Embryonic control of epidermal cell patterning in the root and hypocotyl of Arabidopsis

A fundamental feature of development in multicellular organisms is the specification and patterning of distinct cell types. An important issue in cell specification relates to the developmental origin of the mechanisms that establish cell-type patterns. Of particular interest is the relationship between the origin of cell patterning mechanisms and the development of the anatomy and identity of the resident tissues and organs during embryogenesis (Davidson et al., 1998; Simpson et al., 1999).

The development of a position-dependent pattern of epidermal cell types in the Arabidopsis root and hypocotyl provides a simple model system to explore the control of cell patterning in plants. Epidermal cells in contact with two underlying cortical cells (i.e. cells over an anticlinal cortical cell wall; the ‘H’ cell position) preferentially differentiate as root hair cells in the root and stomata in the hypocotyl, whereas cells in contact with a single cortical cell (i.e. cells over a periclinal cortical cell wall; the ‘N’ cell position) differentiate into non-hair cells in the root and non-stomatal cells in the hypocotyl (Dolan et al., 1993; Dolan et al., 1994; Galway et al., 1994; Berger et al., 1998b; Hung et al., 1998). This cell type pattern implies that the underlying cortical cells provide, or influence the presentation of, an inductive signal that determines epidermal cell fate.

Several genes have been identified that influence the specification of the epidermal cell types in the root and hypocotyl of Arabidopsis. The TRANSPARENT TESTA GLABRA (TTG), WEREWOLF (WER) and GLABRA2 (GL2) genes are each required for proper specification of the N cell (non-hair or non-stomatal cell) fate (Galway et al., 1994; Masucci et al., 1996; Hung et al., 1998; Lee and Schiefelbein, 1999). The WD40 protein encoded by TTG (Walker et al., 1999) and the myb transcription factor encoded by WER (Lee and Schiefelbein, 1999) are positive regulators of the GL2 gene (Hung et al., 1998; Lee and Schiefelbein, 1999). The GL2 gene encodes a homeodomain-leucine zipper (HD-Zip) protein (Rerie et al., 1994; DiCristina et al., 1996) and is preferentially expressed in the N-cell position in the developing epidermis of both the root and hypocotyl (Masucci et al., 1996; Hung et al., 1998). In contrast to these three genes, the CAPRICE (CPC) gene is required to specify the hair cell fate in the root epidermis and encodes a small myb protein that negatively regulates GL2 (Wada et al., 1997). Thus, epidermal cell fate appears to be largely determined by the regulated expression of the downstream N-cell-specific transcription factor GL2.

The similar pattern of epidermal cell types in the root and hypocotyl implies that the patterning mechanism originates during embryogenesis, because the epidermal tissue of the root and hypocotyl is derived from the same set of embryonic protodermal cells (Scheres et al., 1994). To define the embryonic origin of the cell patterning mechanism, we have assessed GL2 gene expression throughout embryogenesis. In a prior study, N-cell-specific expression of an enhancer-trap marker gene was found to initiate in the torpedo stage embryo (Berger et al., 1998a). However, the identity of the presumed gene responsible for this expression pattern is unknown, the possible role of this gene in the epidermal cell fate pathway is unclear, and the embryonic N-cell expression is limited to the root (Berger et al., 1998a). Thus, the reporter gene expression from this enhancer trap line may not accurately reflect the activity of the fundamental patterning mechanism that is shared by the root and hypocotyl.

In the present study, we have employed in situ RNA hybridization and reporter fusions to show that GL2 expression becomes established in a cell-position-dependent pattern during the heart stage of embryogenesis. Further, we find a progressive reduction in GL2 expression during embryonic development of an anatomical transition zone between the root and hypocotyl, which is associated with a unique pattern of epidermal cell differentiation in the seedling. We also show that the TTG and WER genes, but not CPC or GL2, are required for proper establishment of the GL2 expression pattern during embryogenesis.

The ttg-1, wer-1, cpc and gl2-1 mutant alleles used in this study have been described previously (Koornneef, 1981; Lee and Schiefelbein, 1999; Wada et al., 1997; Koornneef et al., 1982). Plants harboring the GL2::GUS reporter construct have been described previously (Masucci et al., 1996). The GL2::GFP and GL2::GUS reporter constructs were introduced into the various mutant backgrounds by genetic crosses. The growth and analysis of Arabidopsis seedlings in vertically oriented agarose-solidified plates has been described (Schiefelbein and Somerville, 1990).

The GL2::GFP translational reporter construct was generated by fusing a 2032 bp 5′ promoter fragment from the GL2 gene (including the first four codons of GL2) to the mGFP5 (ER-localized GFP variant) (Haseloff et al., 1997) coding sequence. Details of the transgene construction are available upon request.

Plant transformation was achieved by electroporating constructs (in the binary vector pBIN19) into the Agrobacterium strain GV3101 followed by introduction into Arabidopsis using the floral dip method (Clough and Bent, 1998). T₁ seeds were collected in separate pools and transgenic plants selected by plating on medium containing kanamycin (50 μg/ml). More than ten lines were generated and all had similar phenotypes, and three independent lines were chosen for detailed characterization.

The number of cells in the ground and epidermal tissue layers was determined from transverse sections from 3-day-old seedlings. Hand sections were obtained from at least 15 independent agarose-embedded seedlings and stained with Fluorescent Brightener 28 (Sigma Chemical) as described previously (Galway et al., 1994). Sections were designated as either root (if no ectopic root hair cells were observed), root-hypocotyl junction (if ectopic root hair cells were present), or hypocotyl (if no root hairs were visible) regions.

Histochemical analysis of whole seedlings containing the GL2::GUS reporter was performed essentially as described (Masucci et al., 1996). The GFP expression in the GL2::GFP embryos and seedlings was examined with a Zeiss LSM510 confocal microscope, with a 488 nm excitation mirror and a 505-530 nm and 530-560 nm emission filter to record images. Seedling roots and some mature embryos were also stained with propidium iodide (10 μg/ml) to visualize cell boundaries.

The in situ localization of GL2 RNA in embryo sections was performed as described by Long et al. (Long et al., 1996). A detailed protocol is available electronically (http://www.wisc.edu/genetics/CATG/barton/protocols.html). Sense and antisense gene-specific GL2 probes (Lee and Schiefelbein, 1999) were synthesized by in vitro transcription reactions with digoxigenin-labeled UTP as previously described (Masucci et al., 1996).

As a first step toward defining the developmental origin of the position-dependent GL2 expression pattern in the seedling epidermis, we analyzed GL2 transcript accumulation during embryogenesis using in situ RNA hybridization. The stages of Arabidopsis embryogenesis that we refer to here have been extensively described (Jürgens and Mayer, 1994). Furthermore, our analysis was aided by the previously determined embryonic fate map for the Arabidopsis root and hypocotyl (Scheres et al., 1994), which depicts the likely developmental fate of cells and cell groups at various stages of embryogenesis.

Using in situ RNA hybridization, we were unable to consistently detect GL2-specific transcripts from embryo sections taken before the heart stage (Fig. 1A). Occasionally, we observed a GL2 hybridization signal in patches of protodermal cells in embryos at the triangular/early-heart stage (unpublished data). At the heart stage, we consistently observed a hybridization signal within protodermal cells in the basal portion of the embryo (Fig. 1B). These GL2-expressing cells are located in the lower lower tier (llt) region of the heart stage embryo, which is the region destined to give rise to the root and hypocotyl in the postembryonic plant (Scheres et al., 1994). The same basic pattern of GL2 RNA accumulation was detected in embryo sections taken at all later stages of embrogenesis, including torpedo stage embryos (Fig. 1C) and mature embryos (Fig. 1D). These results show that GL2 expression is initiated by the heart stage of embryogenesis within protodermal cells that generate the future root and hypocotyl epidermis.

To assess the embryonic pattern of GL2 gene expression in greater detail, we wished to employ a sensitive GL2 reporter gene fusion. An available GUS reporter fusion (GL2::GUS) (Masucci et al., 1996) was deemed unsuitable for this purpose because of diffusion problems associated with the GUS system in Arabidopsis embryos (Y. L. and J. S., unpublished observations). Therefore, we generated a green fluorescent protein (GFP) reporter construct by translationally fusing a 2 kb fragment from the 5′ end of the GL2 gene to the coding region of mGFP5. When employed in complementation constructs, this 5′ GL2 fragment is sufficient to rescue the gl2 root defect (unpublished data). Transgenic plants bearing this GL2::GFP construct exhibited reporter gene expression within the seedling root and hypocotyl in the same cell-specific manner as previously identified for GL2 RNA accumulation and GUS expression in GL2::GUS plants (Fig. 2) (Masucci et al., 1996). In particular, GFP was preferentially detected in epidermal cells located outside periclinal cortical cell walls (the N position; Fig. 2D). This indicates that GFP accumulation in the GL2::GFP plants accurately reflects the transcription pattern of the GL2 gene.

Embryos produced by homozygous GL2::GFP plants were collected at all developmental stages and examined for GFP accumulation. We also examined embryos from non-transformed plants at each stage and employed them as negative controls for these experiments. No GFP signal was detected in GL2::GFP embryos examined at the globular or early triangular stages (Fig. 3A; data not shown). The earliest group of embryos in which green fluorescence could be detected were those at the late-triangular or early-heart stage. Most embryos in this group lacked any detectable GL2::GFP expression (like embryos from non-transformed plants), but approximately one-third of them (12 embryos out of 31 examined) exhibited a low level of GFP either within a few scattered cells (Fig. 3B) or within a larger collection of cells (Fig. 3C). Although generally patchy in their distribution, the GFP-expressing cells sometimes appeared to be arranged in columns (Fig. 3C). Further, the vast majority of the GFP-expressing cells were located in the basal portion of the embryo (the lower tier region)(Scheres et al., 1994), although fluorescing cells were occasionally found in the apical portion (Fig. 3C). In all GFP-expressing embryos, the GFP was restricted to cells in the protoderm (Fig. 3D).

In GL2::GFP embryos at the heart stage, a large number of GFP-expressing cells were present in most of the embryos examined (21 of 24). The GFP was exclusively protodermal and principally located in the basal portion of the embryo (Fig. 3E,F) that is destined to give rise to the epidermis of the seedling root and hypocotyl (the lower lower tier region) (Scheres et al., 1994). GFP expression is evident within cells at the base of the protoderm layer (destined to form epidermal/lateral root cap initials) but is notably absent from derivatives of the hypophyseal cell (Fig. 3F), which give rise to the columella root cap and quiescent center within the lower portion of the root apical meristem (Scheres et al., 1994). In each of the 21 embryos exhibiting GFP, the GFP-expressing cells were arranged in particular columns that were separated by columns of non-GFP-expressing cells (Fig. 3E). By examining and comparing optical sections of the protoderm and underlying ground meristem layer, we determined that the non-GFP-expressing cells were located over an anticlinal wall between ground meristem cells (the H cell position; data not shown).

Embryos at later stages displayed a pattern of GL2::GFP expression similar to the heart stage embryos. At the torpedo stage, embryos exhibit a higher level of GFP expression and a clearer distinction between the GFP-expressing and non-GFP-expressing files of cells (Fig. 3G). The GFP retains its protoderm-specific expression and remains localized in the future root and hypocotyl region (Fig. 3H). In the mature embryo, GFP expression is present in distinct files of cells in the root-hypocotyl axis that mirrors its postembryonic pattern (Fig. 3I). GFP-expressing cells are restricted to the N position and non-GFP-expressing cells are in the H position of both the embryonic hypocotyl (Fig. 3K) and embryonic root (Fig. 3L). In addition to epidermal expression in the root and hypocotyl, GFP is present in epidermal cells on the cotyledon margins (Fig. 3I,J), which mirrors the GL2 reporter expression observed in the corresponding cells of the seedling cotyledon (Fig. 2A,B) (Masucci et al., 1996).

Taken together, our analysis of GL2 transcript accumulation and GL2::GFP reporter expression indicates that epidermal cell patterning is initiated early in embryogenesis and establishes position-dependent GL2 expression by the heart stage.

During the course of our analysis of GL2::GFP embryos, we discovered that GL2::GFP is not uniformly expressed along the length of the epidermal cell files at all embryonic stages. Specifically, there is significantly less GFP accumulation within an apical-basal segment containing four to seven cells per file near the junction of the root and hypocotyl segments in the mature embryo (Fig. 3I). This difference in GL2::GFP expression is first detected in embryos at the torpedo stage (Fig. 4A) and it becomes more apparent as the embryo matures (Fig. 4B,C).

Because of the unusual GL2::GFP expression in this root-hypocotyl junction region during embryogenesis, we analyzed postembryonic GL2 expression and epidermis development in this region. We discovered that this region of sharply diminished GL2 expression persists in the young seedling, and it precisely coincides with an apical-basal segment of approximately four to seven cells that exhibits abnormal epidermal cell patterning (Fig. 4D,E). This unusual region of the Arabidopsis seedling had been previously noted as consisting of shortened epidermal cells that exclusively differentiate as root hair cells (Dolan et al., 1993; Scheres et al., 1994; Cheng et al., 1995). Considering that GL2 acts as a negative regulator of root-hair cell differentiation (Masucci et al., 1996), our finding of diminished GL2 expression within this region is consistent with the unusual epidermal cell phenotype. We also observed that the root hairs formed in this region tend to emerge closer to the center of the cell, rather than the basal end as is typical for hairs in the root proper (Schiefelbein and Somerville, 1990). Thus, the epidermis of the root-hypocotyl junction region is unique in lacking position-dependent GL2 expression and cell-type patterning.

Although the root and hypocotyl exhibit a similar position-dependent pattern of epidermal cell types, they differ in their cellular anatomy. We considered the possibility that the unique GL2 expression and epidermal patterning characteristics of the root-hypocotyl junction region may be related to an anatomical transition within this region. To test this, we analyzed the cell organization of the ground and epidermal tissues within the root, hypocotyl, and root-hypocotyl junction regions from transverse sections of 3-day-old seedlings. As expected from prior studies (Dolan et al., 1993; Scheres et al., 1994), the root sections possessed two concentric layers of ground tissue (one layer of endodermis and one layer of cortex), whereas the hypocotyl sections contained three layers of ground tissue (one layer of endodermis and two layers of cortex; Fig. 4F,I). Sections from the root-hypocotyl junction region were identified by the presence of ectopic root-hair cells (root-hair bearing epidermal cells located outside a periclinal cortical cell wall). Most of these sections (32/38) exhibited an arrangement of ground tissue layers not found in the hypocotyl or root (Fig. 4G,H). Because the identity of individual ground tissue cells in this region is unclear, we employed the terms inner, middle, and outer layer of ground tissue to describe the set of cells that contact the vascular tissue (inner), contact neither vascular or epidermal tissue (middle), or contact the epidermal tissue (outer). Using these criteria, we interpret the root-hypocotyl junction sections as typically containing two complete ground tissue layers (inner and outer layers) and one incomplete layer (middle layer) (Table 1). Considering that the middle layer of ground tissue represents the defining difference between the root and hypocotyl, the presence of an incomplete middle layer in the junction region indicates that a transition in ground tissue anatomy does indeed occur in this region.

We also analyzed the number of cells within the inner and outer ground tissue layers and the epidermal layer in these transverse sections. We found that the inner layer of the root-hypocotyl junction region invariably contains the same number of cells (8) as the inner layer of the root and the inner layer of the hypocotyl (Table 1). However, the outer layer of ground tissue in the root-hypocotyl junction sections contains a variable and intermediate number of cells (range of 8-15 cells), relative to the root and hypocotyl (Fig. 4G,H; Table 1). Because the number of cells in this outer ground tissue layer determines the number of epidermal cells in the H position, the differences in cell number in this layer in the junction region means that there are corresponding differences in the number of H cell files in the epidermis of this root-hypocotyl junction. In addition, we determined that the total number of epidermal cell files in the root-hypocotyl junction region is intermediate to that in the root and hypocotyl (range of 21-35 cells; Table 1). Together, this anatomical analysis shows that the ground and epidermal tissue of the root-hypocotyl junction region is distinct from the root or hypocotyl and effectively represents an anatomical transition zone between these apical-basal segments.

Following the discovery of an embryonic origin for epidermal cell patterning, we wished to define genes that control this process. The TTG, WER and CPC genes are known regulators of postembryonic patterning of the epidermal cell types (Schiefelbein, 2000). To determine whether these might also regulate the embryonic pattern, we examined the effect of mutations in each of these, as well as mutations in gl2, on the position-dependent GL2::GFP expression during embryogenesis.

ttg mutations cause a significant reduction in the level, but no change in the position-dependent pattern, of GL2 expression in the seedling root and hypocotyl (Fig. 5E,F, compare with Fig. 2D)(Hung et al., 1998). In an analogous manner, we found that ttg-1 inhibits the level, but not the basic pattern, of GL2::GFP expression during embryogenesis. In ttg GL2::GFP embryos, GFP expression is generally too weak to be detected until the torpedo stage (Fig. 5A,B). However, the expression pattern is the same as the wild type, with GFP-expressing cells confined to epidermal cells outside periclinal cortical cell walls (the N position) in the future root and hypocotyl regions (Fig. 5B,C; data not shown). This normal pattern persists and intensifies during embryogenesis, but the GFP expression level is significantly lower than wild type throughout all stages, including the mature embryo (Fig. 5C,D).

The wer mutations dramatically alter the postembryonic expression of GL2, eliminating expression in most of the cells and abolishing the position-dependent pattern of the few GL2-expressing cells in the root and hypocotyl (Fig. 5K,L, compare with Fig. 2D)(Lee and Schiefelbein, 1999). We discovered that the wer-1 mutation has essentially the same effect on the embryonic expression of GL2. GFP expression was generally undetectable in the wer-1 GL2::GFP embryos, although some embryos (particularly at the mature stage) possessed a few scattered protodermal cells with a low level of GFP (Fig. 5G-J). These GFP-expressing cells were not restricted to the N position, but were located in both the N and H positions (data not shown).

In contrast to TTG and WER, the CPC gene is proposed to act as a negative regulator of GL2 expression (Wada et al., 1997). Accordingly, the cpc mutation reduces the frequency of root-hair cells, generates a large number of ectopic non-hair cells, and causes GL2 expression in the H-cell position during postembryonic development of the root epidermis (Wada et al., 1997; Lee and Schiefelbein, 1999)(Fig. 5Q). Unexpectedly, we found that the cpc mutation does not alter GL2::GFP expression during embryogenesis. Like the wild type, GFP expression in cpc GL2::GFP embryos is first detected at the late-triangular stage, is established in a position-dependent pattern by the heart stage, and is maintained and enhanced in the same pattern during the rest of embryogenesis (Fig. 5M-P). Further, we did not detect any GFP accumulation in cpc GL2::GFP embryos at the globular and pre-globular stages, which indicates that CPC does not act to restrict GL2 expression during early embryogenesis (data not shown).

We also examined the effect of the gl2-1 mutation on GL2::GFP expression to determine whether the GL2 homeodomain transcription factor may regulate the expression of its own gene during embryogenesis. In prior studies, gl2 mutations did not significantly alter the postembryonic expression of the GL2 (Hung et al., 1998), and we have confirmed this here by showing that position-dependent expression of GL2::GFP in the seedling root epidermis is not altered in the gl2-1 mutant background (Fig. 5V). Similarly, we did not detect any difference in embryonic GL2::GFP expression in gl2-1, as compared with the wild type (Fig. 5R-U). Throughout embryogenesis, a normal level and position-dependent pattern of GL2 expression was observed in the gl2-1 GL2::GFP embryos.

Together, these findings show that the embryonic establishment of the GL2 expression pattern is dependent on the TTG and WER products but does not require the CPC or GL2 proteins.

The basic organization of the plant body plan is accomplished during embryogenesis, including establishment of apical-basal and radial patterns of tissues, organs and regions (Laux and Jürgens, 1997). In this paper, we examined the possible embryonic origin and regulation of the circumferential pattern of cell types that arises in the Arabidopsis seedling epidermis. We employed a cell-type-specific transcription factor gene, GL2, that is expressed postembryonically in a position-dependent fashion during epidermal differentiation in the seedling root and hypocotyl. We discovered that GL2 gene expression initiates within the protoderm during the late-triangular and early-heart stage of embryogenesis. Although this initial expression is weak and sporadic, it rapidly assumes a position-dependent pattern that is clearly visible by the mid-heart stage and matches the postembryonic pattern. This demonstrates that the mechanism responsible for establishing a specific pattern of cell types in the postembryonic plant is already in place by the early-heart stage of embryogenesis.

Many of the basic patterning elements and anatomical features of the Arabidopsis embryo have already been established by the early-heart stage of embryogenesis. In particular, the apical-basal and radial axes are already defined by this stage (Scheres et al., 1994; Laux and Jürgens, 1997), meaning that groups of cells destined to give rise to hypocotyl, root, or cotyledon segments have been defined, as have the cell layers generating the epidermis, ground tissue and vascular tissue. Notably, it is at the early-heart stage when the final number of cell files in the embryonic epidermis and outer ground tissue layer is established in the future root/hypocotyl region. A ring of eight ground meristem cells and a ring of 16 protoderm cells are initially generated in the lower tier region (which gives rise to root and hypocotyl) by the globular stage (Scheres et al., 1994). However, it is not until the early-heart stage that an additional ground meristem layer (destined to be the outer cortical cell layer) is formed and the final number of cells in the protoderm and ground meristem is achieved within the future hypocotyl region (Scheres et al., 1994). Thus, the final cellular anatomy of the future epidermis and outer ground tissue is established at approximately the same embryonic stage that position-dependent GL2 expression is initiated. Considering the established connection between epidermal cell fate and cell position relative to the underlying cortical cells (Berger et al., 1998a), it may be that initiation of the epidermal cell patterning program is in some way triggered by completion of the divisions generating the final anatomy of the protoderm and outer ground tissue layer.

Our finding of the initiation of the epidermal cell patterning mechanism during the early-heart stage of embryogenesis has additional implications. This timing is much earlier than the N-cell expression pattern of the J2301 enhancer-trap line, which first arises at the mid-torpedo stage (Berger et al., 1998a). It is therefore likely that the reporter gene transcription in this line is responding to later-acting factors associated with N-cell specification. It is also interesting to note that the early-heart stage initiation of GL2 expression precedes the establishment of the root and shoot apical meristems in the Arabidopsis embryo (Dolan et al., 1993; Scheres et al., 1994; Laux and Jürgens, 1997). Thus, despite the general postembryonic connection between meristem activity and cell specification in plants, the present study demonstrates that patterning of cell types within a tissue need not require a functional meristem.

One of the key features of the embryonic GL2 expression pattern uncovered by our study is that GL2 largely initiates its expression in a distinct positionally defined set of cells, rather than within a broader set of cells or tissues and subsequent refinement of the expression domain. This feature is reminiscent of the embryonic cell-specific expression of some transcriptional regulators in animals, including the selector genes associated with bristle patterning in Drosophila. In this system, the proneural genes (including members of the achaete-scute complex) act in larger clusters of cells and cause the selector genes (including the homeobox gene cut) to become initially expressed in a particular set of cells (Simpson, 1996; Simpson et al., 1999). In an analogous manner, the precise initiation of position-dependent expression we have found for the GL2 homeobox gene in the early Arabidopsis embryo may be dictated by regulators that exhibit broader domains of expression.

In this study, we identified an apical-basal segment of the Arabidopsis embryo with a unique GL2 expression character. This segment, located near the root-hypocotyl junction, exhibits a progressive reduction in GL2 expression during embryogenesis that culminates in an undetectable level of GL2 expression in newly germinated seedlings (Fig. 4D,E). This region coincides with an apical-basal segment of the seedling previously characterized as producing root hairs from every epidermal cell, and termed the collet or upper part of the embryonic root (Dolan et al., 1993; Scheres et al., 1994; Cheng et al., 1995; Schneider et al., 1997). Although this region has been considered to possess root identity because of the presence of root-hair cells and lack of pigmentation, we discovered that it possesses an intermediate cellular anatomy and therefore, anatomically, represents a transition zone between the root and hypocotyl segments. Further, the lack of GL2 expression and the formation of ectopic root-hair cells in this root-hypocotyl junction region implies that the epidermal patterning mechanism is not employed in this region.

It is notable that the reduction in GL2 expression in this region is first detected at the torpedo stage of embryogenesis (Fig. 4). This is the stage when a set of periclinal divisions generate two layers of ground tissue in the root and three layers in the hypocotyl (Scheres et al., 1994; Scheres et al., 1995), probably creating the incomplete middle layer of ground tissue in the future root-hypocotyl junction (Table 1). Further, although the differences in cell file number in the epidermal and outer ground tissue layer of the future root and hypocotyl is established during the early-heart stage, the apical-basal segment of cells comprising the root-hypocotyl junction arise later, in part by divisions during the torpedo stage and later in embryogenesis. Taken together, our findings show that a close correlation exists between the unique cellular anatomy that develops in this root-hypocotyl region and its unusual GL2 expression and epidermal cell specification features.

At present, it is unclear whether the correlation between the cellular anatomy and cell specification characteristics of this region reflect a cause-effect relationship. It is possible that the unusual epidermal and ground tissue anatomy in this region prevents cell signaling events that are required for epidermal specification. For example, the incomplete middle layer of ground tissue (Table 1) may disrupt the identity and thereby the signaling capacity of the cells in the ground tissue layers, and/or the variation in cell file number in the epidermal (from 21 to 35) and outer ground tissue (from 8 to 15) layers in this short region may inhibit signaling between cells within individual files. Alternatively, it may be that the correlation we observe reflects the influence of a regional regulator that independently controls both cellular organization and epidermal cell specification within this root-hypocotyl junction region.

In addition to defining the embryonic GL2 expression pattern, we examined the role of known postembryonic regulators on embryonic GL2 expression. We discovered that two genes, TTG and WER, are required for the proper establishment of the embryonic GL2 expression pattern. Mutations in ttg significantly reduce the level of GL2 expression throughout embryogenesis, whereas wer mutations effectively abolish embryonic GL2 expression entirely. These embryonic effects of ttg and wer are similar to their postembryonic effects (Hung et al., 1998; Lee and Schiefelbein, 1999), indicating that these genes’ products act in a similar manner during embryonic and postembryonic epidermis development. The disruption of embryonic GL2 expression in the wer-1 mutant is particularly notable because it suggests that the WER protein is a critical regulator of the establishment of the position-dependent patterning mechanism during embryogenesis.

Mutations in two genes, gl2 and cpc, had no detectable effect on the embryonic GL2 expression pattern. The gl2 mutation was previously shown to have no effect on postembryonic GL2 expression (Hung et al., 1998), which together with the findings reported here, implies that the GL2 homeodomain protein does not regulate the transcription of its own gene. The inability of the cpc mutation to alter embryonic GL2 expression was unexpected, because CPC acts genetically as a negative regulator of GL2 and the cpc mutation causes ectopic GL2 expression in the seedling root epidermis (Wada et al., 1997; Lee and Schiefelbein, 1999)(Fig. 5Q). This suggests a difference in the molecular control of embryonic vs. postembryonic epidermal patterning, which may be due to the postembryonic restriction of CPC action or to the presence of a redundant factor acting during embryogenesis.

The reason for activation of the epidermal patterning mechanism during early embryogenesis is not clear. We were not able to detect any cytological difference in the epidermal cells in the N and H positions at any stage of embryogenesis (Y. L. and J. S., unpublished observations), which implies that their final differentiation programs (Dolan et al., 1993; Galway et al., 1994) are not engaged before seed germination. Furthermore, successful epidermal patterning is not essential for embryo development or epidermal tissue identity, because gl2 mutants do not exhibit any detectable abnormalities in cell organization or cytological features during embryogenesis or early postembryonic epidermal development (Masucci et al., 1996)(Y. L. and J. S., unpublished observations). One possibility is that position-dependent patterning during early embryogenesis helps to ensure the correct specification of cells generated later in embryogenesis and during postembryonic development. In a conceptually similar mechanism, the activity of meristem initial cells is influenced by signaling from more mature cells during postembryonic root development (van den Berg et al., 1995).

This research was supported by the US National Science Foundation (grant IBN-0077900). The authors thank Ken Cadigan, Steve Clark, Myeong Min Lee and Christine Bernhardt for helpful comments.