display broad, but distinct, expression patterns.

The tomato MADS-box gene JOINTLESS is expressed in all tissues tested by northern hybridisation, but has a defined role in the establishment of the abscission zone on pedicels only(Mao et al., 2000).

SVP, a negative regulator of flowering time in Arabidopsis, is expressed in reproductive meristems in a pattern similar to INCO (Hartmann et al.,2000). Interestingly, overexpression of SVP or INCO in Arabidopsis prolongs the vegetative phase and prevents flowering suggesting potential functional similarities. INCO, like SVP, could either directly repress the vegetative to floral transition in Arabidopsis, or could indirectly interfere with the function of proteins controlling flowering. As several regulators of Arabidopsis flowering time are MADS-box proteins (reviewed by Becker and Theißen,2003), reinforced interactions in the presence of excess of SVP or INCO might result in their depletion and inactivation. In spite of the potential to interfere negatively with flowering in Arabidopsis and in apparent contrast to svp mutants, inco mutants have no obvious flowering time defects. In fact, as discussed below, the negative role of INCO is masked by competitive interaction with other proteins, such as SQUA, in the wild-type plant.

AGL24 and SVP are close Arabidopsis relatives within the StMADS11 group (Fig. 2B). The AGL24 gene is abundantly expressed in Arabidopsis in all meristems, except for the floral, where its expression is limited to a single cell layer (Yu et al., 2004). When overexpressed, AGL24 promotes flowering(Michaels et al., 2003), in apparent contrast to INCO and SVP, which repress flowering(see above). Furthermore, in the wild type, AGL24 negatively controls floral meristem identity as demonstrated by the rescued, wild-type like phenotypes of double mutants of agl24 with lfy(Yu et al., 2004). By contrast, defects in FLO (the orthologue of LFY) in flo mutants are enhanced by inco, suggesting a positive role of INCO in the control of Antirrhinum floral meristem identity, as discussed below. Thus, the role of INCO in floral meristem identity control is distinctively different from that of either SVP or AGL24.

Last but not least, INCO is a crucial control gene for repression of prophyll development, a function that has not been reported yet for MADS-box genes. Thus, the StMADS11 clade unites members with a variety of different functions: some are unique to individual members, some are shared among them and some others have opposite functions.

The term prophyll is applied to the leaf or leaves at the first (proximal)node on the shoot, distinguishable in shape and arrangement from other leaf organs. Bracteoles, which are small leaf-like organs developing between a subtending bract and the calyx of the flower, are therefore prophylls(Bell, 1991). Prophylls possibly protected floral buds and reproductive organs in ancestors of modern angiosperms, and their function was taken over by the calyx during evolution. The evolutionary ancient origin of prophylls is supported by their presence in the `living fossil' Amborella, where female and male flowers contain two prophylls placed close to, or within, the flower(Endress and Igersheim, 2000). Similar to inco mutant flowers Amborella prophylls can fuse with the nearest sepals.

Prophylls became integrated into the calyx of some species during evolution and in some others prophylls were `lost'(Weberling, 1989). Indicative for the former case are flowers where lateral sepals (the genuine position of the prophylls) initiate first, while evolutionary loss (degeneration by suppression) of prophylls is suggested by the abaxial-dorsal-abaxial sepal initiation pattern, before lateral sepals appear. In Antirrhinum, the order of sepal initiation in the wild type indicates that prophylls were not integrated into the calyx and the development of prophylls in the inco mutant suggests that INCO was recruited during evolution to prevent their development. This interpretation correlates well with the lateral position of inco prophylls beneath the calyx, the emergence of their primordia before those of ventral and dorsal sepals, and the maintenance of the principal order of floral organ initiation in the inco mutant. Alternatively, it is possible that the incomutation affects the timing of lateral sepal initiation and causes a heterochronic homeotic defect. In this scenario, the lateral sepals initiate first, at a time when floral identity is not fully established, and hence they acquire an intermediate bract/sepal fate. The pair of lateral organs that develop subsequent to the ventral and dorsal sepals correspond then to extra sepals, which, in the mature calyx, occupy the position of lateral sepals. The developmental role of inco thus would be to prevent premature lateral sepal initiation. However, it is difficult to relate displacement of the two first initiated (transformed) sepals outside of the calyx and initiation of two extra organs to a homeotic defect alone. Therefore we favour the `prophyll hypothesis', where the extra inco organs are prophylls rather than homeotically transformed lateral sepals.

Several species of the Scrophulariaceae lack prophylls, like Antirrhinum, but their presence is common as well(Heywood, 1998), for example,in two of the English Verbascum species(Stace, 1997) and in the Chinese Mimulus bracteosus. It will be interesting to elucidate whether presence and absence of prophylls in a species is associated with changes at an INCO-like locus.

Various theories assume that organ initiation is regulated by the geometry of the apex and by mechanical forces (tension and compression) that act on meristem surfaces (Reinhardt and Kuhlemeier, 2002). According to the theory of the `first available space', based on microsurgical manipulations, the timing and positioning of organ initiation is regulated by the availability of the minimal free area on the meristem surface, at a minimal distance from the summit and from pre-existing primordia (Snow and Snow,1931; Snow and Snow,1933). In other interpretations, space itself is not decisive; a new primordium emerges at the position of weakest inhibition where the most recently formed primordium is the strongest source of inhibition(Tooke and Battey, 2003; Wardlaw, 1949). Indeed, in inco mutants, development of lateral sepal primordia is significantly delayed, owing to the aberrant initiation of prophyll primordia, but the temporal order and the principal site of their initiation are not affected. Initiation of prophylls thus does not seem to interfere with auxin redistribution, shown to be a decisive factor in the maintenance of phyllotaxis (Reinhardt et al.,2003).

However, the presence of prophylls has severe consequences for the overall architecture of the flower, in that lateral sepal primordia are forced towards the developing petal primordia in the second whorl and, perhaps owing to consumption of cells by the prophylls, lateral sepals are frequently smaller than the corresponding wild-type sepals. Chimeric sepaloid-petaloid organs develop frequently, or, if contact is established to the petal primordia,sepals and petals can fuse. Such anomalies were also observed in sunflower,where applying mechanical stress during development resulted in large bracts instead of the dyad (bract/floret) structure(Hernandez and Green, 1993). The mechanical nature of these alterations in the inco mutant is corroborated by the lower frequency of size reduction of organs and of fusions between first and second whorl organs in inco def double mutants,where the size of second whorl primordia is reduced, owing to their homeotic transformation to sepaloid organs. Thus, repression of prophyll initiation by INCO is a prerequisite for establishment of the wild-type floral architecture.

The phenotype of inco with the lack of repression of prophyll development and the disordered development of floral organs in incomutants resembles the phenotype of rarely forming squa flowers. In fact, we found that during the time of bract initiation, SQUA is a direct or indirect activator of INCO expression(Fig. 8, left).

During flower formation, however, INCO expression is less dependent on SQUA and the relation between their functions becomes more complex. This is revealed by the observation that squa incoplants produce more flowers than do squa plants, suggesting that INCO in the absence of SQUA prevents reproductive axillary meristems to become flowers. INCO, therefore, is a repressor of flower development, although other factors may be involved in addition, as squa mutants can flower,albeit at low frequency. The fact that inco mutants do not flower more abundantly than wild type can be explained when assuming that INCO promotes expression or function of a repressor of flowering whose effect on flower formation is counteracted by SQUA(Fig. 8, thin red lines). This function is most probably performed by INCO as homodimer, the existence of which is supported by yeast two-hybrid experiments and whose negative influence on flowering is manifested in transgenic Arabidopsis plants overexpressing INCO. The observed dependence of flowering on INCO dose in the squa mutant background would thus indicate that the amount of homodimer that can form in an inco/INCOheterozygote is not fully sufficient to promote the function of the putative repressor of flowering. However, the possibility that, in addition to (or instead of) the INCO homodimer, INCO in association with other proteins performs this repressor function cannot be excluded.

Somewhat surprisingly, INCO is also a positive factor in the control of floral meristem identity. This is uncovered in inco flo-662 double mutants, where inco enhances the otherwise not manifested meristem identity defect in the weak flo-662 mutant. In this respect, the influence of INCO is comparable with the enhancement of the flo-662 mutant phenotype in the background of squa [see Introduction and Carpenter et al.(Carpenter et al., 1995)]. Thus, in the presence of SQUA, INCO acts together with FLOto promote flower development.

Given that wild-type plants flower in spite of INCO expression and hence potential repression of flowering, we have to explain how INCO can become a positive factor in flowering. One appealing assumption is that the INCO/SQUA heterodimer (and/or heterodimerisation with proteins whose expression is controlled by SQUA) performs the promoting function, and that in the presence of SQUA heterodimerisation is favoured compared with INCO homodimerisation (Fig. 8, green lines). This would deteriorate the repressive function of the INCO homodimer in the wild type, which is in favour of promotion of flowering. In fact, at least in yeast, the SQUA/INCO interaction (and the interaction of INCO with several other SQUA-controlled proteins) is stronger than INCO homodimerisation. In addition, impaired floral determination of transgenic Arabidopsis plants expressing excess of INCO shows that disturbing the balance in favour of INCO, and hence facilitating homodimer formation,enhances the negative effect of INCO. The overlap of the SQUA and INCO expression patterns in initiating floral primordia is in agreement with the potential for protein-protein interaction between the INCO and SQUA proteins in planta.

Intriguingly, in the presence of sufficient SQUA and FLOfunction in the wild type, the role of INCO in the control of floral meristem identity appears superfluous, and the lack of INCO in the inco mutant manifests itself in prophyll development only. Possibly,suppression of prophyll development by INCO is a relatively novel function acquired by a MADS-box protein with the potential to interfere with the floral transition. The complex relations to SQUA and FLOwere perhaps established to prevent this interference.

We thank Coral Vincent for her kind help with the SEM printing technique,Thomas Münster for helping us in phylogenetic tree reconstruction,Brendan Davies for helpful criticism, and Sybille Richter and Manfred Pohe for their excellent care of plants. S.M. was supported by an EEC Marie Curie postdoctoral fellowship.