Abstract
PETAL LOSS is a new class of flower development gene whose mutant phenotype is confined mostly to the second whorl. Two properties are disrupted, organ initiation and organ orientation. Initiation is frequently blocked, especially in later-formed flowers, or variably delayed. The few petals that arise occupy a wider zone of the flower primordium than normal. Also, a minority of petals are trumpet-shaped, thread-like or stamenoid. Studies of ptl combined with homeotic mutants have revealed that the mutant effect is specific to the second whorl, not to organs with a petal identity. We propose that the PTL gene normally promotes the induction of organ primordia in specific regions of the second floral whorl. In ptl mutants, these regions are enlarged and organ induction is variably reduced, often falling below a threshold. A dominant genetic modifier of the ptl mutant phenotype was found in the Landsberg erecta strain that significantly boosts the mean number of petals per flower, perhaps by reinforcing induction so that the threshold is now more often reached. The second major disruption in ptl mutants relates to the orientation adopted by second whorl organs from early in their development. In single mutants the full range of orientations is seen, but when B function (controlling organ identity) is also removed, most second whorl organs now face outwards rather than inwards. Orientation is unaffected in B function single mutants. Thus petals apparently perceive their orientation within the flower primordium by a mechanism requiring PTL function supported redundantly by that of B class genes.
INTRODUCTION
Floral organs usually occur in concentric whorls within a flower, with each whorl containing a defined number of organs. Typically, there are several whorls of sterile organs, sepals and petals, surrounding and protecting the reproductive organs, stamens and carpels. The arrangement of floral organs within a flower, the floral ground plan, is defined early in flower development when the exact number and placement of organs is established.
The relative constancy of the floral ground plan in Angiosperms implies that it is under strict genetic control. While much has been learned about genes that control the identity of floral organs once they have arisen, less is known about genes that control when and where floral organs arise (Irish, 1999; Meyerowitz, 1997). In Arabidopsis thaliana, mutants of several genes have been described that, when mutated, consistently result in changes to the numbers of organs initiated. Arabidopsis flowers normally contain four outer sepals surrounding four petals in positions alternate to them. Internal to these six stamens are present, two in lateral positions and four inserted medially. The two-carpelled gynoecium occupies the centre of the flower. In mutants of three CLAVATA (CLV) genes, however, there is an increase in the number of all four floral organ types, particularly the inner whorls of stamens and carpels. Two of these genes have been cloned. One (CLV1) encodes a receptor kinase likely to be involved in signalling within the flower meristem that keeps in check the proliferation of the undifferentiated central zone (Clark et al., 1997). Another (CLV3) encodes a small extracellular protein that may act as a ligand in the same process (Fletcher et al., 1999).
On the other hand, mutations of the SHOOT MERISTEMLESS (STM) and WUSCHEL (WUS) genes can generate flowers with the complementary phenotype, reduced numbers of organs especially in the inner whorls. Both STM and WUS encode homeodomain proteins that regulate the expression of other genes, and it has been proposed that they normally promote proliferation of the central zone of the flower meristem (Long et al., 1996; Mayer et al., 1998). For both classes of gene, therefore, mutant changes in floral organ number may reflect the secondary consequence of disruption to the balance between the dampening and promotion of proliferation of the undifferentiated central meristem. Increased or reduced numbers of organs may reflect changes in the amount of material upon which the ground plan genes can act.
One Arabidopsis gene likely to play a more direct role in generating the floral ground plan is PERIANTHIA (PAN) (Running and Meyerowitz, 1996). In this case, mutant flowers frequently demonstrate five-fold symmetry in their outer three whorls rather than the normal bilateral symmetry seen in the wild type. Importantly, unlike the genes described above, no changes to the size of the mutant flower primordium were detected. This is consistent with a role for PAN in defining the bilateral ground plan that defaults to a more regular radial pattern in the absence of PAN function (Chuang et al., 1999).
In the present study, we describe a different category of floral organ mutant. Mutations of the PETAL LOSS (PTL) gene result in the loss of just one type of floral organ. Even so, PTL is not absolutely required for petal initiation as some petals may arise in ptl mutants, especially in the first-formed flowers. In addition, a dominant modifier specific to ptl mutants was discovered that boosts organ production in the second whorl. The PTL gene (and its modifier) may support the action of an inducer that is required above a threshold level for these primordia to be initiated. At the same time, genetic evidence suggests that PTL has another unique function, that of allowing petal primordia to perceive their orientation within the flower primordium. This function is redundantly supported by the products of B function organ identity genes.
MATERIALS AND METHODS
The ptl-1 mutant was isolated by John Alvarez (Alvarez and Griffith, 1994) in a screen of the Columbia strain mutagenised with ethyl methane sulphonate. The ptl-2 allele arose in the C24 strain following regeneration from root tissue transformed with T-DNA, but was unlinked to kanamycin resistance encoded by the T-DNA construct. The ecotypes Columbia (Col-3), Landsberg erecta (Ler-0) and Niederzenz (Nd-1) were obtained from Elliot Meyerowitz, Bensheim (Be-0) and Enkheim (En-2) from the Arabidopsis Information Service, and Kiruna (Kr-2) from Liz Dennis. The floral homeotic mutants agamous-1 (ag-1), apetala3-1 (ap3-1), ap3-3 and pistillata-1 (pi-1), all in Ler background, were available from laboratory stocks (Bowman et al., 1991). Recombinant inbred lines were obtained from Caroline Dean (Lister and Dean, 1993).
Plants were grown in seed-raising mix and perlite, and sub-irrigated with nutrient solution. Most were grown in glasshouse conditions at 22-24°C under daylight combined with continuous fluorescent lighting (Cool White). Some were grown in incubators at 16°C or 29°C using the same artificial lighting conditions. The first 20 flowers on the main inflorescence stem of 4 to 20 plants were scored unless otherwise mentioned. Comparisons were based on plants raised together. Crosses were carried out in duplicate, and genotypes confirmed by progeny testing if appropriate.
For scanning electron microscopy, samples were fixed in 2% glutaraldehyde in 25 mM sodium phosphate buffer (pH 6.8), infiltrated with 0.5% osmium tetroxide, dehydrated through an ethanol series, and critical point dried using liquid carbon dioxide. After coating with gold, samples were viewed using a Hitachi s570 scanning electron microscope. For light microscope sections, material was fixed in glutaraldehyde and dehydrated as above, embedded in LR White resin, sectioned at 2 μm, and sections stained with 0.3% toluidine blue in 1% tetraborate buffer (pH 9.0).
RESULTS
petal loss mutants result in aberrant petal development
The most obvious effect of the ptl-1 and ptl-2 mutations is a reduction in the number of petals per flower (Fig. 1A-C). Wild-type flowers always have four petals, one between each of the sepals. The first formed flowers in ptl-1 mutants may have two or three petals, but the mean number per flower falls progressively such that most later arising flowers have none (Fig. 2A). The regions where petals normally arise are vacant. In the second mutant allele, ptl-2, flowers have more petals than ptl-1 on average initially, but the number falls to a similar baseline level (Fig. 2A). This difference between alleles may reflect the relative strength of the mutational changes, or it may be a consequence of the difference in their genetic background. The number of petals per flower is also influenced by temperature. Both ptl-1 and ptl-2 plants raised at 16°C have about one more petal per flower on average in early flowers compared with those raised at 22-24°C, but plants raised at 29°C have almost no petals in any flower (data not shown). As the two mutant alleles respond similarly, the initiation process they disrupt may itself be influenced by temperature.
Strikingly, the orientation of petals within the mutant flower may also be disrupted. In wild-type petals the concave adaxial surface normally faces inward (Fig. 1B), but in ptl mutants it may face any direction (e.g. Fig. 1C). Control experiments show that the direction of lateral curvature of the petal blade accurately reflects the petal’s orientation. In the wild type (Fig. 1D), epidermal cells of the concave (adaxial) surface are consistently more dome-shaped, with radial ridges that are more evenly spaced (Fig. 1F) than those of the convex (abaxial) outer surface (Fig. 1E). When abnormally oriented ptl petals are examined (Fig. 1G), the absolute association between curvature and surface structure remains. If petals now face outwards, the outer concave surface typically shows ridged conical-shaped cells (Fig. 1H) while cells of the inner convex surface are block-shaped and irregularly sculptured (Fig. 1I). Importantly, close examination of cell files in the claw of mature mutant petals reveals that they do not twist significantly. In ptl mutant flowers, the majority of mutant petals are still normally oriented, but about 15-20% are reversed, and around 20-25% face sideways, mostly medially (see Table 2). The patterns of disorientation are similar in earlier-and later-formed flowers (data not shown).
The form of petals that arise in ptl mutant flowers is also frequently abnormal (Table 1). About one quarter are trumpet-shaped or tubular (Fig. 1J). In these, the outer surface has cells characteristic of the normally abaxial surface, and the inner surface cells are of the adaxial type. Another frequent abnormality is the occurrence of thread-like organs in petal positions. These do not have petalloid surface cells, and they are thinner than the adjacent stamen filaments (Fig. 1K). Less frequently (<10%), petals arise that contain mosaic regions of stamenoid tissue, or occasionally a petal position may be occupied by a stamen. Despite these abnormalities, more than one third of second whorl organs are petals with the normal shape (Table 1), although they are often smaller, especially in later flowers. Overall the proportions of the abnormal types are similar in ptl-1 and ptl-2 mutants, but there are fewer stamenoid organs in the latter (Table 1).
Other organ whorls in ptl mutant flowers are much less affected, if at all. The first whorl sepals arise as the usual four per flower, although compared with the wild type they are more infolded, deeper in profile (Fig. 1A-C), and are sometimes fused along their edges towards the base (Fig. 1L), especially in ptl-2 mutants. Also, in nearly one in five sepals of ptl-1 flowers, carpelloid structures develop from their edges. These structures are mostly found in the central region, and include stigmatic papillae and ovules whose development is often arrested (Fig. 1M). In later-formed flowers, the medial sepal closest to the inflorescence stem occasionally contains patches of petal tissue along its flanks. Third whorl stamens, too, may show some abnormalities. These are manifest as an occasional additional stamen in either lateral or medial positions (seen in 9 out of 398 flowers), or a fused or bifid stamen with a single filament but with two anthers. The presence of additional stamens does not vary acropetally. The fourth whorl gynoecium is apparently normal.
Both ptl mutants are fully recessive, and the PTL locus was mapped at the top of chromosome 5, 0.5 map units above TERMINAL FLOWER1 (4 recombinant chromosomes out of 854).
A dominant modifier of the ptl mutant phenotype, pmd-1d, is present in Landsberg erecta
The ptl-1 mutant allele was isolated in the Columbia genetic background, and ptl-2 in C24. When either was crossed into Landsberg erecta, plants showing a marked increase in the number of petals per flower were observed among the F2 progeny. In the first formed flowers, between five and eight petals per flower were now seen (Fig. 3A,B), although the number decreased to two or three by the twentieth flower on the inflorescence (Fig. 2B).
To deduce the genetic basis of this increase, 275 F2 plants were raised, and 216 had four petals per flower and were indistinguishable from wild type, 38 had significantly more than four petals in the first flowers, while 21 had fewer than four. This is consistent with the increase being caused by an unlinked dominant modifier gene with an effect only in plants homozygous for ptl (the classes are not significantly different from a 12:3:1 ratio (χ2[2]=4.87, 0.05<P<0.1)). To confirm this, 15 of the ptl-1 F2 plants with more than four petals were allowed to self-fertilise. Of these, 7 generated F3 families in which all plants had more than four petals in early flowers, whereas 8 generated F3 progeny one quarter of which now had significantly fewer than four. The mean numbers of petals produced by these 7 presumed F2 homozygotes and 8 presumed heterozygotes had already been recorded for each of the first 20 flower positions, and the distributions were now compared and found not to differ significantly (data not shown). Thus the action of the modifier allele is fully dominant.
The gene was called PETAL LOSS MODIFIER (PMD), and the dominant pmd-1d allele was discovered in the Landsberg erecta strain. The Columbia and C24 strains are homozygous for the recessive PMD allele. To determine the modifier genotype of other strains, ptl-1 (in Columbia) was crossed with the Bensheim, Enkheim, Kiruna and Niederzenz strains. In each case, no modified plants were seen among F2 progeny, indicating that the modifying pmd-1d allele is restricted to Landsberg erecta. This strain also carries the recessive erecta (er) mutation, reducing plant height and pedicel elongation, but PMD is inherited independently of er, and the er mutation had no significant effect on the numbers of petals produced by unmodified ptl-1 flowers. In modified flowers, however, the average number of petals was slightly higher initially, and the rate of loss somewhat faster in er mutant plants compared with ER controls (data not shown).
The PMD locus was mapped utilising recombinant inbred lines derived from a cross between Lansdsberg erecta (pmd-1d) and Columbia (PMD) (Lister and Dean, 1993). The original ptl-1 PMD mutant line was crossed with a selection of 32 highly recombinant lines, and F2 progeny scored in each case to determine if the recombinant inbred line carried the PMD or pmd-1d allele. The results were then matched against the distribution of known molecular markers in the lines. In this way, the PMD locus was assigned to the middle of chromosome 1 between markers m235 and GAPB, and closer to the former.
The modifier boosts the number of petals, and affects their form but not their orientation
Counts of the first twenty flowers in plants from an F2 family segregating for both ptl-1 and pmd-1d revealed that while the mean number of petals is increased in modified ptl pmd flowers, the number falls away acropetally in a similar pattern to that seen in unmodified flowers (Fig. 2B). [Note that the number of petals in unmodifed mutant flowers in this segregating family is higher than reported earlier for the original unmodified mutant (Fig. 2A). This is likely due to the hybrid genetic background (Columbia and Landsberg erecta) of the present plants.] When there are more than four petals in a ptl-1 pmd-1d flower, the additional petals also arise from the alterni-sepalous regions such that some now contain two petals. These arise as inner and an outer petal (Fig. 1G), with the former almost always larger than the latter. Overall, the phenotypes at each of the four petal positions within a flower are independent. For example, when there are four petals in ptl pmd flowers, one position may have two petals whereas another may be vacant.
The modifier also influences the proportions of the various forms of mutant petals. Relatively few thread-like organs are present, and there is a consequent increase in the number of normal-shaped petals (Table 1).
The effect of the modifier is also strongly temperature dependent. The higher the temperature, the fewer petals seen for all flower positions within the inflorescence. For example, the first formed flower of modified mutant plants raised at 16°C had, on average, 7.3±0.3 petals, but 6.6±0.3 when raised at 22-24°C, and 4.8±0.5 when grown at 29°C. By the twentieth flower, the mean number of petals per flower was still 4.9±0.4 for the 16°C plants, whereas at 22-24°C it had fallen to 2.8±0.4 (Fig. 2B), and at 29°C there were only 0.2±0.2 petals per flower. Thus the rate of acropetal reduction is faster at higher temperatures. This is consistent with the process that is influenced by the modifier also being influenced by temperature.
Even though the modifier boosts the number of petals, it does not further disrupt their orientation (Table 2). 60% of petals are still normally oriented, and the other three orientations are represented in the same relative proportions. Because the number of petals that arise in ptl mutants is sensitive to pmd and to temperature, yet their orientation is relatively insensitive to pmd, we tested whether orientation is also temperature independent. This proved to be the case, as the spectrum of orientations in ptl pmd plants raised at 16°C was similar to that seen in plants raised at 22-24°C (Table 2).
Finally, the effects of pmd are apparently specific to the second whorl. Sepals and stamens of modified ptl mutants are similar in number and phenotype to those of unmodified plants, and the carpels remain normal.
Developmental defects in organ initiation and orientation
Next, the development of ptl-1 mutant flowers was examined to determine where and when petals arise, and to identify any changes in their pattern of growth compared with wild type (Hill and Lord, 1989; Smyth et al., 1990). No overt differences were detected up to developmental stage 4 (Fig. 4A,B). At stage 5 in the wild type, four petal primordia appear simultaneously between the sepals. These continue to grow slowly as small mounds through stages 6 (Fig. 4C), 7 and 8 (Fig. 4E,G), before rapid growth as tongue-shaped organs commences in stage 9 (Fig. 4I) and continues through stage 11 (Fig. 4K) to maturity (Fig. 1D).
In ptl-1 mutants, on the other hand, petal primordia are present sporadically, and they are more variable in size throughout their development (e.g. stages 6 (Fig. 4D) and 8 (Fig. 4F)). Counts of the number of primordia per petal position show a progressive increase throughout flower development in samples of both unmodified (ptl-1 PMD) and modified (ptl-1 pmd-1d) mutant flowers (Fig. 5), unlike the wild type where every position is occupied from stage 5. Thus it seems there is a staggered delay in the initiation of petal primordia, a conclusion supported from sections of individual flowers (e. g. Fig. 4M,N). It seems likely that the smaller petals seen in mature flowers arise later than the larger petals, and that their growth and maturation are correspondingly delayed, although they eventually reach the same end point of differentiation. The site of initiation of petal primordia is also somewhat disrupted in ptl mutants, especially in modified plants. For example, they may arise from the base of a sepal rather than between them (Fig. 4F,N). This is especially true of the outer petal in cases where two share the one position.
The abnormal orientation of ptl mutant petals can be seen to result from their abnormal insertion on the floral receptacle (Fig. 4H,J,L), and this can be traced back to at least stage 8 (Fig. 4H). In the wild type, petal primordia are flattened radially in relation to the flower primordium from the time they arise (Fig. 4G), whereas in ptl mutants flattened primordia can be seen in any one of a full range of other orientations from early in their development. Significantly, by careful examination of cell file patterns at the base of petals at all stages of their growth, it can be seen that these abnormal orientations are not the result of twisting or rotation (e.g. Fig. 4H,J,L).
Finally, trumpet-shaped petals apparently arise and grow as radially symmetric domes rather than from flattened primordia. These are apparent from stages 6-8, and from stage 9 some are seen to have begun tubular growth from their apical rim (Fig. 4L). Threads develop in a similar way except they are smaller, maintain solid, cylindrical growth, and do not differentiate petal-like cells.
Loss of organs is position dependent, not identity dependent
To test if disruption to PTL function is specific for petals, or localised to the second whorl where they normally occur, homeotic mutants that change organ identity were used. Firstly, third whorl petals were produced in place of stamens using the agamous-1 (ag-1) mutant (Bowman et al., 1991). Double mutants of ptl-2 with ag-1 (in pmd-1d background) were carefully observed, and among 119 flowers, 111 (93%) had six third whorl petals, closely similar to the number seen in 52 ag-1 single mutant controls (46, or 88%). (The other flowers had five or seven third whorl organs.) On the other hand, second whorl organs still displayed disruptions to their number and shape similar to those seen in single mutants. Thus organs with a petal identity are insensitive to the effects of ptl mutation if they arise outside the second whorl.
The second test involved changing the identity of second whorl organs to sepal-like organs using the apetala3-3 (ap3-3) mutant (Bowman et al., 1991; Fig. 6A). Observations of the first 20 flowers of ptl-1 ap3-3 double mutants (in pmd-1d background) revealed the same acropetal reduction in the number of second whorl organs as seen in matched ptl-1 single mutant flowers. The mean number fell from 7.4±0.2 in the first formed flowers to 2.5±0.4 by flower 20 in ptl-1 ap3-3 double mutants, compared with a reduction from 6.7±0.3 to 3.0±0.5 for equivalent flowers in ptl-1 single mutant plants (ap3-3 mutant flowers always have four second whorl organs). Thus the effects of ptl mutations are specific to organs that arise in second whorl positions, not to petals per se.
Orientation is further disrupted when ptl is combined with B function mutants
While observing double mutants of ptl-1 with the B function mutant ap3-3, a striking observation was made. Sepalloid second whorl organs are now almost always fully reversed (Fig. 6D-H; Table 3). In mature flowers, both the shape of the organs and their epidermal identities were consistently reversed (Fig. 6D-F compared with Fig. 6A-C). Again no evidence of twisting of the organs was seen, an observation supported by the pattern of expression of the normally abaxially expressed gene FILAMENTOUS FLOWER (FIL). In ptl-1 ap3-3 double mutant flowers, FIL is already expressed adaxially in the young second whorl organs by at least stage 8 (Siegfried et al., 1999).
The reversal in second whorl organ orientation seen in ptl-1 ap3-3 mutants is apparently the consequence of removing B function, as a similar change was also seen when ptl-1 mutants were combined with the null B function mutant pistillata-1 (pi-1) (Bowman et al., 1991; Hill and Lord, 1989; Table 3). Both ap3-3 and pi-1 mutants result in aggregation of third and fourth whorl organs into a central gynoecium. To test if this modification in floral architecture is responsible for the further effect on orientation, ptl-1 was combined with the weaker ap3 allele, ap3-1, in which third whorl organs are now mostly separate from those in the fourth whorl (Bowman et al., 1991). Clearly the re-establishment of third whorl organs does not fully ameliorate the B function disruption (Table 3). As a final test of the influence of inner whorls on orientation, they were converted from carpels to sepal-like organs by incorporating the ag-1 mutant into the ptl-1 B function double mutants. In these triply mutant plants, the second whorl organs were still almost always reversed (data not shown) so disruption does not depend upon the internal organs having a carpelloid identity. The same pattern of reversal was seen in internal flowers.
Finally, pairs of second whorl organs that occupy the one position may adopt the same orientation (e.g. Fig. 6H), but they do not necessarily do so. In fact, in ptl-1 pmd-1d flowers they adopted different orientations in 103 cases among 209 pairs observed (54%). The orientation of the smaller, outer organ is less often disrupted, especially in ptl B function double mutants. In 77 doubly occupied positions in ptl-1 pmd-1d ap3-3 flowers, for example, nearly all (98%) of the inner organs but only 57% of the outer organs showed the reversed orientation.
In conclusion, 30% to 40% of second whorl organs apparently arise in a sideways or reversed orientation in ptl mutants, but they are almost all fully reversed when B function is also removed.
DISCUSSION
Two different disruptions of second whorl organ development are seen in petal loss mutants, reduced number and variable orientation. That these have a different mechanistic basis is highlighted by their differential sensitivity to modifying influences. Whether or not an organ arises is dependent upon where the flower arises within the inflorescence (i.e. there is an acropetal effect), the presence of a dominant modifier gene pmd-1d, and the growth temperature. In contrast, organ orientation does not change acropetally, it is not influenced by the modifier, and it is not temperature dependant. However, orientation but not organ number is further disrupted when B function is removed. Thus initiation and orientation are separable phenomena, and they will be discussed separately.
Initiation
PTL regulates induction of second whorl organs
To account for our data, we propose that an inducer of petal initiation, supported by PTL function, is normally active within the floral primordium. A discrete number of cells respond to a threshold level of signal by initiating organs in four zones within the flower (Fig. 7A, left). In unmodified ptl mutants (Fig. 7A, centre), induction is disrupted in two ways. Firstly, cells are unable to respond as effectively to the inducer so that an active threshold is achieved only infrequently, and hence fewer petals are initiated. Secondly, the inducer’s sphere of action is somewhat enlarged, accounting for the observed disturbance to the sites of petal initiation. In modified ptl mutants (Fig. 7A, right), the only change proposed is a uniform boost to the level of induction without any changes to its variability or site of action. Thus the threshold is more often reached, and, owing to the expanded zone, two primordia may sometimes arise within the one zone. Weaker induction at the periphery of the flower would account for the apparently reduced number and later initiation of the smaller outer primordia in ptl pmd plants. Finally, the level of induction is markedly influenced by growth conditions, in that it falls progressively in later-arising flower primordia, and it is also reduced in flowers developing at higher temperatures.
Influence of differential growth within the floral meristem
Changes to the size and shape of the floral primordium in various mutants may result in changes to the number of floral organs that arise (see Introduction). In ptl mutants, however, if there are any changes to flower meristem growth, they are relatively subtle as preliminary size measurements revealed no major differences. Even so, the occasional congenital fusion of adjacent sepals in ptl mutant flowers may be the consequence of a slightly smaller floral meristem. The development of aberrant structures, particularly threads, in the second whorl of ptl mutant flowers, too, might be the result of them arising from a constrained input of cells. On the other hand, fusion of stamens primarily occurred when there were seven stamens in the flower, and this part of the flower meristem may be slightly larger.
One way ptl mutants may reduce second whorl initiation is by allowing ectopic AG function in this region. AG normally inhibits growth in the centre of the floral meristem. In apetala2 (ap2) mutants, however, it is ectopically expressed in the outer whorls (Drews et al., 1991), resulting in carpelloidy of the first whorl and the loss of almost all second whorl organs. That the latter is the result of AG function is inferred from organ restoration when AG is also inactive in ap2 ag double mutant plants (Bowman et al., 1991). However, this is unlikely to be the primary cause of petal loss in ptl mutants. When the null mutant ag-1 was bred into ptl-1 or ptl-2 mutant plants, the mean number of second whorl organs per flower increased somewhat, but it was not restored to the normal four per flower, the number of organs decreased acropetally, and small petals and tubular petals continued to arise (results not shown). In addition, we could detect no differences in the pattern of AG expression in young ptl mutant flowers compared with the wild type (M. E. G., unpublished).
Role of the PETAL LOSS MODIFIER gene
Two properties of pmd-1d requiring explanation are its dominance and the lack of detectable effects in wild-type plants. The pmd-1d allele is restricted to the Landsberg erecta background, being absent in the other six ecotypes tested, suggesting that the dominant form is a mutant or variant allele. If so, it may result from a gain of function (although we cannot exclude the possibility that its function has been lost in all strains except Landsberg). In the context of our model (Fig. 7), PMD may be a component of the induction process, with the frequency of the outcome (petal initiation) being amplified by the presence of one or more pmd-1d alleles. The dominance of pmd could result from its now constitutive activation of a signalling pathway, for example. Amplification might also occur in wild-type PTL plants, but the threshold of petal initiation would already have been met, and the site of initiation restricted such that only one organ can arise at each location. It is also possible that once a petal primordium has begun to form, PTL function may directly prevent the triggering of further primordia.
Role of boundaries within the floral meristem
Studies of cell lineages have shown that boundaries between whorls are established relatively early in the development of Antirrhinum flowers (Vincent et al., 1995). One gene that may play a specific role in setting up petal whorl boundaries is FIMBRIATA (FIM) of Antirrhinum and its Arabidopsis ortholog UNUSUAL FLORAL ORGANS (UFO; Ingram et al., 1995). After activating B function organ identity genes, UFO expression is subsequently limited to a small region immediately between the sepals at the base of each developing petal. The number of petals in ufo mutant flowers is reduced somewhat, petal/sepal and petal/stamen mosaics are frequent, and filamentous organs are common in petal positions. Thus there are some parallels between ufo and ptl mutant phenotypes, although the orientation of second whorl organs is apparently not disrupted in ufo flowers.
One gene apparently involved in setting boundaries between adjacent organs within a whorl is the NO APICAL MERISTEM (NAM) gene of Petunia (Souer et al., 1996). In developing flowers, NAM expression encircles the stamen primordia, and in nam mutants additional petals arise adjacent to each stamen primordium as well as between them. The setting of boundaries between organs may also be the function of two CUP-SHAPED COTYLEDON (CUC) genes in Arabidopsis. In cuc1 cuc2 double mutants, fusion of adjacent sepals and adjacent stamens is frequent (Aida et al., 1997), and, interestingly, CUC2 also encodes a protein related to NAM. It will be important to observe the phenotype of mutants that combine ptl with ufo or cuc to test if PTL shares organ boundary functions with these genes.
Orientation
PTL regulates organ orientation
In the present study, we distinguish two aspects of organ development – polarity and orientation. We follow the conventional definition of polarity as the abaxial-adaxial differentiation that occurs within flattened organs. However, we define organ orientation as the angle adopted by flattened, polarised organs in relation to the centre of the floral meristem. While a range of genes has been identified that regulate organ polarity (see below), to our knowledge PTL is the first to be described that influences organ orientation.
In ptl B function double mutants, second whorl organs are usually reversed, and it could be argued from this that PTL simply controls polarity. That is, such organs may arise in the normal orientation but their abaxial and adaxial surfaces are interchanged. However, the presence of organs showing any one of a continuous 360° range of abnormal orientations, as is seen in ptl single mutants and in ptl ap3-1 double mutants, strongly suggests that orientation rather than polarity is disrupted.
Among genes known to regulate polarity, the PHANTASTICA (PHAN) gene of Antirrhinum is apparently involved in establishing adaxial identity (Waites and Hudson, 1995). In phan mutants, leaves and petals are filamentous, and the surface cells are usually only of the abaxial type. The PINHEAD/ZWILLE and ARGONAUTE1 genes of Arabidopsis, too, may share functions in controlling adaxial tissue fate, at least within petals (Lynn et al., 1999). On the other hand, in the dominant phabulosa (phb) mutant of Arabidopsis, the leaves and petals are also radialised but are now covered with adaxial cells suggesting that abaxiality has been affected (McConnell and Barton, 1998). Recently, a group of new abaxial identity genes has been described in Arabidopsis (Siegfried et al., 1999). Several YABBY family genes, including FIL, are preferentially expressed in abaxial regions of all lateral organ primordia from an early age, and the effects of loss or gain of their function are consistent with an abaxial identity role.
Even though orientation and polarity can be distinguished, they could arise from a common mechanism. For example, Waites and Hudson (1995) propose that the lateral outgrowth of an organ primordium is the consequence of the sensing of an already-established adaxial-abaxial boundary within it. This accounts for the radial growth of organs whose adaxial (or abaxial) identity has been disrupted. Under this scenario, PTL would control organ orientation by defining the position of the abaxial-adaxial boundary in the petal anlagen before flattened growth begins. Furthermore, the occasional production of tubular petals in ptl mutants might reflect displacement of such a boundary.
Signalling of petal orientation within the flower meristem
In general, to account for our observations we propose that the orientation of petal primordia is somehow signalled to precursor cells (Fig. 7B, left), and that the PTL gene product plays a role in the perception of, or response to, this signal. In ptl mutants, the signal is perceived to be generally weaker and variable, so that the orientation of some petals is distorted (Fig. 7B, centre left), perhaps partly as a consequence of expansion in the zone of cells within which primordia may arise.
In B function single mutants, where sepals arise in place of petals, signalling may be reduced but it remains effective (Fig. 7B, centre right). In ptl B function double mutants, however, the signal is usually ineffective, and most second whorl organs are now reversed (Fig. 7B, right). Thus the role of B function in signalling is fully covered by that of PTL function, although not vice versa. This reversal in organ orientation may reflect the influence of another signal, or it may be the default orientation adopted in the absence of signal. Such a default orientation has also been proposed to account for the reversed orientation of bristles in Drosophila mutants that lack Frizzled activity (Shulman et al., 1998).
The existence of signals that control leaf orientation has already been deduced from surgical experiments on the shoot apical meristem (e.g. Sussex, 1955). A range of different cuts disturbs the orientation of newly arising leaves in a pattern consistent with a controlling signal emanating from the meristem apex. It is possible that an equivalent signal controls the orientation of second whorl organs and originates from the centre of the flower meristem. If so, we have shown that it is unaffected if the third and fourth whorl organs are fused together, or if the central organs are no longer carpelloid.
In conclusion, discovery of the PTL gene has provided us with an entry point into gaining a fuller understanding of two fundamental processes in floral architecture, initiation of individual organs within a floral whorl, and specification of their orientation within the flower. The roles of a proposed inducer of petal initiation, and of possible signals that determine their orientation, will be important aspects of future work.
ACKNOWLEDGEMENT
We thank John Alvarez for providing the ptl-1 mutant and for his invaluable help, John Bowman for communicating information about YABBY genes before publication, and Gunta Jaudzems for excellent microscope facilities. M. E. G. was supported by an Australian Postgraduate Research Award. A. da S. C. was the recipient of a fellowship from CAPES (4647/88-12) and CNPq (260147/92.3) of the Brazilian Government, and is grateful to Dr Enno Krebbers and Dr Natasha Raikhel for support and encouragement and Dr Bart den Boer for his advice at early stages of the work. This work was funded in part by Australian Research Council Grant A19801319, and an ARC Small Grant (1996).