Genetic analysis of the floral initiation process (FLIP) in Arabidopsis

The growth of any plant shoot occurs through the production of nodes by the shoot apex. During the plant life cycle, the shoot apex proceeds through a series of developmental phases (Poethig, 1990; McDaniel et al., 1992). Each phase is reflected by the changing fate of the organ or lateral meristem borne at the nodes. Within a species, the sequence and timing of developmental phases is remarkably constant. Moreover, the timing of phase switching is responsive to the environment (for example, time to flowering reviewed by Napp-Zinn, 1969). Based on these observations, the control of phase switching requires an environmentally sensitive mechanism that detects developmental time and in response activates appropriate morphological programs.

Development of the Arabidopsis shoot can be divided into four phases (Fig. 1), based on node morphology at maturity: (1) juvenile rosette, (2) mature rosette (Medford et al., 1992), (3) early inflorescence, and (4) late inflorescence, (Schultz and Haughn, 1991). Within the rosette, nodes are closely appressed and bear a leaf and a lateral meristem which may develop into an inflorescence. Elongation of internodes (bolting) designates the beginning of the inflorescence. Otherwise, early inflorescence nodes, which produce a leaf subtending a lateral inflorescence (coflores-cence, Weberling, 1989), are similar to the late rosette nodes, and the two node types may be considered to occur within a single developmental phase. In the late inflorescence phase, nodes lack leaves and lateral meristems develop as flowers instead of coflorescences.

The most dramatic morphological change within the inflorescence, which requires an alteration of both organ and lateral meristem fate, is the transition from nodes bearing coflorescences with subtending leaves to nodes bearing flowers. Coflorescences are indeterminate shoots which produce several lateral coflorescences before a number of lateral flowers, thus reiterating the inflorescence program (Figs 1, 2). In contrast, flowers are determinate shoots, producing a set of specialized organs in an invariant sequence and arrangement. Thus, the change from coflorescence to floral program requires the suppression of internode elongation, the suppression of lateral shoot development, the loss of indeterminate growth, and the initiation of floral-specific organ type and arrangement.

Recent molecular and genetic analysis in both Arabidopsis and Antirrhinum majus have resulted in a model for the control of floral organ identity (Haughn and Somerville, 1988; Schwarz-Sommer et al., 1990; Bowman et al., 1991; Coen and Meyerowitz, 1991). The two basic tenets of the model are firstly that three gene classes with overlapping fields of expression control whorl identity. Class A genes, such as APETALA2 (AP2) in Arabidopsis, act in whorls one and two, Class B genes, such as PISTILLATA (PI) and APETALA3 (AP3) act in whorls two and three, and Class C genes, such as AGAMOUS (AG) act in whorls three and four. Secondly, the model states that the expression of Class A and C genes are mutually exclusive, such that each gene class restricts the expression of the other. Class A activity also appears to influence Class B expression, since carpels can be found in the second and third whorls of strong AP2 alleles (Kunst et al., 1989; Haughn et al., 1993) and AP3 transcript levels are reduced in some AP2 alleles (Jack et al., 1992). The down regulation of Class B genes in the absence of AP2 may be the result of direct regulation by AP2, or may be the result of ectopic AG expression (Schultz et al., 1991; Jack et al., 1992). Finally, AP2 function is also required for normal development of the gynoecium (Kunst et al., 1989; Haughn et al., 1993).

Although several genes have been identified as playing a role in the inflorescence phase switch, no comprehensive model for their interactions has yet been proposed. Mutations in TERMINAL FLOWER 1 (TFL1) (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992) result in both an earlier bolting time and an earlier formation of flowers, such that coflorescences are replaced by flowers. Moreover, the total number of inflorescence nodes is extremely reduced, and the apical meristem itself develops into a flower. Mutations in a second gene LEAFY (LFY) (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992) result in inflorescences in which there is a gradual transition from nodes bearing leaves and coflorescences to nodes bearing flower-like structures. Such a phenotype suggests that LFY functions in lateral meristems and subtending organ primordia to repress development of coflorescence and leaf and to activate floral development. A third gene, APETALA1 (AP1) (Irish and Sussex, 1990; Bowman, 1992; Mandel et al., 1992), is considered to function in perianth organ identity, since in Ap1 flowers, leaves subtending flowers develop in place of sepals and petals. As in Lfy, the phenotype becomes less extreme acropetally. Ap1 flowers may also be described as a partial transformation of the floral meristem to a coflorescence meristem, an interpretation which would suggest that like LFY, AP1 is involved in sup-pressing the coflorescence program. This explanation is consistent with results from double mutant analysis (Huala and Sussex, 1992; Weigel et al., 1992; Shannon and Meeks-Wagner, 1993). A fourth gene, AP2, has been considered a Class A organ identity gene, involved predominately in perianth whorl identity (Komaki et al., 1988; Bowman et al., 1989; Kunst et al., 1989). However, there is increasing evidence that it has additional roles; certain alleles of AP2 may result in flowers that form leaves in the first whorl (Bowman et al., 1989) and develop tertiary meristems under specific conditions (Komaki et al., 1988). Moreover, analysis of Ap1/Ap2 double mutants suggests that the two genes have overlapping functions (Irish and Sussex, 1990; Shannon and Meeks-Wagner, 1993). These observations suggest that, like AP1, AP2 may have some role in activating the floral- and deactivating the coflorescence-program in lateral meristems.

In order to understand further the regulation of inflorescence morphogenesis, we have isolated new alleles of TFL1, LFY and AP1, examined mutant phenotypes under different growth conditions, and constructed double and triple mutants to identify gene interactions. On the basis of this analysis we propose a model for the initiation of flower development. This model accounts for most of the available data, including several previously unexplained aspects of both wild-type and mutant development.

All novel mutant lines described in this analysis were isolated from ethyl methanesulfonate-mutagenized populations of Arabidopsis ecotype Columbia. Before being used for analysis, lines were back-crossed to wild type as follows: Tfl1-14, 3 times; Tfl1-11, 2 times; Tfl1-12, 2 times; Tfl1-13, 2 times; Lfy-2, 3 times; Ap1-10, 3 times; Ap1-11, 1 time; Ap1-12; 3 times; Ap1-13, 3 times; Ap1-14, 1 time. Lines used for construction of double mutants were SAS 1-2-6 (homozygous for ap2-5, backcrossed three times to wild-type Columbia), SAS 1-3-7 (homozygous for ap2-6, backcrossed two times to wild-type Columbia; Kunst et al., 1989), SAS 1-13-0 (heterozygous for ag-1), SAS 1-0-0 (homozygous for ap2-1), SAS 1-10-0 (homozygous for ap1-1; gifts from Maarten Koornneef).

Plants were grown at 22°C under Grow-Lux fluorescent light (Sylvania) on Tera-lite Redi-earth prepared by W.R. Grace & Co. Canada Ltd., Ajax, Ontario, Canada. Plants were exposed to three photoperiodic regimes: continuous light (CL) -continuous illumination, 100-120 μE m⁻² second⁻¹ PAR; long day (LD) -16 hours illumination, 120-150 μE m⁻² sec⁻¹ PAR; short day (SD) -10 hours illumination, 150-180 μE m⁻² sec⁻¹ PAR. Plants grown under shorter day conditions but with equivalent light intensities were unhealthy, likely as a result of reduced photosynthates; therefore, we increased the light intensity as the daylength decreased.

Approximately equal numbers of seeds were sown in 4 inch diameter pots and kept for 3 days at 4°C to synchronize germination. Time to bolting (BT) was counted as the number of days from removal from vernalization to the elongation of the inflorescence shoot to 1 cm. Number of rosette nodes was counted at this time.

At least 50 2° shoots were taken from plants of each genotype at various positions within the inflorescence for morphological characterization using the dissecting microscope and scanning electron microscope. Samples were prepared for SEM as described in Schultz et al., 1991.

The doubly and triply mutant lines listed in Table 1 were isolated from F₂ populations generated by cross-pollinating parental lines homozygous for individual mutations. In the case of the Ag-1 line, which is both male and female sterile, a heterozygote was used. The F₂ frequency observed for each of the double mutants was consistent with that expected on the basis of Mendelian segregation (Table 1). In addition, unless both parental phenotypes were self sterile (eg. Lfy-1, Ag-1) progeny of several F₂ plants with a parental phenotype were examined. The appearance of the parental phenotype and the novel phenotype in a 3:1 ratio among the progeny of two thirds of the selected F₂ plants (Table 1) supported the conclusion that the novel phenotype represents the double or triple mutant.

The inflorescence structure of Arabidopsis is a potentially infinite series of branching shoots since each indeterminate shoot has the capacity to initiate additional indeterminate shoots. We will refer to the shoot produced by the main apical meristem as primary (1°). We will divide the 1° inflorescence into two halves, the basal 1° inflorescence, and the upper 1° inflorescence. Any lateral shoot developing from the primary shoot will be referred to as secondary (2°) and one developing from a 2° shoot will be referred to as tertiary (3° ) (Fig. 1).

Although wild type 2° shoots are either coflorescences or flowers, many of the mutants we describe here have 2° shoots with characteristics of both coflorescences and flowers (Fig. 2). Because any of the shoots may end in a fused pistil-like structure and few of the shoots have completely normal floral organs, presence of floral organs and determinacy cannot be used to distinguish shoot types. To simplify the description, unless otherwise stated, we will refer to shoots as coflorescence-like if they have 3° shoots and internode elongation, and as flower-like if they have neither 3° shoots nor internode elongation. However, it is important to note that the distinction between coflorescence-like and flower-like is artificial, since a complete spectrum of shoots with varying degrees of floral and coflorescence features can be observed.

The nomenclature of leafy organs occurring within the Arabidopsis shoot needs to be clarified. Leaves formed on the 1° shoot early in development and not separated by internode elongation are consistently called rosette leaves. Leaves formed on the 1° shoot and separated by internode elongation have been called either bracts (Schultz and Haughn, 1991) or cauline leaves (Huala and Sussex, 1992; Weigel et al., 1992). Leaves formed on 2° shoots have consistently been called bracts (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992). An analysis of shape and epidermal cell types in leaves (Martinez-Zapater et al., 1993) indicates a continuum of gradual change occurs throughout shoot development, and therefore suggests that distinctions among leafy organs is somewhat arbitrary. Thus, in the following discussion, we will refer to all leafy organs simply as leaves.

Four allelic lines having similar phenotypes, one of which we have described under the name TERMINATOR (Schultz and Haughn, 1991), were isolated from EMS mutagenized Arabidopsis populations of the Columbia ecotype. Except for the early-flowering aspect of the phenotype, which is partially dominant, all phenotypic characteristics are recessive. Similarities between the Terminator and Terminal Flower 1 phenotypes (Tfl11; Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992, Shannon and Meeks-Wagner, 1993) led to complementation analyses that confirmed that both phenotypes are due to alleles of the same gene (data not shown). Consequently, we have renamed our alleles tfl1-11, tfl1-12, tfl1-13 and tfl1-14. Because of the similarity of the four mutant phenotypes (Table 2), we will describe in detail only that of Tfl1-14.

The Tfl1-14 phenotype is similar to strong alleles described previously (Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992). Like other alleles in Columbia ecotype (Shannon and Meeks Wagner, 1991), plants homozygous for tfl1-14 have an earlier BT than wild type under continuous light, and a correspondingly fewer number of rosette leaves (Table 2). The early flowering Landsberg erecta genetic background (Table 2) may have obscured a similar early BT in other tfl1 alleles (Alvarez et al., 1992).

Tfl1 plants develop a highly modified inflorescence (Fig. 3). Compared to wild type, Tfl1-14 inflorescences produce slightly fewer leaf-bearing nodes usually having flower-like 2° shoots rather than coflorescences. Studies of these 2° shoots in early stages of development reveal that organs are initiated in the whorled pattern typical of flowers (data not shown). Most often (63%), flowers develop that are wild type except for an elongated pedicel. Occasionally, 2° shoots are coflorescence-like such that a leaf-like organ develops on the pedicel below the flower (32%), or a cluster of two to three flowers develop (5%). Before initiation of 2° shoots ceases, the inflorescence forms 1-2 flowers that are not subtended by any leafy organ. Thus, the terminal flower 1 mutation reduces the number of nodes produced in each phase of shoot development.

The terminal region of the Tfl1-14 1° inflorescence may consist of a single flower (42%) or of 2-3 clustered flowers separated by little internode elongation (58%). Examination of early developmental stages of this terminal region suggests that the apical meristem itself develops as the terminal flower(s) (data not shown; Shannon and Meeks-Wagner, 1991); therefore, the Tfl1-14 inflorescence may be described as determinate. While the 2° flowers produced in the Tfl1-14 inflorescences are completely wild type both in organ type and arrangement, organs in the three outer whorls of the 1° flower may be missing or may be mosaic organs (Alvarez et al., 1992; Shannon and Meeks-Wagner, 1991; Schultz and Haughn, data not shown). However, in all flowers examined, the gynoecium appears to be wild type. The phenotype of these terminal flowers is similar to a weak Lfy phenotype (this study, Huala and Sussex, 1992; Weigel et al., 1992).

Time to formation of flowers is highly influenced by photoperiodic conditions. As daylength decreases, the number of nodes in each developmental phase increases. Thus, wild-type plants grown under short day conditions have increased numbers of leaf-bearing nodes (rosette and early inflorescence) compared to plants grown under CL (Fig. 3 and Table 2, Koornneef et al., 1991). In order to determine if our alleles, like those of Shannon and Meeks-Wagner (1991), retain daylength sensitivity, we compared the four new Tfl 1 phenotypes (Table 2) under three photoperiodic conditions, CL, LD and SD. As described in the methods, it was necessary to change light intensity as well as the photoperiod, but we will refer to both changes simply as photoperiodic changes. Under LD (data not shown), BT is increased, plants produce more rosette nodes, and 2° shoots show more coflorescence-like tendencies than under CL. Under SD conditions the Tfl1 phenotype is more obviously altered (Fig. 3 and Table 2, Shannon and Meeks-Wagner, 1991). BT is significantly longer, more rosette leaves are formed, and the first nodes of the inflorescence bear coflorescences, followed by a greater number of flowers than in CL. As under CL, both 1° and 2° inflorescence shoots terminate in a group of closely appressed flowers. Quite often (45%), the 1° meristem senesces before 2° shoots of the upper inflorescence have matured beyond the primordial stage. Although more frequent in tfl1 alleles, this premature senescence occurs frequently in all genotypes examined (for example, 29% of wild-type plants).

A number of recessive alleles of the gene LFY have previously been described (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992). The phenotypes of the alleles are not identical, but all result in a transformation of floral nodes to coflorescence and leaf bearing nodes. Here we describe in detail the phenotype of a very weak LFY allele, lfy-2.

Plants homozygous for the lfy-2 allele bolt slightly earlier than wild type under CL conditions (Fig. 3 and Table 2), but with no decrease in the number of rosette nodes. As do other Lfy phenotypes (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992), Lfy-2 varies as one proceeds acropetally through the inflorescence, such that basal nodes bear both a leaf and a coflorescence-like 2° shoot, nodes from the mid-inflorescence bear only a coflo-rescence-like 2° shoot, and upper nodes bear flower-like 2° shoots (Table 2, Figs 2, 3). The 1° inflorescence of both Lfy-1 and Lfy-2 ends in a number of partially fused carpelloid organs (Fig. 4F), as described by Huala and Sussex (1992).

Unlike wild-type coflorescences and as in other Lfy phenotypes (Huala and Sussex, 1992, Schultz and Haughn, data not shown), all Lfy-2 coflorescence-like 2° shoots end in a number of carpel-like organs. Early in the development of these 2° shoots, organs and 3° meristems are initiated in the spiral phyllotaxy typical of wild-type coflorescences (Fig. 4A). Later organs produced by the same 2° shoots are initiated in a pattern which is somewhat more whorled (Fig. 4B,C). As they mature, these organs often overgrow the apical meristem, develop carpel cell types, and partially fuse to one another (Fig. 4D) to form a pistil-like structure (Fig. 4E).

As one proceeds acropetally in the inflorescence, organ type and arrangement in 2° meristems becomes increasingly like wild-type flowers (Fig. 5). As in wild type, four organs are typically initiated in the outermost whorl, and only rarely are they associated with 3° shoots (Fig. 5A). These may develop into sepal or sepal-carpel intermediate organs (Table 3, Fig. 5F,G). Numbers of primordia initiated in second and third whorls are variable, and the pattern of initiation often deviates from wild type (Fig. 5B-D). Although we examined a large number of flowers, no consistent pattern of primordial initiation or number was evident. Because the fate of these organs is also variable (Table 3 and Fig. 5E-G), assignment of organs to either second or third whorl is inconclusive (Table 3). Most frequently, a normal gynoecial cylinder develops in the fourth whorl (Fig. 5B-D,F-G). Infrequently, the fourth whorl may consist of 3 or 4 fused carpels, or a number of unfused carpels or carpel-stamen intermediate organs (Fig. 5H).

The Lfy phenotype is known to be affected by photoperiodic conditions (Huala and Sussex, 1992); therefore, we grew both Lfy-1 and Lfy-2 under CL, LD and SD conditions. Like wild type, Lfy plants have both an increased BT and a greater number of rosette nodes under LD compared to CL (data not shown), although the inflorescence phenotype is not significantly different from that seen under CL (Huala and Sussex, 1992). Under SD, the pheno-types of the strong Lfy-1 and the weak Lfy-2 are indistinguishable (Fig. 3 and Table 2). As under CL, both alleles bolt slightly earlier than wild type, but with no fewer rosette nodes. They then produce an inflorescence that consists only of leaves subtending coflorescences (Fig. 3).

The phenotype of only a single Ap1 (Ap1-1) mutant has been described in detail (Irish and Sussex, 1990). We have isolated five new AP1 mutants from EMS mutagenized Columbia populations. Segregation and complementation analyses (data not shown) suggest that all five phenotypes are the result of recessive alleles of the AP1 gene (data not shown). To distinguish our alleles from others newly isolated (Mandel et al., 1992; Bowman, personal communication), we have designated the alleles ap1-10, ap1-11, ap1-12, ap1-13, ap1-14. Because Ap1-12, -13 and -14 lines were isolated from a single population, they may not represent independent mutations. All five phenotypes are quite distinct from Ap1-1.

Under CL, plants homozygous for any ap1 allele flower slightly earlier than wild type, with a correspondingly reduced number of rosette nodes, and then produce a slightly reduced number of coflorescences subtended by leaves (Table 2). The fate of early inflorescence nodes is not altered by AP1 mutations; therefore in the following analysis we have considered only 2° shoots formed in the late inflorescence phase. As in Lfy, the phenotype of 2° shoots in Ap1 inflorescences are more coflorescence-like in the basal than the upper inflorescence. In addition, those 2° shoots in upper nodes have homeotic conversions, which suggest changes in organ identity gene expression.

The phenotypes caused by the five alleles vary consider-ably, the least extreme producing a number of wild-type flowers and the most extreme being similar to Ap1-1. We have grouped the alleles into three phenotypic classes, with the strongest phenotype having the most coflorescence-like basal 2° shoots, and the most severe floral organ transformations in upper 2° shoots (Tables 1 and 2): weak (ap1-13, ap1-14, and ap1-12), intermediate (ap1-11) and strong (ap1-1, ap1-10). Sequence analysis of both ap1-10 and ap1-11 is consistent with their phenotypic strengths, Ap1-10 being the result of a stop codon, and Ap1-11 being the result of an intron donor site mutation (Martin Yanofsky and Alejandra Mandel, personal communication).

The phenotypes of the three weak AP1 alleles are very similar, and we will describe only that of Ap1-12 in detail. Within the first whorl of 2° shoots, the trend from more to less coflorescence-like characteristics is apparent as one moves acropetally through the inflorescence. Most commonly, four organs arise in what appears to be a whorled arrangement (Fig. 6A), although in basal-most 2° shoots, first whorl organs in lateral positions may be missing or replaced by filaments or small protrusions of tissue (Fig. 6G). Identity of first whorl organs varies with the position of the 2° shoot in the inflorescence, from leaves, to leaf-sepal intermediate organs, and finally to sepals and sepal-carpels (Table 3). Any of the organ types may be flanked by stipules (Fig. 6D-G), although these become less frequent as one moves acropetally through the inflorescence. The frequency of elongation between lateral and axial organs (Fig. 6G) and formation of 3° shoots in the axils of the organs (Fig. 6D,G) decreases from basal to upper inflorescence. The 3° meristems reiterate the phenotype of the 2° shoots, but are always less coflorescence-like than the 2° shoots from which they develop.

As in Lfy-2, alteration in organ fate and position, combined with organ fusion, loss and/or gain makes assign-ment of organs to either whorl 2 or 3 inconclusive (Table 3). In basal-most 2° shoots (Fig. 6G), only stamens are seen at maturity. Second whorl organs are usually initiated normally in less basal 2° shoots, and form petals or filaments (Fig. 6H), or rarely petal-stamens and stamens (Table 3). Organs in the position of lateral stamens are commonly missing (Fig. 6E), and organs in axial positions usually form stamens, although filaments or petal-stamens may occur. In some cases, these intermediate organs may be the result of fusion between an organ of the second whorl and a organ arising in the position of a lateral stamen (Fig. 6F). In the fourth whorl a bicarpellate gynoecium develops in a manner indistinguishable from wild type (Fig. 6D-H).

Development of the first whorl of Ap1-11 flowers follows the same trend as is observed in the weak alleles. More commonly than in Ap1-12, organ primordia in lateral positions and occasionally in axial positions form only small protrusions of tissue or small filaments (Fig. 6M-P). More frequently, axial organs develop into leaf-like organs in basal-most 2° shoots (Fig. 6M), or into leaf-carpel or sepal-carpel intermediate organs in upper-most 2° shoots (Fig. 6O,P). These sepal-carpel organs are similar in structure to those seen with strong AP2 alleles (Kunst et al., 1988), having an central region of sepal tissue, and margins bearing ovule-like structures and stigmatic papillae. Elongation between lateral organs and axial organs (Fig. 6M,O), and development of 3° meristem in axils of lateral organs (Fig. 6M), occur more frequently than in Ap1-12, but show a similar decrease in frequency as one proceeds acropetally through the inflorescence.

Similar organ arrangements and types occur in the second and third whorl organs of Ap1-11 flowers as in Ap1-12, although with greater consistency. In most 2° shoots examined, second whorl primordia appear to be initiated at the correct time and position (Fig. 6I,J), and typically form stamens (Fig. 6J-L,N), although filaments, petal-stamens (Fig. 6N), or petal-sepals may also develop. In the third whorl, the short lateral stamens are often missing (Fig. 6I-K), and the number of axial stamens may be reduced (Fig. 6J). Occasionally, a lateral stamen appears to be transformed into a stamen-carpel intermediate organ. Most commonly in the fourth whorl a normal bicarpellate pistil develops. Occasionally the carpels may be incompletely fused (Fig. 6P), or a three or four carpellate gynoecium may develop. This mul-ticarpellate pistil is always accompanied by lateral stamen loss, suggesting that the extra carpels may be the result of third whorl stamen to carpel transformation. However, many flowers missing lateral stamens have only a bicarpellate gynoecium, indicating that incorporation into the gynoecium does not account for all absent lateral stamens.

In the most basal 2° shoots, as many as six nodes bearing 3° meristems, typically with no leafy organ, are initiated in spiral phyllotaxy (Fig. 6Q). 2° shoots taken from the mid-inflorescence initiate organs in an arrangement similar to that seen in basal Ap1-11 2° shoots (Figs 6I,W). In uppermost 2° shoots, lateral organs are missing or form filamentous organs, while axial organs are sepal-carpels (Fig. 6X).

Development of the second and third whorls in Ap1-10 2° shoots is similar to that described for Ap1-11. Four primordia, initiated in positions consistent with their representing second whorl organs, develop into stamens, stamen-petals or filaments. In the third whorl, organs in lateral positions are consistently missing, and numbers of axial third whorl organs is also reduced (Fig. 6R). The remaining two to three third whorl organs most frequently develop into stamens. Thus, a set of six stamens (Fig. 6S) may arise from both second and third whorls. Third whorl organs may also form stamen-carpel organs (Fig. 6T,V), which fuse, partially or completely, to the developing gynoecium, resulting in a multicarpellate structure (Fig. 6T).

Finally, in the fourth whorl, a gynoecial cylinder typically forms (Fig. 6S,V-W). Occasionally the carpels may remain unfused (Fig. 6U), or may be replaced by a number of stamens. Often the fertility of these flowers is reduced, with stamens shedding little pollen, and ovules appearing to degenerate.

The Ap1-1 phenotype has been described as a replacement of sepals by a whorl of leaves subtending 3° shoots, and an absence of the petal whorl (Irish and Sussex, 1990). That organs are produced in whorled rather than spiral phyl-lotaxy suggests that ap1-1 should be considered a weak AP1 allele. However, unlike the weak phenotype Ap1-12, Ap1-1 produces neither normal sepals nor petals. Thus, we suggest that the whorl-like arrangement of the organs is the result of the Ler genetic background rather than a weak AP1 allele, and have classified ap1-1 as a strong AP1 allele.

To determine if, like the LFY alleles, the AP1 alleles are influenced by photoperiod, we grew Ap1-1, Ap1-10, Ap1-13 and Ap1-12 plants under SD conditions. Ap1-1, Ap1-10 and Ap1-13 plants have a similar BT with no fewer rosette nodes than the wild-type ecotype (Table 2). However, Ap1-12 have an earlier BT with fewer rosette nodes than wild type (Table 2). Like wild type, all alleles then produce a number of leaves subtending coflorescences. Following this stage, allelic differences become apparent. In the most extreme case, Ap1-10, all 2° meristems in the inflorescence form coflorescence-like structures (Fig. 3). In Ap1-1 plants, the first 2° shoots not subtended by leaves are transformed into coflorescences. However, following this basal region, 2° shoots are initiated, which are similar to those formed in Ap1-1 under CL. We suggest that the phenotypes of the two strong alleles, Ap1-1 and Ap1-10, differ from one another under SD as a result of differences in Ler and Columbia backgrounds. With weak alleles such as Ap1-12 the phenotype of the basal 2° shoots are more coflorescence-like than under CL such that they produce 4-8 leaves subtending 3° laterals and separated by internode elongation (Fig. 3). Following this region, six stamens and a bicarpellate pistil form normally. The upper 2° shoots of Ap1-12 plants form flowers similar in structure to those in CL.

Certain alleles of AP2 have some characteristics which suggest that, like AP1, AP2 may have some function in suppressing the coflorescence program. For example, Ap2-1 inflorescences produce 2° shoots in which the sepal whorl is replaced by leafy organs or leaf-carpel organs (Bowman et al., 1991). As well, the phenotypes of certain alleles appear to be affected by photoperiod (Komaki et al., 1988).

We grew plants homozygous for ap2-1, ap2-5 and ap2-6 under SD conditions and analyzed the resulting phenotypes. Neither number of rosette nodes nor number of coflores-cence nodes were different than in wild-type plants grown under SD (data not shown). Flowers of the weak Ap2-5 allele were not significantly different from those seen under CL. However, the phenotypes of both Ap2-1 and Ap2-6 were significantly altered. As under CL, the first whorl of Ap2-1 flowers consists of leafy organs, but under SD these frequently subtend 3° shoots (data not shown). Whereas under CL the second whorl of Ap2-1 flowers consists of petal-stamens, under SD, leafy organs arise. Under CL, the first and second whorls of Ap2-6 flowers are transformed to carpels (Kunst et al., 1989); under SD, both whorls are trans-formed to leafy organs, which often subtend 3° shoots and may be separated by internode elongation (data not shown).

The homeotic transformations of 2° meristems within the inflorescence of Tfl1-14 plants may be described as opposite to those in Lfy and Ap1 inflorescences. The photoperiodic responses of the mutants are also opposite. SD conditions suppress the Tfl1-14 phenotype but enhance the phenotype of Lfy and Ap1. To determine how TFL1, LFY and AP1 interact to regulate the transition from early to late inflorescence, we constructed doubly and triply mutant lines. As well, because homeotic transformations of organs in Lfy and Ap1 2° shoots suggest interactions of these genes with the whorl identity genes, we constructed lines doubly mutant for mutations in either Classes A and C organ identity genes and either LFY or AP1.

Lines doubly mutant for either lfy-1 or lfy-2 and tfl1-14 form an inflorescence after a shorter time but with no fewer rosette nodes than the Tfl1-14 single mutant. Both doubly mutant lines produce fewer coflorescence and leaf bearing nodes than the corresponding Lfy single mutant (Table 2, Fig. 3), followed by nodes bearing flower-like 2° shoots. Thus, the 2° shoots of the double mutants are similar in form to those of the corresponding Lfy single mutant, but occur at an earlier node on the 1° shoot. The total number of inflorescence nodes produced is greater than that in Tfl1-14 single mutants, but significantly fewer than Lfy or wild type (Table 2, Fig. 3). In both doubly mutant lines, 1° and 2° inflorescence shoots end in a number of partially fused carpels or leaf-carpel organs.

In order to determine if loss of TFL1 activity has any effect on the Ap1 phenotypes, we generated lines doubly mutant for tfl1-14 and both the strong alleles ap1-1 and ap1-10 and the weak alleles ap1-13 and ap1-14. Because of the similar phenotypes between Tfl1-14/Ap1-13 and Tfl1-14/Ap1-14, as well as between Tfl1-14/Ap1-1 and Tfl1-14/Ap1-10, we will describe only Tfl1-14/Ap1-10 and Tfl1-14/Ap1-13.

Plants doubly mutant for tfl1-14 and either ap1 allele have the same BT and number of rosette nodes as the Tfl1-14 mutant. They then produce a number of Ap1-like 2° shoots, which become less coflorescence-like as one proceeds acropetally; the first 1-2 are subtended by a leaf, and the remaining 2-3 by no leaf. The phenotypes of the 2° shoots are essentially similar to 2° shoots formed in the corresponding Ap1 plant, but are less coflorescence-like than those formed at a corresponding node in the Ap1 single mutant. In the Tfl1-14/Ap1-13 double mutant, the 1° meristem forms a flower similar in phenotype to an Ap1-13 flower. In contrast, the 1° meristem of Tfl1-14/Ap1-10 double mutants forms a number of incompletely fused carpelloid organs.

The LFY mutation affects the fate of both the subtending leaf and 2° meristem, while Ap1 affects only the fate of 2° meristems. Moreover, LFY mutations result in 2° shoots that are more coflorescence-like than any AP1 allele. Together, these observations suggest that LFY has a more profound role in the switch from coflorescence to floral fates than AP1. Previous double mutant constructions suggest that loss of AP1 results in a more extreme Lfy phenotype (Huala and Sussex, 1992; Weigel et al., 1992). To determine if allelic differences would influence the double mutant phenotypes, we constructed lines doubly mutant for strong and weak alleles of both LFY and AP1, generating the following lines: Lfy-1/Ap1-1; Lfy-1/Ap1-10; Lfy-2/Ap1-1; Lfy-2/Ap1-10; Lfy-2/Ap1-12.

Rather surprisingly, the double mutant phenotypes of all allelic combinations are similar, resulting in inflorescences consisting only of coflorescence-like structures. Because all allelic combinations give similar phenotypes, we will describe only Lfy-2/Ap1-10. Doubly mutant inflorescences produce the same number of nodes with leaves and coflo-rescences as Lfy-2. However, no flower-like structures are ever produced in the doubly mutant lines, although as in the single mutants, there is an acropetal decrease in coflores-cence-like characteristics (compare Fig. 7A with 7C). All 2° shoots appear to end in a malformed pistil-like structure (Fig. 7B,C) similar to those seen in Lfy mutants.

Plants triply mutant for TFL1, LFY and AP1 form an inflo-rescence at the same time as Tfl1-14 mutants, and with the same number of rosette nodes. They then produce the same number of leaves as does the Tfl1-14/Lfy-1 double mutant, before the 1° meristem forms a pistil-like structure similar to that in the Tfl1-14/Lfy-1 double mutant. In the axils of these leaves, 2° shoots develop (Fig. 8), which are similar in form to the 2° shoots of a Lfy-1/Ap1-1 double mutant. Basal 2° shoots are coflorescence-like (Fig. 8A), but 2° shoots become rapidly less coflorescence-like as one moves acropetally through the inflorescence (Fig. 8B), such that uppermost 2° shoots are more flower-like than those at a corresponding node in the Lfy-1/Ap1-1 double mutant, consisting only of fused leaf-carpel organs (Fig. 8C).

We have generated lines doubly mutant for lfy-1 or lfy-2 with ap2-1 and ap2-6. The phenotype of Lfy-1/Ap2-1 and Lfy-2/Ap2-1 are more similar to lines doubly mutant for LFY and AP1 than those double mutant combinations described by Huala and Sussex (1992). Inflorescences fail to produce flower-like 2° shoots, but rather continue to form 2° shoots having leaves and leaf-carpel intermediate organs subtending 3° meristems and separated by elongated internodes (data not shown). In contrast to lines constructed by Huala and Sussex (1992), the 2° shoots of either Lfy-1/Ap2-1 or Lfy-2/Ap2-1 end in a number of partially fused leaf-carpel intermediate organs.

We have presented a brief description of Lfy-1/Ap2-6 double mutants previously (Schultz and Haughn, 1991). Doubly mutant plants produce flower-like 2° shoots much more rarely than Lfy-1 inflorescences. Instead, the inflorescence consists almost entirely of coflorescence-like shoots similar to those in Lfy-1 single mutants, having organ types ranging from leaves, through sepals and carpels (Fig. 9A-C). Lines doubly mutant for lfy-2 and ap2-6 do eventually form flower-like 2° shoots (Fig. 9F), although at a later node than either Ap2-6 or Lfy-2 single mutant inflorescences. The coflorescence-like 2° shoots (Fig. 9D,E) are similar to those formed in Lfy-1/Ap2-6 doubles, while the more flower-like uppermost 2° shoots are composed entirely of carpels (Fig. 9F).

Like AP2, both LFY (Huala and Sussex, 1992) and AP1 (Mandel et al., 1992) also appear to influence Class C gene expression, since in Lfy-1, Lfy-2, Lfy/Ap1 and Lfy/Ap2-1 2° shoots, increasingly carpel-like organs are initiated (leaf, leaf-carpel, carpel). Moreover, analysis of Lfy/Ap2-6 double mutants suggests that absence of AP2 does not greatly alter the pattern of Class C gene expression in Lfy coflorescence-like 2° shoots. In order to determine if the carpel-like organs that form in Lfy coflorescence-like 2° shoots require AG, we generated lines doubly mutant for lfy-1 or lfy-2 and ag-1. Doubly mutant lines with ag and other LFY alleles have been described previously (Huala and Sussex, 1992; Weigel et al., 1992).

Organ types within Lfy-1/Ag-1 and Lfy-2/Ag-1 coflorescence-like 2 ° shoots do not show a dramatic alteration from those in the Lfy single mutant plants, having similar organ types, and ending in partially fused carpels or leaf-carpels (Fig. 10E). The degree of carpelloidy appeared to be somewhat reduced in these structures, especially in Lfy-2/Ag-1, such that fusion was rarely as complete as in the Lfy single mutants.

Double mutant 2° shoots formed at a node corresponding to the most flower-like 2° shoots in the Lfy single mutant are similar to those described by Huala and Sussex (1992). Lfy-1/Ag-1 flower-like 2° shoots consist of sepals and sepal-carpel intermediate organs, and Lfy-2/Ag-1 of sepals, rare sepal-carpel intermediate organs and petals or petal-sepal intermediate organs. In both Lfy-1/Ag-1 (Fig. 10B) and Lfy-2/Ag-1 (Fig. 10E), organs frequently subtend 3° meristems. Moreover, although internodes between the floral organs are compressed early in development, as the plant matures, the internodes elongate. Thus, loss of AG appears to increase coflorescence-like characteristics as well as decreasing numbers of reproductive organs.

We generated double mutants between ap1-1 and each of ap2-1, ap2-5 and ap2-6. As reported by Irish and Sussex (1990), plants doubly homozygous for ap1-1 and ap2-1 were found to produce very coflorescence-like 2° shoots in place of flowers (data not shown). These are especially coflorescence-like at the base of the inflorescence, and become less so as one proceeds acropetally. Such coflorescence-like 2° shoots were also seen in the basal-most 2° shoots of Ap1-1/Ap2-6 double mutants, although at a low frequency (data not shown). Except for these basal-most 2° shoots, the phenotypes of the outer whorls of Ap1-1/Ap2-6 and Ap1-1/Ap2-5 were essentially additive of the single mutant phenotypes, while the phenotype of the inner whorls was more extreme than either single mutant. 2-4 3° shoots form first in Ap1-1/Ap2-6 2° shoots, and may be subtended by no organ or by leaf-carpel organs, and be separated by internode elongation (data not shown). Next, several stamens, stamen-carpel intermediate organs and finally carpels develop. In some flowers, two carpels fuse to form a typical bicarpellate pistil. However, more frequently than in either single mutant, carpels appear abnormal in morphology and remained unfused.

During development of the Arabidopsis shoot, the primary apical meristem switches from producing leaf- and coflorescence-bearing nodes to producing flower-bearing nodes. Three genes, TFL1, LFY and AP1 are known to affect this process (Irish an d Sussex, 1990; Schultz and Haughn, 1991; Shannon and Meeks-Wagner, 1991; Alvarez et al., 1992; Huala and Sussex, 1992; Mandel et al., 1992; Weigel et al., 1992; Shannon and Meeks-Wagner, 1993). In this manuscript we have (1) described four new mutant alleles of TFL1, one new allele of LFY, and five new alleles of AP1 genes and (2) analyzed interactions between TFL1, LFY, AP1, AP2 and AG by constructing double and triple mutants. On the basis of these studies we propose a working model for the control of the transition from coflorescence-bearing to flower-bearing nodes.

The transition from vegetative to reproductive phases of Arabidopsis shoot development is facultative and involves changes in leaf shape, internode elongation and lateral shoot type. The most dramatic morphological change occurs when the primary meristem switches from producing a leaf- and coflorescence-bearing node to a flower-bearing node. The number of leaf-bearing nodes that develop prior to the formation of the first flower is strongly regulated by environmental conditions such as photoperiod and temperature (recent review Martinez et al., 1993). As has been suggested for other species (Lang, 1965; Murfet, 1989; Poethig, 1990; McDaniel et al., 1992) Arabidopsis must have a mechanism that monitors developmental time and activates morphological programs associated with phase transition in response to the number of nodes produced. In addition, the mechanism must be responsive to the environment (a broader discussion will be given in a subsequent publication: Haughn et al., unpublished data). A simple working model consistent with all the above observations is graphically illustrated in Fig. 11. We suggest that the activity level of a factor(s) changes with node production, and that at critical levels, morphological programs associated with a phase change are activated. We will refer to the factor(s) as Controller(s) of Phase Switching (COPS). Moreover, variations in environmental conditions such as photoperiod must accelerate or decelerate COPS change (compare Fig. 11A with B), resulting in a concomitant decrease or increase, respectively, in the time to phase transitions. For simplicity throughout this discussion, we present the change in COPS activity as a decrease (Fig. 11), a hypothesis favoured by the phenotype of the early flowering mutant Embryonic Flower (EMF) (Sung et al., 1992) (see following section). Support for the concept that COPS change is gradual, occurs throughout shoot development and is lower in the lateral shoot than in the shoot from which the lateral arose, is discussed in following sections.

An interesting aspect of plant development is that at any point in development the fate of the apical meristem may be distinct from that of the lateral meristems it generates (Poethig, 1990; McDaniel et al., 1992). For example, in the late inflorescence phase of Arabidopsis, lateral meristems follow the floral program, while the 1° meristem pursues an inflorescence program. This reduction in inflorescence-like characteristics in lateral as compared to apical meristems is reiterated in all mutants we have analyzed. In Tfl1 inflorescences, 2° meristems adopt a floral fate several nodes before the 1° meristem is converted to the floral program. In Lfy and Ap1 inflorescences, 3° shoots are always more flower-like than the 2° shoot from which they are derived.

We have suggested that the switching between phases is the result of changing COPS activity. The observations described above suggest that phase changing is accelerated in lateral meristems relative to the apical meristems. This suggests that COPS activity is reduced in lateral meristems relative to the apical meristem from which it originated.

The Tfl1 shoot has a reduced number of nodes in rosette, early inflorescence and late inflorescence developmental phases. At the same time, there seems to be no absolute requirement for TFL1 activity during any developmental phase, since in various double mutant combinations and environmental conditions, all shoot developmental phases can be observed (coflorescence-bearing nodes in Tfl1-14/Lfy double mutants or in Tfl1-14 under SD conditions). Moreover, developmental phases are expressed in the correct sequence, although at an altered rate. These data suggest that the TFL1 gene product is involved in establishing the timing of phase transition during shoot development, and thus influences COPS activity (Fig. 11C). All Tfl1 mutants are responsive to environmental conditions. Assuming that at least some of the TFL1 alleles are null or close to null alleles, this indicates that TFL1 is not required for the photoperiodic response of COPS.

The development of a terminal flower on the Tfl1 1° shoot demonstrates that the primary meristem can itself enter the floral program although it does not do so in the wild-type plant. We suggest that the wild-type plant normally senesces before COPS activity is low enough to induce the floral program in the primary meristem (Fig. 11).

The phenotypes of the early- and late-flowering loci suggest that, like TFL1, they influence COPS activity. Mutations in late-flowering loci (for example FCA, FY, FD, GI) result in increased numbers of leaf-bearing nodes relative to wild type (Martinez-Zapater and Somerville, 1990; Koornneef et al., 1991), indicating that the wild-type gene products accelerate COPS activity change (Fig. 11B). Mutations in EARLY-FLOWERING (ELF1, ELF2) loci (Zagotta et al., 1992) and in EMF (Sung et al., 1992), result in a decreased number of leaf-bearing nodes, implying that they retard COPS change (Fig. 11C). Mutations in EMF result in the most dramatic phenotype, such that a determinate inflorescence forms immediately following germination. In addition, the EMF gene is required for photoperiodic and vernalization response of flowering. Such a phenotype suggests that EMF may be central to the COPS mechanism.

The available evidence suggests that LFY, AP1 and AP2 gene products all have a role in switching meristems from an inflorescence program to a floral program (suppressing inflorescence program, activating floral program), a process we will refer to as FLIP (Floral Initiation Process). It is also clear that the contribution of the three genes to FLIP is not equal.

The requirement for LFY in FLIP is well documented (Schultz and Haughn, 1991; Huala and Sussex, 1992; Weigel et al., 1992; Shannon and Meeks-Wagner, 1993). In addition, we demonstrate here that LFY is required for the FLIP that occurs in the 1° meristem of Tfl1 inflorescences. However, even in plants homozygous for the lfy-1 allele, which DNA sequence analysis suggests is a null allele (Weigel et al., 1992), flower-like 2° shoots form in the upper inflorescence. This gradual replacement of coflorescences by flowers in Lfy-1 inflorescences suggests that the FLIP can occur even in the absence of LFY activity.

Observations made previously have suggested that besides LFY, at least two other genes, AP1 and AP2, are involved in FLIP (Komaki et al., 1988; Irish and Sussex, 1991; Huala and Sussex, 1992; Weigel et al., 1992; Shannon and Meeks-Wagner, 1993). Our analysis has confirmed and extended these observations. 2° shoots in both Ap1 and Ap2 inflorescences have many coflorescence-like characteristics which, as in Lfy, vary both in an inflorescence- and photoperiodic-dependent manner.

Analyses of Lfy/Ap1, Lfy/Ap2 and Ap1/Ap2 doubly mutant lines support the hypothesis that the three genes have redundant roles in FLIP. Absence of either AP1 or AP2 in a Lfy background results in an increased tendency towards coflorescence-like 2° shoots, suggesting that in Lfy plants, the eventual formation of flowers in the upper inflorescence is dependent at least in part on AP1/AP2 expression. Ap1/Ap2 double mutants form very coflorescence-like structures in the basal-most 2° shoots (Irish and Sussex, 1990; this study), suggesting that LFY activity alone is insufficient to completely initiate FLIP at the appropriate time. Similar types of functional redundancy have been well-documented in animal development (Tautz, 1991).

Because of their common role in determining 2° meristem fate, we designate LFY, AP1 and AP2 as FLIP genes. The single mutant phenotypes suggest that while all FLIP genes are necessary to suppress the coflorescence and activate the floral program in meristems, their roles are not equivalent. The transformation of flowers to coflorescences is more complete in Lfy inflorescences than in either Ap1 or Ap2 inflorescences. Similarly, Ap1 2° shoots are more coflorescence-like than those of Ap2 inflorescences. Based on these observations, we suggest that the requirement of the genes for FLIP is LFY > AP1 > AP2.

It is notable that ap2-1, which is known to have reduced ectopic AG expression compared to other AP2 alleles (Drews et al., 1991), has a more coflorescence-like phenotype in both singly and doubly mutant combinations. As well, more extreme coflorescence-like structures form in Lfy/Ag double mutants compared to Lfy single mutants. These observations suggest that AG activity has a role in suppressing coflorescence characteristics.

As well as activating FLIP in 2° meristems, LFY acts in another aspect of the early to late inflorescence phase change, affecting the fate of the subtending leaf. Both leaf and 2° meristem are thought to be derived from a common anlage (Weigel et al., 1992), and, in wild type, have tightly coupled fates (coflorescences subtended by a leaf, flowers subtended by no leaf). However, in both Lfy and Tfl1 inflorescences, their fates become uncoupled, such that flowers may be subtended by a leaf or coflorescences subtended by no leaf. This suggests that although the structures may arise from a common anlage, their fates become uncoupled very early in development.

Successive 2° shoots in inflorescences of Lfy, Ap1 and Ap2 single and double mutant combinations become gradually less coflorescence-like as one moves acropetally through the inflorescence. Thus, activation of the FLIP genes is coordinate, independent of one another and progressively stronger as inflorescence development proceeds. Moreover, the rate at which FLIP activation occurs is dependent on TFL1 activity and daylength. Taken together, these data suggest that COPS coordinately activates the FLIP genes, resulting in the morphological changes associated with the late to early inflorescence phase change (Fig. 11).

The phenotype of 1° structures in Tfl1, Lfy, Ap1-10 and doubly mutant combinations suggests that COPS determines the timing of FLIP gene expression in 1° as well as 2° meristems. The phenotype of the Tfl1 terminal flower is similar to Lfy flower-like 2° shoots, suggesting that LFY expression in the terminal flower may be reduced compared to 2° flowers. In Lfy/Tfl1-14 and Ap1-10/Tfl1-14 the main inflorescence forms a pistil-like structure, indicating that ectopic LFY and AP1 expression are required for the Tfl1 terminal flower. Thus, in wild-type plants, COPS may maintain an inflorescence program in the 1° meristem by ensuring that the FLIP genes are not expressed prior to senescence.

As discussed above, the COPS mechanism determines the time of FLIP gene activation, resulting in the coflorescence to floral transition in meristems. In wild-type or Tfl1 plants, the transition is abrupt. In contrast, loss of function of any one of the FLIP genes results in a COPS-dependent, gradual transition from basal coflorescence-like shoots to apical flower-like shoots (Fig. 3). Thus, combined expression of all FLIP genes is required for an abrupt FLIP, indicating that the FLIP genes must in some way act cooperatively. Following initial coordinate activation by COPS, several possible models for such cooperative action are consistent with the data: (1) the FLIP genes up-regulate one anothers expression; (2) the FLIP genes negatively regulate an inhibitory function, such as COPS; (3) the FLIP gene products act in a cumulative fashion to activate downstream genes.

The requirement for AP1 in perianth organ identity is well established (Irish and Sussex, 1990; Mandel et al., 1992). Our analysis of several new AP1 alleles suggests that, like AP2, AP1 may be considered a Class A organ identity gene.

The flower-like 2° shoots formed in the upper inflorescence of all AP1 alleles show transformations of perianth organs to reproductive organs. In stronger alleles, third whorl organ number is reduced and gynoecial abnormalities often occur. Such organ transformations are similar to those observed in flowers mutant for AP2. The organ transformations suggest that like AP2, AP1 has some role in restricting Class C gene expression to the inner whorls, and that in its absence, Class C genes are ectopically expressed in the perianth. Molecular analyses (Mandel et al., 1992) indicate that AP1 is strongly expressed in all whorls of Ag flowers, suggesting that AG restricts AP1 expression to the perianth. Thus, AP1 seems to share both aspects of AP2-AG regulation, being negatively regulated by AG, and serving to negatively regulate AG. For these reasons, we believe AP1 should be considered a Class A whorl identity gene.

We suggest that AP1 and AP2 gene products act together as Class A genes specifying organ identity. An alternative possibility for the similarities in Ap1 and Ap2 phenotypes is that one gene is necessary for activation of the other. However, Ap2/Ap1 double mutants have a phenotype distinct from either single mutant suggesting that each gene is expressed in the absence of the other (Irish and Sussex, 1990; this study); in support of this, molecular data suggest that young Ap2 floral primordia express AP1 (Mandel et al., 1992).

A Class A gene equivalent has not yet been isolated in Antirrhinum majus. Interestingly, the sequence of SQUAMOSA (Huijser et al., 1992) bears significant sequence identity to AP1 (Mandel et al., 1992). An attractive possibility is that SQUAMOSA represents Class A gene activity in Antirrhinum.

Transformations of sepals to carpels, petals to stamens, and stamens to sepals in Lfy flower-like 2° shoots suggest an altered pattern of Class C gene expression (Huala and Sussex, 1992; Weigel et al., 1992). Though similar to those associated with loss of Class A gene expression, the organ transformations are not identical. Thus, while LFY seems to have properties of a Class A gene, its role is slightly different from AP1 and AP2. The apparent Class A LFY activity may be a reflection of cooperative function among LFY, AP1, and AP2. In support of this hypothesis, organ transformations become less extreme in upper 2° shoots of weak Lfy alleles, suggesting that the transformations in lower 2° shoots may be due in part to reduced AP1/AP2 expression.

Transformations of petals to sepals and stamens to carpels in Lfy single mutants indicate reduction in PI/AP3 expression, a hypothesis supported by double mutant analysis (Huala and Sussex, 1992; Weigel et al., 1992). As in Ap2 and Ap1 phenotypes, the reduced Class B gene expression may be the result of ectopic Class C gene expression. The rare occurrence of stamen-carpels in the fourth whorl suggests that Class B organ identity genes are ectopically expressed in the fourth whorl. Thus LFY may have some role in restricting PI/AP3 expression, perhaps through activation of the FLO10 gene (Schultz et al., 1991; Bowman et al., 1992).

The analysis of Huala and Sussex (1992), as well as our own, suggests that LFY has some function in the apical meristems of both coflorescences and the main inflorescence. In wild-type plants, both coflorescences and the main inflorescence are indeterminate, never forming differentiated organ structures. In contrast, the inflorescence-like shoots of Lfy mutants (including the 1°) always terminate in pistil-like structures consisting of fused carpels and carpel-leaves. Such structures form in Lfy/Ap2, Lfy/Ap1, Lfy/Tfl1-14, and, to a lesser extent, in Lfy/Ag double mutants. Their development suggests that LFY is required to repress Class C organ-identity gene activation late in inflorescence development. Since carpels develop to a certain extent even in Lfy/Ag double mutants, Class C gene activities besides AG must be involved.

The fact that the LFY gene product not only activates the floral program early in floral meristem development but also represses class C organ-identity gene expression late in indeterminate inflorescence development seems paradoxical. These data could be explained if (1) COPS weakly activates floral specific genes late in inflorescence meristem development (2) low levels of LFY gene product repress Class C organ-identity gene expression but are insufficient for floral activation. Through most of plant development, COPS activity remains too high in the apical meristems to allow FLIP gene expression. Late in plant development, COPS activity is reduced sufficiently to allow low level expression of the floral genes including LFY and Class C organ-identity genes. The expression of LFY is not sufficient to activate the floral pathway, but is sufficient to inhibit Class C gene expression. This hypothesis predicts that LFY be expressed late in the development of inflorescence meristems. Weigel et al. (1992) did not detect LFY transcript in inflorescence meristems; however, it is not clear if inflorescence meristems very late in development were examined.

Along with Fig. 12, the following points summarize our conclusions regarding the control of the early to late inflorescence phase switch.

1. COPS activity decreases gradually throughout shoot development (Fig. 11).

2. At a critical COPS level, the FLIP genes are activated in nodal anlage that will give rise to 2° shoots (all FLIP genes activated) and leaves (only LFY activated).

3. The FLIP genes together initiate FLIP by repressing the inflorescence program, repressing Class C organ identity gene expression and activating perianth development. In addition, LFY represses leaf development.

4. Late in flower development, a decrease in FLIP gene expression allows the expression of Class C organ identity genes which, as well as activating reproductive organ development, continues maintenance of the floral program by repressing the inflorescence program.

While this manuscript was being reviewed, an analysis of genetic interactions between TFL1, LFY, AP1, AP2 and CLV1 by Shannon and Meeks-Wagner (1993) was published. Despite the fact that these authors used different alleles than ourselves for many of the genes, we were gratified to find that, in those areas where we overlapped, our data and overall conclusions were similar.

We disagree with the interpretation of Shannon and Meeks-Wagner on only a few points but two of these warrant comment. First, they discuss TFL1 function as if it were unique to and required solely for the maintenance of the inflorescence meristem. However, Tfl1 bolting occurs markedly earlier (with respect to both time and number of rosette leaves) than wild type. These data suggest that TFL1 functions even during the vegetative phase promoting the formation of the inflorescence at the expense of the rosette. For these reasons we believe that the proposed COPS mechanism, in which TFL1 plays a role, more accurately represents the available data.

Second, Shannon and Meeks-Wagner suggest that two separate pathways exist for FLIP; one represented by LFY and the other represented by AP1 and AP2. In light of the additional data that we present, it is clear that all three genes act independently of one another and have differential roles. There seems no compelling reason to separate out LFY while grouping AP1 and AP2 together. Further, all three genes work toward a common end and are required for FLIP to function properly. Thus we prefer to group the three genes together. We believe that the evocation of separate pathways at this point is unnecessary and somewhat misleading.

We are grateful to Jose Martinez-Zapater for many insightful discussions on the phase-switching mechanism in Arabidopsis. Thanks to Drs Thurston Lacalli and Susanne Kohalmi, Zora Modrusan and Mark Wilkinson for critical reading of the manuscript. We thank Marilyn Martin for generously providing Ap1/Ap2 double mutant lines, Susan Shannon and Ry Meeks-Wagner, and Alejandra Mandel and Marty Yanofsky for sharing unpublished data, and Dennis Dyck for excellent advice on graphics and photography.