Mutations at the CLAVATA2 (CLV2) locus of Arabidopsis result in enlarged shoot and flower meristems, as well as alterations in the development of the gynoecia, flower pedicels, and stamens. The shoot and flower meristem phenotypes of clv2 mutants are similar to weak clv1 and clv3 mutants. We present genetic analysis that CLV2 may function in the same pathway as CLV1 and CLV3 in the regulation of meristem development, but function separately in the regulation of organ development. We also present evidence that clv2 phenotypes are altered when the mutants are grown under short-day light conditions. These alterations include flower-to-shoot transformations, as well as a nearly complete suppression of the flower phenotypes, indicating that the requirement for CLV2 changes in response to different physiological conditions. The stm-1 mutation dominantly suppresses clv2, and clv2 mutations suppress the strong stm-1 allele, but not the weak stm-2 allele.
All above-ground organs are derived from the shoot meristem. The shoot meristem accomplishes continuous organ formation by maintaining a pool of undifferentiated cells at the center of the meristem and directing appropriately positioned descendant cells toward organ formation and eventual differentiation (Clark, 1996). Several genes have been described that appear to specifically regulate shoot meristem development in Arabidopsis. Mutations at the CLV1 and CLV3 loci accumulate a massive population of undifferentiated cells at the shoot meristem, while mutations at the WUS and STM loci result in shoot meristems that fail to maintain organogenesis (Barton and Poethig, 1993; Clark et al., 1993, 1995; Laux et al., 1996). The flower meristem has long been thought to be a modified shoot meristem, and clv1, clv3, stm and wus mutations exert similar effects on flower meristem development as they do on shoot development (Weigel and Clark, 1996). Genetic interactions between clv1, clv3, stm and wus mutants suggest that these genes may all function in a related pathway (Clark, 1997). CLV1 and CLV3 appear to function in the same pathway (Clark et al., 1995), while wus mutants are epistatic to both clv1 and clv3 mutants (Laux et al., 1996). STM appears to function in a balanced, competitive manner with CLV1 and CLV3 (Clark et al., 1996), raising the possibility that these genes act on a common downstream target, perhaps the WUS gene.
Two models have been proposed for the action of the CLV1 and CLV3 loci. One model predicts that they regulate the rate of division of the central undifferentiated cells. For example, the clv1 mutant phenotype could be explained by postulating that the undifferentiated cells divide more rapidly in the mutant, implying that the role of CLV1 is to repress cell division. Another hypothesis is that these genes regulate the undifferentiated/differentiated state of the shoot meristem cells. Under this hypothesis, CLV1 would act to promote the transition of cells from an undifferentiated state toward organ formation. Thus, in a clv1 mutant, cells on the flanks of the meristem that would normally contribute to organ primordia would often remain undifferentiated, enlarging the shoot meristem.
CLV1 codes for a putative receptor-kinase and is specifically expressed at the shoot and flower meristem (Clark et al., 1997). Thus, CLV1 likely relays positional information to regulate the development of cells at the shoot meristem. STM codes for a homeodomain-containing protein and likely acts as a transcription factor (Long et al., 1996). STM is also expressed in a central region of the shoot meristem.
Understanding the control of meristem development requires detailed analysis of critical regulators. clv2 mutants have been superficially described (McKelvie, 1962; Koornneef et al., 1983; Griffin, 1994), indicating a possible role for CLV2 in regulating meristem development. We have carried out a detailed phenotypic and genetic analysis to determine the role that CLV2 plays in meristem development, and whether CLV2 works on the same pathway as CLV1 and CLV3.
MATERIALS AND METHODS
The clv2-1 and clv2-4 alleles have been previously described (Koornneef et al., 1983; Pogany et al., 1998). The clv2-3 allele was identified in a screen of fast neutron mutagenized wild-type Columbia seeds purchased from the Lehle Seed company. The clv2-2 and clv2-5 alleles were identified in screens of T-DNA transformed wild-type Wassilewskija (Ws) (Forsthoefel et al., 1992). The clv2-5 allele was kindly provided by Robert Williams.
Seeds were sown on a 1:1:1 mix of top soil:perlite:vermiculite and imbibed for 7 days at 4°C. Plants were grown at 22°C under approximately 800 foot-candles of constant cool white fluorescent light. Plants were fertilized once a week. Plants in short-day conditions were grown at 22°C under approximately 800 foot-candles of constant cool white fluorescent light for 8 hours each 24 hour period.
Tissue and image processing
Scanning electron microscopy (SEM) was carried out as described (Bowman et al., 1989), except the Hitachi S3200N SEM allowed images to be collected digitally. In situ mRNA hybridization was carried out as described (Clark et al., 1997), except that light-field and dark-field images were taken separately and combined in Adobe Photoshop 4.0 after digitization. Slides were scanned and digitized using a Polaroid SprintScan35. Brightness, contrast and color balance were adjusted using Photoshop.
We have examined the phenotypes of five clv2 alleles (Table 1; Materials and methods). clv2-1 and clv2-4 were isolated from the Landsberg erecta (Ler) background (Koornneef et al., 1983; Pogany et al., 1998). clv2-2, which was isolated from the Ws background, and clv2-3, which was isolated from the Columbia background, were backcrossed into Ler before analysis. All alleles were recessive. Overall, clv2-1, clv2-3 and clv2-4 exhibit very similar phenotypes, while clv2-2 exhibited a weaker phenotype. clv2-5, which was isolated from T-DNA mutagenesis in the Ws ecotype, was not backcrossed into Ler and thus could not be directly compared to the other alleles. Linkage of clv2 alleles to CLV1, STM and AP1 were consistent with the previously published map position (Koornneef et al., 1983).
clv2 mutants affect the shoot and flower meristem
All clv2 alleles exhibited an enlarged inflorescence shoot meristem when compared to wild type (Fig. 1A-I). While not remarkably larger in circumference, the clv2 shoot meristems were much taller than those observed at the apices of wild-type plants. clv2 shoot meristems are enlarged during vegetative development as well (Griffin, 1994; data not shown). Under normal growth conditions (Materials and methods) fasciation of the meristem was not observed in clv2 mutants (but see short-day growth conditions below). This phenotype is similar to that observed in weak clv1 (clv1-6, clv1-5) and clv3 (clv3-3) alleles. As with clv1 and clv3 mutations, the size of the clv2 shoot meristems varied, with the shoot meristems of the weak clv2-2 allele often observed similar to wild type (data not shown), and the shoot meristems of the other clv2 alleles occasionally enlarging significantly in circumference (Fig. 1I). clv2 mutations similarly affected the flower meristem. The number of organs generated by the flower meristem was increased (Fig. 2). Under standard growth conditions, the number of carpels generated in each clv2 flower was almost twice that of wild type, while the number of stamens was also slightly increased. The sepals and petals were only rarely affected. This is similar to the number of organs observed in weak clv1 mutants (Clark et al., 1993). In addition, clv2 flowers developed an additional whorl of organs in the center of the flower meristem. At the time of carpel initiation (stage 6; all stages according to Smyth et al., 1990), the carpel primordia in clv2 mutant flowers initiated in a ring around the center of the flower meristem, instead of terminating the flower meristem as in wild type (Fig. 1K; see also Smyth et al., 1990). This results in the development of a gynoecium interior to the gynoecium formed by the whorl 4 carpels (Fig. 3J). All of these features are also similar to observations of weak and intermediate clv1 and clv3 mutants (Clark et al., 1993).
clv2 mutants affect organ development
Mutations at clv2 also affected the development of several organ types. The flower pedicels in clv2 mutants were about 50% longer than wild type (Fig. 4). While wild-type pedicels were approximately 5 mm in length, pedicels in clv2 plants averaged approximately 7.5 mm for the stronger alleles. These measurements were all performed in the Landsberg erecta background. erecta mutations have been demonstrated to reduce pedicel length (Torii et al., 1996), indicating that clv2 suppressed this erecta phenotype.
The morphology of the gynoecia were also altered in clv2 mutant backgrounds. Strikingly, the portion of the gynoecium covered by valves with developing ovules can be greatly reduced in clv2 flowers. In wild-type gynoecia, the valves extend from shortly below the stigmatic tissue to the base of the gynoecium, with only a very short segment between the bottom of the valves and the attachment site of sepals, petals and stamens (Clark and Meyerowitz, 1994; Fig. 3A). In the most severely affected clv2 flowers, the gynoecium completely lacked valves (Fig. 3B). Some clv2 flowers lacked valves only on the basal portion of the gynoecium (Fig. 3D), while other clv2 flowers developed multiple valves that extended to the base of the gynoecium, with a valveless region extending between these valves from the base to the apex of the gynoecium (Fig. 3C,E). The morphology of the cells in the valveless region is similar to the morphology of cells between the valves in wild-type gynoecia (Fig. 3F-H). The cells typical of the style region, which is normally found between the top of the valves and the stigmatic tissue in wild-type gynoecia (Bowman, 1994), are only found in the region immediately adjacent to the stigmatic tissue in the mutants as well (Fig. 3H,I). The extent of the valveless phenotype was quite variable in clv2 flowers, with many flowers not exhibiting the phenotype. The percentage of flowers with significant valveless phenotypes are indicated in Table 2.
The valveless phenotypes was only rarely observed in weak clv1 or clv3 mutants (Table 2, and data not shown). However, a significant number of gynoecia of strong clv1 and clv3 alleles did lack valves (Table 2). This is discussed in more detail under Genetic Interactions, below. Mutations at the ETTIN locus also result reduced coverage of the gynoecium with valve tissue (Sessions and Zambryski, 1995). In many clv2 mutant flowers, stamens developed abnormally. Wild-type stamens develop an anther with two locules on the end of a stalk or filament (Clark and Meyerowitz, 1994; Fig. 5A). clv2 mutant flowers developed anthers in which the locules were partially fused (Fig. 5C), or only contained a single locule (Fig. 5B). clv2 mutants also developed stamens that lacked anthers. 11% of clv2-1 stamens, 12% of clv2-2 stamens and 10% of clv2-3 stamens were antherless (over 250 stamens analyzed per genotype). This compared with previously published analysis of clv1-1, clv1-5 and clv1-4 flowers in which 4%, 5% and 6% of stamens, respectively, developed without anthers (Clark et al., 1993). The antherless stamens often contained anther-like cells at their apex whether they were found in clv1, clv2, clv3 single mutants or various clv double mutants (Fig. 5D), suggesting that perhaps these are extremely reduced anthers.
A number of novel features of the development of clv2 plants was revealed when the mutants were grown under short-day light conditions (SD; 8 hours of light per 24 hour period). Wild-type Ler plants and most clv2 alleles flower very late under SD conditions when compared to growth under continuous light (Martinez-Zapater et al., 1994; Table 3). The weak clv2-2 allele, however, flowered significantly earlier than wild type. However, clv2-2 was originally isolated in the Ws genetic background and the difference in flowering time may be the result of Ws-specific genetic factors that were not lost during the backcrosses of clv2-2 into the Ler background.
One clear effect of growth under SD conditions was the suppression of the clv2 flower phenotype (Fig. 6). The number of organs in the flowers of most clv2 alleles grown under SD were almost identical to wild type. No additional organs were formed interior to the whorl 4 gynoecium, and the valveless phenotype was not observed (data not shown). The weak clv2-2 allele was not as strongly suppressed, with many flowers developing additional carpels (Fig. 6). However, not all clv2 phenotypes were suppressed under SD conditions. The shoot meristem remained enlarged compared to wild type. In fact, the shoot meristems of clv2-1 and clv2-3 plants occasionally fasciated under SD conditions (Fig. 1J). In addition, the length of pedicels in clv2 plants grown under SD conditions were identical to those of clv2 plants grown under continuous light (Fig. 6).
An equally dramatic phenotype was observed in clv2-2 and clv2-5 mutants grown under SD conditions. In these plants, flowers along the inflorescence were occasionally converted into shoot meristems. In approximately half of the clv2-2 plants examined, at least one flower was converted into a shoot meristem. In plants that were affected, usually multiple flowers were converted to shoot meristems (Fig. 7A). These numbers do not include plants in which the first several flowers produced after the cauline leaves were transformed into shoots, as we have observed that this phenomenon occurs, albeit rarely, even in wild-type plants grown under continuous light. Instead, we observed that clv2 flowers at any length along the inflorescence could become transformed into shoots, even after the production of ten or twenty normal flowers. Flowers on secondary and tertiary inflorescences could also be transformed into shoots. The flowers that did develop as shoots were not subtended by a leaf; however, they did produce several cauline leaves before initiating flowers. The clv2-5 allele, which is interestingly also from the Ws background, is the only other clv2 allele that exhibited this phenotype, but it did so at a lower frequency. Only 3 of 36 clv2-5 plants examined exhibited flowers transformed into shoots. This phenotype was never observed in wild-type plants.
The other clv2 alleles (clv2-1, clv2-3 and clv2-4) did not exhibit the dramatic transformation seen occasionally in clv2-2 and clv2-5 plants. They did, however, display a more subtle transformation of flowers into shoots when grown under SD conditions. The flowers that were produced shortly after the transition to flowering exhibited phenotypes similar to weak leafy mutants (Weigel et al., 1992). This consisted of the failure of organs to initiate in clear whorls, especially for the sepals and petals, and additional sepals were formed at the expense of petals (Table 4). This was only apparent on the earliest flowers formed after the transition to flowering, and the affect was reduced acropetally. leafy and apetala1 mutants also exhibit a similar acropetal reduction in shoot-to-flower transformation (Weigel et al., 1992; Bowman et al., 1993). Interestingly, clv2-1 enhances the partial flower-to-shoot transformation observed in the weak leafy-5 allele grown under continuous light (data not shown).
clv1-1 and clv1-6 mutants, which exhibit defects in shoot meristem size and flower organ number similar to that observed in clv2 alleles, were also grown under SD conditions. These clv1 alleles exhibited none of the changes in phenotype that clv2 mutants underwent. The number of organs initiated by the flower was unchanged (Fig. 6, and data not shown), and no flower-to-shoot transformations were observed. clv3-1 and clv3-2 mutants also did not exhibit the suppression of flower organ number or flower-to-shoot transformations when grown under SD conditions(data not shown).
To determine the developmental period during which clv2 flowers are sensitive to SD conditions, clv2-3 and clv2-4 plants were shifted from SD to continuous light after the transition to flowering. Thus, the earliest flowers formed on these plants developed entirely under SD conditions, while much later arising flowers developed entirely under continuous light. We observed a very marked transition from the suppressed phenotype to the clv2 phenotype, indicating that the developmental period during which clv2 flowers are sensitive to the light conditions is brief (Fig. 7B). This presumably corresponds to stages 2 to 5, when the flowers are initiating organs. Interestingly, the flowers formed during the transition appeared to be very strongly affected and displayed strong valveless phenotypes.
clv2 mutants result in expanded CLV1 expression
CLV1 expression in wild-type plants is restricted to a central region of the shoot meristem and stage 2 through stage 5 flower meristems (Clark et al., 1997). Because clv2 mutants result in phenotypes similar to clv1 mutants, we considered the possibility that clv2 mutant phenotypes were the result of reductions in the pattern or level of CLV1 expression. In situ RNA hybridization experiments were performed on sections of both wild-type and clv2-1 inflorescence tissue. The results as shown in Fig. 8 revealed an expanded pattern on CLV1 expression in the enlarged clv2 shoot meristem. As the signal in these images was achieved under identical conditions, it would appear that the level of CLV1 mRNA accumulation is not reduced in the clv2 mutant plants.
clv2 displays complex interactions with clv1 and clv3 mutants
An important question is whether CLV2 acts in the same pathway as CLV1 and CLV3. clv2 mutants affect the development of the shoot and flower meristem in a manner similar to that of weak clv1 and clv3 mutants, yet clv2 mutants also display a number of phenotypes not previously observed in clv1 or clv3 mutants. To test whether this resulted from CLV2 functioning in a separate pathway from CLV1 and CLV3, we generated doubles mutants of clv1 and clv2 as well of clv2 and clv3.
We generated double mutants of the strong clv1-4 and clv3-2 mutations with various clv2 alleles. While meristem size in these strong alleles is variable, we did not detect any difference in meristem size between the double mutants and the strong single mutants (Fig. 9). However, mutating clv2 in a weak or intermediate clv1 or clv3 mutant background (clv1-1, clv1-6, clv1-5, clv1-7, clv3-1) did lead to an enlargement of the shoot meristem (Fig. 9, and data not shown). clv1-7 is the weakest clv1 allele and develops a shoot meristem even smaller than that of clv1-6 (data not shown). The clv1-7 clv2-2 double mutant lead to the development of massively enlarged shoot meristems comparable to strong clv1 or clv3 single mutants.
The phenotypes of clv1 clv2 and clv2 clv3 flowers revealed a complex interaction in the flower. The number of organs initiated by the double mutant flowers were similar to (or perhaps less than) those initiated by strong clv1 or clv3 single mutants (Table 5). In this aspect, clv1-4 and clv3-2 were epistatic to clv2. However, the effects of clv2 on organ development was additive with clv1 and clv3. The pedicels of clv1 and clv3 plants were found to be slightly longer than those of wild type and this was unchanged in the clv1 clv3 double mutant. clv2 clv3 double mutants developed dramatically longer pedicels than in clv2 or clv3 alone. Similarly, strong clv1 and clv3 mutant alleles also give rise to the valveless phenotype observed in clv2 mutants. However, the clv1 clv2 and clv2 clv3 double mutant plants displayed a higher frequency and more extensive valveless phenotype compared to either single mutant, with nearly every gynoecium exhibiting the phenotype (Table 2). In fact, the slight reduction in mean carpel number in the clv1 clv2 and clv2 clv3 flowers compared to clv1 and clv3 single mutants, respectively, may be due to the high frequency of missing valves on the double mutants, as we use the number of valves to determine carpel number.
clv2 displays complex interactions with stm mutants
STM and CLV1/CLV3 appear to function in a competitive manner in the regulation of shoot meristem development. This is based on the genetic interaction displayed between clv1/clv3 and stm mutations (Clark et al., 1996). clv1/clv3 mutations dominantly suppress the strong stm-1 allele, while stm-1 dominantly suppresses clv1/clv3 alleles. The clv1/clv3 stm double mutants show a loss of meristem homeostasis (i.e., the meristems vary greatly in size, they often become enlarged, but always terminate prematurely). clv1/clv3 mutations also suppress the weak stm-2 allele: the clv3-2 stm-2 double mutant, for example, develops dramatically fasciated stems and each inflorescence produces hundreds of organs. The reduction of organs in the stm-2 flowers is also partially suppressed by clv1/clv3 mutations.
We tested the ability of the stm-1 allele to dominantly suppress clv2, as well as the ability of clv2 to suppress both the strong stm-1 and weak stm-2 allele. To determine whether stm-1 can partially suppress clv2-2 in a dominant manner, we scored the progeny of a plant homozygous for clv2-2 and heterozygous for stm-1. As expected, one-quarter of the progeny germinated with the stm phenotype and were presumably homozygous for stm-1 (however, see below for the adult phenotypes of these double mutants). The remaining progeny were scored for potential suppression and the genotypes of these plants at the STM locus (stm-1 / + or + / +) were determined by testing progeny (Table 6). While the results indicate that stm-1 dominantly suppresses the clv2-2 phenotype, the correlation of stm-1 heterozygosity and clv2-2 suppression is not complete. This indicates that stm-1 suppression of clv2-2 is not as effective as stm-1 suppression of clv3-1, where the phenotype was completely correlated with genotype (Clark et al., 1996).
The clv2-2 stm-1 and clv2-3 stm-1 double mutants were also analyzed. These plants frequently developed the ‘rescued’ phenotype characteristic of clv1 stm-1 and clv3 stm-1 double mutants (data not shown). Namely, postembryonic growth consisting of rosettes of leaves and inflorescences bearing several flowers were observed. These structures were indicative of the development of meristems in the clv1 and clv3 double mutants with stm-1. Thus, clv2 partially suppressed the stm-1 phenotype. However, no suppression was observed in the clv2-1 stm-2 double mutants. These plants were indistinguishable from stm-2 single mutants and developed similar numbers of organs per flower (data not shown). Thus, stm-2 was epistatic to clv2 mutants.
We have examined the phenotypes of clv2 mutants in order to understand the role that CLV2 plays in regulating meristem development in Arabidopsis. Our observations indicate that CLV2 plays a role not only in regulating shoot and flower meristem development, but also the development of several organ types.
clv2 phenotypes/CLV2 functions
The effects of clv2 mutations on meristem development are apparent in the enlarged shoot meristems, as well as the additional organs and the additional whorl of organs that are initiated in clv2 flowers. These phenotypes are similar to that observed in weak clv1 and clv3 mutants. Thus, like CLV1 and CLV3, CLV2 is required to prevent the accumulation of undifferentiated cells at the shoot and flower meristem. The same models proposed for CLV1/CLV3 action (inhibition of undifferentiated cell division, and promotion of organ formation) could also be proposed for CLV2. clv2 mutant plants also develop abnormal organs, namely, elongated pedicels, gynoecia lacking valves and reduced anthers. Evidence outlined below suggests some of these organ defects are not indirect consequences of alterations in meristem structure.
The requirement for CLV2 during flower meristem development is dependent on physiological conditions. During growth in continuous light, CLV2 is required for normal development of the flower meristem. However, when grown under short-day (SD) light conditions, clv2 mutant flowers are nearly identical to wild type. This could be because the change in physiological state associated with SD growth conditions results in an altered flower meristem that no longer requires CLV2 function. An alternative explanation is that a redundant or compensating factor is active, or active at a higher levels, only under SD conditions. The fact the clv2 mutants develop normal flower meristems with elongated pedicels under SD conditions is direct evidence that CLV2 regulation of pedicel length is independent of its regulation of flower meristem development. clv2 flowers under SD conditions, however, do not exhibit valveless phenotypes or reduced stamens, indicating that an alteration of the flower meristem is at least a prerequisite for these developmental abnormalities.
clv2 mutants also exhibit a novel flower-to-shoot transformation specifically under SD growth conditions. For the Ws-derived clv2-2 and clv2-5 alleles, this consists of late-arising flowers occasionally converted completely into functional shoot meristems. We are unaware of any other mutation that gives rise to a similar phenotype. While mutations such as leafy and apetala1 covert flowers into shoots, they affect all of the flowers of the inflorescence, and generally exhibit partially conversion. For the other clv2 alleles, a more subtle flower-to-shoot transformation, reminiscent of weak leafy alleles, affect the flowers that are initiated shortly after the transition to flowering. That these flower identity phenotypes are only observed under SD conditions is likely the result of the reduced level of flower induction generally believed to exist under SD conditions.
A critical question is how does CLV2 fit into the hierarchy of genes known to regulate shoot and flower meristem development. clv2 mutants, overall, are similar to clv1 and clv3 mutants, although there are several differences. First, all clv2 alleles have phenotypes relatively weak compared to clv1 and clv3 alleles. One possible explanation is that the clv2 alleles are all partial-loss-of-function alleles. However, three alleles all have very similar phenotypes (clv2-1, clv2-3 and clv2-4), while there is often great variability between partial-loss-of-function alleles (see for example, the various partial-loss-of-function clv1 alleles; Clark et al., 1993, 1997). In addition, the clv2-5 allele is linked to a T-DNA insertion (S.-H. Jeong and S. E. C., unpublished) and many, albeit not all, T-DNA insertions result in null alleles. Another difference between clv2 mutants and clv1/clv3 mutants is that clv2 alleles exhibit more dramatic organ defects (pedicel elongation, valveless gynoecia, reduced stamens) than comparable clv1 alleles. Finally, clv1/clv3 mutants are not suppressed by SD growth conditions, nor do they exhibit flower-to-shoot transformations. Thus CLV2 could function in a separate pathway from CLV1/CLV3 entirely, or could function with CLV1/CLV3 to regulate meristem development and separately to regulate organ development.
To distinguish between these possibilities, clv1 clv2 and clv2 clv3 double mutants were generated and analyzed. The data are consistent with CLV2 functioning with CLV1/CLV3 in the regulation of shoot and flower meristem development. This is based on the observations that the strong clv1 and clv3 alleles were epistatic to clv2 in terms of the size of the shoot meristem and the number of organs initiated by the flower meristem. However, the clv2 clv3 double mutant was additive in terms of pedicel length and both the clv1 clv2 and clv2 clv3 double mutants were additive in terms of valveless gynoecia, consistent with a model in which CLV2 functions in a separate pathway in the regulation of organ development. If CLV2 functions in the same pathway as CLV1/CLV3, then it should exhibit similar interactions with stm mutations. clv1/clv3 and stm dominantly suppress each other’s phenotypes, and the same is true for clv2 and stm. However, clv2 is unable to suppress a weak stm allele. This is perhaps related to the weak clv2 phenotypes. As CLV1 likely acts as a signaling molecule and STM likely functions as a transcription factor (Long et al., 1996), it will be interesting to determine how the CLV2 gene product fits into this developmental hierarchy.
This work is supported by grant IBN-9506952 from the National Science Foundation – Developmental Mechanisms Program. The scanning electron microscope used was acquired under grant BSR-83-14092 from the National Science Foundation. We thank David Bay for photographic assistance, and Keiko Torii for critical reading of the manuscript. We thank Sang Ho Jeong for determining the frequency of flower-to-shoot transformations in clv2-5.