Developmentally distinct MYB genes encode functionally equivalent proteins in Arabidopsis

New traits are thought to arise during evolution via gene duplication and divergence. The divergence of gene function may formally result either from changes in regulatory sequences and/or by alterations in protein-coding regions. The relative importance of these mechanisms in generating novel gene functions during evolution is presently unclear (Doebley and Lukens, 1998; Carroll, 2000).

The MYB family of transcriptional regulators in plants provides an excellent model for defining the molecular mechanisms underlying the divergence of gene function during the evolution of multicellular organisms. Although small in animals and fungi, the MYB gene family is large and diverse in flowering plants, and its members regulate a wide array of developmental and biochemical processes (Martin and Paz-Ares, 1997; Rabinowicz et al., 1999; Riechmann et al., 2000). Furthermore, the proliferation of MYB genes during plant evolution coincided with the acquisition of new cellular functions and morphological forms, prompting suggestions that MYB gene divergence played an important role in the diversification of plant developmental mechanisms (Martin and Paz-Ares, 1997; Romero et al., 1998; Rabinowicz et al., 1999).

To explore the factors contributing to morphological diversification, we analyzed two paralogous MYB genes from Arabidopsis, WEREWOLF (WER) and GLABROUS1 (GL1), which participate in different developmental processes. The WER and GL1 proteins share 57% amino acid sequence identity, including 91% identity in their MYB domains, and are members of a subclass of R2R3 MYBs in Arabidopsis (Fig. 1A; Lee and Schiefelbein, 1999; Oppenheimer et al., 1991). Despite their sequence similarity, the WER and GL1 genes influence distinct aspects of Arabidopsis development. The WER gene is required to specify the non-hair cell type in the root epidermis and the non-stomatal cell type in the hypocotyl, and, accordingly, WER gene expression is limited to a specific subset of developing epidermal cells in the root and hypocotyl (Lee and Schiefelbein, 1999). The wer mutations alter the fates of these root and hypocotyl epidermal cells, causing them to adopt the alternate root-hair or stomatal cell fate (Lee and Schiefelbein, 1999). The GL1 gene specifies the production of trichomes (single-celled hairs) on the surface of leaves and stems, and its expression is limited to the epidermal cells in developing shoot tissue (Koornneef et al., 1982; Larkin et al., 1993). The gl1 mutations inhibit trichome formation, which generates a glabrous (hairless) shoot phenotype, but they do not affect other epidermal cell types (Koornneef et al., 1982; Esch et al., 1994).

In this study, we defined the functional relationship between WER and GL1 by exploiting the distinct mutant phenotypes exhibited by these two related genes. We show that these two genes encode proteins with equivalent functions and that their regulatory sequences are entirely responsible for their distinct roles in plant development. We also identify specific domains of these proteins that confer common biochemical activities. Further, we find that a third Arabidopsis MYB gene, which is distantly related to WER and GL1, encodes a functionally distinct protein. These results provide direct support for the notion that changes in the cis-regulatory sequences of transcription factor genes played a major role in morphological diversification during eukaryotic evolution.

The EMS-induced wer-1 and gl1-1 mutant alleles used in this study have been described previously (Koornneef et al., 1982; Lee and Schiefelbein, 1999). The wer-1 gl1-1 double mutants were obtained by crossing the single mutant lines, generating an F₂ population, and confirming putative double mutants by backcrossing to each single mutant. Plants harboring the GL2::GUS reporter construct have been described previously (Masucci et al., 1996). The GL2::GUS was introduced into each of the mutant and transgene lines by genetic crosses.

The selection of the WER and GL1 regulatory sequences used in these constructs was based on previously published work (Larkin et al., 1993; Lee and Schiefelbein, 1999). Constructs designed to encode the WER, GL1, or MYB2 proteins were generated by making a gene cassette that included the entire transcriptional unit of each gene from the start to the stop codon. The WER regulatory sequences included a 4.0 kb fragment 5′ to the start codon and a 1.1 kb fragment 3′ to the stop codon. The GL1 regulatory sequences included a 1.4 kb fragment 5′ to the start codon and a 1.8 kb fragment 3′ to the stop codon. The 3′ region of these genes was included in these constructs because the 3′ region of GL1 is known to be essential for proper expression (Larkin et al., 1993). Details of the transgene constructions 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 for each transgene and all had similar phenotypes, and either two or three independent lines of each were chosen for detailed characterization.

The proportion of cell types in the root epidermis was determined from at least ten 5-day-old seedlings from each strain as previously described (Hung et al., 1998). An epidermal cell was scored as a root-hair cell if any protrusion was visible. Ectopic root-hair cells were identified from transverse hand-cut sections of at least five 4-day-old roots from each strain (>50 hair cells scored per line) as described previously (Galway et al., 1994). 4-day-old roots were embedded in JB-4 resin as described previously (Masucci et al., 1996) and transverse sections were cut.

The distribution of epidermal cell types in the hypocotyl was analyzed as described previously (Hung et al., 1998). Trichome production was determined by counting the total number of trichomes on the surface of the first two leaves of 12-day-old plants. For scanning electron microscopy, specimens from 10-day-old plants were prepared and examined as described previously (Yu et al., 2000).

Histochemical analysis of whole seedlings or plant tissue containing the GL2::GUS reporter was performed essentially as described by Masucci et al. (Masucci et al., 1996). 4-day-old roots and hypocotyls were subjected to the GUS histochemical assay, followed by embedding in JB-4 resin as described previously (Masucci et al., 1996) and transverse sections were cut.

To assess transcriptional activation, the yeast one-hybrid assay was employed as described by Sadowski et al. (Sadowski et al., 1992). Fragments from the WER coding region were fused to the GAL4 DNA-binding domain in the vector pGBT9 and, following transformation into yeast strain HF7c, lacZ and HIS3 gene expression was assessed.

Although their distinct gene expression patterns and mutant phenotypes indicate independent roles for WER and GL1 in plant development, it was conceivable that they possess some degree of functional overlap. Therefore, as a first step toward understanding the relationship between WER and GL1 gene function, we generated plants homozygous for mutations in each gene and examined them for the presence of novel phenotypes. These wer-1 gl1-1 plants displayed the phenotype expected from the simple additive effect of each single mutation; that is, they lack trichomes in the shoot epidermis and produce a small number of non-hair cells in the root (Table 1). Aside from these attributes, the plants are vigorous, fertile and exhibit no visible abnormalities (data not shown). The lack of a synergistic effect between these mutations confirms that WER and GL1 have non-overlapping functions in Arabidopsis.

To define the relationship between the WER and GL1 functions, we wished to examine the effect of each protein when controlled by the other gene’s regulatory sequences. As a first step, we identified the cis-regulatory sequences necessary for WER gene function. Based on results from a prior study (Lee and Schiefelbein, 1999), we first tested a construct harboring the WER protein-coding region fused to a 4.0 kb fragment 5′ to the start codon and a 1.1 kb fragment 3′ to the stop codon. This construct (designated WER::WER; Fig. 1B) was introduced into the wer-1 mutant background, and the resulting transgenic plants were able to produce a normal proportion of epidermal cell types in the root (Fig. 2A; Table 1). The ability of this WER::WER construct to rescue the wer-1 mutant phenotype indicates that the cis-regulatory sequences employed in this construct are sufficient to direct normal WER gene function.

To determine whether the GL1 protein can functionally substitute for the WER protein, wer-1 mutant plants were transformed with a construct (designated WER::GL1) containing the GL1 coding region fused to the same WER cis-regulatory sequences employed in the WER::WER construct (Fig. 1B). These WER::GL1 wer-1 plants were able to generate a normal proportion of epidermal cell types in the root (Fig. 2A, Table 1), providing the first evidence for functional equivalence between the GL1 and WER proteins.

The two epidermal cell types in the Arabidopsis root are normally distributed in a position-dependent pattern. Epidermal cells located outside a single underlying cortical cell (i.e. outside a periclinal cortical cell wall) are specified as non-hair cells, whereas cells located outside two cortical cells (i.e. outside an anticlinal cortical cell wall) are specified as hair cells (Dolan et al., 1994; Galway et al., 1994; Berger et al., 1998). This pattern is disrupted in the wer-1 mutant, which produces 47% of its hair cells in the non-hair position (termed ectopic hair cells; Lee and Schiefelbein, 1999; Fig. 2A; Table 2). In the WER::GL1 wer-1 plants, we discovered that the normal cell pattern was restored, with a low frequency of ectopic hair cells that is similar to the wild type and to the WER::WER wer-1 plants (Fig. 2A, Table 2). This shows that, when controlled by WER regulatory sequences, the GL1 protein is able to direct root epidermal patterning in the same manner as WER.

In addition to its effect on root epidermis development, the WER gene influences cell-type specification in the hypocotyl, where it contributes to a position-dependent pattern of stomatal and non-stomatal cells that is similar to the hair/non-hair pattern in the root epidermis (Lee and Schiefelbein, 1999). The wer-1 mutations alter the normal pattern of hypocotyl cell types, causing a large proportion of epidermal cells outside a periclinal cortical cell wall to differentiate into stomatal complexes (termed ectopic stomata; Table 2). We found that the WER::GL1 transgene restores the normal pattern of hypocotyl cell types to the wer-1 mutant (Table 2), which shows that GL1 (when controlled by the WER regulatory sequences) is able to provide the WER function necessary for appropriate epidermal development in the hypocotyl.

To further define the molecular basis for the effect of the WER::GL1 transgene on root and hypocotyl development, we examined the expression pattern of a presumptive target gene, GLABRA2 (GL2), which encodes a homeodomain protein (Rerie et al., 1994). The WER gene is known to be required for appropriate position-dependent regulation of GL2 gene expression, directing GL2 expression in differentiating epidermal cells of the root and hypocotyl that are located outside periclinal cortical cell walls (Fig. 3; Lee and Schiefelbein, 1999). Using a GL2 promoter::GUS reporter construct, we analyzed GL2 expression during root and hypocotyl development in the wer-1, WER::WER wer-1, and WER::GL1 wer-1 lines. In the root, we discovered that GL2 was expressed in an appropriate temporal manner (during the period of cell differentiation) and spatial pattern (within epidermal cells located outside periclinal cortical cell walls) in the WER::GL1 wer-1 and WER::WER wer-1 plants (Fig. 3A). Similarly, the normal GL2 expression pattern was restored in the hypocotyl of the WER::GL1 wer-1 plants (as well as the WER::WER wer-1 plants), relative to the wer-1 mutant plants (Fig. 3B). Together, these results indicate that the GL1 protein (produced by the WER::GL1 transgene) utilizes the same molecular mechanisms as WER to direct GL2 expression and epidermal cell fate in an appropriate position-dependent manner during root and hypocotyl development.

In a set of experiments that are complementary to those described above, we sought to determine whether the WER protein can functionally substitute for GL1. We first generated a construct harboring the GL1 protein-coding region fused to a 1.4 kb fragment flanking the 5′ end and a 1.8 kb fragment flanking the 3′ end of the GL1 gene. This construct (designated GL1::GL1; Fig. 1B) was introduced into the gl1-1 mutant background, and the resulting transgenic plants were found to produce a normal number of trichomes (Fig. 2; Table 1). This shows that a sufficient amount of 5′ and 3′ cis-regulatory sequences were employed in this construct to enable normal GL1 gene function during leaf epidermis development.

A GL1::WER fusion construct was generated, which included the WER coding sequence fused with the GL1 regulatory sequences used in the GL1::GL1 construct (Fig. 1B). Following transformation of gl1-1 mutant plants with this GL1::WER construct, we discovered that the GL1::WER gl1-1 plants possess a wild-type number of morphologically normal trichomes on their leaves and stems (Fig. 2B; Table 1, data not shown). Further, although trichome spacing is not determined by a position-dependent mechanism like the epidermal cells of the root and hypocotyl, we observed a general distribution of trichomes (and a lack of trichome clusters) in the GL1::WER gl1-1 leaves that is similar to the non-random distribution found in the wild-type and GL1::GL1 gl1-1 plants (Fig. 2B, data not shown).

The GL1 gene is a positive regulator of GL2, specifically directing GL2 expression in differentiating trichome cells (Fig. 4; Szymanski et al., 1998). Using the GL2::GUS reporter construct, we monitored the spatial and temporal expression of GL2 in the GL1::WER gl1-1 plants during leaf development. We discovered that the GL1::WER construct (like the GL1::GL1 construct) restores the normal pattern of GL2::GUS expression in the developing leaf, relative to the gl1-1 mutant (Fig. 4). This demonstrates that the WER protein (when directed by GL1 cis-regulatory sequences) can replace the GL1 function and appropriately direct GL2 expression and trichome specification during leaf development.

In addition to the epidermal phenotypes described above, we did not observe any abnormalities in the WER::GL1 wer-1 and GL1::WER gl1-1 plants (data not shown). Taken together, our analysis of these plants shows that the WER and GL1 proteins are functionally equivalent. Either one can appropriately specify the non-hair cell type in the root and the non-stomatal cell type in the hypocotyl (when expressed under control of WER regulatory sequences) or the trichome cell type in the shoot (when expressed under the control of the GL1 regulatory sequences).

More than 100 MYB genes of the R2R3 type exist in the Arabidopsis genome (Romero et al., 1998; Riechmann et al., 2000). One extreme possibility raised by our findings with WER and GL1 is that all of these MYB genes may encode functionally equivalent proteins. To test this possibility, we determined whether a distantly related R2R3 MYB protein from Arabidopsis (MYB2) can functionally substitute for the WER or GL1 protein. The MYB2 protein is a member of a different MYB subclass than WER and GL1, it possesses approximately 34% amino acid identity with WER and GL1, and it is associated with plant hormone action (Fig. 1A) (Urao et al., 1993). We fused the coding region of the MYB2 gene to the cis-regulatory regions of the WER gene and the GL1 gene to generate the WER::MYB2 and GL1::MYB2 constructs (Fig. 1B). After introducing these constructs into appropriate mutant backgrounds (WER::MYB2 into the wer-1 mutant and GL1::MYB2 into the gl1-1 mutant), we discovered that the phenotypes of the transgenic plants were indistinguishable from the phenotypes of the original mutant lines (Table 1). The reciprocal experiment, testing the ability of WER or GL1 to functionally substitute for MYB2, was not possible because no known myb2 mutant line exists. Nevertheless, the inability of the WER::MYB2 or GL1::MYB2 constructs to rescue the phenotype of the wer-1 or gl1-1 mutant indicates that the MYB2 protein is not functionally equivalent to either WER or GL1.

The functional equivalence of WER and GL1 means that these proteins must possess the same biochemical activities. We sought to define the domain that confers one of these biochemical activities, transcriptional activation, by employing the yeast one-hybrid assay. A series of WER-derived segments were fused to the GAL4 DNA-binding domain and GAL4-responsive gene expression was assayed. We determined that a protein segment containing the C-terminal 24 amino acids of WER is sufficient for transcriptional activation in yeast (Fig. 5). Interestingly, this 24-residue segment represents the portion of the protein, outside the MYB domains, that is most similar between WER and GL1 (14 of the 24 residues are identical in WER and GL1; Fig. 1A). This suggests that WER and GL1 employ an evolutionarily conserved C-terminal domain to activate gene transcription.

This study exploited the unique availability of two related genes that have distinct developmental roles, to define the functional relationship between members of an important family of transcriptional regulators. The most striking finding is that the WER and GL1 genes, which specify distinct cell types during Arabidopsis development, actually encode functionally equivalent MYB proteins. Either protein can fully direct the cell specification events normally controlled by the other when expressed under the other gene’s regulatory controls. This provides an elegant demonstration of the importance of cis-regulatory sequences in specifying developmental gene function. Together with previous comparative evolutionary studies that correlated changes in regulatory sequences of genes with changes in developmental features (Doebley and Lukens, 1998; Wang et al., 1999; Carroll, 2000), our functional analysis of these MYB genes supports the view that modification of cis-regulatory sequences of transcriptional regulators has played a key role in morphological diversification during the evolution of multicellular organisms.

To our knowledge, this represents the first direct in vivo demonstration of functionally equivalent proteins in plants. Furthermore, unlike studies in animal systems that uncovered functionally equivalent proteins encoded by partially redundant genes or genes with overlapping expression patterns (Bouchard et al., 2000; Greer et al., 2000; Malynn et al., 2000), the present study employed genes with entirely non-overlapping expression domains, which uniquely demonstrates that functional equivalence can be maintained even after complete divergence of regulatory elements.

The ability of the WER and GL1 proteins to fully substitute for one another in different developmental processes shows that, despite sharing only 57% amino acid sequence identity, they possess common biochemical activities. The common activities for these transcription factor proteins are likely to include equivalent DNA binding specificities, equivalent partner protein interactions, and equivalent transcriptional activation properties. As outlined below, each of these activities appears to be conferred by specific domains conserved between the WER and GL1 proteins.

The DNA-binding activity of MYB proteins is known to be determined by the Myb repeats themselves, which form α-helices that contact DNA in a sequence-specific manner (Thompson and Ramsay, 1995). Consistent with the notion that WER and GL1 possess equivalent DNA binding properties, these proteins exhibit a high degree of amino acid sequence identity (91%) within their two (R2R3) Myb repeats (Fig. 1A).

The view that WER and GL1 possess equivalent partner protein interactions is further supported by previous studies demonstrating that each protein physically interacts with a basic helix-loop-helix (bHLH) transcription factor related to the maize R protein (Larkin et al., 1997; Lee and Schiefelbein, 1999). In a separate study, the specific interaction between the R protein and an R2R3 MYB protein from maize (C1) was shown to be mediated by residues in the first helix of the second Myb repeat of C1 (Grotewold et al., 2000). Significantly, the WER and GL1 protein sequences are identical within this 12-amino-acid first-helix region of their second Myb repeats, implying that these two proteins possess the same bHLH interaction domain.

In the present study, we showed that a 24-residue C-terminal domain, which is similar in the WER and GL1 proteins, confers transcriptional activation properties on GAL4 fusion proteins in yeast. Therefore, equivalent transcriptional activation by WER and GL1 is likely to be mediated by this conserved C-terminal domain.

Our identification of functional equivalence between WER and GL1 extends our understanding of the close relationship between the molecular mechanisms controlling cell specification in the root, hypocotyl, and leaf of Arabidopsis. Each of these processes is dependent upon an identical WD-repeat protein (TRANSPARENT TESTA GLABRA; Koornneef, 1981; Galway et al., 1994; Hung et al., 1998; Walker et al., 1999), a similar, if not identical, bHLH transcription factor related to the maize R protein (Lloyd et al., 1992; Galway et al., 1994; Hung et al., 1998; Payne et al., 2000), and a functionally equivalent MYB transcription factor (WER or GL1; this study). Together, these findings suggest that the same transcriptional machinery may be employed to specify epidermal cell fate in each of these three organs.

The ability of these common transcriptional components to specify different cell types (non-hair cells, non-stomatal cells, and trichome cells) implies that as yet unidentified organ-specific factors also influence epidermal cell fate in the root, hypocotyl, and leaf. One general possibility is that organ-specific regulatory proteins exist and act within the MYB/bHLH pathway to generate distinct sets of target proteins in the three differentiating epidermal cell types. An alternative hypothesis is that the MYB/bHLH pathway is identical in each of the cells and generates the same suite of proteins, but the presence of additional organ-specific proteins cause these three differentiating cells to complete distinct developmental programs.

In addition to identifying WER and GL1 as functionally equivalent proteins, we have shown that the distantly related MYB2 protein is not equivalent to either WER or GL1. This demonstrates that alterations in protein-coding sequences has also played a role in generating distinct gene functions during the evolution of the Arabidopsis MYB family.

The inability of MYB2 to functionally substitute for WER or GL1 may be due to differences in its DNA binding activity, partner protein interactions, or transcriptional activation properties. The Myb repeats in MYB2 exhibit only 58% identity with the WER Myb repeats, which suggests that these proteins have different DNA binding properties. Also, the first α-helix of the second Myb repeat in MYB2 possesses only 5 of the 12 residues present in the equivalent region of WER, implying that MYB2 is unlikely to interact with an R-like bHLH protein. Finally, MYB2 differs significantly from WER and GL1 in its C-terminal domain (only 6 of 24 residues are identical in MYB2 and WER), which implies that MYB2 may lack the transcriptional activation ability of WER and GL1.

Considering the functional and structural relationships we have discovered between the WER, GL1 and MYB2 proteins, an attractive possibility is that the extensive proliferation of MYB genes that occurred during plant evolution generated specific subclasses of MYB proteins (like the GL1/WER subclass) composed of functionally similar (or equivalent) members, due in part to constraints placed on their divergence by their common biochemical activities. Clearly, the functional characterization of additional MYBs in Arabidopsis and other plant species will be necessary to rigorously test this hypothesis. From this perspective, it is significant to note that among the more than 100 R2R3 MYBs in Arabidopsis (Romero et al., 1998; Riechmann et al., 2000) there is a third member of the GL1/WER subclass. This third protein, AtMYBrtf (GenBank accession number Z68158), has an unknown function in Arabidopsis but possesses a high level of overall sequence identity to WER and GL1 (approximately 63%), including extensive identity within the Myb repeats (95% identical to WER), a perfect match within the putative bHLH interaction domain described above (the same 12 residues as in WER and GL1), and a similar C-terminal domain (16 of the 24 residues identical to WER). Thus, it will be interesting to determine whether AtMYBrtf is functionally equivalent to WER and GL1 and whether it represents a third MYB that participates in epidermal cell specification during Arabidopsis development.

This research was supported by the US National Science Foundation (grant IBN-0077900). The authors thank Steve Clark and Michael Frohlich for helpful comments.