Leaf polarity and meristem formation in Arabidopsis

The shoot apical meristem (SAM) of angiosperm plants is the site at which new leaves and stem are made. The primary SAM is formed during embryogenesis and gives rise to the main axis of the plant. New SAMs, axillary SAMs, develop on the main axis and give rise to branches. These SAMs usually develop in the axils of leaves; the axil is the junction between leaf and stem. The subtending leaf may be required in some species for axillary SAM formation: when the subtending leaf is surgically removed the associated axillary SAM often fails to form (Snow and Snow, 1942). In Arabidopsis, two observations support a connection between the subtending leaf and the axillary meristem. First, the axillary SAM appears to arise on the adaxial leaf base (Talbert et al., 1995). Second, the subtending leaf and its associated axillary bud are clonally related (Furner and Pumfrey, 1992; Irish and Sussex, 1992).

One model for the development of axillary SAMs (reviewed by Steeves and Sussex, 1989) proposes that fragments of the main meristem, so called detached meristems, remain associated with each axil. These detached meristems become activated and form buds when the leaf and its associated axil are some distance from the main SAM. In this model, SAM fate is acquired only once in the development of the plant. Alternatively, axillary SAMs may form from cells that have lost SAM identity, partially differentiated, and then been instructed to regain SAM fate.

The first indication of leaf primordium formation in Arabidopsis occurs while the presumptive leaf primordium resides entirely within the SAM. Loss of SHOOT MERISTEMLESS (STM) expression in the presumptive leaf distinguishes it from other regions of the SAM (Long et al., 1996). Subsequently, these groups of cells grow outward from the SAM as small bumps, or leaf primordia. As the leaf develops, the adaxial side of the primordium (the side towards the center of the plant) grows more than the abaxial side. This unequal growth causes the leaf to bend outward and away from the long axis of the plant such that the adaxial side of the leaf primordium becomes the top of the leaf and the abaxial side of the leaf primordium becomes the bottom of the leaf.

The Arabidopsis leaf consists of a short petiole, connecting the leaf to the stem, and an entire, flattened blade. The leaf blade is polarized along its ad/abaxial axis (Telfer and Poethig, 1994). The adaxial epidermis is glossy, dark green and trichome-rich while the abaxial epidermis is matte, grey-green and, especially in the early leaves, trichome-poor. Internal tissues are also polarized: a layer of closely packed palisade cells underlies the adaxial epidermis while a loosely packed layer of spongy mesophyll cells lies adjacent to the abaxial epidermis. In addition, within the vascular strand, xylem is located adaxial to the phloem.

Waites and Hudson (1995) have proposed a model for leaf development in which the juxtaposition of ad- and abaxial leaf cell fates is required for the development and outgrowth of the leaf blade (Fig. 1). This model is based on observations of snapdragons homozygous for the recessive phantastica (phan) mutation; phan mutants possess radially symmetric leaves with abaxial characters around their circumference. An important prediction of this model is that in a mutant with the opposite phenotype (one in which abaxial leaf fates are transfomed to adaxial leaf fates) the leaf will again fail to form a blade and develop with radial symmetry.

We have recently isolated an Arabidopsis mutant the phenotype of which satisfies this prediction. The dominant phabulosa-1 (phb-1d) mutation causes both sides of the leaf to develop with adaxial characters. phb-1d leaves fail to develop blades and are frequently radially symmetric. Additionally, the phb-1d mutant provides evidence that adaxial leaf identity plays a critical role in the development of axillary SAMs.

Plants were grown in soil (Metromix 200) under 24 hour cool white flourescent light at either 24°C or 17°C (all plants used as pollen donors were grown at 17°C). All phenotypes were scored at 24°C unless otherwise noted.

Seeds of ecotype Landsberg erecta (Ler) were mutagenized by soaking them in 0.4% EMS in 100 mM phosphate buffer (pH 7) for 8 hours. The self-progeny of individual plants (M¹ plants) grown from mutagenized seed (n=1800) were harvested, planted in soil and screened for mutants defective in vegetative growth. A single M¹ plant segregated phb-D mutants in its progeny. The mutant was recovered by crossing a phb-1d mutant male to Ler. At least five backcrosses to Ler were performed prior to any genetic or phenotypic analysis. To determine the phenotype of plants doubly mutant for stm-1 and phb-1d, stm-1/+; phb-1d/+ individuals were crossed to stm-1/+; +/+ individuals. In the cross progeny 21 plants had the stm single mutant phenotype (presumed genotype stm-1/stm-1), 52 plants had the phb-1d single mutant phenotype (presumed genotype stm-1/+ or +/+; phb-1d/+), 57 plants were wild-type (presumed genotype stm-1/+ or +/+; +/+) and 20 plants were both phb-1d (as judged by their cotyledons) and stm (presumed genotype stm-1/stm-1;phb-1d/+). The observed progeny are consistent with the presumed genotypes (χ²=0.684; P<0.8).

STM-GUS/phb-1d transgenic plants were generated by crossing phb-1d/+ mutants to individuals homozygous for the STM-GUS construct. phb-1d/+ individuals in the cross progeny were then examined for GUS expression.

CAPS mapping was carried out as described by Konieczny and Ausubel (1993). phb-1d/+ individuals (Ler background) were crossed to wild-type Columbia (Col) individuals. Mutant cross progeny were crossed back to Col individuals. 2/32 mutant progeny of this second cross were found to carry recombinant chromosomes between the PHB locus and the m429 CAPS marker corresponding to a recombinant frequency of approximately 6%. 2/34 mutant progeny from the same cross carried recombinant chromosomes between the PHB locus and the GPA1 marker corresponding to a recombinant frequency of approximately 6%. This map data places the PHB locus at an approximate position of 59.8 on chromosome 2.

Samples were prepared and visualized by scanning electron microscopy as described by McConnell and Barton (1995). Tissues for thin sections were fixed overnight in a solution containing: 3% glutaraldehyde, 1.5% acrolein, 1.6% paraformaldehyde, and postfixed in 1% osmium tetroxide. Fixed material was dehydrated through an ethanol series and embedded in Spurr’s medium. Approximately 1 μm sections were cut and stained with 0.2% toluidine blue in 2.5% sodium carbonate (pH 11). Histological staining of β-glucuronidase (GUS) activity was performed overnight on seedlings at 37°C in 2 mM 5-bromo-4-chloro-3-indolyl β-D-glucuronic acid, 100 mM NaPO₄, pH 7, 0.1% Triton X-100, 1 mM FeCN and 10 mM EDTA. Samples were then washed with 70% ethanol for 2 days and visualized using a dissecting microscope.

The dominant phb-1d mutation causes a transformation of abaxial cell fates to adaxial cell fates in leaves and floral organs and alters organ shape. The mutation was isolated in a screen for EMS-induced mutants altered in SAM development and maps to chromosome II between CAPS markers m429 and GPA1. Crosses of phb-1d/+ pollen onto Ler carpels resulted in 47.1% phb-1d mutants (n=191: χ₂=0.6335; P<0.5). The wild-type siblings were never observed to segregate phb-1d mutants in their self-progeny (n=50) indicating that the phb-1d mutation is completely penetrant. Since the phb-1d/+ mutants are sterile as females at 24°C, the phb-1d mutation is propagated through the pollen (except at low temperatures as described below).

phb-1d heterozygotes have leaves with adaxial epidermal characters (dark green, glossy, trichome-rich surfaces) around their circumference and exhibit varying degrees of radial symmetry and loss of blade outgrowth (Fig. 2). The expressivity of the phb-1d mutant phenotype is variable. Some leaves are rod-like while others are shaped like trumpets. The latter have adaxial tissue on their outer surfaces and abaxial tissue on the inside of the ‘bell’ of the trumpet. phb-1d leaves grow nearly vertically rather than bending away from the plant. The upright posture of mutant leaves is consistent with equal growth rates of both sides of the leaf, as expected if identical cell fates are specified on both faces of the developing primordium.

Wild-type epidermal leaf cell shape varies according to the surface upon which it is located as well as its position along the leaf. Fig. 3 shows ad- and abaxial epidermal cells from three positions on the leaf: petiole, midrib midway along the length of the blade, and blade midway up the length of the blade and midway between the margin and the midrib. Wild-type petiole cells are long and rectangular with straight anticlinal walls. This is true of both surfaces though the adaxial petiole surface also contains stomates along its margin (Fig. 3A,D). Continuing up the midrib, midway along the length of the blade, ad- and abaxial cell morphologies differ significantly. The abaxial epidermal cells at this position are long and rectangular, similar to petiole cells. The corresponding adaxial cells are more complex and resemble blade cells (Fig. 3B,E). Epidermal blade cells are jig-saw shaped on both leaf surfaces, but adaxial blade cells are larger, more uniform in size, and less complex than the corresponding abaxial cells (Fig. 3C,F).

Comparing wild-type and phb-1d epidermal leaf cell shapes further supports the conclusion that phb-1d leaves are adaxialized. Trumpet-shaped leaves have larger, less complex cells on the outside of the ‘bell’ than on the inside (Fig. 3H-J). This is similar to the difference seen between the wild-type adaxial and abaxial blade epidermis. The cells are smaller overall in the phb-1d mutant than in the wild type. This is likely due to less organ expansion in the mutant as phb-1d organs are smaller than wild-type organs (Fig. 2A-C). It is more difficult to assign adversus abaxial fates to the epidermal cells of rod-shaped leaves based on cell shape. Rod-shaped leaves have cells that are similar in size to blade epidermal cells but they are less jigsaw-shaped and more rectangular (Fig. 3G). Stomates are distributed throughout the rod-shaped leaves as in wild-type blade tissue. Epidermal cells in young, wild-type leaf primordia are simple in shape and only achieve their complex, jigsaw shapes upon leaf expansion. The rod-shape of the extreme phb-1d leaves does not allow for normal blade expansion and this may affect cell shape.

Internal tissues are also affected in phb-1d mutants (Fig. 2I-L). The more extremely radialized leaves either entirely lack a vascular strand or possess single xylem elements. In such leaves, phloem (normally found in the abaxial pole of the vascular strand) was not seen. In less extremely affected leaves, vascular tissue in which xylem surrounds phloem tissue has been observed; the latter arrangement is the opposite of what is seen in the abaxialized snapdragon phan mutant where phloem surrounds xylem (Waites and Hudson, 1995).

All four types of floral organs – sepals, petals, stamens and carpels – are affected in phb-1d heterozygotes to some degree (Figs 2, 4). Both sepals and petals of phb-1d/+ individuals frequently develop with what appears to be radial symmetry (Fig. 4B,H). Sepals may be filamentous, trumpet-shaped, or may appear wild-type in shape. When trumpet-shaped, cells on the outside exhibit adaxial epidermal fates while cells on the inside exhibit abaxial epidermal fates (Fig. 4C-E). Petals are most commonly filamentous and have cell types characteristic of the wild-type adaxial epidermis around their circumference (cone-shaped cells with straight cuticular ridges) (Fig. 4F-I).

The reproductive organs are also affected in phb-1d mutants. In wild-type stamens, the pollen sacs form laterally and rotate to the adaxial side through unequal growth of the ad- and abaxial surfaces (Fig. 4J,K) (Bowman, 1994). In phb-1d/+ mutants, the pollen sacs often remain in lateral positions indicating a lack of unequal growth consistent with a failure to specify different cell types on the two surfaces (Fig. 4L,M). The ovules, which normally originate from the interior (adaxial side) of the carpel, develop ectopically in the phb-1d mutant from the abaxial base of the carpel (Fig. 4N). Furthermore, instead of displaying the curvature seen in the wild-type ovule where the unequal growth of the integuments and the nucellus generates a curved structure, phb-1d/+ ovules are frequently linear (Fig. 4O,P).

The phb-1d mutation affects sepals developing in the abaxial region (relative to the inflorescence axis) more than sepals in the adaxial region of the flower. The sepal closest to the inflorescence meristem is least likely to develop as a filament (Fig. 5). The shape of the adaxial sepals is most frequently either wild-type or trumpet-shaped. While other floral organs did not show any obvious position-dependent variation in phenotype, we cannot rule out minor differences since quantitative comparisons were not made.

The phb-1d mutant phenotype is temperature sensitive. At low temperatures (17°C), the phb-1d phenotype is somewhat alleviated and phb-1d/+ plants are weakly self-fertile. This has allowed us to recover homozygous phb-1d plants. Heterozygous phb-1d plants raised at 17°C produce wild-type (+/+; n=67), moderately affected (phb-1d/+; n=123) and severely affected seedlings (presumed phb-1d/phb-1d; n=59) in their self-progeny. The presumed homozygotes have more fully radialized and adaxialized leaves than the heterozygotes (Fig. 2C). This is especially striking in the cotyledons which are often only weakly affected in the heterozygote. The presumed homozygotes are very small and grow slowly. These individuals are sterile as the floral organs produced are completely radialized.

Normally, axillary SAMs develop on the adaxial leaf base, initially oriented toward the stem (Talbert et al., 1995; Fig. 7C). Later growth of the developing bud obscures this early relationship making the branch appear to emanate from the stem side of the axil. Although axillary buds normally develop on the adaxial leaf base, it is unclear what aspects of position are assessed by these cells in making developmental decisions. If adaxial leaf fate is a critical determinant of axillary SAM formation we would expect phb-1d mutants to develop SAMs ectopically. Indeed, ectopic SAMs form on the undersides of phb-1d mutant leaves (Fig. 6). To determine the origin of the ectopic buds in phb-1d mutants, we observed the expression of an STM promoter-GUS reporter construct in phb-1d mutants. This construct is expressed in the SAM (Long and Barton, unpublished). Regions of dark staining indicative of early steps in SAM formation were found on the undersides of leaves in phb-1d mutant seedlings but not in wild-type seedlings (Fig. 7A,B). In the most severely adaxialized mutant leaves, reporter gene expression could be observed in a ring around the entire base of the leaf (Fig. 7D). Thus, the development of the axillary bud behaves like other adaxial characters in that additional buds develop on the underside of leaves in phb-1d mutants.

During embryogenesis the SAM develops at the bases of and between the adaxial surfaces of the cotyledons. This position in some ways parallels the environment in which the axillary meristem develops. If adaxial cotyledon fate is responsible for promoting the development of the embryonic SAM, we might expect the phb-1d mutation to affect the embryonic SAM. We have detected a positive effect of phb-1d on the development of the embryonic SAM in two situations. First, the SAM of phb-1d/+ mutant embryos is enlarged relative to wild-type indicating that a stronger ‘meristem promoting’ signal may be present in phb-1d mutant embryos (Fig. 8A,B). Second, the phb-1d mutation partially suppresses the stm mutant phenotype. In mutants homozygous for stm-1, a strong but non-null allele of the STM gene, the SAM fails to form and cells at the site of the presumptive SAM terminally differentiate (Barton and Poethig, 1993). Thus, stm-1 homozygotes entirely lack any structures at the site normally occupied by the SAM. A week or two following germination, stm-1 mutants form leaves ectopically from a region below the point of cotyledon fusion (Barton and Poethig, 1993; Fig. 8C). In stm-1; phb-1d double mutants, an abnormal determinate structure is present at the site normally occupied by the SAM at germination (Fig. 8D). The position of this structure and its early appearance are consistent with it forming directly from the presumptive SAM.

The model for leaf development put forth by Waites and Hudson (1995; Fig. 1) proposes that the juxtaposition of ad- and abaxial cell fates is required for outgrowth of the leaf blade. The phenotypes of two leaf mutants described to date support the model. In the snapdragon phan mutant, adaxial characters are transformed into abaxial characters (Waites and Hudson, 1995). In the Arabidopsis phb-1d mutant, the opposite transformation of cell fate occurs: abaxial characters are replaced with adaxial characters. In neither mutant are ad- and abaxial cell fates juxtaposed; under this model lack of juxtaposition causes the leaves of both mutants to develop as radially symmetric, bladeless organs.

The snapdragon phan mutation is recessive and is therefore likely caused by a loss-of-function mutation in the PHAN gene. If so, the wild-type role of the PHAN gene is to specify adaxial cell fates since these are missing in the phan mutant. Since the phb-1d mutation is dominant, the role of the wild-type PHB gene in the specification of ad/abaxial leaf cell fates is uncertain. One possibility is that phb-1d represents a gain-of-function mutation in a gene normally required to specify adaxial fate, and thus plays a similar role in Arabidopsis to that played by the PHAN gene in snapdragon. Interestingly, both phan and phb-1d mutant phenotypes are temperature sensitive, exhibiting more abaxialized phenotypes at low temperature and more adaxialized phenotypes at high temperature. Like PHAN, the action of phb-1d is cell autonomous, or at least limited in its range, because we have obtained wild-type sectors on otherwise phb-1d mutant plants after treatment of phb-1d/+ seeds with EMS (McConnell and Barton, unpublished). Taken together these observations are consistent with PHAN in snapdragon and PHB in Arabidopsis controlling similar processes in leaf development.

However, it is also possible that the wild-type PHB gene plays no role in the development of leaf polarity. The dominant phb-1d mutation may have altered the pattern of PHB expression or the activity of the PHB gene product to allow it to take on an activity it does not normally possess. Even if this is true, the results presented here still show a correlation between adaxial transformation and lack of blade formation and a correlation between adaxial transformation and ectopic SAM formation.

Our results show a correlation between the acquisition of adaxial leaf fate and the development of ectopic buds. While it is possible that the phb-1d mutation acts to generate these two phenotypes independently of one another, a simpler explanation is that the ectopic SAMs on the underside of the phb-1d leaf are a consequence of the transformation of this tissue into adaxial leaf tissue. In this model, adaxial leaf fate is a major factor in determining whether an axillary SAM forms or not. We propose that, in Arabidopsis, leaf and meristem development are linked through a cycle in which SAMs make leaves, the adaxial sides of which in turn induce the development of new SAMs from their bases (Fig. 9A). This is in contrast to a model in which axillary SAMs are derived from remnants, so-called detached meristems, of the primary SAM left behind in the leaf axil. For the detached meristem model to hold, such remnants of the SAM would have to be left behind on the underside of the leaf in the phb-1d mutant and, in the extremely radialized phb-1d leaves, around their entire circumference. This would require a more complex model for PHB action than proposed here.

In dicot species, numerous observations indicate a link between adaxial leaf cell fate and competence to develop axillary meristems. For instance, begonia leaves regenerate SAMs from the adaxial but not the abaxial leaf surface. Transgenic tobacco plants that ectopically express the knotted1 gene (Sinha et al., 1993) or ectopically synthesize cytokinin (Estruch et al., 1991) develop SAMs on the adaxial but not the abaxial leaf surfaces. Also, after surgical manipulation, Epilobium and potato leaf primordia developed as abaxialized, radially symmetric organs. In these situations, axillary meristems often did not develop (Snow and Snow, 1959; Sussex, 1955). These observations suggest that the adaxial leaf environment is required for the development of an axillary SAM. Consistent with this, functional axillary meristems are lacking in the axils of the most severely affected leaves in snapdragon phan mutants (A. Hudson, personal communication). Our observations of the phb-1d mutant phenotype suggest that adaxial, basal leaf fate is also sufficient to direct axillary SAM formation.

There are many possible mechanisms through which the adaxial leaf environment might be responsible for the development of the axillary bud. For instance, a cell type present in the adaxial portion of the leaf may promote axillary SAM development; or a cell type present in the abaxial portion of the leaf may inhibit axillary SAM development; or cells in the adaxial portion of the leaf may express a molecule, for instance a receptor or a transcription factor, that confers competence to respond to SAM-inducing signals.

Once the proposed leaf-meristem-leaf cycle is initiated, it can be propagated indefinitely – but where does it start? The position at which the primary SAM develops, at the adaxial bases of the cotyledons, is reminiscent of the position at which the axillary SAM develops, at the adaxial leaf base. The phb-1d mutation positively affects the development of the embryonic SAM in two situations. It partially suppresses the phenotype of mutants homozygous for the non-null stm-1 allele, causing the development of determinate structures at the site of the presumptive SAM. Since the stm-1 allele retains some activity, the suppression by phb-1d could act by boosting the levels of this residual activity. In addition, phb-1d/+ mutant embryos have larger SAMs than wild-type embryos. These results can be explained by a model in which the adaxial sides of the presumptive cotyledons positively influence development of the presumptive SAM (Fig. 9B). While the evidence for this model of embryonic SAM development is limited, it provides a novel conceptual framework for apical embryonic development.

Adaxial leaf fate need not play a similar role in SAM formation in other types of vascular plants. For instance, in several grass species, the axillary SAM develops not on the adaxial leaf base but in closest association with the stem side of the axil (Sharman, 1945). Furthermore, the axillary SAM is clonally related to the leaf and stem segment above it (Johri and Coe, 1983; Poethig et al, 1986) suggesting that positional signals related to stem fate rather than adaxial leaf fate may be important in axillary SAM formation in this group.

Branch formation is a fundamental process in the development of plant form. The finding that a subregion of the leaf may play a critical role in the formation of SAMs, and therefore branches, sheds light on the way that plants make buds and also may change the way we think about the development of the embryonic SAM.

This work was supported by the National Science Foundation and the US Department of Agriculture. This is paper number 3506 from the Laboratory of Genetics. We thank A. Hudson and M. Evans for helpful discussions and M. Evans for critical reading of the manuscript. We also thank R. Fisher, E. Moan, G. Heck and J. Long for providing the STM-GUS reporter line and Heidi Barnhill for excellent technical assistance with the SEMs.