The bHLH genes GL3 and EGL3 participate in an intercellular regulatory circuit that controls cell patterning in the Arabidopsis root epidermis

The determination of distinct cell fates in appropriate patterns is a crucial feature of development in multicellular organisms. In plants, the formation of the root hair and non-hair cells in the root epidermis has been used extensively as a simple and experimentally tractable model for studying cell fate patterning (Dolan and Costa,2001; Larkin et al.,2003). The root epidermal cell types are not essential for growth under laboratory conditions, they are easy to examine, and they arise continuously during development.

In Arabidopsis, root hair cells are specified in a position-dependent manner, such that all cells located in a cleft between two underlying cortical cells (designated the H position) develop as hair cells and cells located outside a single cortical cell (designated the N position)adopt the non-hair fate (Dolan et al.,1994; Galway et al.,1994). Molecular genetic studies have shown that a suite of putative transcription factors regulates the patterning of root hair cells in Arabidopsis. These factors include a homeodomain protein, GLABRA2(GL2) (Masucci et al., 1996; Rerie et al., 1994); a WD-repeat protein, TRANSPARENT TESTA GLABRA (TTG)(Galway et al., 1994; Walker et al., 1999); an R2R3 MYB-type transcription factor, WEREWOLF (WER)(Lee and Schiefelbein, 1999);two closely related basic helix-loop-helix proteins, GLABRA3 and ENHANCER OF GLABRA3 (Bernhardt et al.,2003); and three small MYB proteins, CAPRICE (CPC), TRIPTYCHON(TRY) and ENHANCER OF TRIPTYCHON AND CAPRICE (ETC1)(Kirik et al., 2004; Schellmann et al., 2002; Wada et al., 2002; Wada et al., 1997). The GL2,TTG, WER, GL3 and EGL3 appear to have a primary role in promoting the non-hair fate, whereas the CPC, TRY and ETC1 are most important in specifying the hair cell fate. The cell pattern is proposed to result from a lateral inhibition mechanism that is mediated by CPC, TRY and ETC1(Larkin et al., 2003; Schiefelbein, 2003). The transcription of CPC (and presumably of TRY and ETC1) is promoted by a putative complex of TTG, WER, GL3 and EGL3 in the N cell position, and these small MYB proteins inhibit the neighboring H cells from adopting the non-hair fate, possibly by directly moving from cell-to-cell and interfering with the WER function(Schellmann et al., 2002; Schiefelbein, 2003; Wada et al., 2002; Wada et al., 1997).

Although the importance of the putative complex containing GL3, EGL3, WER and TTG is clear, the mechanisms that regulate the accumulation of these components is poorly understood. In this study, we sought to analyze the expression and regulation of GL3 and EGL3 during root epidermis development. In prior work, we have analyzed mutants and overexpression lines to show that GL3 and EGL3 are likely to act redundantly to help specify both the hair and non-hair cell fates(Bernhardt et al., 2003). The gl3 egl3 double mutant was shown to have excessive root-hair cells,while the overexpression of these genes caused an increased frequency of non-hair cells. Furthermore, these bHLH proteins are required for the positive transcriptional control of GL2 (which specifies the non-hair fate)and CPC (which helps specify the hair cell fate), and they interact with the WER and the CPC MYB proteins in yeast(Bernhardt et al., 2003). The prediction from this work was that these bHLH genes might be expressed (and their gene products accumulate) in both the H and N cells. Here, we use RNA hybridization, promoter reporter fusions and genetic analyses to show that the GL3/EGL3 genes are preferentially expressed in the H position, and furthermore, that this expression pattern is controlled by several of the known cell fate regulators, including the GL3/EGL3 proteins themselves. Using a YFP translational fusion, we find that GL3 accumulates in the nuclei of the N cells. These results suggest a new feedback loop in the epidermal regulatory network that helps establish and reinforce the cell fate pattern.

The isolation of the mutant alleles used in this study has been described: gl3-1 and gl3-2, both in the Ler background(Koornneef, 1982; Payne et al.,2000); egl3-1, in the Ler genetic background(Zhang et al., 2003); wer-1, in the Columbia ecotype(Lee and Schiefelbein, 1999); ttg-1, in the Ler genetic background(Galway et al., 1994); cpc-1, in the WS genetic background(Wada et al., 1997); try-82, in the Ler genetic background(Hulskamp et al., 1994); gl2-1, in the Ler genetic background (Koornneef, 1982); rhd6-1, in the WS ecotype(Masucci and Schiefelbein,1994).

The following transgenic lines have been described previously: GL3::GUS and EGL3::GUS(Zhang et al., 2003), GL2::GUS (Masucci et al.,1996), CPC::GUS (Wada et al., 2002), 35S::GL3(Payne et al., 2000), 35S::EGL3 (Zhang et al.,2003), 35S::CPC (Wada et al., 1997) and 35S::TRY(Schellmann et al., 2002).

Lines homozygous for multiple mutations and/or transgenes were constructed by crossing single mutant or transgenic plants, examining the F2 progeny for putative mutant phenotypes, and confirming the desired genotype in subsequent generations by backcrossing to single mutants and/or PCR-based tests.

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

The histochemical analysis of plants containing the GUS reporter constructs was performed on at least 20 four-day-old root tips for each strain essentially as described (Masucci et al.,1996). Root epidermal cells were deemed to be in the N position if they were located outside a periclinal cortical cell wall, whereas cells in the H position were located outside a radial wall between adjacent cortical cells.

The distribution of epidermal cell types in the hypocotyl was analyzed in 25 nine-day-old seedlings for each strain by determining the number of stomata formed along the hypocotyl in two adjacent epidermal cell files, one located over anticlinal cortical cell walls and one located over periclinal cortical cell walls.

The whole-mount in situ RNA hybridization procedure has been described(de Almeida Engler et al.,1994). The RNA probe was designed to hybridize to both GL3 and EGL3 transcripts, so it included bp 550-1120 and bp 1400-1850 downstream from the start site of the EGL3-coding sequence corresponding to the most similar region of the GL3 and EGL3 proteins but excluding the bHLH signature region to eliminate the possibility of cross-hybridization to other bHLH proteins.

To construct the GL3::GL3-YFP translational fusion, pD2L-2(Payne et al., 2000) was used to provide a GL3 genomic DNA fragment containing the entire GL3 gene, including ∼1 kb upstream of the start codon and 1 kb downstream of the stop codon. The existing SacI and SalI sites of pD2L-2 were destroyed and new SacI and SalI sites were generated by inverse PCR at the 3′ end of the GL3-coding region. The EYFP-coding region was amplified from pEYFP (Clontech)and fused in-frame to the GL3 3′ end. A BamHI fragment from this vector, containing the entire GL3::EYFP fusion, was subcloned into the BglII site of the T-DNA vector pAL47(Lloyd and Davis, 1994). The GL3-YFP fusion protein is predicted to be 96 kDa, whereas the predicted size of the GL3 and EGL3 proteins is 70 kDa and 66 kDa, respectively. Plant transformation was performed by the floral dip method(Clough and Bent, 1998).

Our earlier results led to the suggestion that the GL3 and EGL3 genes are expressed in both the developing hair cells and non-hair cells (Bernhardt et al.,2003). To test this, we generated and analyzed promoter-GUS fusions for each gene (GL3::GUS and EGL3::GUS). In each line, the maximum GUS activity was observed in the meristematic region of the developing root tip (Fig. 1). This temporal pattern is similar to that previously reported for WERand earlier than for GL2 (Lee and Schiefelbein, 1999; Masucci et al., 1996), which is consistent with the view that the bHLH genes act at the same stage as WER and regulate GL2 expression(Bernhardt et al., 2003). However, the GL3::GUS and EGL3::GUS showed significantly higher levels of GUS activity in cells located in the H position than the N position (Fig. 1), although longer incubation times also caused some GUS staining in the N cells. The cells preferentially expressing these reporters were determined to be in the H position because the stained cell files overlie radial cortical cell walls,they show a higher cell division rate than unstained files (see magnified view in Fig. 1), and they can be traced to mature cells that are hair forming. Thus, despite the important role of GL3 and EGL3 in specifying the non-hair fate in the N position, these genes are preferentially expressed in the H cell position.

To determine whether the unexpected bHLH promoter activity pattern was associated with a similar bHLH RNA accumulation pattern, we conducted in situ RNA hybridization experiments. Because of the high degree of sequence similarity throughout the GL3 and EGL3 genes, it was not possible to generate viable gene-specific antisense RNA probes. Rather, we designed an antisense RNA probe corresponding to a conserved region in GL3/EGL3, so that the probe would recognize both genes, but not other related bHLH genes in Arabidopsis. Using whole roots from four-day-old wild-type seedlings, we found that this probe preferentially hybridized to cells in the H position (Fig. 2A). Again, the position of the cell files was determined by their location relative to the underlying cortical cell walls and their differential cell division rate (Fig. 2A). This result was reinforced by the analysis of bHLH RNA accumulation in rare epidermal clones (Fig. 2A, last panel). These clones form when occasional longitudinal anticlinal divisions in cells in the H position create two cell files; one located over the anticlinal cortical cell wall (the H position) and one overlying a periclinal cortical cell wall (the N position) (Berger et al.,1998a). In these clones, we found that cells in the N position no longer accumulate GL3/EGL3 RNA(Fig. 2B). As a control, roots hybridized with the sense RNA strand as a probe yielded no signal(Fig. 2A, first panel).

To confirm that the antisense RNA probe indeed recognizes both gene sequences and to determine whether both RNAs accumulate in the same pattern,we also tested roots from the gl3-2 and egl3-1 single mutants and the gl3-2 egl3-1 double mutant. Both single mutants showed a similar but weaker hybridization pattern as the wild-type roots,while gl3-2 egl3-1 double mutant roots displayed a very low level of signal (Fig. 2B). This suggests that the antisense probe primarily detects GL3 transcripts in the egl3 mutant and EGL3 transcripts in the gl3 mutant. This conclusion is supported by gene-specific RT-PCR experiments that show a reduced level of GL3 RNA in the roots of gl3-2 and gl3-2 egl3-1 relative to egl3-1 and a reduced level of EGL3RNA in the roots of egl3-1 and gl3-2 egl3-1 relative to gl3-2 (data not shown). Together, these data imply that the GL3 RNA and EGL3 RNA exhibit the same pattern of preferential accumulation in the H cell. These results support the reporter fusion results and suggest that the GL3 and EGL3 RNAs do not move from cell-to-cell.

The position-dependent patterning of root epidermal gene expression is known to be established during early stages of embryogenesis(Costa and Dolan, 2003; Lin and Schiefelbein, 2001). To explore further the unexpected expression pattern of GL3 and EGL3, we examined the GL3::GUS and EGL3::GUSactivity during embryogenesis. We found that GUS activity accumulated in the same H cell pattern in the epidermis of the mature embryonic root as in the postembryonic root (Fig. 3A).

Considering this embryonic expression, we wished to determine whether the GL3 and EGL3 genes act in a similar manner in the embryonic and postembryonic root. Specifically, we examined whether embryonic GL2 and CPC gene expression (using the GL2::GUS and CPC::GUS reporters) is reduced in a gl3 egl3 mutant, as they are in the seedling root (Bernhardt et al.,2003). Indeed, we observed significantly less GUS activity in the root of the gl3 egl3 GL2::GUS mature embryos(Fig. 3B). We were unable to detect any GUS expression in the mutant or wild-type embryonic root bearing the CPC::GUS reporter, which prevented an assessment of the role of GL3/EGL3 on CPC (Fig. 3B).

The hypocotyl is known to be patterned by a similar network of transcription factors as the root to define two position-dependent cell types:cells capable of producing stomata and non-stomata cells(Berger et al., 1998b; Hung et al., 1998). In gl3 egl3 embryos bearing either the GL2::GUS or the CPC::GUS reporters, we detected a significantly reduced level of GUS activity in the embryonic hypocotyl relative to the wild type(Fig. 3B). This dependence of GL2 and CPC expression on the presence of GL3/EGL3 in the embryonic hypocotyl suggested a possible role for both proteins during stomatal patterning in the hypocotyl. To explore this further, we analyzed stomata distribution in the hypocotyls of nine-day-old seedlings of wild type and gl3 egl3 double mutants. In contrast to the wild type, which has stomata predominantly form over a radial cortical cell wall (analogous to the H position of the root epidermis), the gl3 egl3 hypocotyls have an increased number of stomata and they are present in both positions(Table 1). This shows that GL3 and EGL3 are required for proper stomata patterning, and it strengthens the close relationship between the mechanism for patterning the hypocotyl and root epidermis in Arabidopsis.

We wished to determine whether any of the transcription factors in the epidermal cell fate network might regulate the expression of the GL3and EGL3 genes. Because the EGL3::GUS line exhibited a greater expression level than the GL3::GUS line, thus allowing for better analysis of weaker effects, we first focused our efforts on studying the control of the EGL3 expression pattern.

We first examined the possible role of the upstream regulators in this system, WER and TTG. The EGL3::GUS transgene was introduced into the wer-1 and ttg-1 mutant backgrounds, and we found that each of the homozygous mutants exhibit ectopic GUS expression in the N cell position (Fig. 4A). This suggests that WER and TTG negatively regulate EGL3 expression in the N cells.

Next, we tested the possible effects of the lateral inhibitors CPC and TRY. We found that neither cpc nor try mutants significantly altered EGL3::GUS expression, but the cpc try double mutant exhibited essentially no GUS activity (Fig. 4B). We also introduced the EGL3::GUS reporter construct in both the 35S::CPC and 35S::TRY background and found that overexpression of each gene induced ectopic EGL3::GUS expression in the N cell position (Fig. 4B). As CPC and TRY are related genes known to act in a partially redundant fashion (Schellmann et al.,2002; Wada et al.,2002), this indicates that the CPC/TRY function is required for EGL3 gene expression in the H position.

We next tested the effect of gl2 and rhd6 mutants on EGL3 expression. Results from mutant and overexpression analyses indicate that GL3/EGL3 act at an early stage of root epidermis development(Bernhardt et al., 2003), so we expected no effect of mutations in the later acting genes GL2 and RHD6. Consistent with this, the EGL3::GUS pattern in gl2 and rhd6 was indistinguishable from the wild type(Fig. 4A).

Last, we examined the possibility that the GL3/EGL3 proteins themselves might regulate EGL3 promoter activity. Although roots of the gl3 EGL3::GUS and egl3 EGL3::GUS lines exhibit only weak ectopic GUS activity, the gl3 egl3 EGL3::GUS double mutant line possessed strong GUS expression in the N position (Fig. 4C). Furthermore, overexpression of either bHLH gene (via the 35S::GL3 and 35S::EGL3 constructs) caused a weak or modest reduction in EGL3::GUS expression. These results suggest an autoregulation of EGL3 expression: the EGL3 protein (together with GL3) is able to inhibit its own gene's promoter activity.

To determine whether the GL3 promoter activity is regulated in a similar manner as EGL3, we introduced the GL3::GUS reporter into selected genetic backgrounds. We found that the wer mutant, the gl3 egl3 double mutant, and the 35S::CPC overexpression construct caused ectopic GL3::GUS expression, whereas the cpc try double mutant exhibited a much lower level of GL3::GUSexpression (Fig. 5). In addition, GL3::GUS activity was not effected by the cpcsingle mutant or the gl2 mutation (data not shown). These results are all consistent with the effects of these factors on EGL3::GUS, and it suggests that both genes are regulated similarly.

In addition to assessing the role of the cell fate transcription factors on promoter activity, we wished to examine their effects on GL3/EGL3 RNA accumulation. Using whole-mount RNA in situ hybridization, we analyzed the hybridization of an antisense probe (which can hybridize to both GL3 and EGL3 RNA) to roots from the wer, cpc tryand 35S::CPC. We found, relative to the wild type, ectopic signal in the wer mutant and the 35S::CPC line, and very low level of signal from the cpc try double mutant(Fig. 6). These results provide support for the promoter-reporter fusion studies above, and show that WER negatively regulates GL3/EGL3 RNA accumulation in the N position and CPC/TRY positively regulates GL3/EGL3 RNA accumulation in the H position.

To determine the location of the GL3/EGL3 bHLH proteins during epidermis development, we constructed a GL3-YFP translational fusion under the control of the GL3 promoter (GL3::GL3-YFP). This construct was introduced into both gl3-1 and gl3-2 mutant plants. These transgenic plants showed normal root hair patterning and trichome formation (data not shown), indicating that the GL3-YFP fusion was functional. Transgenic plants expressing this construct displayed YFP expression predominantly in the nuclei of developing epidermal cells in the N position (as determined by their relative cell length, Fig. 7). We also verified this cell location by tracing the cells in the YFP-expressing files back to non-hair cells in the mature portion of the root. This suggests that the GL3 protein accumulates in the N cell, although the GL3 promoter activity and GL3 RNA accumulation is primarily in the H cell.

The GL3 and EGL3 genes encode bHLH proteins essential for the proper specification of both epidermal cell types in the Arabidopsis root. Prior work showed that they probably act together with the WER MYB and perhaps TTG WD-repeat protein to promote transcription of GL2 (to specify non-hair fate) and CPC in the N cells of the seedling root (Bernhardt et al.,2003). In the H cells, they appear to interact with the CPC (and possibly other related small MYBs such as TRY and ETC1), rather than WER, and cause these cells to adopt the hair fate(Bernhardt et al., 2003). Here,we extend the role of GL3 and EGL3 by showing that they are also required for proper stomata patterning in the hypocotyl, because the gl3 egl3mutant lacks an appropriate stomata distribution. Furthermore, we find that GL3 and EGL3 probably begin to act during embryogenesis, by inducing GL2 and CPC expression in the N cell position to help establish the cell pattern.

Despite their important role in specifying both epidermal cell types, we used promoter reporter fusions to show that GL3 and EGL3 are preferentially expressed in the developing hair cells. How might GL3/EGL3 affect specification of the N cells while being expressed predominantly in the H cells? One possibility is that the GL3 and EGL3 proteins may not directly function in the N cells but rather act through another yet unidentified factor involved in a lateral signaling from H to N. However, this possibility is not supported by yeast two-hybrid data, which suggest a direct interaction between the WER and GL3/EGL3 (Bernhardt et al.,2003). In a second scenario, GL3 and EGL3 could act in a non-cell autonomous manner in that either the proteins themselves or their RNA is moving from cells in the H position to cells in the N position. Our results support this second scenario. Although in situ hybridization analysis indicated that GL3 and EGL3 RNAs accumulate preferentially in cells in the H position (similar to results from the promoter reporter fusion analysis), a GL3-YFP construct was detected predominantly in the N cell position. Thus, the requirement for GL3/EGL3 activity in the developing N cells may be fulfilled by the movement of these proteins from H cells to N cells.

It is likely that the H-cell specific expression pattern of the GL3 and EGL3 genes is due, at least in part, to negative autoregulation at the transcriptional level. We found that functional GL3 and EGL3 genes are required to inhibit GL3/EGL3gene transcription and RNA accumulation in the N position. We also found that overexpression of GL3 or EGL3 causes modest reduction in EGL3 transcription in the H position. Furthermore, we show that the putative partners of GL3/EGL3 action, WER and TTG, are also required to inhibit GL3/EGL3 in the N cells. Taken together, these data indicate that the GL3 and EGL3 gene transcription is negatively regulated by the putative WER/GL3/EGL3/TTG complex, which is likely to be most abundant in the N cell position.

However, the GL3 and EGL3 genes were found to be positively regulated by the CPC and TRY proteins, which act in the H cell. Functional CPC/TRY genes are necessary for GL3/EGL3expression in the H position, and overexpression of the CPC/TRY genes cause ectopic GL3/EGL3 promoter activity and RNA accumulation in the N position. This may be due, in part, to the ability of CPC (and possibly TRY)to inhibit WER gene expression in the H position(Lee and Schiefelbein, 2002),which would reduce the abundance of the WER complex and thereby indirectly increase GL3 and EGL3 transcription.

Together, these results lead us to make the following proposal(Fig. 8). In the N cell, the abundant WER protein accumulation (presumably owing to the action of positional cues) leads to the formation of sufficient WER/GL3/EGL3/TTG complex to induce expression of the N-cell fate-promoting factor GL2 and the lateral inhibitor CPC (Bernhardt et al., 2003; Lee and Schiefelbein, 2002). The CPC protein (and possibly TRY) moves to neighboring cells in the H position leading to the formation of an inactive CPC/GL3/EGL3/TTG complex that prevents activation of GL2, thus allowing for the specification of the hair cell fate(Bernhardt et al., 2003; Lee and Schiefelbein, 2002; Wada et al., 2002). At the same time, while the accumulation of CPC (and possibly TRY) in H cells leads to a reduction in WER (and CPC) expression, it also leads to an increase in GL3 and EGL3 expression (this study). The GL3(and likely EGL3) protein then acts in a lateral feedback loop by moving to the neighboring N cells (possibly through plasmodesmata). This process is likely to be efficient (perhaps driven by the constant removal of free bHLH protein out of the equilibrium by binding to WER), as the GL3-YFP fusion protein is found predominantly in N cells rather than being distributed in equal intensity in all cell files. The additional GL3/EGL3 protein would generate more of the WER/GL3/EGL3/TTG complex in the N cells, inducing additional GL2 and CPC expression and also repression of GL3 and EGL3 expression. In the end, this would mean that essentially all of the GL3/EGL3 protein used in the N-cell complex formation would come from the H cells.

An important aspect of the proposal above is that CPC (and probably TRY)effectively acts as a positive regulator of GL3/EGL3transcription in the H position. Prior work has shown that CPC gene expression requires GL3/EGL3 action in the N position(Bernhardt et al., 2003). Taken together, this implies that expression of GL3/EGL3 and CPC is regulated by a reciprocal positive feedback circuit across adjacent cells. This mechanism of mutual positive feedback is likely to reinforce their gene expression patterns and help enhance the distinction between the N and H cells during epidermal development.

The explanation above suggests that bi-directional signaling, from the N to the H cell (via CPC and possibly also TRY/ETC1) and from the H to the N cells(via GL3 and possibly EGL3), is required for appropriate accumulation of GL3/EGL3 in the N cell position during root epidermis development. The `back and forth' signaling between cells proposed here is conceptually similar to the kinds of bi-directional signaling identified in other cell specification models, including the forespore/mother cell fate decision in Bacillus,embryonic midgut and larval wing patterning in Drosophila, and vulval cell specification in C. elegans(Bondos and Tan, 2001; Losick and Dworkin, 1999; Yoo et al., 2004). However,the signaling mechanism used in the Arabidopsis root epidermis differs from these others because it involves the intercellular movement of transcription factors, rather than receptor-mediated signaling, thereby directly influencing gene expression and cell fates.

We thank Myeong Min Lee, Marissa Simon and Su-Hwan Kwak for helpful comments. This work was supported by USDA Grants (2001-35304-10134 and 2004-35304-14924) to J.S. and a US National Science Foundation Grant(IBN-9986391) to A.L.