The gene regulatory network for root epidermal cell-type pattern formation in Arabidopsis

Abstract

A fundamental aspect of multicellular development is the patterning of distinct cell types in appropriate locations. In this review, the molecular genetic control of cell-type pattern formation in the root epidermis of Arabidopsis thaliana is summarized. This developmental system represents a simple and genetically tractable example of plant cell patterning. The distribution of the two epidermal cell types, root-hair cells and non-hair cells, are generated by a combination of positional signalling and lateral inhibition mechanisms. In addition, recent evidence suggests that reinforcing mechanisms are used to ensure that the initial cell fate choice is adopted in a robust manner.

Introduction

The patterning of cell types in the root epidermis of plants has been used for many years as a model for studying pattern formation in plant development (Haberlandt, 1887; Leavitt, 1904; Cormack, 1935; Sinnot and Bloch, 1939; Cutter, 1978). There are only two cell types produced in the root epidermis, root-hair cells and non-hair cells, so the problem of pattern formation becomes the study of a simple ‘either-or’ decision. Although many flowering plants have been analysed for their root epidermal patterning, most recent attention has been focused on using the model plant species Arabidopsis thaliana to study root epidermal pattern formation. The reasons for this emphasis are abundant: the small size of Arabidopsis makes the study of the early growth of these plants to be conducted in a small amount of space on defined media, which enables high-throughput genetic, developmental, and physiological analyses; the root hairs are easily visible on the root surface and appear rapidly (within 3 d) after seed germination, making them one of the most convenient post-embryonic phenotypes (Fig. 1); the entire developmental history of the Arabidopsis epidermis has been defined, from its origin in the embryo through its mature state (Dolan et al., 1993, 1994; Scheres et al., 1994); the root epidermal cells are generated and differentiate in a file-specific manner, which enables all stages of development to be analysed along the root at any time; the root hair cells are not required for plant viability or fertility, so, in principle, any type of mutant can be isolated and analysed; and molecular biology resources are available in Arabidopsis for the isolation and analysis of new genes.

Fig. 1.

Photographs of roots from wild-type, wer mutant, and cpc mutant Arabidopsis plants. The wer mutant produces excessive root hair cells; whereas the cpc mutant has a reduced number of root hair cells. The bar length is 200 μm.

The position-dependent pattern of Arabidopsis root epidermal cells

Root-hair cells arise from trichoblasts; cells that have been specified to be root-hair cells but have not yet exhibited morphological evidence of hair formation. Similarly, non-hair cells arise from atrichoblasts. During its morphogenesis, each root-hair cell produces a long tubular shaped extension (‘hair’) that is involved in water and nutrient uptake as well as anchorage and rhizosphere interactions; the non-hair cells lack this cellular extension. Depending on the plant species, the pattern of root-hair cells and non-hair cells is determined by an asymmetric division mechanism, a random mechanism, or a position-dependent mechanism (Clowes, 2000). In the asymmetric division mechanism, used by some grasses, the smaller of the daughter cells from an epidermal cell precursor adopts the root-hair cell fate and the larger daughter becomes a differentiated non-hair cell. For the random mechanism, used by many species including rice, the final fate of a given cell cannot be predicted by its history because it adopts its fate at a late stage, probably due to environmental factors.

The position-dependent mechanism is used by many plant species, including Arabidopsis (Cormack, 1935; Bunning, 1951; Dolan et al., 1994; Galway et al., 1994). In this mechanism, root-hair cells arise over the crevice between two underlying cortical cells (designated the ‘H’ cell position), whereas non-hair cells develop over the position of a single cortical cell (designated the ‘N’ position). This cell-type arrangement implies that positional cues play an important role in cell fate determination in this mechanism. It is interesting to note that, prior to hair formation, immature Arabidopsis epidermal cells in the H position can be distinguished from those in the N position by a variety of criteria, including a greater rate of cell division (Berger et al., 1998), reduced cell length (Dolan et al., 1994; Masucci et al., 1996), enhanced cytoplasmic density (Dolan et al., 1994; Galway et al., 1994), reduced vacuolation rate (Galway et al., 1994), variant cell surface features (Dolan et al., 1994; Freshour et al., 1996), and distinct chromatin organization (Costa and Shaw, 2005; Xu et al., 2005). These observations show that the root-hair cells and the non-hair cells undergo extensive differences during their development, and they do not merely differ in their final cell shape. Furthermore, the analysis of promoter–reporter lines and in situ RNA hybridization shows that cells in the two epidermal positions already exhibit differential gene expression during embryogenesis (Berger et al., 1998; Lin and Schiefelbein, 2001; Costa and Dolan, 2003). Thus, epidermal cells begin to assess their position and adopt their appropriate fate at an early stage, during both embryogenesis and post-embryonic root development. However, despite early differential cellular activities in H and N cells, a relatively late change in the position of an immature epidermal cell (through human manipulation or aberrant cell division planes) can induce a change in its developmental fate (Berger et al., 1998). Together, these results suggest that epidermal cell fate is not fixed at an early stage in Arabidopsis; rather, positional signalling apparently acts continuously during post-embryonic root development to ensure appropriate cell-type patterning.

Arabidopsis genes controlling root epidermal cell patterning

Molecular genetic approaches have led to the identification of many genes that influence early events in epidermal cell patterning in Arabidopsis (Grierson and Schiefelbein, 2002; Pesch and Hulskamp, 2004; Xu et al., 2005; Schellmann et al., 2007). Most of these encode transcription factors that influence the expression of genes involved in epidermal cell differentiation. The emerging view is that these act at an early stage in a complex regulatory network involving the mechanism of lateral inhibition with feedback (Lee and Schiefelbein, 2002; Larkin et al., 2003; Schiefelbein, 2003). Four genes, TRANSPARENT TESTA GLABRA (TTG), GLABRA3 (GL3), ENHANCER OF GLABRA3 (EGL3), and WEREWOLF (WER) are required to specify the non-hair fate because mutations in these (alone or in combination) cause plants to produce “hairy” roots (roots that have root-hair cells in place of non-hair cells) (Galway et al., 1994; Masucci et al., 1996; Lee and Schiefelbein, 1999; Bernhardt et al., 2003) (Fig. 1). Three genes, CAPRICE (CPC), TRIPTYCHON (TRY), and ENHANCER OF TRY AND CPC (ETC1), appear to specify the root-hair cell fate because mutations in them (alone or in combination) cause plants to produce ‘bald’ roots (roots that have non-hair cells develop in place of hair cells) (Wada et al., 1997; Schellmann et al., 2002; Kirik et al., 2004; Simon et al., 2007) (Fig. 1). The current evidence suggests that TTG (a small protein with WD40 repeats: Walker et al., 1999), GL3 and EGL3 (related basic helix-loop-helix (bHLH) transcription factors (Bernhardt et al., 2003)), and WER (an R2R3 MYB-domain transcription factor: Lee and Schiefelbein, 1999) act in a core transcriptional complex to promote the non-hair cell fate (Schiefelbein, 2003; Pesch and Hulskamp, 2004; Ueda et al., 2005). This complex preferentially accumulates in cells located in the N position and positively regulates the expression of the GLABRA2 (GL2), which encodes a homeodomain-leucine-zipper (HD-Zip) transcription factor protein (Rerie et al., 1994) required for non-hair cell differentiation (Di Cristina et al., 1996; Masucci et al., 1996; Costa and Shaw, 2006) (Fig. 2). The GL2 protein negatively regulates root-hair-specific genes and positively regulates non-hair cell specific genes to promote the non-hair cell fate (Masucci et al., 1996; Lee and Schiefelbein, 1999, 2002).

Fig. 2.

A model to explain cell-type patterning in the Arabidopsis root epidermis. The SCM receptor preferentially represses WER transcription in the H cell position to instigate the pattern formation. The two distinct cell types arise by the differential accumulation of the WER-GL3/EGL3-TTG transcriptional complex, which directs non-hair cell differentiation. (This figure is available in colour at JXB online.)

Likely targets of the GL2 transcription factor include ROOTHAIRDEFECTIVE6 (RHD6, which is required for hair initiation: Masucci and Schiefelbein, 1994, 1996; Menand et al., 2007), and PLDξ1 (which encodes a phospholipase D protein involved in hair outgrowth and elongation). GL2 has been shown to bind to the PLDξ1 promoter in gel shift assays and it prevents expression of PLDξ1 in planta, supporting the idea that GL2 acts directly upstream of this gene in non-hair cells (Ohashi et al., 2003). Root hair initiation is also dependent on auxin and ethylene signalling, although the precise relationship between these hormones and the known transcription factors controlling root epidermis cell fate has not been defined (Masucci and Schiefelbein, 1994; Fischer et al., 2007; Stepanova et al., 2007). A complex interaction is likely because ethylene acts, at least in part, through changes to auxin biosynthesis and transport (Ruzicka et al., 2007; Swarup et al., 2007). It is tempting to speculate that these hormones may be used to modulate the final epidermal cell-type pattern in response to environmental factors.

The TTG–GL3/EGL3–WER core transcription complex is also involved in lateral inhibition through its positive regulation of the transcription of CPC, TRY, and ETC1 in the N cells (Lee and Schiefelbein, 2002; Wada et al., 2002; Kirik et al., 2004; Koshino-Kimura et al., 2005; Ryu et al., 2005; Simon et al., 2007). These three genes encode small one-repeat MYB proteins that lack transcriptional activation domains and appear to move to the adjacent H cells (possibly via plasmodesmata) to inhibit the WER/MYB23 MYB proteins to participate in complex formation and/or promoter binding (Wada et al., 1997; Schellmann et al., 2002; Wada et al., 2002; Esch et al., 2003; Kirik et al., 2004; Kurata et al., 2005; Ryu et al., 2005). Recent evidence suggests that this inhibition is due to competition between the WER and the CPC/TRY/ETC1 proteins for binding to the GL3/EGL3 proteins (Tominaga et al., 2007; Ishida et al., 2008).

The reduced accumulation of the TTG–GL3/EGL3–WER complex in the H cells is correlated with establishment of hair-specific gene expression. These genes include ones encoding secretory proteins, cytoskeletal elements, and cell wall synthesis components that are probably important in expanding the cell and generating the hair structure (Jones et al., 2006). In a comparative analysis, one of the root-hair specific genes from Arabidopsis, expansin A7, has been found to share similar functional promoter elements with the expansion A7 orthologue from rice (Kim et al., 2006). This suggests that there is a common set of downstream root-hair morphogenetic genes that are controlled by different upstream fate specification mechanisms in plant species with different hair distribution patterns, which provides a framework for developing hypotheses concerning the evolutionary origin of the multiple fate specification mechanisms.

In addition to promoting GL2 and CPC/TRY/ETC1 gene transcription, the TTG–GL3/EGL3–WER complex represses transcription of the GL3/EGL3 bHLH genes and, as a result, the bHLH genes are preferentially expressed in the H cells (Fig. 2). This implies that these bHLH proteins move from the H cell (where they are produced) to the N cell (where they act for non-hair cell specification) to complete a regulatory circuit involving two loops of communication between the N and H cells (Bernhardt et al., 2005). It is likely that these intercellular events result in mutual dependence of cell identities between neighbouring cells, and thus help to reinforce the cell fate decisions and ensure that distinct cell fates arise in a robust fashion (Schiefelbein and Lee, 2006).

The positional control of the root epidermal cell pattern

The preferential accumulation of the central transcription complex in the N position has been proposed to result from a positional signalling pathway (Larkin et al., 2003; Schiefelbein and Lee, 2006). For a long time, the molecular basis of this positional influence was unknown, and it was speculated to be due either to a cortical cell signal or to an underlying signal. Recently, a leucine-rich repeat receptor-like kinase (LRR-RLK), named SCRAMBLED (SCM), was discovered that is likely to mediate this positional signalling process (Kwak et al., 2005). Homozygous recessive scm mutations cause the developing root epidermis to exhibit a distribution of GL2, CPC, and WER expression and epidermal cell types that is not strictly determined by the cell's position. Nevertheless, there is a tendency for the cells to adopt fates based on their position (even in presumed scm null mutants), which suggests that the SCM signalling pathway is not the only mechanism mediating positional patterning. Detailed genetic and expression studies indicate that SCM's effect is probably due to its preferential inhibition of WER transcription in the H cell position (Kwak and Schiefelbein, 2006). The predicted SCM protein possesses the structural features of a typical LRR receptor-like protein kinase, of which there are more than 200 in Arabidopsis (Initiative, 2000; Shiu and Bleecker, 2001, 2003; Torii, 2004). It possesses a putative extracellular domain with six tandem copies of a 24-residue leucine-rich repeat (LRR), a single predicted transmembrane domain near its centre, and a C-terminal putative intracellular kinase domain. These results suggest that SCM enables immature epidermal cells to detect a positional signal and adopt an appropriate fate.

The biochemical effect of SCM may be to influence epidermal cell type differentiation through remodelling the chromatin around the transcription factor genes. In support of this view, three-dimensional fluorescence in situ hybridization (FISH) has shown that the chromatin around the GL2 gene is ‘open’ in the differentiating N cells, whereas it is ‘closed’ in the differentiating H cells (Costa and Shaw, 2006). Further, the GL2 chromatin status appears to be reset during mitosis, which makes the epidermal cells in the meristematic zone vulnerable to chromatin modification by a change in cell position (Costa and Shaw, 2006). Also, two studies suggest that histone modification influences the cell fate decision. Inhibition of histone deacetylation, using trichostatin A, increased expression of the CPC and GL2 genes, decreased expression of WER, and led to an increase in root hair cell formation (Xu et al., 2005). Further, the GEM1 protein, which interacts with the cell division machinery, influences histone H3 modifications at the GL2 and CPC promoters (Caro et al., 2007). Thus, there appears to be a link between cell division, chromatin status, and cell fate decisions in the root epidermis.

A proposed model for Arabidopsis root epidermis patterning

The molecular genetic results to date suggest a possible model for cell-type pattern formation in the root epidermis of Arabidopsis (Fig. 2). At its core, this model proposes that SCM signalling influences, in a position-dependent manner, the accumulation of the core transcription factor complex (TTG–GL3/EGL3–WER) which promotes the non-hair cell fate (via GL2) as well as the hair cell fate (by lateral inhibition via CPC/TRY/ETC1). More specifically, the SCM pathway is proposed negatively to regulate WER transcription in the H cells, which causes these cells preferentially to succumb to the CPC/TRY/ETC1-dependent lateral inhibition and adopt the hair cell fate. In the absence of SCM signalling, the random resolution of the mutual lateral inhibition by the H and N cells would yield root-hair cells and non-hair cells in a non-position-dependent pattern.

Several key aspects of this model are supported by experimental evidence. For example, it is known that a functional WER–GFP fusion protein localizes to the N cell nucleus (Ryu et al., 2005), whereas a CPC–GFP fusion protein accumulates in the nuclei of both N and H cells (Wada et al., 2002). This supports the postulate of differential mobility of the WER and CPC during early root epidermis development, which is crucial for the model's mechanism. Also, WER has been shown to regulate CPC expression directly, possibly via distinct WER-binding sites in the CPC promoter (Ryu et al., 2005), and the action of the WER-equivalent MYB protein GL1 is disrupted by TRY in yeast (Esch et al., 2003), which supports the ability of the small MYBs to inhibit the core transcriptional complex.

An examination of this model has recently been conducted, using a combined experimental and mathematical modelling approach (Savage et al., 2008). This analysis uncovered the importance of the reciprocal movement of the CAPRICE and GLABRA3 proteins between the developing N and H cells, rather than a putative local activation of the WER gene function, in establishing stable differences in the adjacent cells (Savage et al., 2008). In this way, the developing N and H cells may be considered to mutually support the specification of one another as distinct cell types.

Future research will be necessary to address several unclear issues in this regulatory system. For example, the precise way SCM alters the transcription factor network needs to be determined. Although the WER gene seems to be the ultimate target, the components or the action of the putative signal transduction chain is not known. Also, other mechanisms that influence position-dependent epidermal patterning are not known, although they are likely to exist, because scm mutants do not entirely abolish the cell pattern (Kwak and Schiefelbein, 2006). Also, it will be essential to define the putative SCM ligand to understand the origin of the position-based pattern fully. According to many current models, this ligand is hypothesized to be asymmetrically distributed around the epidermis, perhaps due to the cortical cell arrangement.

Another area of interest for future research will be to define the relationship between this patterning mechanism and the response of the root epidermis to environmental factors. For many plants, it is known that the length and density of root hairs are modified in response to changes in the environment, including phosphorus and iron deficiency (Bates and Lynch, 2000; Muller and Schmidt, 2004). In Arabidopsis, these environmental effects appear to act after the cell patterning mechanism described above and to be associated with the action of plant hormones, particularly ethylene (Zhang et al., 2003; Muller and Schmidt, 2004). It is possible that these represent epigenetic effects, acting at the level of chromatin modification, which have been shown to play a role in Arabidopsis root epidermal cell specification (Costa and Shaw, 2006; Caro et al., 2007). Future work will be required to resolve this potential mechanistic link and to determine whether these environmental response pathways provide the major specification process for those plant species that produce a variable distribution of root hairs.

It is worth mentioning that some features of this model are present in the patterning mechanisms for other epidermal cell types. Like the root, the Arabidopsis hypocotyl epidermis possesses two files of different cell types that arise in a position-dependent pattern (Wei et al., 1994; Gendreau et al., 1997; Berger et al., 1998; Hung et al., 1998). Cells in one file preferentially adopt the stomatal fate (located outside an anticlinal cortical cell wall, like the ‘H’ position of the root), whereas cells in the other file adopt the non-stomatal fate (located outside a periclinal cortical cell wall, like the ‘N’ position). With the exception of SCM, almost all of the genes associated with root epidermal cell fate also affect the hypocotyl epidermis in a similar way (Berger et al., 1998; Hung et al., 1998; Lee and Schiefelbein, 1999). Considering that the root and hypocotyl epidermis derives from embryonic epidermal tissue, it seems likely that the same patterning mechanism is used throughout the embryonic axis to generate position-dependent specification of a common pattern of two different cell types. Also, some of the root epidermal specification genes (including TTG and GL2) also affect trichome (hair) formation in the shoot epidermis and seed coat mucilage production in the developing seed (Koornneef, 1981; Koornneef et al., 1982; Galway et al., 1994; Masucci et al., 1996; Larkin et al., 1997; Schellmann and Hulskamp, 2005). Recent studies of trichome pattern formation, using reaction-diffusion based modelling with a stochastic component, have provided hypotheses to explain the distribution of these cell types (Szymanski et al., 2000; Digiuni et al., 2008). However, because the patterning of trichomes in the shoot tissues appears to be different from patterning of root epidermal cells, with no cell-position-dependent influence yet identified, these models may not be directly applicable to the root-hair patterning problem. Interestingly, the common genes control hair formation in the opposite ways in the root and leaf; they are required to specify the non-hair cell type in the root but they specify the hair-bearing (trichome) cell type in the leaf. Indeed, this may reflect the evolutionary origins of these systems; these transcriptional regulators may have been recruited during angiosperm evolution to participate in a common mechanism that is able to define distinct cell fates during epidermis development in different organs.