Mutations in the APETALA1 gene disturb two phases of flower development, flower meristem specification and floral organ specification. These effects become manifest as a partial conversion of flowers into inflorescence shoots and a disruption of sepal and petal development. We describe the changes in an allelic series of nine apetala1 mutants and show that the two functions of APETALA1 are separable. We have also studied the interaction between APETALA1 and other floral genes by examining the phenotypes of multiply mutant plants and by in situ hybridization using probes for several floral control genes. The results suggest that the products of APETALA1 and another gene, LEAFY, are required to ensure that primordia arising on the flanks of the inflorescence apex adopt a floral fate, as opposed to becoming an inflorescence shoot. APETALA1 and LEAFY have distinct as well as overlapping functions and they appear to reinforce each other’s action. CAU - LIFLOWER is a newly discovered gene which positively regulates both APETALA1 and LEAFY expression. All functions of CAULIFLOWER are redundant with those of APETALA1. APETALA2 also has an early function in reinforcing the action of APETALA1 and LEAFY, especially if the activity of either is compromised by mutation. After the identity of a flower primordium is specified, APETALA1 interacts with APETALA2 in controlling the development of the outer two whorls of floral organs.
Most angiosperm species are induced to flower in response to environmental conditions such as day length and temperature, and internal cues including age. Following germination, the shoot meristem produces a series of leaf meristems on its flanks. However, once floral induction has occurred flower meristems arise instead. These in turn produce a determinate number of floral organ primordia, which develop individually into sepals, petals, stamens or carpels. Thus, flower formation can be thought of as a series of distinct developmental steps - floral induction, the formation of flower primordia and the production of floral organs. Mutations disrupting each of the steps have been isolated in a variety of species, suggesting that a genetic hierarchy directs the flowering process (for review see Weigel and Meyerowitz, 1993).
Recently, studies of two distantly related dicotyledons, Arabidopsis thaliana and Antirrhinum majus, have shed light on some of the molecular genetic mechanisms underlying the flowering process. For example, the specification of floral organ identity depends on the products of at least three classes of homeotic genes, acting alone and in combination (Bowman et al., 1989, 1991; Carpenter and Coen, 1990; Schwarz-Sommer et al., 1990). Several of these genes are transcription factors whose conserved DNA-binding domain has been named the MADS box (Schwarz-Sommer et al., 1990; Coen and Meyerowitz, 1991).
Earlier-acting genes controlling the identity of floral meristems have also been characterized. Flower meristems are derived from the flanks of the indeterminate inflorescence meristem in both Arabidopsis and Antirrhinum. Two of the factors that determine that these meristematic cells will develop as flowers are known. In Arabidopsis, they are the products of the LEAFY gene (Weigel et al., 1992) and the APETALA1 gene (Mandel et al., 1992). When inactivated by mutation, the flanking primordia develop into structures combining the properties of flowers and inflorescence shoots in both leafy mutants (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992) and, in different ways, in apetala1 mutants (Irish and Sussex, 1990). In Antirrhinum, the homologue of the Arabidopsis LEAFY gene is FLORICAULA (Coen et al., 1990) and that of the APETALA1 gene is SQUAMOSA (Huijser et al., 1992). The latter pair contain MADS box domains.
An allelic series of strong, intermediate and weak leafy mutants has recently been characterized, presumably containing progressively higher levels of LEAFY activity. This series has been useful in deducing the wild-type LEAFY function (Huala and Sussex, 1992; Weigel et al., 1992). Further, the LEAFY gene product seems to interact with that of APETALA1 as the phenotype of the apetala1 leafy double mutant deviates more severely from wild type than do the phenotypes of either single mutant (Huala and Sussex, 1992; Weigel et al., 1992).
Here we describe the phenotypes of an allelic series of nine apetala1 mutants and compare their morphology and development. We have also examined interactions between apetala1 mutants and those of five other floral control genes. One of these is a new gene, CAULIFLOWER, which greatly enhances the ap1 mutant phenotype when in homozygous recessive form. The results extend those of an earlier study of the apetala1-1 mutant (Irish and Sussex, 1990) and provide insights into the various roles APETALA1 plays in controlling the identity of flower primordia and of floral organs that arise on them.
MATERIALS AND METHODS
Origin of mutant alleles
The origins of the nine mutant ap1 alleles studied are summarized in Table 1.
The cauliflower-1 allele was identified as an enhancer of ap1-1 in the F2 of a cross between an ap1-1 plant of Landsberg erecta background and a wild-type Wassilewskija (Ws) ecotype plant (Feldmann and Marks, 1987). Selfing of plants homozygous for ap1-1 and heterozygous for cal-1 resulted in a segregation of 498 ap1-1 to 175 ap1-1 cal-1 plants in the next generation. The enhanced phenotype was identified in five out of five F2 families, indicating that it is homozygous in the Ws background. Two lines of evidence suggest that the apparently recessive cal-1 allele is specific to Ws. Firstly, no enhancement of the ap1-1 mutant phenotype was observed after crosses to different ecotypes (Columbia, Niederzenz, Bensheim and Kiruna, a late flowering ecotype). Secondly, mutant ap1 alleles isolated in four different ecotypes, Landsberg erecta, Columbia, Nossen and either Estland or Limburg (Table 1), display the non-enhanced phenotype. For phenotypic analyses, the recessive cauliflower allele was back-crossed into the Landsberg erecta ap1-1 background five times. The cal locus is on chromosome 1 in the vicinity of ga4 and dis2.
Mutants of other genes used in this study have been described previously (Table 2).
Genetic and phenotypic analyses
Plants were raised in a mixture of equal parts of perlite and a commercial seed-raising mix and irrigated with nutrient solution. Except where noted, phenotypic analyses were done on plants grown in a greenhouse maintained at close to 25°C under constant illumination with fluorescent lights (cool white) at approximately 130 μE m−2 sec−1 in addition to natural daylight. For other temperatures, phenotypic analyses were performed on plants grown under continuous fluorescent lighting at 15±1°C or 30±1°C.
All double mutants segregated close to the expected ratio of 1:15 in F2 families. For ap1-1 ap2-2, ap1-4 lfy-5 and ap1-5 ap2-1, seeds were harvested from individual ap1 or ap2 F2 plants, and the novel double mutant phenotype was observed in a 1:3 ratio in the F3 generation. The ap1-1 cal-1 lfy-6 and ap1-1 cal-1 tfl-2 triple mutants were identified by harvesting seed from individual ap1-1 cal-1 F2 plants and observing the novel phenotype segregating 1:3 in the F3 generation. The heterozygous effect of cal-1 in the ap1-1 background was shown by crossing ap1-1 cal-1 plants with ap1-1 plants and observing a phenotype intermediate to that of the two parents in the F1 plants. The heterozygous effect of cal-1 in the ap1-5 and ap1-1 tfl-2 backgrounds was shown similarly. The triple mutant ap1-1 ap2-2 ag-1 was identified as follows. A novel phenotype was recorded in the F2 of a cross between an ap1-1 plant and a plant homozygous for both ap2-2 and pi-1 and heterozygous for ag-1. Seeds were collected from individual F2ap1-1 plants. Some of the F3 families segregated the novel phenotype at the ratio of about 1:7. Each of these families homozygous for ap1-1 segregated ap2-2 and ag-1 but not necessarily pi-1. This segregation ratio (ap2 and ag are linked and show about 25% recombination) and the segregating phenotypes indicate that the novel phenotype corresponds to the ap1-1 ap2-2 ag-1 triple mutant. That no different phenotype was noted in those families also segregating pi-1 suggests that the ap1-1 ap2-2 ag-1 pi-1 quadruple mutant is indistinguishable from the ap1-1 ap2-2 ag-1 triple mutant.
For phenotypic analyses, usually the first fifteen structures produced by the primary inflorescence meristem after cauline leaf production were examined.
Scanning electron microscopy (SEM) was performed as described in Alvarez et al. (1992) except that an accelerating voltage of 20 kV was used and images were recorded on either Kodak Plus-X Pan 220 professional film or Ilford FP4 film.
In situ hybridization
Fixation of tissue, sectioning, hybridization and washes were carried out as described by Drews et al. (1991). Briefly, inflorescences were fixed for 3 hours in 3.7% formaldehyde, 5% acetic acid and 50% ethanol. Fixed tissue was dehydrated with ethanol, cleared with Histo-Clear (National Diagnostics) and embedded in paraffin (Paraplast Plus). Embedded tissue was sliced into serial 8 μm sections with a Reichert-Jung 2040 microtome and attached to microscope slides coated with poly-L-lysine (Sigma). [35S]UTP probes were generated by run-off transcription with T3, T7 or SP6 RNA polymerase (Promega). Probes were synthesized from templates described by Weigel et al. (1992) and Mandel et al. (1992). In the case of AP1, probes were synthesized from a region of the gene that did not cross-hybridize with any other cloned genes (Mandel et al., 1992). Slides were dipped in nuclear emulsion, exposed for from one to two weeks, developed for 2.5 minutes in Kodak D19 and stained with toluidine blue.
The apical meristem of Arabidopsis thaliana generates leaves until the plant becomes florally induced. Induction occurs relatively rapidly under the influence of long days or much later under short-day conditions (Napp-Zinn, 1985). It results in the apical meristem generating the primary inflorescence shoot and bractless flowers are now produced on its flanks instead of leaves (Fig. 1). The leaves and flowers arise in a continuous phyllotactic spiral, and the number of flowers produced is indeterminate. The plant bolts soon after the first flower primordia develop due to increased internode elongation between several of the later-formed leaves (cauline leaves) and between flowers. A cluster of earlier-formed rosette leaves remains at the base of the stem. Each of the cauline leaves contains an axillary inflorescence meristem that reiterates the development of the primary inflorescence. The inflorescence is a raceme with the development of individual flowers proceeding acropetally (from basal to apical positions; Müller, 1961).
Arabidopsis flowers are composed of four concentric whorls of organs, with four sepals in the outermost whorl, four petals in the second whorl, six stamens in the third whorl (four medial and two lateral) and two fused carpels occupying the center of the flower (Fig. 2B; Müller, 1961; Hill and Lord, 1989; Smyth et al., 1990). Their early development has been described in detail and divided into twelve stages (Smyth et al., 1990; Fig. 2A).
We have characterized nine recessive mutant ap1 alleles (Table 1), one of which, ap1-1, has been described in detail previously (Irish and Sussex, 1990). The mutant alleles can be arranged into a phenotypic series (strong, intermediate and weak) based both on the extent of inflorescence-like character exhibited by ap1 flowers and the severity of alterations in floral organ development. These two aspects will be discussed separately.
ap1 mutations cause partial transformations of flowers into inflorescences
The production of leaves is not altered in ap1 mutants. However, those meristems that would give rise to a single flower in wild type instead often give rise to a determinate, branched structure comprised of several individual flowers (Figs 1A, 2C). These structures consist of a central ‘primary’ flower with the extra flowers usually arising slightly later in the axils of the first whorl organs of the primary flower. In turn these secondary flowers may also have flowers arising in the axils of their first whorl organs, and so on such that tertiary and quaternary flowers may be formed. The overall result is a complex branched structure. In some cases, axillary flowers may not be subtended by a fully developed organ (Fig. 2C). However, close inspection reveals that there is usually an aborted organ primordium with flanking stipules (Fig. 3K).
In ap1 mutant plants, the total number of flowers in the branched structures varies acropetally (from bottom to top of the inflorescence shoot) and with growth temperature (Table 3). The most basal positions in relatively strong mutants (alleles 1, 9, 7 and 2) may have more than twenty branched flowers arising from the position a single flower would occupy in wild type, while the more apical positions may be occupied by a single flower. Similarly, growth at lower temperatures (15°C) results in more flowers per branch while growth at 30°C results in less. In flowers of intermediate ap1 alleles (4, 6 and 8), the determinate structures are seldom composed of more than eight individual flowers, while in weak ap1 alleles (5 and 3) no axillary flowers were observed at 25°C (Table 3).
Ectopic inflorescences may sometimes develop directly on the pedicels of ap1 flowers. They arise below the first whorl organs of the primary flowers of the determinate branched structures (Fig. 1B). The ectopic structures are usually not associated with any subtending organ or aborted primordia, although occasionally they arise from the axil of a cauline leaf-like organ. In strong alleles, these ‘pedicellar inflorescences’ are relatively common. For example, 39 were observed on 20 ap1-1 plants and 18 on 6 ap1-7 plants.
In stronger alleles the main ‘flower’ meristem itself may sometimes behave as a full inflorescence meristem, producing an indeterminate number of flowers in a phyl-lotactic spiral (Figs 1C, 2F). We term this a second order inflorescence meristem (the first order is the normal apical inflorescence meristem). Even third order inflorescence meristems (Fig. 1D) may sometimes develop. These higher order apices develop in a manner that is indistinguishable from that of the apical inflorescence meristem (Fig. 2F), suggesting that the transformation of flower to inflorescence meristem is complete. Such conversions are almost always restricted to the most basal positions of the main inflorescence stem. At 25°C, there are only one or two per plant (if any) for strong and intermediate alleles, and none at all for weak alleles (Table 3). However, the number varies markedly with temperature. For example, ap1-1 plants grown at 15°C had an average of 11 secondary inflorescences in the first 15 ‘flower’ positions, while at 30°C only one was seen on 15 ap1-1 plants (Table 3).
Thus strong and intermediate mutant ap1 alleles show partial, and occasionally complete, conversions of flowers to inflorescences, especially at basal positions.
ap1 mutations alter the development of floral organs
First whorl effects
Strong alleles. The first whorl organs display features not found in either floral or vegetative organs of wild-type plants. They have been referred to as bract-like since they resemble the bracts that subtend flowers of many other species (Irish and Sussex, 1990). The shape of the organs is neither sepal-like nor leaf-like, being elongate and rather pointed and lacking a defined petiole (Fig. 3A). They do have some leaf-like properties including epidermal cells resembling those of leaves, stipules and late senescence, while sepal-like characteristics include elongate cells on their outer surface and the absence of stellate trichomes. Other first whorl positions of strong mutants are not occupied by fully developed organs but aborted primordia, often flanked by stipules, are frequently present (Fig. 2I, Table 4). In the more acropetal flowers, organs often fail to fully develop in all four first whorl positions (Fig. 2I).
Observations of developing flowers show that the four first whorl primordia are usually produced in a whorled, cruciform pattern (Fig. 3B). However, the pattern is often slightly twisted and, unlike wild type, the lateral organ primordia are usually initiated much lower on the receptacle than the medial primordia. In addition, later internode elongation between organs, especially those of the first whorl, results in their marked spatial separation (Fig. 2C).
In the more acropetal flowers (positions 20 and higher, Fig. 1) of strong alleles grown at 25°C, the medial first whorl organs develop carpelloid characteristics such as stigmatic papillae at their tips and rudimentary ovules along their margins. This is accentuated when ap1-1 plants are grown at 30°C where the medial first whorl organs develop as solitary carpels. The lateral organs are less affected in each case.
Intermediate alleles. The first whorl organs of basal flowers in ap1 alleles 2, 4 and 6 more closely resemble leaves (Table 4). They are triangular in shape, flanked by stipules (Fig. 3K), and have stellate trichomes present on both adaxial and abaxial surfaces. They become more bract-like acropetally (Table 4), and in flowers above about the tenth to fifteenth position, they usually resemble those of basal flowers of strong mutants.
Weak alleles. There is a marked difference between the morphology of the medial and lateral first whorl organs in ap1-5 and ap1-3. On basal flowers the medial organs superficially resemble wild-type sepals but are usually mosaics of central phylloid tissue and marginal petaloid tissue (Fig. 3C). The boundaries between the two types of tissue are usually sharp. On more apical flowers, the medial organs become leaf-like (flowers 6 to 10) and then bract-like (flowers 11 to 15 and above, Table 4), although they are often mosaics including sectors of bract-like, carpel and stamen tissue (Fig. 3D). By contrast, the lateral first whorl organs are more often bract-like, are not usually mosaics and are often absent in the more apical positions (Table 4).
All first whorl organ primordia of weak alleles arise in a cruciform pattern, but they are narrower and more pointed than wild-type sepal primordia and the lateral primordia may be lower on the receptacle compared to wild type (Figs 3E, 3F). There is little or no aberrant internode elongation between the lateral and medial first whorl organs.
Second whorl effects
Strong alleles. The most common case for ap1-1, ap1-7 and ap1-9 flowers is that no organs occupy the second whorl positions (for ap1-1 data see Table 5; Irish and Sussex, 1990). This is due to the failure to initiate any second whorl organ primordia (Fig. 2D,E). However, a small fraction of the second whorl positions are occupied by either nearly wild-type petals, morphologically wild-type stamens, or mosaic organs with distinct sectors of petal and either stamen or bract-like tissue. Although no petals were observed in the sample of plants scored for Table 5, near-wild-type petals are sometimes seen (e.g. Fig. 3I).
Intermediate alleles. Second whorl organs are often absent, especially in apical positions. When present, their structures differ markedly between the three intermediate alleles scored (Table 5). For ap1-2, stamens and leaf-like organs are often present, especially in basal flowers. The leaves closely resemble those of the first whorl both in mature morphology and in ontogeny. Their early growth is rapid compared with normal second whorl petal primordia (Fig. 3J). By contrast, ap1-4 plants often have petals and petaloid leaves in early flowers, later changing to filamentous organs (Fig. 3K; Table 5). In the second whorl of ap1-6, petals and petaloid stamens predominate (Table 5).
When organs with properties of two organ types develop, they may appear either as distinct mosaics or as blends. Distinct sectors occur in organs composed of petal and stamen tissue, whereas epidermal cells in petal/leaf organs seem to combine the properties of petals (domed-shaped with radial ridges) and leaves (rectangular with some stomata present).
Weak alleles. In basal flowers, the second whorl is most frequently occupied by morphologically wild-type petals as in wild type (Table 5). However, in the more apical flowers, such organs are often absent. When present, they are either mosaic organs consisting of distinct stamen and petal tissue (Fig. 3G,H) or morphologically wild-type stamens.
Third and fourth whorl effects
ap1 mutations have only minor effects on the third whorl. Even in strong alleles there are often six stamens present. However, the mean number per flower is slightly reduced (5.2 in ap1-1 compared with 5.9 in wild type), and the reduction is more marked in apical positions and at higher temperatures (a mean of near 4 for ap1-1 plants raised at 30°C) Observations on developing ap1 flowers show that the boundary between the second and third whorl primordia is often distorted (Fig. 3L), especially when fewer than six third whorl primordia develop. In these cases, it was sometimes difficult to assign an organ to the second or third whorl. Relevant to this, second and third whorl organs are occasionally fused (2 cases in 60 ap1-3 flowers; 1 case in 60 ap1-5 flowers, e.g. Fig. 3M), and mosaic organs consisting of petal and stamen tissue may arise from intermediate positions.
Finally, the fourth whorl gynoecium of ap1 flowers is usually normal although occasionally the two carpels are not completely fused, primarily in flowers arising higher on the main stem.
It is worth pointing out that the structures of second and higher order flowers within determinate branched flowers (Fig. 1A) are at least as severely affected as the primary flower and often more so. In the extreme, they may consist of a solitary two-carpelled gynoecium (Fig. 2H).
In conclusion, study of floral organs in nine mutant ap1 alleles has revealed the following generalizations. (i) The outer two whorls of organs are preferentially affected. (ii) The lateral first whorl organs tend to be bract-like or absent. (iii) The medial first whorl organs are highly variable and can range from being bract-like to leaf-like to carpelloid. (iv) Second whorl organs tend to be absent in strong mutants, staminoid in intermediate mutants and petaloid in weak mutants. (v) Mosaic organs tend to arise in the medial first whorl and all second whorl positions, especially in the weaker alleles. (vi) Later-produced apical flowers are more severely affected than earlier produced basal flowers. (vii) Higher temperatures result in a strengthening of the mutant phenotype for each of the alleles.
cauliflower enhances the ap1 phenotype
An apparently recessive allele of the CAULIFLOWER (CAL) locus, cal-1, significantly enhances the ap1 phenotype (see Materials and Methods for the origin of cal-1). In ap1-1 cal-1 double mutants, each meristem that in wild type would give rise to a single flower, consistently behaves instead as an inflorescence meristem. These meristems arise in a phyllotactic spiral on the flanks of the main apex in positions where flowers would normally arise (Fig. 4B). In turn these second order inflorescence meristems immediately produce further meristems in a phyl-lotactic spiral that also behave as inflorescence meristems, i. e. third order inflorescence meristems (Figs 1E, 4B-D). This process may be repeated several times resulting in the production of fourth, fifth and higher order inflorescence meristems (Fig. 4E-H). The transformed inflorescence meristems are indistinguishable from the primary apical inflorescence meristem.
After a large number of undifferentiated meristems has developed, individual flowers may eventually differentiate from a few of the last-formed meristems on any one branch (Fig. 4I). They may arise from high order meristems on early formed branches or from second or third order meristems on later formed branches. The meristems develop into individual flowers that have a phenotype like that of apical ap1-1 flowers.
The number of orders of inflorescence meristem that are produced varies acropetally and with environmental conditions. When grown at 25 ° C, the basal positions of ap1-1 cal-1 plants may have up to ninth order inflorescence meristems, while the more apical positions may have only second order inflorescence meristems. When grown under unfavorable conditions (such as 30 ° C) the basal positions may have only second or third order inflorescence meristems, whereas in the apical positions flowers develop without any proliferation of higher order inflorescence meristems. Perhaps the most striking phenotype occurs when ap1-1 cal-1 plants are grown at 15 ° C. In this case no differentiation of flowers was observed after four months, by which time twelfth order inflorescence meristems were present. Additionally, such plants produce cauline leaves at many positions within the second, third and higher order inflorescence meristems (Fig. 4J).
ap1-1 plants heterozygous for cal-1 also display a slightly enhanced mutant phenotype. Such plants have an increased number of pedicellar inflorescences (Figs 1B, 4K; 87 pedicellar inflorescences were scored on 11 plants grown at 25 ° C). In addition, the frequency of second and third order inflorescence meristems (Fig. 1D) developing in the basal positions is increased. Thus the cal-1 allele is not completely recessive, but weakly semi-dominant.
The cal-1 allele also modifies the phenotype of weak ap1 alleles. For example, the phenotype of ap1-5 cal-1 plants is enhanced so that it now resembles that of the strongest mutant, ap1-1. Such double mutants have pedicellar and second order inflorescence meristems in their basal positions, structures that are absent in ap1-5 alone (Table 3). Also, the floral organs that develop in ap1-5 cal-1 flowers resemble those of ap1-1 plants in number and type.
Plants homozygous for cal-1 alone closely resemble wild type although the first position above the cauline leaves is often occupied by an inflorescence or a determinate branched structure, a phenomenon that occurs in wild type, but at a much lower frequency. Thus a major effect of cal-1 is only seen when it is combined with ap1 mutants.
Expression pattern of AP1 in cauliflower plants
Since AP1 expression is associated with floral meristems but not inflorescence meristems (Mandel et al., 1992), we examined its expression patterns in ap1-1 cal-1 plants at different ages and grown under different environmental conditions.
In wild-type flowers, AP1 mRNA is first detected in stage 1 flowers and is uniformly expressed at a high level through-out stage 1 and stage 2 flower primordia (Fig. 5A,B; Mandel et al., 1992). In stage 3-4 flowers, AP1 mRNA becomes restricted to the first and second whorls of the wild-type Ara - bidopsis flower, with no expression detectable in the presumptive third and fourth whorls. Expression is maintained until past stage 12, throughout developing sepals and petals, with no expression detectable in the stamens and carpels. In addition, AP1 mRNA is present throughout the pedicel during flower development but is not detectable in the inflorescence meristem.
The expression of AP1 in ap1-1 mutant flowers occurs at normal levels in stage 1-2 flowers (data not shown). In contrast, its expression in ap1-1 cal-1 is markedly different. A very low level of AP1 mRNA is detected in young ap1-1 cal-1 inflorescences (Fig. 5C,D). However, in older inflorescences a level of AP1 expression similar to that seen in stage 1-2 of wild-type flowers is observed in some of the higher order meristems (Fig. 5E,F). AP1 expression appears to occur in the higher order meristems first, regions associated with the later formation of flowers but prior to any morphological evidence of flower development. Also there is an acropetal effect. In branches arising higher up the main stem, AP1 mRNA accumulates in meristems that have been through only a few orders of branching. In addition, a range of AP1 expression is observed in the peduncles and pedicels of the supernumerary meristems, with low expression in those carrying low order meristems and increasing expression in those of higher order meristems. These results are all from plants grown at 25°C. Very little AP1 expression is observed in the supernumerary meristems of ap1-1 cal-1 plants grown at 16°C (data not shown).
Thus the expression of AP1 in ap1-1 cal-1 plants is initially very low in primordia formed on the flanks of inflorescence apices. However, depending on environmental conditions, levels gradually increase in later formed primordia until they reach near wild-type levels. At this stage the primordia are apparently committed to develop as flowers.
Interactions between ap1, cal and terminal flower mutants
In terminal flower (tfl) mutant plants, the primary apical inflorescence meristem becomes determinate and its growth terminates in a flower composed of more than the normal number of floral organs (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992). A few normal flowers may be produced before termination. The inflorescence meristems arising in the axils of the cauline leaves usually produce only a single terminal flower (Alvarez et al., 1992). The tfl phenotype has been interpreted as a conversion of the inflorescence meristem into a flower meristem (Shannon and Meeks-Wagner, 1991) or an invasion of the inflorescence apex by flower primordia (Alvarez et al., 1992).
Since ectopic flower meristems are formed in tfl plants, we examined the pattern of AP1 expression in these meristems. The apical inflorescence meristem, which develops into a terminal flower in tfl-2 mutant plants, exhibits ectopic AP1 expression early in its development (Fig. 5G,H). Later expression is present in the outer floral whorls (Fig. 5I,J). In addition, ectopic expression is observed in the inflorescence stem itself, which, in tfl plants, acts as the pedicel of the terminal flower. A high level of ectopic AP1 expression is also detected very early in meristems in the axils of cauline leaves which will develop into individual flowers (Fig. 5G,H). Thus in tfl-2 plants AP1 expression is initiated at the time when the identity of ectopic flowers is established
Since tfl-2 and ap1-1 cal-1 mutant plants display seemingly opposite effects in the interconversion of flower and inflorescence meristem identities, we constructed all mutant combinations. Firstly, cal-1 tfl-2 plants were apparently indistinguishable from tfl-2 singles in families segregating for both, further evidence that cal-1 influences development only when ap1 is also mutant. Secondly, ap1-1 tfl-2 doubles display an additive phenotype except that the terminal flower on the main inflorescence apex has few if any axillary flowers. Overall the phenotype is close to that of ap1-1 tfl -1 plants recorded by Shannon and Meeks-Wagner (1991). Finally, the triple mutant, ap1-1 cal-1 tfl-2, displays an unexpected phenotype. Its morphology is indistinguishable from that of the ap1-1 tfl-2 double mutant (Fig. 4L), meaning that tfl is fully epistatic to cal-1 in ap1-1 mutant background. Furthermore, even heterozy-gosity for tfl-2 reduces the enhancement by cal-1 of the ap1-1 mutant phenotype. In ap1-1 cal-1 plants heterozygous for tfl-2 a maximum of three rounds of meristem proliferation is observed, in basal positions, before flowers differentiate.
Interactions between ap1, cal and leafy mutants
Like ap1, mutations of the LEAFY (LFY) gene also result in the partial transformation of flower meristems into inflorescence meristems (Schultz and Haughn, 1991; Weigel et al., 1992; Huala and Sussex, 1992). The basal positions in strong lfy mutants exhibit complete transformation of flowers to inflorescence shoots, while the more apical flowers are only partially converted (Fig. 6A). However, a major difference from ap1 is that the basal transformed inflorescences are usually subtended by a cauline leaf and the upper ‘flowers’ by a bract in lfy mutants. Also, the organs that arise in lfy ‘flowers’ are more leaf-like, are often initiated in a spiral, leaf-like manner (interior to a cruciform first whorl) and significant internode elongation may occur between the organs (Weigel et al., 1992). Finally, secondary flowers are less frequent than in ap1 mutants.
The AP1 and LFY gene products apparently interact to specify floral meristem identity. Evidence for this comes from studies of double mutants. A significant enhancement of the lfy phenotype occurs when strong or weak lfy alleles were placed in double mutant combination with the strong mutant ap1-1 (Weigel et al., 1992; Huala and Sussex, 1992).
We have found a similarly dramatic interaction between the intermediate ap1-4 allele and the weak lfy-5 allele, indicating that an interaction does not depend on a complete loss of AP1 function. While petals and stamens are common in flowers of ap1-4 (Table 5; Fig. 6F) and in apical flowers of lfy-5 (Fig. 6D; Weigel et al., 1992), ap1-4 lfy-5 flowers consist entirely of leaf-like, carpelloid leaf-like and carpel-loid organs (Fig. 6G,H). The phyllotaxy of the organs is often intermediate between spiral and whorled (Fig. 6G, compare Fig. 6C and Fig. 6E). In addition, flowers now develop in the axils of some of the leaf-like organs. The axillary flowers are much reduced, often consisting of only two or three carpels. The primary inflorescence meristem of ap1-4 lfy-5 plants usually terminates in a mass of carpelloid tissue as in strong lfy plants (Weigel et al., 1992).
To examine if lfy and cal interact when in mutant form in the same way as do ap1 and cal, plants were bred that carried both lfy-6 and cal-1. Strikingly, their phenotype could not be distinguished from that of lfy-6 single mutants indicating a major difference between AP1 and LFY in their relationship with CAL. This was reinforced when the phenotype of the triple mutant ap1-1 cal-1 lfy-6 was examined (Fig. 6B). Such plants were indistinguishable from ap1-1 lfy-6 double mutants (Weigel et al., 1992), showing that lfy-6 is fully epistatic to cal-1 in an ap1-1 background.
Expression pattern of LEAFY in cauliflower plants
Because the ap1 cal mutant combination apparently results in a prolonged loss of the ability to specify floral meristem identity at 25 ° C, we analyzed the expression of the LFY gene product in ap1 cal inflorescences grown at this temperature. In wild-type flowers, LFY mRNA is first detected at a low level in cauline leaf primordia and then in the flanks of the inflorescence meristem in the floral anlagen (precursor) cells (Fig. 7A,B; Weigel et al., 1992). It is uniformly expressed at a high level throughout stage 1 and stage 2 flower primordia, but in flowers later than stage 4 LFY mRNA becomes preferentially localized to particular differentiating organs such as the carpels and stamen filaments. Unlike AP1, LFY expression is not significant in floral pedicels.
Whereas early LFY expression is normal in ap1-1 single mutants (Weigel et al., 1992), expression in ap1-1 cal-1 double mutants is, like AP1, drastically altered. Only low levels of LFY transcript are observed in the basal-most primordia produced by young ap1-1 cal-1 inflorescence meristems (Fig. 7C,D). These levels range from those observed in wild-type cauline leaf primordia to those in floral anlagen. However, a level of LFY expression comparable to that observed in stage 1-2 wild-type flowers is seen in older ap1-1 cal-1 inflorescences (Fig. 7E,F). As in AP1, there is a marked acropetal effect. The expression of LFY is greatest in the higher order primordia, those from which mature flowers are likely to soon develop (Fig. 7E,F).
Strikingly, as with AP1, very little LFY expression is detected in any of the supernumerary meristems of ap1-1 cal-1 plants grown at 16 °C (Fig. 7G,H).
Interactions between ap1 and floral homeotic mutants
apetala1 apetala2 double mutants
Mutations in apetala2 (ap2), like ap1, affect the outer two whorls. In flowers of strong ap2 mutants (Fig. 8A), the medial first whorl positions are occupied by solitary carpels, while the lateral positions either lack organs or are occupied by leaf-like organs. No second whorl organs are present and the number of third whorl stamens is reduced to an average of less than one per flower (Bowman et al., 1991). Flowers of weak ap2 mutants have a first whorl of leaf-like organs, with occasional axillary flowers in the basal positions (10 occurrences in 20 flowers at 25°C). Organs that are intermediate between petals and stamens occupy the second whorl (Fig. 8D; Bowman et al., 1989; 1991). The severity of changes in organ identity of all ap2 alleles increases acropetally and with environmental conditions (Bowman et al., 1989; 1991) as in ap1 alleles.
In terms of organ identity, strong ap2 alleles are largely epistatic to both strong and weak ap1 alleles. For example, in ap1-1 ap2-2 flowers (Fig. 8B,C), the medial first whorl organs, when present, are carpels and the lateral first whorl positions are generally lacking fully developed organs, although there may be aborted primordia in these positions. There are no second and few third whorl organs. When present third whorl organs are either stamens or carpelloid stamens, and the fourth whorl carpels often fail to fuse properly and may also have sectors of stamen tissue. In the more apical positions, the primary flower of ap1-1 ap2-2 double mutants consists of only a single gynoecium composed of 2 to 4 carpels. This is also observed in ap2-2 single mutant flowers. Flowers of other double mutant combinations, ap1-5 ap2-2, ap1-1 ap2-9 and ap1-6 ap2-9, also resemble those of the respective ap2 single mutants (results not shown).
In contrast, a dramatic interaction is observed in double mutant combinations of ap1 with the weak allele ap2-1. For example, when ap1-5 (a weak allele) and ap2-1 are combined, the resulting double mutant flower (Fig. 8E) closely resembles that of the strong ap2-2 allele (Fig. 8A). The medial first whorl organs are now most often carpels, although in the single mutants these positions are occupied by organs that are bract-like (ap1-5, Fig. 3D) or leaf-like (ap2-1, Fig. 8D). No second whorl organs form, even in the basal flowers. Their development (Fig. 8F) is also similar to that of ap2-2 (Bowman et al., 1991). Much stronger mutant phenotypes are also seen when ap2-1 is combined with ap1-1 or ap1-2 (results not shown).
Weak ap2 mutants also enhance the inflorescence character of the branched floral structures of ap1-1 mutants. In ap1-1 ap2-1 double mutants, determinate structures composed of upwards of 50 flowers have been observed developing from positions that in wild type would be a single flower (Table 6; Irish and Sussex, 1990). In contrast, in double mutant combinations involving strong ap2 mutants, such as ap1-1 ap2-2, this effect is reversed such that the determinate branched structures are seldom composed of more than ten flowers (Table 6). However in this case the frequency of second order inflorescence meristems is increased (Table 6).
apetala1 agamous double mutants
agamous (ag) mutants affect the identity of the third and fourth whorl. Third whorl organs develop into petals instead of stamens and cells that would normally give rise to a fourth whorl gynoecium behave instead like another flower meristem. Thus ag flowers are indeterminate and consist of a large number of whorls of sepals and petals (Fig. 8G)
Flowers of double mutant plants carrying strong ap1 and ag alleles such as ap1-1 and ag-1 often appear to consist of an indeterminate number of whorls of bract-like organs with flowers arising in the axils of some of the organs of the first and internal whorls (Irish and Sussex, 1990). The bract-like organs are occasionally carpelloid (about 1 occurrence per plant). However, further inspection reveals that some of the internal organs are often petaloid (Fig. 8K) and near wild-type petals may be common (Fig. 8L).
The development of the first whorl of ap1-1 ag-1 flowers resembles that of ap1-1 flowers, with bract-like organs and axillary flowers occupying those positions (Fig. 8H,I). The pattern of organ primordia interior to the first whorl, in the region of the second and third whorls, is irregular (Fig. 8H,I). This developmental program may repeat itself several times, similar to the pattern seen in ag-1 flowers. However, unlike ag1, the floral meristem may soon revert to an inflorescence meristem and produce floral meristems on its flanks in a phyllotactic spiral. This is related to another feature of the ap1-1 ag-1 phenotype, conversion of the flower meristem to a second order inflorescence meristem (Fig. 1D). This occurs at a higher frequency in ap1-1 ag-1 double mutants (26 occurrences in 16 plants) than in ap1-1 alone (7 in 20 plants).
apetala1 apetala2 agamous triple mutants
Because AP1 seems to interact with both AP2 and AG, and the latter also interact (Bowman et al., 1991; Drews et al., 1991), we were interested to examine the phenotype when all three were in mutant form. In ap1-1 ap2-2 ag-1 triply mutant plants (all strong alleles) the striking observation is that ‘flowers’ proliferate extensively, producing an indeterminate number of primordia which remain relatively undifferentiated. Interspersed among them are leaf-like organs which are occasionally carpelloid (Fig. 9A). No structure recognizable as a floral organ was ever seen to differentiate, apart from the slightly carpelloid leaf-like organs. On the basis of the size, shape and pattern of initiation, the primordia appear to be a combination of inflorescence meristems, flower meristems and organ primordia (Fig. 9D-I).
This is seen more clearly in an analysis of developing structures (Fig. 9B,C). The pattern in which the primordia arise is sometimes spiral (as in inflorescence meristems), sometimes whorled (as in flower meristems) and sometimes intermediate (Fig. 9D-I). Those meristems that produce whorled primordia initially generate a first whorl of bract-like organs, each with an associated axillary meristem. It then either produces another whorl of similar organs (Fig. 9H), reverts to producing meristems in a phyllotactic spiral (i. e. it acts like an inflorescence meristem, Fig. 9E), or behaves intermediately between these two extremes (Fig. 9I).
AP1 controls two stages of floral development
Our study of an allelic series of ap1 mutants demonstrate that the APETALA1 gene is required early in the determination of flower meristem identity, and later during initiation and development of floral organs. Disruption of the early function results in the partial transformation of flowers into inflorescence shoots. Disruption of the later function causes failure of flower organ initiation and changes in floral organ identity, predominantly in the outer two whorls. These two effects are separable in ap1 mutants. Inflorescence properties are weaker in later formed flowers on a plant whereas floral organ changes are stronger. Also the weakest mutant alleles display only the late phenotype. The two-fold activity of AP1, as inferred from genetic analysis, correlates well with the biphasic expression pattern of AP1 RNA (Mandel et al., 1992). We will discuss the two stages in turn.
AP1 interacts with LFY in controlling floral meristem identity
Early effects of ap1 mutants result in flower meristems taking on some of the properties of inflorescence meristems. The stronger the mutant allele, the more often the ‘flowers’ grow in an indeterminate, inflorescence-like pattern, and the more often secondary and higher order flowers arise in the axils of the outer floral organs (Table 3). The strongest mutant alleles studied here (alleles 1, 7 and 9) do not result in a complete transformation. Even so it seems likely that they result in full loss of AP1 function. They have closely similar phenotypes as expected of null mutants. Also the ap1-1 allele has been sequenced and it carries a splice site acceptor mutation in the third exon (Mandel et al., 1992). This would result in loss of part of the K box, a conserved domain downstream of the MADS box with sequence similarity to the coiled-coil domain of keratins (Ma et al., 1991). While not necessarily producing an inactive product, the ap1-1 mutation would seem likely to severely disrupt its function (Yanofsky et al., 1990; Ma et al., 1991).
Much more complete flower-to-inflorescence transformations can be obtained if strong ap1 mutants are combined with strong lfy mutants, which independently cause partial flower-to-inflorescence conversions (Huala and Sussex 1992; Weigel et al., 1992). Even in combinations of the intermediate ap1-4 mutant and the weak lfy-5 mutant, we have found that flower-to-inflorescence transformations are greatly enhanced. This suggests that the products of AP1 and LFY normally interact to reinforce each others activity to specify flower meristem identity. The other four genes studied here also interact with ap1 and/or lfy when in mutant form. However, none has a significant effect on the identity of flower meristems when mutant alone. It seems that AP1 and LFY are the major players in the specification of floral meristem identity and that the other genes have secondary or partially redundant roles.
Basis of early actions of AP1 and LFY
Firstly, it appears that the initial activation of AP1 and LFY is coincident with floral induction. In Arabidopsis, factors that mediate floral induction seem to be promoted by long days and higher temperatures (Napp-Zinn, 1985). Both AP1 and LFY appear to be responsive to (i.e. activated by) these same factors. The evidence for this is as follows. In ap1 single mutants, where LFY is still functional (Weigel et al., 1992), transformation of flower meristems to inflorescence meristems is much more complete if flowering is held in check by lower temperatures (Table 3) or short days (Huala and Sussex, 1992). This suggests that LFY expression also remains at relatively low levels under these conditions, allowing prolonged, inflorescence-like growth. Likewise in lfy mutants, where AP1 is still functional, lower temperatures (Weigel et al., 1992) and short days (Huala and Sussex, 1992) result in more complete inflorescence-like transformations. Activation of the two genes by those factors that mediate floral induction could also explain the fading of inflorescence-like properties in later-arising flowers of both ap1 and lfy mutants. This is expected if levels of the factors controlling floral induction progressively increase as an induced plant ages.
Secondly, it seems that a threshold level of a factor (or factors) regulated by both AP1 and LFY must be exceeded for the complete specification of floral meristem identity to occur. It seems that the AP1 and LFY gene products act in a partially redundant manner in ensuring that this threshold is reached. If the activity of either gene is compromised by mutational change, the other can work to a limited extent to allow the threshold level to be obtained. The meristem then takes on some of the properties of a flower, the particular properties dependent on which downstream functions are lacking (see below). If the threshold level is not reached (as in the strong double mutant), the meristem behaves as an inflorescence meristem. This threshold proposal, combined with the proposed progressive strengthening of the factors promoting floral induction, also allows a relatively simple explanation for the overlapping patterns of acropetal variation shown by the strong, intermediate and weak allelic series of both ap1 and lfy mutants. The weaker the mutational change, the sooner the threshold is reached and the earlier any inflorescence-like properties are lost.
Thirdly, it seems likely that, once present, the wild-type AP1 and LFY gene products mutually enhance each other’s activity (Fig. 10). This is proposed to account for the considerably greater disruption of phenotypes seen when weak or intermediate ap1 and lfy mutants are combined. The enhancement seems to be primarily at the level of activation of target genes rather than through directly influencing each other’s transcription because each is transcribed when the other is in mutant form (see above). Such mutual enhancement, in combination with the requirement that threshold levels of activity be reached, provides a basis for the sharp transition between the production of inflorescence meristems and flower meristems seen in wild type, and the gradual transition observed in ap1 and lfy mutants.
Finally, although AP1 and LFY act in combination to specify the identity of the floral meristem, their early roles are not equivalent. In lfy but not ap1 mutations the main apex continues to produce cauline leaves/bracts. It could be that LFY but not AP1 has a role in suppressing bract formation (Schultz and Haughn, 1991; Weigel et al., 1992). This may depend on the low level of LFY expression observed in the anlagen of floral meristems even before stage 1 (Figs 7, 10). Another difference between AP1 and LFY functions lies in development of the stems of individual flowers. AP1 but not LFY is expressed in developing pedicels (Figs 5, 7; Mandel et al., 1992; Weigel et al., 1992). The phenotype of ap1 mutants indicates that AP1 activity is needed to identify this tissue as floral stem, suppressing its potential to develop ectopic floral and inflorescence meristems (Fig. 1B) and preventing its elongation.
The role of CAULIFLOWER
Neither AP1 nor LFY is expressed significantly in the early primordia of ap1-1 cal-1 double mutants (although both are in ap1-1 singles). This formally suggests that the expression of both genes is positively regulated, either directly or indirectly, by the CAL gene product.
A simple scheme for CAL function is that, in combination with AP1, it elevates the expression of LFY and AP1 in early flower primordia (Fig. 10). LFY expression is greatly reduced compared to wild type in the first primordia produced by the main inflorescence meristem in ap1-1 cal-1 plants (Fig. 7C,D). Even so LFY seems to be at a high enough level to suppress bract (cauline leaf) development although not enough to allow any floral characters to develop. This leads to the proliferation of multiple, undifferentiating inflorescence meristems. However, LFY activity seems to gradually accumulate in the higher order meristems of ap1-1 cal-1 plants (Fig. 7E,F), eventually reaching a level required for floral meristem specification. Flowers then develop and mature. We have proposed above that LFY expression is relatively low at low temperatures because of the reduced strength of factors inducing floral development. The phenotype of ap1-1 cal-1 plants raised at 15 ° C is consistent with this. The level of LFY is not sufficient any more to suppress bract formation consistently and, apparently, it never reaches the threshold at which flowers are specified, even in apical positions (Fig. 7G,H)
The expression of AP1 parallels that of LFY in ap1-1 cal-1 plants (Fig. 5), suggesting that AP1 is also positively regulated by CAL. These proposals also account for the fact that the ap1-1 cal-1 lfy-6 triple mutant plants are indistinguishable from ap1 lfy doubles. If both ap1 and lfy are inactive, it is irrelevant whether or not CAL is present to boost their activity.
It may be that the role of CAL is fully encompassed by AP1, since cal-1 reveals a phenotype only in an ap1 mutant background. The enhancement of the phenotype of weaker ap1 alleles by cal-1 such that they now resemble stronger ap1 alleles suggests that CAL acts truly redundantly with AP1 and that the strong ap1-1 cal-1 phenotype is the consequence of none of the AP1 targets being activated. Even though they act redundantly, CAL can substitute for only a proportion of AP1 functions. If it could replace them all, the ap1 CAL phenotype would be wild type.
Mutants with phenotypes closely similar to that of ap1-1 cal-1 have been seen in other species. Most familiar is that of Brassica oleracea var botrytis, the cultivated cauliflower of our meal table (Sadik, 1962). In the same family as Ara - bidopsis, the origin of its abnormal phenotype is apparently different. Its genetic basis depends upon one semi-dominant mutation rather than a combination of two mutations. The flowers that may break out of the inflorescence of cultivated cauliflowers are wild type and not ap1-like (Yarnell, 1956). The anantha mutant of tomato also results in the proliferation of undifferentiated inflorescence primordia, but in this case a single recessive gene is responsible (Helm, 1951). Thus it seems a range of different genetic alterations can independently lead to the threshold conferring floral identity not being reached.
Hyper-induction in terminal flower mutants
Several lines of evidence suggest that the tfl mutant phenotype arises because such plants are ‘over-induced’ to flower. Firstly, tfl plants are early flowering as compared to wild type (Shannon and Meeks-Wagner, 1991). Secondly, the number of normal flowers produced before the apical inflorescence meristem terminates in a floral meristem is influenced by environmental conditions (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992). The same non-inductive environmental conditions that enhance the phenotypes of ap1 and lfy mutants, short days and low temperatures, reduce the severity of the tfl phenotype. Thirdly, tfl-2 ap1-1 cal-1 triple mutants look just like tfl-2 ap1-1 doubles. That is, loss or reduction of CAL activity has no effect in tfl-2 ap1-1 plants. This suggests that factors responsible for floral induction are already at high enough levels in tfl ap1 mutant plants such that LFY expression reaches levels sufficient to determine floral meristem identity without the stimulus provided by CAL. Fourthly, in tfl-2 plants, both AP1 (Fig. 5G-J) and LFY (Weigel et al., 1992) are ectopically expressed in the transformed terminal meristems. That is, these genes are activated in inappropriate parts of the inflorescence meristem perhaps in response to the very high levels of floral induction factors in tfl plants. Finally, terminal flowers can be produced by the normally indeterminate inflorescence meristem in Sinapis alba, another member of the Brassicaceae, if plants are given extreme inductive growth conditions (Bernier, 1992).
Double mutant combinations of tfl and ap1 (Results; Shannon and Meeks-Wagner, 1991) and of tfl and lfy (D. W. and J. A., unpublished results) are apparently close to additive, suggesting that the TFL gene product does not interact directly with either of these main players. It may be that the normal TFL function is to inhibit one or several factors that promote floral induction (Koornneef et al., 1991) and that all the effects of tfl mutations are direct or indirect consequences of this release from inhibition. Of course, it is possible that TFL function may negatively regulate AP1 and LFY more directly in the apex of the inflorescence meristem. In the absence of TFL function, AP1 and LFY would now have the potential to become activated ectopically in these apices.
The early role of APETALA2
AP2 also appears to have a role in the specification of floral meristem identity in addition to its later role in the specification of floral organ identity (see below). This seems likely even though flowers of ap2 single mutants reveal few if any inflorescence-like properties (Bowman et al., 1989; 1991). The evidence comes from double mutants involving ap2 alleles and either ap1 or lfy mutations. In each case, the flowers have enhanced inflorescence properties suggesting that the AP2 product positively interacts with both AP1 and LFY (Fig. 10). For example, combination of the strong ap2-2 mutant with the strong ap1-1 mutant produces a larger number of flower-to-inflorescence conversions (second order inflorescences, Table 6). If the weak ap2-1 allele is involved with ap1-1, the flowers produced are much more branched than in either single mutant (Table 6; Irish and Sussex, 1990). [The fact that there is much less branching in flowers of the ap1-1 ap2-2 double (Table 6) may result from ectopic expression of AGAMOUS (see below).] ap2 mutants also interact with lfy mutants in a parallel manner (Huala and Sussex, 1992). Similar intensified changes have been observed by us in weak ap2-1 lfy-5 doubles (unpublished).
These effects can be accounted for if, in the absence of either AP1 or LFY activity, AP2 activity is now required for the floral identity threshold to be reached in primordia developing on the flanks of the inflorescence apex (Fig. 10).
Later roles of AP1 in floral organ development
Recent models of Arabidopsis flower development propose that the identities of the four whorls of floral organs are specified by the products of at least four floral homeotic genes APETALA2, APETALA3, PISTILLATA and AGAMOUS (Fig. 10; Bowman et al., 1991). The genes are suggested to function alone and in combination in three overlapping fields. Field A (whorls 1 and 2) is represented by AP2, field B (whorls 2 and 3) is represented by AP3 and PI, and field C (whorls 3 and 4) is represented by AG. In addition, it is proposed that the A and C functions are mutually antagonistic (Bowman et al., 1991; Fig. 10).
Mandel et al. (1992) found that, from late stage 3, AP1 is expressed in regions of flower primordia corresponding to field A and proposed that AP1 is involved in determining development of the outer two whorls. Observations on our allelic series of ap1 mutants support this, as disruption is predominantly confined to sepal and petal development. It seems likely that AP1 promotes or acts as a partner in the activity of AP2, the other known field A gene, in specifying the identity of outer whorl organs (Fig. 10). This is because double mutant combinations involving weak alleles, such as ap1-5 ap2-1, exhibit a much enhanced change in organ identity toward that of strong ap2 mutant alleles (Fig. 8E).
Mandel et al. (1992) also suggested that AP1 expression may be negatively regulated by AG in whorls 3 and 4 (Fig. 10). This was proposed because AP1 is not normally expressed in the inner whorls once AG comes on, but it is if ag is in mutant form. Further, this negative regulation of AP1 by AG may extend to the outer two whorls in ap2 mutants where AG is ectopically expressed in whorls one and two (Drews et al., 1991). Consistent with this, strong ap2 alleles are epistatic to strong and weak ap1 alleles with respect to disruptions to the identity of floral organs (Fig. 8B).
Proposed interactions between AP1, AP2 and AG from stage 3 can be summarized thus: AP2 negatively regulates AG in field A, AG negatively regulates AP1 in field C, and both AP1 and AP2 are necessary for A function (Fig. 10). Mutant phenotypes of the outer whorl organs can be accounted for as follows. In ap2 mutants, AG comes on and inactivates AP1 in field A leading to carpelloid first whorl organs. By contrast, in ap1 mutants expression of AP2 and AG is largely normal but A function is disrupted resulting in bract-like organs developing in place of sepals. In ag mutants AP1 and AP2 act as usual in field A and the outer whorl organs are normal sepals. However, these genes are also ectopically active in field C and stamen to petal conversion is one result. AP1 thus appears to be, along with AP2, a component of the A function that regulates organ identity in whorls one and two. This follows both from its action in specifying organ identity in these whorls, and from the regulatory interaction between AP1 and the C function gene AG.
One effect of ap1 mutations may be an ectopic expansion of field C activity into areas of the flower primordium where field A function is now compromised. This could account for why stamens sometimes arise in positions normally occupied by second whorl organs and why second whorl organs themselves do not develop. The latter could be related to ectopic AG expression and the role that AG normally plays in suppressing growth in the center of the determinant flower primordium (Bowman et al., 1991). Consistent with this, petals may reappear in second whorl positions of ap1 flowers when AG is also inactivated (as in ap1 ag doubles (Fig. 8K,L)). Finally, petal-stamen mosaic organs seen in the second whorl of ap1 mutants (Fig. 3G,H) could develop from primordia that straddle the A/C boundary.
The phenotype of ap1 ap2 ag triple mutants
An observation that needs to be accounted for is the inability of ‘floral’ meristems in ap1-1 ap2-2 ag-1 triple mutants to develop mature organs of any type. Indeed, such meristems seem to revert frequently to inflorescence-like meristems. This is perhaps unexpected given that the two earlier acting genes which control flower meristem identity, AP1 and AP2, do not lead to such a phenotype when in double mutant combination. The additional inactivation of the later acting AG seems to be responsible.
It seems that, in the triple ap1 ap2 ag mutant, the ectopic flower that normally develops inside an ag flower frequently loses its floral identity and reverts back to inflorescence-like growth. When a primordium first develops on the flank of the triply mutant inflorescence meristem, sufficient activities of those factors that specify floral meristem identity are apparently present to allow it take on floral identity and to develop first whorl primordia (stage 3). However, perhaps these activities fall away in the central cells of the flower primordium and they revert to inflorescence-like growth. A similar explanation might apply in ap1-1 ag-1 double mutants where we found that many flowers also revert to inflorescence growth patterns. The fact that loss of AG seems to be responsible for the dramatic reversionary phenotype of the triple mutant suggests that AG plays a role in maintaining floral identity. Additional evidence for this role is that in ag single mutants secondary flowers occasionally arise in the axils of the sepals and such flowers are more common in ag ap2 double mutants (unpublished observations).
AP1 function in Arabidopsis resembles SQUA function in Antirrhinum
AP1 has a cognate homolog in Antirrhinum majus, the SQUAMOSA gene (Huijser et al., 1992). AP1 and SQUA share 68% identity in their predicted amino acid sequence and their patterns of expression in flower development are closely similar (Mandel et al., 1992). The corresponding mutant phenotypes also share some features. There is an incomplete conversion of flower meristems into inflorescence meristems in squa mutants, even in null mutants (Huijser et al., 1992). The range of floral organ types in squa mutant plants is also somewhat similar to that in ap1 mutants (Huijser et al., 1992).
Flower development in squamosa mutants is also environmentally sensitive so it is difficult to compare directly the relative severity of squa and ap1 mutational effects. However, the phenotypes reported by Huijser et al. (1992) reflect more frequent and more complete conversions of floral meristems to inflorescence meristems. Conversely, if flowers arise in squa plants, they usually seem to be less severely affected than in ap1. These differences indicate that AP1 plays a less important role than SQUA early in establishing floral meristem identity, but a more important role later in the developing flower.
It seems likely that a core of conserved regulatory genes underpin the control of floral development across flowering plants. Even though such genes seem to have retained their basic roles, modification of their expression patterns and their interactions may account for at least some of the great diversity in inflorescence and flower structure that has arisen since flowering plants first appeared.
Special thanks go to Marty Yanofsky for making the AP1 probe available prior to publication and for constructive comments on the manuscript. We also thank Tom Jack, Gerd Bossinger, Megan Griffith and Alan Neale for helpful discussion of the manuscript. We are indebted to Maarten Koornneef and Luca Comai for supplying genetic material. Gunta Jaudzems provided excellent microscopy facilities. J. L. B. was supported by a Short Term Fellowship from the Human Frontiers Science Program and an Australian Research Council Postdoctoral Research Fellowship. J. A. held a Monash Postgraduate Scholarship. D. W. was supported by a EMBO Long Term Fellowship and a Senior Fellowship from the American Cancer Society, California Division. This work was supported by an Australian Research Council Grant A19131181 to D. R. S. and a US Department of Energy, Division of Energy Biosciences Grant DE-FG03-88ER13873 and a US National Science Foundation Grant MCB-9204839 to E. M. M.
Note added in proof
While this manuscript was under review, a related study appeared reporting mutant combinations of apetala1 with terminal flower, leafy and apetala2 mutants (Shannon, S. and Meeks-Wagner, D. R. (1993). Genetic interactions that regulate inflorescence development in Arabidopsis. Plant Cell5, 639-655.).