The Arabidopsis FILAMENTOUS FLOWER gene is required for flower formation

During early flower development in Arabidopsis, a primordium forms on the flanks of the apical inflorescence meristem and adopts floral fate. As the floral primordium enlarges, cells in different positions within the primordium adopt different fates. The cells at the floral primordium’s distal end give rise to a flower, while those at the proximal end differentiate into a pedicel (Fig. 1A). During development of the distal end into a flower, floral organ primordia form at specific positions, adopt specific fates, and develop into specific organ types (Smyth et al., 1990).

Much progress has been made in identifying genes involved in several aspects of this process. At least five groups of genes have been identified: floral meristem identity genes, floral meristem size genes, floral organ pattern genes, cadastral genes and floral organ identity genes. The floral meristem identity genes APETALA1 (AP1), APETALA2 (AP2), CAULIFLOWER (CAL), LEAFY (LFY), and UNUSUAL FLORAL ORGANS (UFO) act early to specify the identity of the young floral primordium (Mandel et al., 1992; Weigel et al., 1992; Bowman et al., 1993; Jofuku et al., 1994; Kempin et al., 1995; Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995). The floral meristem size genes CLAVATA1 (CLV1), CLAVATA2 (CLV2), CLAVATA3 (CLV3), and WIGGUM (WIG) regulate the size of the floral meristem and also affect organ number (Clark et al., 1993, 1995; Kayes and Clark, 1998; Running et al., 1998). The floral organ pattern gene PERIANTHIA (PAN) acts to establish floral organ primordia in specific numbers and positions (Running and Meyerowitz, 1996). The cadastral genes LEUNIG (LUG) and SUPERMAN (SUP) act to define the expression patterns of the floral organ identity genes (Schultz et al., 1991; Bowman et al., 1992; Liu and Meyerowitz, 1995; Sakai et al., 1995). Finally, the floral organ identity (homeotic) genes APETALA1 (AP1), AP2, AP3, PISTILLATA (PI), and AGAMOUS (AG) specify the identities of the floral organ primordia (Bowman et al., 1989; Yanofsky et al., 1990; Jack et al., 1992; Mandel et al., 1992; Bowman et al., 1993; Goto and Meyerowitz, 1994; Jofuku et al., 1994).

Relatively little is known about the steps between specification of floral primordium identity and initiation of flower formation. Because the proximal and distal ends of the floral primordium give rise to different structures (pedicel and flower, respectively), one of these steps is likely to include subdividing the floral primordium into a proximal domain, which gives rise to the pedicel, and a distal domain, which develops into the flower. The simplest model for this developmental step is the establishment of a proximal-distal (P-D) axis within the floral primordium as depicted in Fig. 1A.

One approach to investigate this step is to search for mutants with defects in floral primordium P-D patterning. A failure to establish a P-D axis within the floral primordium would be expected to affect subdivision of the floral primordium into the proximal pedicel-forming domain and the distal flower-forming domain. In theory, two classes of P-D patterning mutants could exist. One mutant class could fail to form the proximal structure (pedicel), which would result in flowers forming directly on the peduncle (sessile flowers; Fig. 1C). Alternatively, mutants could fail to form the distal structure (flower), which would result in the formation of flowerless pedicels on the peduncle (Fig. 1B). To the naked eye, flowerless pedicels would appear as filament-like structures. Although Arabidopsis mutants with sessile flowers have not been reported, several mutants containing filament-like structures along the peduncle have been described including ufo (Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995), revoluta (rev) (Talbert et al., 1995), and filamentous flower (fil) (Komaki et al., 1988). Although descriptions of the ufo (Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995), rev (Talbert et al., 1995) and fil (Komaki et al., 1988; Okada and Shimura, 1994a,b; Sawa et al., 1999) mutants have been reported, the possible roles of the UFO, REV and FIL genes in P-D patterning has not been investigated.

We describe the results of our morphological and genetic analysis of the fil mutant and report the isolation of six new fil alleles. We show that the fil mutation causes some floral primordia to develop into flowerless pedicels, and that the rev mutation dramatically enhances this phenotype. These data suggest that FIL and REV are required for an early step of flower formation, perhaps for establishment or perception of P-D positional information within the floral primordium. We also show that the fil mutation causes organ number, organ position, organ identity, and organ development defects during flower development; and that FIL acts as a cadastral gene in controlling the expression patterns of AG, AP3 and SUP.

Plant growth was as previously described (Christensen et al., 1997). All strains except those containing the ap3-1 and ap2-1 mutations were grown at 20°C. The ap3-1 and fil-5 ap3-1 mutants were grown at 29°C, and the ap2-1 and fil-5 ap2-1 mutants were grown at 27°C.

We identified the fil-3, fil-4 and fil-5 mutants in a screen of Landsberg erecta (Ler) plants that were mutagenized with ethyl methanesulfonate (EMS). Mutagenesis was carried out as previously described (Liu and Meyerowitz, 1995) except that 0.3% ethyl methanesulfonate was used.

fIn crosses of the fil-3, fil-4 and fil-5 mutants with wild type, all F₁ progeny were wild type and 25% of the F₂ generation was mutant, indicating that the mutant phenotypes were caused by single recessive mutations. To determine the genetic map position of the fil locus, we crossed fil-3 in the Ler genetic background with wild-type plants in the Columbia genetic background. Segregation data from 500 F₂ plants homozygous for the fil-3 mutation showed that the fil locus is located on the lower part of chromosome 2, between markers TAMU8M12 and GBF3 (1.5 cM from each marker).

We scored fil-5 flowers at different positions along the inflorescence and referred to these as early, mid and late flowers. Early flowers were defined as the first 10 flowers produced by the inflorescence meristem. Mid flowers were flowers 21-30 above the reduced flower cluster (thereby avoiding the severe floral phenotype near that region). Late flowers were flowers 51-60 above the reduced flower cluster.

Tissue was fixed in 3% glutaraldehyde in 6.25 mM sodium phosphate buffer (pH 7.0) at 4°C for at least 24 hours. The tissue was then rinsed in water and further fixed in 1% osmium tetroxide for at least 48 hours. Specimens were then dehydrated through an ethanol series, critical point dried with carbon dioxide, mounted on stubs, and shadowed with gold and palladium (6:4). A Hitachi S450 was used. Images were documented using Polaroid P/N 55 film.

Seed was collected from individual F₂ plants exhibiting the fil phenotype and, when possible, from those exhibiting the phenotype of the other mutant. F₃ seed was planted to identify seed stocks segregating the double mutant. Phenotypic analysis was carried out with these F₃ plants. Whenever possible, crosses were carried out to verify double mutant genotype. This was not possible for sterile double mutants (i.e., fil lfy, fil ufo, fil rev, fil ag, and fil lug). Because the fil-5 floral phenotype becomes stronger at more acropetal positions along the peduncle, for consistency, we compared only flowers 1-10 in all single and double mutants.

Nonradioactive in situ hybridization experiments were carried out as previously described (Klucher et al., 1996) utilizing digoxigenin-labeled RNA probes. A detailed description of this protocol can be found at http://genome-www.stanford.edu/Arabidopsis/cshl-course. Antisense probes were made from pCIT565 for AG (Yanofsky et al., 1990), pDW124 for LFY (Weigel et al., 1992), and pD793 for AP3 (Jack et al., 1992). Histochemical analysis of GUS expression was carried out as previously described (Hill et al., 1998). The AP3::GUS strain used was 890-7 (Jack et al., 1994).

For in situ hybridization and GUS-staining images, photographic slides were made using a Zeiss Stemi SV11 camera system and Kodak Elitechrome 200 film. For scanning electron microscope images, photographic prints were made using Polaroid P/N 55 film. Photographic images were digitized using an Agfa DuoScan scanner. Image brightness and contrast were adjusted, and background was darkened using Adobe Photoshop 5.0. Figures were assembled using Deneba Canvas 5.0.3.

To identify genes required for P-D patterning within the floral primordium, a screen for mutants in which floral primordia give rise to flowerless pedicels was carried out (Fig. 1B). From this screen, three new fil alleles were identified. Also three additional fil alleles were obtained from other sources and the six new alleles have been named fil-2–fil-7 (Table 1). A preliminary phenotypic analysis of the fil-1–fil-7 mutants was carried out and they were then divided into three phenotypic categories as listed in Table 1. The fil-5 mutant was selected for further analysis because it exhibited the strongest overall phenotype. The fil-5 phenotype is summarized in Fig. 2.

In wild type, peduncles have only flowers above the coflorescences (Fig. 2A). By contrast, fil-5 peduncles contained numerous filament-like structures (Fig. 3A). Several lines of evidence suggest that these structures are flowerless pedicels. First, most (>80%) of the epidermis of these structures consisted of elongate, irregularly shaped cells with interspersed stomata (Fig. 3D), which is an epidermal morphology similar to that of wild-type pedicels (Fig. 3C) but not to that of any other plant organ (including peduncle, leaf and all floral organs; data not shown). Second, the distalmost 10-20% of filament-like structures was composed of sepal tissue (Fig. 3E, F), indicating partial floral identity of the apices of these structures. Third, filament-like structures arose on the flanks of the inflorescence meristem in the positions and patterns typical of wild-type flowers (Fig. 3I; discussed below). Fourth, both the cuticular wax pattern and the internal cellular anatomy of filament-like structures was similar to that of wild-type pedicels (data not shown). Taken together these data suggest very strongly that the filament-like structures found along the peduncle of fil-5 mutants are flowerless pedicels. These observations indicate that fil floral primordia have reduced capacity to form flowers.

In addition to flowerless pedicels, fil-5 peduncles also contained another unusual structure (Fig. 3B) that consisted of a pedicel (Fig. 3C,G) with an appendage at its distal end. We refer to the appendages as ‘curled sepals’ because they resembled wild-type sepals (Fig. 3E) but were curled in at the margins (Fig. 3B,H). Developmental analysis showed that the curled sepals arose in the position of the first whorl abaxial medial organ (Fig. 3I,J), which in wild type, is the first organ to form during flower development (Smyth et al., 1990). These data suggest that stalked curled sepals arise when flower development is interrupted just after initiation of the first organ (i.e., the abaxial medial organ of whorl 1). These observations support the conclusion that fil floral primordia have reduced capacity to form flowers.

The frequency of flowerless pedicels and stalked curled sepals varied at different positions along the peduncle. Beginning at about flower 11, fil-5 plants produced a series (20±8 structures; n=30 plants) of flowerless pedicels and stalked curled sepals. We refer to this series as the ‘reduced flower cluster’. Within the reduced flower cluster, flowerless pedicels and stalked curled sepals occurred at approximately equal frequency. Within the reduced flower cluster, floral primordia that gave rise to flowerless pedicels and stalked curled sepals were produced in normal phyllotaxy (Fig. 3I). Above and below the reduced flower cluster, flowerless pedicels and stalked curled sepals were infrequent (20%).

When fil-5 floral primordia gave rise to flowers, the associated pedicels were longer than those of wild type. Pedicel length was 3.5 ± 0.8 mm in wild type (n=233 pedicels) and 7.8 ± 2.1 mm in pollinated fil-5 flowers (n=174 pedicels). These data indicate that in fil-5 mutants, an increased proportion of the floral primordium is recruited to form a pedicel, possibly due to reduced capacity to form flowers.

When fil-5 mutants produced flowers, the associated pedicels frequently (51%) had organs along their abaxial surfaces. Below the reduced flower cluster (flowers 1-10), pedicel organs generally (87%) were filamentous structures comprised mostly of sepal tissue (Fig. 3K and data not shown), and above the reduced flower cluster, pedicel organs generally (>88%) were curled sepals (Fig. 3L). Developmental analysis showed that pedicel organs arose in the position of the medial abaxial first whorl organ (Fig. 3M). These data suggest that pedicel organs arise when the first floral organ (i.e., the medial abaxial first whorl organ) is initiated and then, subsequently, the decision to make the remainder of the flower is delayed.

In summary, fil-5 floral primordia give rise to a variety of structures including flowerless pedicels and stalked curled sepals, as well as flowers with pedicel organs and long pedicels. The existence of these structures point to a reduced capacity of fil-5 floral primordia to form flowers. These observations suggest that FIL acts to promote flower formation within developing floral primordia, possibly by providing or perceiving distal positional information within the floral primordium.

To determine whether the fil-5 mutation also affects later steps of flower development, we analyzed floral morphology in this mutant. SEM micrographs of fil-5 flowers are shown in Fig. 4B-D.

In mature fil-5 flowers, floral organs were present in incorrect numbers and positions (Table 2; Fig. 4A-D). These defects became stronger at more acropetal positions, hindering attempts to assign organs to specific whorls in mid and late flowers (Tables 2 and 3; see Materials and Methods for definition of early, mid and late flowers).

To determine the developmental basis for altered floral organ pattern, we analyzed developing fil-5 flowers. In wild type, first whorl organs are positioned equidistant to one another and at specific positions relative to the inflorescence meristem (Fig. 4E,G). By contrast, in fil-5 flowers, first whorl organ primordia arose in incorrect numbers and at irregular positions relative to each other and to the inflorescence meristem (Fig. 4F,H). In addition, first whorl organs in some late flowers arose in a spiral pattern data not shown). Organ initiation in the second and third whorls also was irregular in number and position (data not shown). Reduced organ number in whorls 2 and 3 did not result from reduced floral meristem size because at the time these organs initiate (stage 5), fil-5 floral meristems had a size and shape similar to or larger than that of wild type (Fig. 4G,H).

Defects in organ identity (Table 3) and/or development were exhibited by all floral whorls in fil-5 flowers. In whorl 1, normal sepals were absent and, most commonly, were replaced by curled sepals (Fig. 3H) containing petal tissue. Generally, the petal tissue was confined to the margins (Fig. 4I,K); however, sometimes it occupied a larger proportion of the organ. Rarely, carpelloid curled sepals or sepalloid filamentous organs (Fig. 4L) were present (Table 3). First whorl organs generally were misshapen and/or twisted, and often had narrowed or stalked basal regions (Fig. 4B-D). In whorl 2, petals generally were present; however, these petals usually were misshapen and/or twisted (Fig. 4M,N). Occasionally, petal-sepal mosaic organs were present, and rarely, filamentous organs or staminoid petals were found (Table 3).

In whorl 3, normal stamens (Fig. 4P) were rare (<1%). Most often, filaments lacking anthers were present (Fig. 4B-D). Filaments resembled wild-type stamen filaments at their bases (data not shown) and generally had elongate anther-like cells at their tips (Fig. 4J,O). When anthers were present, they generally (>90%) were deformed and sterile (Fig. 4Q). Rarely, anthers were replaced by anther-sepal, anther-petal, or anther-carpel mosaic organs (Table 3).

In whorl 4, organ identity defects were not observed (Table In summary, the fil-5 mutation affects many aspects of flower patterning including establishment of organ position, organ number, and organ identity. The broad effects of the fil-5 3). However, carpel development was abnormal in one or more ways. In all fil-5 carpels, the style was elongated relative to wild type (Fig. 4R,S). The ovary of fil-5 carpels generally had a lumpy appearance due to having different carpel number at the base and tip (Fig. 4T). In addition, carpels occasionally failed to fuse with each other during pistil development (Fig. 4U). mutation suggests that FIL plays a central role in flower patterning such as providing or perceiving positional information within the developing flower.

In addition to the above, fil-5 mutants exhibited several other defects. First, in contrast to wild type, the first flower often (59%; n=140 plants) was subtended by a cauline leaf (as compared to 0% in wild type; n=100 plants).

Second, fil-5 mutants flowered earlier than wild type. fil-5 plants produced a mean of 7.9 vegetative leaves prior to flowering, and the opening of the first flower occurred at an average of 28.1 days following germination (n=33 plants). By contrast, wild-type plants produced a mean of 9.0 vegetative leaves prior to flowering, and the first flower opened at an average of 31.9 days following germination (n=16 plants).

Third, the entire inflorescence gradually became more carpelloid at more acropetal positions along the peduncle. For example, both stalked curled sepals on the peduncle (Fig. 3B) and curled sepals on the pedicel (Fig. 3L) contained well-developed stigma and ovules in the region occupied by late flowers (data not shown).

Fourth, fil-5 plants terminated inflorescence development with a mass of carpelloid tissue (Fig. 4X). However, in hand-pollinated fil-5 plants, inflorescence development terminated as in wild type (Fig. 4W), indicating that this defect was not a direct consequence of the fil-5 mutation.

Fifth, the arrangement of flowers along the peduncle was altered. In wild type, flowers are arranged in a spiral phyllotaxy and are separated from one another by regular internode lengths (Fig. 5A) (Smyth et al., 1990). In fil-5 mutants, by contrast, flowers were not arranged in a spiral phyllotaxy, and internode length was irregular with a proliferation of both very short (<1.0 mm) and very long (>10 mm) internodes (Fig. 5B). Both floral angle (the angle between two adjacent flowers) and internode length occurred in no apparent pattern (Fig. 5B). This defect was weak at basipetal positions and became increasingly stronger at more acropetal positions along the peduncle.

The irregular floral arrangement suggested that the fil-5 mutation may affect the production of floral primordia by the inflorescence meristem. To investigate this, we analyzed inflorescence meristem structure in fil-5 mutants. In wild type, each flower primordium arises on the flanks of the inflorescence meristem with an angle to the prior primordium of between 130° and 150° (Fig. 4E) (Smyth et al., 1990). In fil-5 inflorescence meristems shortly after floral induction, the pattern of floral primordium formation was essentially normal (Figs 3I, 4F). However, as fil-5 inflorescence meristems aged, the ordered pattern of floral primordia formation was gradually lost, and, eventually, a regular pattern was not apparent (Fig. 5V). We also compared the pattern of leaf primordium formation by the vegetative shoot apical meristem in fil-5 mutants to that in wild type and found no difference (data not shown).

In summary, fil-5 mutants exhibited a variety of defects in inflorescence organization. These observations suggest that FIL plays a role in patterning along the peduncle.

In fil-5 mutants, few floral primordia give rise to flowerless pedicels (Fig. 2) and those that do produce flowerless pedicels with partial floral character (Fig. 3F). This weak effect could be due to the existence of genes acting redundantly with FIL. Likely candidates are REV and UFO, which, when mutated, produce filament-like structures along the peduncle (Levin and Meyerowitz, 1995; Talbert et al., 1995; Wilkinson and Haughn, 1995). To test this, we analyzed fil rev and fil ufo double mutants. In contrast to the single mutants, all or most floral primordia in fil-5 rev-tj72 and fil-5 ufo-2 double mutants gave rise to filament-like structures (Fig. 6A). Filament-like structures in these two double mutants had different identities. In fil-5 ufo-2 mutants, filament-like structures were leaf-like in epidermal morphology and had stellate trichomes (Fig. 6D,E), suggesting that FIL and UFO interact to confer floral fate upon the floral primordium. By contrast, in fil-5 rev-tj72 mutants, all filament-like structures were composed entirely of pedicel tissue (Figs 3C, 6F,G), indicating that in this double mutant, all floral primordia gave rise to flowerless pedicels. Taken together, these data suggest that FIL and REV interact to promote flower formation within developing floral primordia.

The fil ufo phenotype discussed above suggests that FIL promotes floral fate within the floral primordium. To test this further, we analyzed strains doubly mutant for fil and the meristem identity mutations lfy and ap1. fil-5 lfy-6 mutants exhibited a phenotype similar to that of fil-5 ufo-2 mutants (Fig. 6A); most floral primordia gave rise to filament-like structures that were leaf-like in epidermal morphology (Fig. 6B,C). fil-5 ap1-1 flowers were inflorescence-like; floral organs were not organized in a whorled pattern (Fig. 7M), and occasionally (5%), flowers were converted into elongate shoots (Fig. 7N). Taken together, these data suggest that FIL interacts with AP1, LFY and UFO to promote floral fate within the floral primordium.

The inflorescence defects in fil mutants suggests that FIL could function upstream of the meristem identity genes. To test this, we characterized the expression of the meristem identity gene LFY in fil mutants. In wild type, LFY RNA first becomes detectable in the anlagen of the floral primordia and is uniformly distributed in developing floral primordia (stages 1 and 2) until initiation of sepal primordia (Fig. 5C). In developing fil flowers, early expression of LFY is identical to that in wild type (Fig. 5D). These data suggest that FIL acts in a pathway that is either downstream or independent of the LFY meristem identity gene.

Carpelloidy in whorl 1 and staminoidy in whorl 2 of fil-5 flowers (Table 3) suggests ectopic AG activity in those whorls (Bowman et al., 1991). To test this, we analyzed AG expression in fil flowers. In wild type, AG RNA is first detected shortly after initiation of sepal primordia (stage 3) in the cells that will give rise to whorls 3 and 4, and expression continues in these two whorls throughout development (Fig. 5E,G) (Drews et al., 1991). In fil flowers, the pattern of AG expression was, in general, similar to that in wild type (Fig. 5F). In contrast to wild type, however, AG RNA often was detected in whorls 1 and 2 (Fig. 5F), and in the peduncle (Fig. 5H).

To determine whether carpelloidy and staminoidy in whorls 1 and 2, respectively, result from ectopic AG activity in those whorls, we analyzed fil ag double mutants. In ag-1 flowers, whorls 1 and 2 are normal, whorl 3 organs are petals (six organs), and the cells of whorl 4 give rise to another ag flower (Fig. 7A) (Bowman et al., 1989). In fil-5 ag-1 flowers (Fig. 7B), whorl 1 was similar to that in fil-5 flowers with regard to both identity and number, except that carpelloidy never occurred. In whorls 2 and 3, organs were petals, sepalloid petals, and filamentous structures that were never staminoid. Organ number in whorl 2 was increased relative to that in fil-5 flowers (≥4 organs in fil-5 ag-1). The cells of whorl 4 gave rise to another flower that was composed entirely of curled sepals, petals and curled sepal-petal mosaic organs. These data suggest that ectopic AG activity in whorls 1 and 2 of fil flowers causes carpelloidy in whorl 1, and staminoidy and reduced organ number in whorl 2.

In addition to the above, fil-5 ag-1 mutants exhibited several other features. First, flowers were extremely elongate and occasionally (∼10%) produced secondary flowers (Fig. 7C,D). Second, floral phenotype in early, mid, and late flowers was similar (data not shown), suggesting that unregulated AG activity is responsible for the increased severity of fil-5 floral defects at more acropetal positional along the peduncle (Fig. 4B-D). Third, the peduncle phenotype of fil-5 and fil-5 ag-1 mutants was similar, suggesting that ectopic expression of AG in the peduncle (Fig. 5H) had no phenotypic consequences.

Ectopic AG expression in whorls 1 and 2 of fil flowers suggests that FIL may interact with the class A genes (AP2, LUG, and AP1), which suppress AG expression in whorls 1 and 2 (Bowman et al., 1991; Liu and Meyerowitz, 1995). To test this, we analyzed fil ap2, fil lug, and fil ap1 flowers.

In fil-5 ap2-2 mutants, flowers typically consisted entirely of 2-3 unfused carpels (Fig. 7F). Because whorl 1 in ap2-2 mutants is extremely carpelloid (Fig. 7E) (Bowman et al., 1991), it was difficult to assess whether addition of the fil-5 mutation increased the carpelloidy in that whorl. We, therefore, analyzed lines doubly mutant for fil-5 and the weaker ap2-1 allele.

In ap2-1 flowers, whorl 1 organs are cauline leaves (four organs); whorl 2 organs are petals, stamens, or stamen-petal mosaic organs (reduced organ number); and whorls 3 and 4 are normal (Fig. 7G) (Bowman et al., 1989). In fil-5 ap2-1 flowers, first whorl organs were carpels (organ number similar to that of fil-5); second and third whorl organs were carpelloid, staminoid, and/or filamentous structures (0-2 organs); and whorl 4 consisted of a gynoecium that contained extra carpels and/or was extremely distorted (Fig. 7H). This is in dramatic contrast to fil-5 (Fig. 4B-D) and ap2-1 (Fig. 7G) flowers, which rarely exhibited carpelloidy in the first whorl. These data suggest that FIL and AP2 interact to repress AG activity in whorls 1 and 2.

In lug-1 flowers, whorl 1 organs are sepals or petaloid/staminoid sepals (four organs), whorl 2 organs are petals or staminoid petals (reduced organ number), whorl 3 organs are stamens (reduced organ number), and whorl 4 consists of improperly fused carpels (Fig. 7I) (Liu and Meyerowitz, 1995). In fil-5 lug-1 mutants, few flowers were produced. Flowers 1-3 were the most normal and consisted of a misshapen central gynoecium, and 3-5 outer organs that were extremely narrow and occasionally (approx. 10%) carpelloid or staminoid (Fig. 7J). At more acropetal positions, flowers became more carpelloid and reduced; by flower 6, flowers were composed entirely of carpel tissue, and after approx. 10 flowers were produced, the inflorescence terminated in a filamentous, carpelloid mass (Fig. 7K). Thus, addition of the lug mutation caused a dramatic increase in carpelloidy, suggesting that FIL and LUG interact to repress AG activity in whorls 1 and 2.

In ap1-1 flowers, whorl 1 organs are bracts (four organs), whorl 2 organs are petals (reduced organ number), and whorls 3 and 4 are normal (Irish and Sussex, 1990; Bowman et al., 1993) (Fig. 7L). fil-5 ap1-1 flowers typically consisted of a gynoecium (similar to that in fil-5 flowers) with 4-5 outer organs (Fig. 7M). Outer organs were either bracts or modified bracts that were misshapen, narrowed, stalked, staminoid, or carpelloid (Fig. 7M). Thus, carpelloidy in whorl 1 was not dramatically increased in the double mutant, suggesting that FIL does not interact with AP1 to repress AG activity in whorls 1 and 2.

In fil-5 flowers, whorl 1 petalloidy (Table 3; Fig. 4K), whorl 2 sepalloidy (Table 3), and the abnormal stamens in whorl 3 (Table 3; Fig. 4O,Q) suggest ectopic class B gene (AP3 and PI) activity in whorl 1, and reduced class B gene activity in whorls 2 and 3 (Bowman et al., 1991). To test this, we analyzed AP3 expression in fil mutants. In wild type, AP3 RNA is first detected shortly after initiation of sepal primordia (stage 3) in

the cells that give rise to whorls 2 and 3 (Fig. 5I), and late in development (stage 12), strongest expression is in the anther (Fig. 5K) (Jack et al., 1992). In fil flowers during organ initiation (stages 3-6), AP3 expression, in general, was not detected in whorls 2 and 3 (Fig. 5J). Late in development, AP3 expression in the anther was either absent or patchy (Fig. 5L). In whorl 1 of fil flowers, expression was consistently observed in the cells along the margins of these organs (Fig. 5J). These data indicate that in fil flowers, AP3 expression is absent or reduced in whorls 2 and 3 and ectopically activated in whorl 1.

To determine whether petalloidy in whorl 1 results from ectopic PI and AP3 activity in that whorl, we analyzed fil-5 pi-1 and fil-5 ap3-1 flowers. In pi-1 flowers, whorls 1 and 2 each contain four sepals, and the cells of whorls 3 and 4 are incorporated into a large and irregularly-shaped gynoecium (Fig. 7O) (Bowman et al., 1989; Hill and Lord, 1989; Bowman et al., 1991). In fil-5 pi-1 flowers, whorl 1 was similar to that in fil-5 flowers with regard to organ number and identity except that organs were not petaloid and were not twisted, misshapen, or stalked (Fig. 7P,Q). In whorls 2-4, the phenotype of fil-5 pi-

1 flowers was additive. fil-5 ap3-1 flowers were similar to fil-5 pi-1 flowers (data not shown). These data suggest that the petalloidy and twisted/stalked appearance of whorl 1 organs in fil-5 flowers is due to ectopic class B activity in this whorl.

SUP is a negative regulator of class B activity in whorl 4 (Schultz et al., 1991; Bowman et al., 1992). Thus, reduced class B activity in whorls 2 and 3 of fil-5 flowers could be caused by ectopic SUP activity in those whorls. To test this, we analyzed fil-5 sup-1 flowers. In sup-1 flowers, whorls 1-3 are normal, but the fourth whorl is partially converted to stamens (Fig. 7R) (Schultz et al., 1991; Bowman et al., 1992). In fil-5 sup-1 flowers, whorls 1 and 2 were similar to fil-5 flowers (Fig. 7S), suggesting that SUP activity is absent in those whorls in fil-5 flowers. The cells at the center of the floral meristem produced a distorted gynoecium similar in morphology to that of sup-1 single mutants (Fig. 7S). The cells between whorl 2 and the central gynoecium produced a large number of stamens (12.9±2.6 stamens per flower), which, in contrast to fil-5 stamens (Fig. 4Q), had essentially normal morphology (Figs 4P, 7T). Thus, addition of the sup-1 mutation suppressed the third-whorl effect of the fil-5 mutation. These data suggest that in whorl 3 of fil-5 flowers, the SUP gene becomes active and reduces class B activity, resulting in loss of normal stamen development.

Based on our analysis, the FIL gene appears to perform three functions during flower development: (1) Establishment of floral meristem identity, (2) promotion of flower formation within the floral primordium, and (3) establishment of homeotic gene expression patterns within the developing flower.

Several lines of evidence suggest that FIL promotes floral fate within the floral primordium. First, in fil flowers, organs often are not organized in a whorled pattern (Fig. 4B-D). Second, the first flower of fil mutants often is subtended by a cauline leaf, suggesting partial conversion of the first flower to an inflorescence shoot. This feature is similar to the meristem identity mutants lfy (Weigel et al., 1992) and ufo (Levin and Meyerowitz, 1995; Wilkinson and Haughn, 1995). Third, combination of fil with ap1, a meristem identity mutation (Weigel et al., 1992), causes a dramatic enhancement of the inflorescence phenotype (Fig. 7N). Fourth, fil lfy and fil ufo mutants produce leaf-like structures (Fig. 6C,E), suggesting a partial conversion of flowers to cauline leaves or bracts. Taken together, these data suggest that FIL may interact with AP1, LFY, and UFO to promote floral identity within the young floral primordium.

The developing floral primordium gives rise to different structures at its proximal and distal ends (pedicel and flower, respectively; Fig. 1A). Thus, to differentiate properly, the cells of the floral primordium must know where they lie within the P-D axis. The mechanism whereby the floral primordium cells acquire P-D positional information is not known. One possibility is that the floral primordium contains a morphogen gradient along the P-D axis. A second possibility is that position is specified by cell age. Both mechanisms have been proposed to account for patterning during vertebrate limb bud development (Gilbert, 1997).

Mutations in genes required to provide or perceive distal positional information may produce flowerless pedicels (Fig. 1B). fil floral primordia occasionally develop into such structures (Fig. 3A). Although this defect is weak in fil single mutants (Fig. 2), it is dramatically enhanced in fil rev double mutants; every floral primordium gives rise to a flowerless pedicel in fil rev mutants (Fig. 6F,G). These data suggest that FIL and REV interact to promote flower formation within the floral primordium. One possibility is that these two genes act redundantly to provide or perceive distal positional information within the floral primordium. Alternatively, FIL and/or REV may act at steps immediately downstream of distal-domain-establishment such as cell division and/or expansion in the lateral dimension.

The filamentous phenotype of fil mutants is also enhanced by several other mutations including lfy, ufo, and clv1 (Fig. 6A) (Clark and Meyerowitz, 1994; Levin and Meyerowitz, 1995). However, in fil lfy, fil ufo and fil clv1 mutants, filament-like structures are leaf-like and/or carpelloid (Fig. 6C,E) (Clark and Meyerowitz, 1994; Levin and Meyerowitz, 1995; Levin et al., 1998). Thus, most likely, LFY, UFO and CLV1 interact with FIL at a different developmental step such as establishment of floral meristem identity (Clark and Meyerowitz, 1994; Levin and Meyerowitz, 1995; Levin et al., 1998). Other double mutants that produce filament-like structures in place of flowers include clv1 lfy, clv3 han and clv3 ufo (Clark et al., 1993; Clark and Meyerowitz, 1994; Levin and Meyerowitz, 1995). The filament-like structures in these double mutants are leaf-like or carpelloid (Clark et al., 1993; Clark and Meyerowitz, 1994; Levin and Meyerowitz, 1995), suggesting that the CLV1, CLV3, HAN, LFY, and UFO genes are not involved in providing or perceiving distal positional information within the floral primordium.

Reduced capacity to provide or perceive distal positional information within the floral primordium may explain other features of the fil phenotype. First, fil mutants produce long pedicels, which could result as a consequence of the distal flower-forming domain being either smaller or becoming established at a later time than wild type. Second, fil pedicels often contain a floral organ on the abaxial surface (Fig. 3K,L). Pedicel organs could result if the first floral organ is initiated but then, subsequently, the decision to make the remainder of the flower is delayed. Consistent with this, the pedicel organ always arises in the position of the abaxial medial first whorl organ (Fig. 3M), which in wild type, is the first floral organ to be initiated (Fig. 4E,G). Third, fil floral primordia produce ‘stalked curled sepals’ (Fig. 3B), which could result if the first floral organ is initiated but then, flower formation fails to occur. Consistent with this, the curled sepal always arises in the position of the abaxial medial first whorl organ (Fig. 3I,J). Fourth, the arrangement of flowers along the peduncle is abnormal in fil mutants (Fig. 5B). Abnormal flower arrangement could result if factors required for phyllotaxy and/or internode elongation are produced by structures arising from the distal domain.

The FIL gene is required for several aspects of flower development. First, FIL is required for floral organ formation in the correct numbers and positions (Table 2; Fig. 4F,H). Second, FIL regulates the activities of several floral genes: FIL is a positive regulator of class B activity in whorls 2 and 3; and a negative regulator of class C activity in whorls 1 and 2, of class B activity in whorl 1, and of SUP activity in whorl 3. In this regard the FIL gene shares activity with several other floral genes: LFY and UFO are positive regulators of class B activity in whorls 2 and 3 (Weigel et al., 1992; Levin and Meyerowitz, 1995); AP1, AP2, and LUG repress class C activity in whorls 1 and 2 (Bowman et al., 1991; Liu and Meyerowitz, 1995); and LUG represses class B activity in whorl 1 (Liu and Meyerowitz, 1995). Some of the apparent gene regulatory effects of the FIL gene may be caused by changes in organ position. For example, AP3 expression in whorl 1 could be caused by whorl 1 organs being positioned slightly into the whorl 2 expression domain, and whorl 3 SUP activity could be caused by whorl 3 being positioned into the whorl 4 domain (Sakai et al., 1995).

Finally, the FIL gene is required for floral organ development. This is most evident in whorl 4, in which carpels are always present but the gynoecium is deformed in one or more ways (Fig. 4S-U). Organ development defects also occur in whorls 1-3 in fil mutants; however, these may be due to ectopic expression of one or more floral genes. For example, misshapen whorl 1 organs are due to ectopic class B activity in that whorl, and deformed anthers in whorl 3 are due to ectopic SUP activity (and, probably, associated reduced class B activity) in that whorl.

One of the most striking features of the fil phenotype is the reduced flower cluster, within which the flowerless pedicel phenotype of fil is dramatically enhanced (Fig. 2). The existence of the reduced flower cluster suggests that genes acting redundantly with FIL decrease in activity during this time. One candidate is REV (Talbert et al., 1995) because the rev mutation dramatically enhances the flowerless pedicel phenotype of fil mutants (Fig. 6F). One possible explanation for the reduced flower cluster is that during this time, the plant undergoes a physiological change that results in decreased REV and/or FIL activity. Alternatively, REV and/or FIL activity may respond to signals that are in gradients along the peduncle. Regardless of the mechanism, it is clear that the reduced flower cluster represents a special region of the peduncle that becomes revealed in the sensitized fil mutant background. Thus, the fil mutant represents an entry point to identify the factors acting in this region.

The ap1-1, ap2-1, ap2-2, ap3-1, ag-1, lfy-5, lfy-6, lug-1, pi-1 and ufo-2 mutants were obtained from the Meyerowitz lab. The AP3::GUS line was obtained from Tom Jack. We acknowledge Jessica Gold for isolation of the rev-tj72 mutant. We thank Leslie Sieburth for many helpful discussions and critical review of this manuscript. This research was supported by a grant from the USDA Plant Growth and Development Program (No. 93-37304-9371).