The bHLH genes GLABRA3 (GL3) andENHANCER OF GLABRA3(EGL3) specify epidermal cell fate in the Arabidopsis root

The specification and patterning of distinct cell types is a crucial feature of development in multicellular organisms. In plants, the formation of the hair and non-hair cells in the Arabidopsis root epidermis has been used extensively as a relatively simple and experimentally tractable model for studying cell fate specification(Dolan and Costa, 2001; Larkin et al., 2003). These cell types are non-essential under laboratory conditions, they are easy to examine, and they arise continuously during root development. Furthermore, the hair and non-hair cell types are patterned in a predictable manner in the Arabidopsis root, with the hair cells located in a cleft between two underlying cortical cells (the H position) and the non-hair cells present outside a single cortical cell (the N position)(Dolan et al., 1994; Galway et al., 1994).

Molecular genetic studies have led to the identification of a suite of putative transcription factors that regulate the epidermal cell pattern. These include gene products required for specification of the non-hair cell type,such as the homeodomain protein GLABRA2 (GL2)(Masucci et al., 1996; Rerie et al., 1994), the R2R3 MYB-type transcription factor WEREWOLF (WER)(Lee and Schiefelbein, 1999),and the WD-repeat protein TRANSPARENT TESTA GLABRA (TTG)(Galway et al., 1994; Walker et al., 1999). Other regulators are involved in specifying the hair cell fate, including CAPRICE(CPC) and TRYPTICHON (TRY), which are small one-repeat MYB proteins that lack a transcriptional activation domain and exhibit partially redundant functions(Schellmann et al., 2002; Wada et al., 2002; Wada et al., 1997). The final cell pattern appears to result from positive and negative regulatory interactions between these components(Schiefelbein, 2003). Specifically, WER promotes transcription of GL2 and CPC (and probably TRY) in the N position, GL2 inhibits hair cell specification in the N position, and CPC (and probably TRY) act in lateral inhibition by moving to the H cell and repressing transcription of WER, GL2 and its own gene (Lee and Schiefelbein,2002; Schellmann et al.,2002; Schiefelbein,2003; Wada et al.,2002).

Several lines of indirect evidence have suggested that a basic helix-loop-helix (bHLH) transcription factor may also be a component of the cell specification pathway in the root epidermis. This evidence includes (1)the ability of the maize R bHLH protein to induce ectopic non-hair cells when expressed in wild type Arabidopsis(Galway et al., 1994); (2) the ability of the maize R bHLH protein to restore non-hair cell production when expressed in the hairy ttg mutant(Galway et al., 1994); (3) the ability of the maize R bHLH protein to promote GL2 gene expression(Hung et al., 1998); (4) the ability of the wer mutations to block the effect of the maize R bHLH protein on non-hair cell specification(Lee and Schiefelbein, 1999);and (5) the ability of the R bHLH protein to interact with the WER protein and with the CPC protein in the yeast two-hybrid assay(Lee and Schiefelbein, 1999; Wada et al., 2002). Together,these findings have led to the suggestion that an Arabidopsis bHLH protein is likely to exist that interacts with WER to induce cells in the N position to adopt a non-hair fate (Larkin et al., 2003; Lee and Schiefelbein, 1999).

In this study, we confirm this long-standing hypothesis. We find that two Arabidopsis bHLH genes, GLABRA3 (GL3) and ENHANCER OF GLABRA3 (EGL3), are important regulators of root epidermal cell specification. These two genes encode bHLH proteins related to the maize R protein and influence trichome development and other TTG-related processes in Arabidopsis(Koornneef et al., 1982; Payne et al., 2000; Zhang et al., 2003). We show that GL3 and EGL3 have largely redundant functions in the specification of the non-hair cell fate and also participate in specifying the hair cell fate. We propose that the GL3 and EGL3 bHLH proteins act as binding partners for the WER or the CPC MYB proteins and thereby mediate the cell fate decision during root epidermis development.

The isolation of the mutant alleles used in this study has been described: cpc-1 (Wada et al.,1997), egl3-1 (Zhang et al., 2003), egl3-2(Zhang et al., 2003), gl2-1 (Koornneef,1981), gl3-1(Koornneef et al., 1982), gl3-2 (Hulskamp et al.,1994) and wer-1 (Lee and Schiefelbein, 1999), and these are likely to represent loss-of-function alleles. The GL2::GUS, CPC::GUS, 35S::EGL3 and 35S::GL3 constructs and transgenic lines have been previously described (Lee and Schiefelbein,1999; Masucci et al.,1996; Wada et al.,2002; Zhang et al.,2003). Lines homozygous for multiple mutations and/or transgenes were constructed by crossing single mutant or transgenic plants, examining the F2 progeny for putative double mutant/transgene phenotypes, and confirming the desired genotype in subsequent generations by backcrossing to single mutants,examining reporter gene expression, and/or PCR-based tests.

Arabidopsis seeds were surface sterilized and grown on agarose-solidified nutrient medium in vertically oriented petri plates as previously described (Schiefelbein and Somerville, 1990).

Root hair cell production and cell type pattern analysis were determined from Toluidine Blue-stained roots as previously described(Lee and Schiefelbein, 2002)from at least 24 four-day-old seedling roots for each strain. The upper region of the root was defined as the segment containing 10 epidermal cells whose upper boundary is four cells below the hairy collet region. The lower region of the root is a larger zone representing approximately the lower half of a four-dayold root and occupied by epidermal cells that differentiate during days 3-4. An epidermal cell was scored as a root-hair cell if any protrusion was visible, regardless of its length.

Plastic transverse sections were obtained from four- to five-day-old roots embedded in JB-4 resin and stained with 0.05% Toluidine Blue O, as previously described (Masucci and Schiefelbein,1996). The relative cell division rate in the H and N cell positions of the epidermis was determined by counting the number of cells in clones derived from rare longitudinal divisions, using a method previously described (Berger et al.,1998a), and by counting the number of cells in adjacent N and H cell files.

The histochemical analysis of plants containing the GUS reporter constructs was performed essentially as described(Masucci et al., 1996).

For RT-PCR assays, tissue of wild-type (Columbia) plants was ground in liquid nitrogen and total RNA was extracted as described(Weigel and Glazebrook, 2002). Tissue from roots and hypocotyl/cotyledons was obtained from four-day-old seedling grown on nutrient plates as described above. All other tissues were obtained from soil grown plants. RT-PCR was performed using the Superscript One-Step RTPCR Kit (Invitrogen) according to manufacturer instructions. Total RNA template (500 ng) was used for each reaction and a total of 40 PCR cycles was performed. UBQ10 gene-specific primers(Weigel and Glazebrook, 2002)were used in control reactions. The length of the gene-specific products obtained for GL3, EGL3 and UBQ was 581 bp, 516 bp and 483 bp, respectively.

Yeast two-hybrid assays were performed essentially as described(Lee and Schiefelbein, 1999). The entire coding regions of the GL3 or EGL3 cDNA were joined as C-terminal fusion to the yeast GAL4 DNA-binding domain in pGBT9 to generate the in-frame protein fusions BD-GL3 and BD-EGL3. The GAL4 transcriptional activation domain in pGAD424 was fused to the full-length WER-coding region to generate AD-WER(Lee and Schiefelbein, 1999)and it was fused to the full-length CPC coding region to generate AD-CPC. After transformation into yeast strain HF7c, the β-galactosidase assays were performed on at least six individual transformants for each combination of constructs.

Among the 133 predicted Arabidopsis proteins that possess bHLH-like motifs (Heim et al.,2003), the GL3 and EGL3 proteins are members of a subfamily with the greatest similarity to the maize R protein. The GL3 bHLH protein (At5g41315) contains 637 amino acids and displays 33% identity with maize R (Payne et al., 2000),whereas the EGL3 gene (At1g63650) encodes a predicted protein containing 596 amino acids that is 74% identical to GL3(Zhang et al., 2003). Because the maize R is able to influence root epidermis development when expressed in Arabidopsis (Galway et al., 1994), the GL3 and EGL3 genes were candidates to represent Arabidopsis bHLH proteins that participate in root epidermal cell specification.

To determine whether either of these R-like bHLH genes from Arabidopsis are normally expressed in the developing root, we conducted RT-PCR analyses using GL3- and EGL3-specific primers on RNA isolated from roots and other organs. GL3 and EGL3 amplified fragments were detected from each RNA sample(Fig. 1), indicating that each bHLH gene is expressed in all of these plant organs. This is consistent with the recent finding that GL3 and EGL3 participate in multiple pathways in the above-ground organs (Zhang et al., 2003). Furthermore, the substantial amplification of GL3 and EGL3 from root RNA samples(Fig. 1) suggests that these genes are expressed in developing Arabidopsis seedling roots.

To examine the possible role of GL3 and EGL3 in root epidermis development, we tested their effect when expressed under the control of the CaMV35S promoter. Plants bearing either the 35S::GL3 or 35S::EGL3 transgene (Payne et al., 2000; Zhang et al.,2003) produced roots with only a small number of root hairs(Fig. 2), owing to the misspecification of cells in the H position to adopt a non-hair cell fate(Table 1). This result is similar to the effect of the 35S::R construct in Arabidopsis(Galway et al., 1994), and it suggests that expression of a high level of GL3 or EGL3 bHLH protein throughout the epidermis can overcome the effects of the hair cell specification pathway in the H position.

Although the general phenotypic effect of the 35S::GL3 and 35S::EGL3 transgenes is similar, we found that the effect of the 35S::GL3 is not as strong as the 35S::EGL3(Fig. 2, Table 1). In particular, the 35S::GL3 line produces more hair cells in the lower region (near the apex) of the 4-day-old root than in the upper region (near the root-hypocotyl junction) (Fig. 2; Table 1). These differences in the 35S::GL3 and 35S::EGL3 effects may be due to different transgene expression levels or to differences in the developmental timing or interactions of the two bHLH genes.

To further explore the effect of the 35S::GL3 and 35S::EGL3 transgenes in relation to the previously characterized 35S::R, we introduced the 35S::GL3 and 35S::EGL3into the ttg-1 mutant. The ttg-1 root specifies hair cells in nearly every root epidermal cell, and this defect can be overcome by the 35S::R transgene (Galway et al.,1994). We found that the 35S::GL3 ttg-1 and 35S::EGL3 ttg-1 roots exhibited a significant reduction in hair cell production,when compared with the ttg mutant(Fig. 2; Table 1). This indicates that the overexpression of either gene can restore non-hair cell production in the ttg-1 mutant, which is similar to the effect of 35S::R on ttg (Galway et al.,1994). Interestingly, each transgene exhibited a difference in their effect on the upper and lower region of the root, with the 35S::EGL3 having its greatest impact on the lower region and the 35S::GL3 on the upper region, which is similar to their effects in the wild-type background. Because neither of the transgenes was able to induce non-hair cell specification in the ttg-1 mutant to the same extent as they do in the wild-type background (Table 1), it is likely that TTG is required for the full effect of the 35S::GL3 and 35S::EGL3. To determine whether this partial TTG dependency can be diminished by expressing both GL3 and EGL3 in the ttg mutant, we constructed a 35S::GL3 35S::EGL3 ttg-1 line. These roots had an enhanced non-hair cell phenotype when compared with either single transgene(Table 1), suggesting that increased expression of these bHLH genes can overcome the effect of the ttg-1. Furthermore, the lack of a synergistic effect implies that the GL3 and EGL3 provide largely similar functions, rather than interdependent functions.

To directly assess the involvement of the GL3 and EGL3genes in root epidermis development, we analyzed plants bearing homozygous mutations in one or both of these genes. We employed gl3 mutant lines(gl3-1 and gl3-2) and egl3 mutant lines(egl3-1 and egl3-2) with mutations that cause premature stop codons and probably represent null alleles(Payne et al., 2000; Zhang et al., 2003). We found that each of the single mutant lines produced a normal number and pattern of epidermal cell types in the lower region of the root, but they show a slight(egl3-1 and egl3-2) or moderate (gl3-1 and gl3-2) increase in hair cell production in the upper region of the root, owing to the misspecification of hair cells in the N (ectopic) position(Fig. 3; Table 2). Thus, GL3and EGL3 are necessary to specify fully the non-hair cell fate and generate the proper epidermal pattern in the upper region, but not the lower region, of the root.

To test the possibility of partial functional redundancy between the GL3 and EGL3 genes, we generated and analyzed all four possible gl3 egl3 double mutants using the gl3-1, gl3-2,egl3-1 and egl3-2 lines. Each double mutant combination produced an extremely hairy root (Fig. 3) because of a dramatic reduction in the frequency of the non-hair cell type throughout the root(Table 2; data not shown). These findings show that GL3 and EGL3 act in a largely redundant manner to specify the non-hair cell fate.

The outgrowth of a root hair from an epidermal cell represents a relatively late event in epidermal cell differentiation. At earlier stages, immature epidermal cells in the H and N positions may be distinguished from one another by their differential vacuolation rate and cytoplasmic density, and these characteristics are controlled by WER and TTG but not GL2(Galway et al., 1994; Lee and Schiefelbein, 1999; Masucci et al., 1996; Schellmann et al., 2002). To determine whether the altered root hair production in the gl3 egl3mutant and the 35S lines were associated with cell fate abnormalities at an early developmental stage, we examined vacuole formation and cytoplasmic density in developing epidermal cells from transverse sections taken from the meristematic region of the root. In contrast to the wild type, which displayed a greater vacuolation rate and reduced cytoplasmic density in the N cell position relative to the H cell position, all epidermal cells in the gl3 egl3 exhibit characteristics of developing hair cells, whereas all epidermal cells in 35S::EGL3 exhibit characteristics of developing non-hair cells (Fig. 4).

Another early characteristic of epidermal cell specification is differential cell division rate, whereby the developing hair cells achieve a greater rate of division than developing non-hair cells in the meristematic zone (Berger et al., 1998a). We assessed the relative cell division rate in the mutant and overexpression lines by comparing cell number in the H and N positions. We discovered a significant reduction in the relative division rate in the gl3 egl3,the 35S::GL3 and the 35S::EGL3 lines when compared with their respective wild-type lines, but no significant difference was detected in the single mutant lines (Table 3). The reduction was similar to the previously documented effect of the wer-1 and ttg-1 mutations on relative cell division rate (Galway et al., 1994; Lee and Schiefelbein, 1999)(Table 3). Together, these results show that alterations in GL3 and EGL3 gene function affect cell specification characteristics at an early stage in root epidermis development, similar to the stage affected by WER and TTG.

To better define the role of GL3 and EGL3 in the specification pathway, we examined the effect of the mutants and transgenes on expression of the GL2::GUS reporter construct. The GL2 gene is required for non-hair cell specification and, accordingly, the GL2::GUS reporter is preferentially expressed in the N cell position of the developing root epidermis (Masucci et al., 1996). In the gl3-1 mutant, GL2::GUSexpression is reduced, but the appropriate pattern of GUS activity remains(Fig. 5). By contrast, the egl3-1 mutant has no detectable effect on GL2::GUS reporter expression (Fig. 5). A strong reduction in GL2::GUS expression is present in the gl3-1 egl3-1 double mutant, although extended incubation time shows that the proper pattern persists (Fig. 5). This indicates that GL3 and EGL3 act redundantly to positively regulate the level, but not the position-dependent pattern, of GL2transcription.

In a corresponding fashion, the 35S::EGL3 transgene caused GL2::GUS expression to expand throughout the developing epidermis and the surrounding lateral root cap (Fig. 5; data not shown). The 35S::GL3 caused a weaker effect on GL2::GUS expression (Fig. 5), consistent with its weaker effect on cell fate specification(Table 1).

Given these results, we wished to investigate the possibility that GL3 and EGL3 promote the non-hair fate by acting through GL2. Therefore, we generated and analyzed 35S::GL3 gl2-1 and 35S::EGL3 gl2-1 plants. Each of these lines possessed a `hairy' root phenotype that is essentially the same as the gl2-1 mutant (Table 1). This suggests that a functional GL2 gene is required for the 35S overexpression constructs to induce non-hair epidermal cells, and therefore, is consistent with the notion that the GL3/EGL3 genes act through GL2.

The CPC gene is expressed in the N cell position, and it is required for hair cell specification through a lateral inhibition mechanism(Lee and Schiefelbein, 2002; Wada et al., 2002). To determine whether CPC is regulated by GL3 and EGL3,we introduced the CPC::GUS reporter construct into the various mutant and transgene backgrounds. Like the GL2::GUS expression, the CPC::GUS expression was reduced in the gl3-1 mutant,unchanged in the egl3-1 mutant, and virtually eliminated in the gl3-1 egl3-1 double mutant (Fig. 6; data not shown). Furthermore, the CPC::GUS reporter was expressed throughout the epidermis in the 35S::EGL3, and to a weaker extent in the 35S::GL3(Fig. 6). Thus, the GL3 and EGL3 genes act in a redundant manner to promote expression of both the non-hair-cell-specification gene GL2 and the hair-cell-specification gene CPC in the N cell position.

Like GL3 and EGL3, WER is a positive regulator of GL2 and CPC (Lee and Schiefelbein, 2002). To determine whether WER is required for GL3 or EGL3 function, we constructed and examined plants bearing the genotype 35S::GL3 wer-1 or 35S::EGL3 wer-1. Roots from each of these lines produced abundant root hair cells, similar to the wer-1 mutant (Table 1). Thus, GL3 or EGL3 action requires a functional WER gene, which implies that WER is either acting downstream or at the same step as GL3/EGL3 in the non-hair specification pathway.

In prior studies, the maize R bHLH protein has been found to interact with both the WER (Lee and Schiefelbein,1999) and with the CPC (Wada et al., 2002) MYB proteins. To examine the possibility that the GL3 or EGL3 proteins physically associate with WER or CPC, we employed the yeast two-hybrid assay (Fields and Sternglanz, 1994). First, we found that fusions of the GAL4 DNA binding domain (BD) to either the GL3 or EGL3 protein alone were sufficient to induce a significant level of lacZ reporter expression(Table 4). This `one-hybrid'assay shows that the GL3 and EGL3 proteins possess transcriptional activation domains that are functional in yeast.

Next, we discovered that yeast cells co-expressing either the BD-GL3 or BD-EGL3 fusions together with the AD-WER fusion exhibited a higher level of lacZ reporter expression than either single fusion(Table 4). This indicates that the WER protein can physically interact with either the GL3 or EGL3 proteins in yeast cells and is consistent with the possibility that they interact in Arabidopsis.

We also discovered that GL3 and EGL3 can each associate with the CPC protein in the yeast two-hybrid assay. Yeast cells expressing an AD-CPC fusion together with the BD-GL3 or the BD-EGL3 produce an increased level of lacZ reporter expression (Table 4). This implies that CPC may also interact with GL3 or EGL3 in Arabidopsis, which suggests a possible competition model for the opposite action of WER and CPC in root epidermis cell specification.

In this study, the Arabidopsis GL3 and EGL3 bHLH genes have been shown to participate in root epidermis development in a largely redundant manner and consistent with expectations from earlier experiments with the heterologous maize R bHLH(Galway et al., 1994). Results from mutant, overexpression and reporter analyses suggest that the major role of GL3/EGL3 is to promote the non-hair cell fate via transcriptional activation of the downstream gene GL2. In addition,these genes are required for appropriate expression of the CPC gene,which promotes the hair cell fate. Thus, GL3 and EGL3 are essential for the specification of both cell fate pathways that generate the normal cell type pattern during root epidermis development.

These results suggest a simple model for the action of the GL3 and EGL3 bHLH proteins in specifying cell fates in the root epidermis(Fig. 7). This model is largely consistent with earlier predictions based on results with the heterologous maize R protein (Galway et al.,1994; Lee and Schiefelbein,1999; Masucci et al.,1996). First, the GL3 and EGL3 bHLH proteins are likely to act as transcriptional regulators in concert with the WER MYB protein. This proposal is supported by the similar effects of WER and GL3/EGL3 on GL2 and CPC expression, and their similar effects on early stages of cell differentiation. Furthermore, it is consistent with the yeast two-hybrid results showing that GL3 and EGL3 interact with WER, and it is supported by the WER-dependent nature of the 35S::GL3 or 35S::EGL3induction of non-hair cells. In plants, it is common for bHLH proteins to act in a combinatorial fashion with MYB-related proteins to regulate gene transcription (Singh, 1998). The best characterized example is the control of anthocyanin production in maize, where the tissue-specific activation of the structural genes of the anthocyanin pathway requires the expression of a bHLH protein encoded by the maize R or B loci as well as a MYB-related protein encoded by the C1 or Pl loci (Ludwig and Wessler,1990; Mol et al.,1998).

Another aspect of our model (Fig. 7), is that the GL3/EGL3 proteins interact with the TTG WD repeat protein and rely on TTG activity, in part, for their function. Although the precise role of TTG is not clear, other WD-repeat proteins are involved in protein-protein interactions (Neer et al.,1994) which implies that TTG may be a component of a signal transduction pathway or may interact with transcription factors (e.g. GL3/EGL3) that specify epidermal cell fate. A close connection between the GL3/EGL3 bHLH proteins and the TTG protein is suggested by several lines of evidence. First, both are essential at an early stage of development, because they each alter all aspects of non-hair cell differentiation (including early developmental characteristics like cell division rate, cytoplasmic density and vacuolation rate) (Berger et al.,1998a; Galway et al.,1994) (Fig. 4). Second, the ttg mutations and the gl3 egl3 double mutant both exhibit a dramatic reduction in GL2 expression, but retain its position-dependent pattern, which implies that neither function is essential for the generation of the GL2 expression pattern but both are required for the proper expression level. In addition, prior studies showed that GL3 and EGL1 can interact with TTG in yeast(Payne et al., 2000; Zhang et al., 2003). Finally,the ability of the 35S::GL3 or 35S::EGL3 to largely complement the N cell defect in the ttg mutant suggests that TTG may enhance the abundance, activity or localization of these bHLH proteins, and this influence can be overcome by overexpression of the bHLH genes. However,the 35S::GL3 and 35S::EGL3 were unable to induce non-hair cell production in the ttg mutant to the same extent as in the wild type (Table 1), so TTG is likely to be partially required for the function of these transgenes.

Another part of our model (Fig. 7) is that the GL3/EGL3 bHLH proteins interact with the CPC protein in the H position, and this interaction leads to specification of the hair cell fate. This proposal is derived from two lines of evidence. First,CPC physically interacts with GL3 or EGL3 in yeast(Table 4). Second, the ability of the 35S::GL3 and 35S::EGL3 lines to induce GL2expression, CPC expression and the non-hair cell fate in the H position implies that a high concentration of GL3 or EGL3 is sufficient to alter the fate of the H cells and convert them into non-hair cells. Considering that WER is required for this effect(Table 1) and therefore must be produced/available in the H position, it is possible that, in wild-type roots,the concentration of GL3/EGL3 available for interaction with WER is low because most of it is bound to CPC. The 35S::GL3 and 35S::EGL3 phenotypes may then be explained because the excess supply of these bHLH proteins enables a significant accumulation of the functional WER-bHLH complex even in the presence of the CPC inhibitor. This explanation is consistent with a competition mechanism for epidermal patterning that is essentially similar to one proposed earlier(Lee and Schiefelbein, 1999). Accordingly, the epidermal pattern is determined by the relative concentration of a functional two-repeat MYB (WER) versus an incomplete one-repeat MYB that lacks a transcriptional activation domain (CPC and probably also TRY). Each of these MYBs is envisioned to compete for binding to a limited supply of the GL3/EGL3 bHLH proteins, with the WER-bHLH interaction leading to a functional transcriptional complex that activates GL2 and CPC, whereas the CPC-bHLH interaction generates a non-functional complex that leads to hair cell specification by default (Fig. 7). The cell-type pattern may then result from positional cues and gene regulatory networks that generate a relatively high concentration of WER in the N position and a relatively high concentration of CPC (and probably TRY) in the H position (Fig. 7). We are currently testing various predictions of this model.

Although we have focused our attention on the seedling root, it is likely that the action of the GL3 and EGL3 genes is initiated during embryonic root development, because the epidermal pattern is known to be established during embryogenesis and each of the other regulators is active during that period (Costa and Dolan,2003; Lin and Schiefelbein,2001). It is also likely that GL3 and EGL3 help to establish epidermal cell fate in the hypocotyl because, to date, all of the regulators that have been examined alter epidermal patterning in the root and hypocotyl (Berger et al.,1998b; Hung et al.,1998; Lee and Schiefelbein,1999). Future studies will be aimed at testing these predictions.

In this study, we detected a significant difference in the epidermal cell pattern in the upper and lower regions of the 4-day-old seedling roots in several of the bHLH mutants and transgenic lines (Tables 1, 2). The cells that comprise the upper region are largely formed during embryogenesis and have been termed the`embryonic root' (Dolan et al.,1994; Lin and Schiefelbein,2001; Scheres et al.,1994). It is therefore possible that the GL3 and/or EGL3 genes or gene products have a different role in epidermal patterning during embryonic versus post-embryonic development. For example,the GL3 and EGL3 may differ in their degree of redundancy or their putative partner proteins in a developmentally dependent manner. Alternatively, it is possible that epidermal patterning in this region of the root is generally less `tightly regulated' by the position-dependent mechanism and therefore more sensitive to genetic perturbation, owing to the proximity of this region to the root-hypocotyl junction (collet), where every epidermal cell adopts the hair fate (Dolan et al.,1994; Lin and Schiefelbein,2001; Scheres et al.,1994).

The patterning of epidermal cells in the root appears to employ a mechanism similar to the one used in the shoot to control trichome distribution(Larkin et al., 2003; Schiefelbein, 2003). This similarity extends to the use of the GL3 and EGL3 proteins, which have been shown to participate in trichome specification in a regulatory network resembling the one described here (Payne et al., 2000; Zhang et al.,2003). Furthermore, these bHLH proteins participate with TTG in seed coat development and anthocyanin production(Zhang et al., 2003), which suggests that a common transcriptional cassette operates in all of these processes and confirms predictions made from studies with the heterologous maize R protein in Arabidopsis(Galway et al., 1994; Lloyd et al., 1992).

Our work shows that the GL3 and EGL3 bHLH genes act in a largely redundant fashion to influence epidermal cell specification in the root. The lack of a major effect of either single homozygous mutant and the lack of a synergistic effect in the 35S::GL3 35S::EGL3 line indicates that the GL3 and EGL3 proteins function in a similar manner. There are two other Arabidopsis bHLH genes, MYC1(Urao et al., 1996) and TT8 (Nesi et al.,2000), that are related to the maize R and in the same subgroup as GL3 and EGL3(Heim, 2003). In preliminary studies, we have found that at least the MYC1 gene probably participates in root epidermal patterning (C.B., M. Sridharan and J.S.,unpublished). Thus, an unexpectedly large collection of bHLH genes may play a role in the specification of epidermal cell fate in the Arabidopsisroot. This probably reflects the importance of genetic redundancy and the complex regulatory nature of cell specification in higher plants.

We thank M. Simon, M. D. Marks and J. Larkin for helpful comments. We acknowledge the Arabidopsis Biological Resource Center (Ohio State University,Columbus, OH, USA) for providing the gl3-1 mutant. This work was supported by grants from the United States Department of Agriculture(IBN-0077900) to J.S. and from the US National Science Foundation(IBN-9986391) to A.M.L.