Control of GL2 expression in Arabidopsis leaves and trichomes

Trichomes are specialized cell types that are present on the surface of nearly all land plants (Johnson, 1975). However, the morphology and pattern of trichomes vary greatly depending on the species. In Arabidopsis, leaf trichomes are unicellular and stellate. The process of trichome development is complex and involves genes that regulate their spacing, density, and morphology. Over twenty mutations that affect various aspects of trichome development have been identified (reviewed by Marks, 1997). The analysis of these genes will aid in understanding the molecular mechanisms that control cell fate and differentiation in plants.

The GLABROUS1 (GL1) and TRANSPARENT TESTA GLABRA (TTG) genes are required for trichome initiation; the surface of gl1 and ttg mutant leaves are glabrous. Mutations in TTG also affect anthocyanin synthesis, integument development (Koornneef, 1981), and root hair patterning (Galway et al., 1994). The GL1 gene has been cloned, and shown to encode a putative myb-class transcription factor (Oppenheimer et al., 1991). GL1 is expressed in fields of cells containing initiating trichomes, and commitment to the trichome cell fate is accompanied by an initial rise in GL1 expression in the differentiated cell (Larkin et al., 1993). Genetic experiments showed that both GL1 and TTG activity are required for trichome initiation, but the ectopic expression of GL1 with the cauliflower mosaic virus 35S promoter (35S) does not lead to increased trichome formation nor does it bypass the requirement for TTG (Larkin et al., 1994). The maize R gene can supply an additional required activity because plants that ectopically express both GL1 and R initiate ectopic trichomes (Larkin et al., 1994). Although the R gene can compliment all aspects of the ttg mutation (Galway et al., 1994; Lloyd et al., 1992), the TTG and R genes do not share significant deduced amino acid similarity (A. Walker and J. Gray, personal communication). Therefore the precise relationship between TTG and R is not known. Nonetheless, the available evidence is consistent with a strict requirement for both GL1 and TTG for trichome initiation (Larkin et al., 1994).

Once an epidermal precursor is specified to enter the trichome pathway, an elaborate morphogenetic transformation occurs (Fig. 1). In wild-type trichome development, a focus of cell expansion centered on the external face of an epidermal precursor cell appears and develops into an elongating stalk. Additional foci of cell expansion are positioned on the growing stalk and produce branches (Hülskamp et al., 1994; Marks, 1994). After trichome expansion ceases the cell wall thickens and numerous papillae form on the outer surface of the trichome.

The GLABRA2 (GL2) gene is required for normal trichome morphogenesis (Koornneef et al., 1982; Rerie et al., 1994; Marks and Esch, 1994). On gl2 plants, expansion of trichomes is aberrant. Most gl2 trichomes are either enlarged abortive epidermal cells that expand in the plane of the leaf or unbranched spikes.The walls of gl2 trichomes do not appear to thicken or acquire papillae. The morphological defects in gl2 trichome cell expansion and cell wall maturation suggest that GL2-dependent activity is required throughout trichome development.

Genetically, the GL2 gene is downstream from GL1 and TTG, as gl2 gl1 and gl2 ttg double mutants lack trichomes (Hülskamp et al., 1994). Similar to ttg plants, gl2 plants lack seed coat mucilage and form ectopic root hairs (Koornneef, 1981; Masucci et al., 1996). In roots, GL2 is expressed in files of cells that become atrichoblasts, which suggests that GL2 limits hair formation in the root (Masucci et al., 1996). The level of GL2 expression in ttg roots is reduced, suggesting that TTG has a role in up-regulating GL2 expression (Di Cristina et al., 1996). A key question centers on the mechanism by which GL2 integrates positional information to regulate aspects of cell fate and differentiation in both roots and shoots.

GL2 has been sequenced and shown to have sequence similarity to homeodomain proteins (Rerie et al., 1994). Homeodomain proteins are transcription factors that often function to coordinate the expression of several target genes during diverse developmental processes (reviewed by Kessel and Gruss, 1990; Gehring et al., 1994). In Arabidopsis, there are known to be at least fifteen homeodomain proteins (Rerie et al., 1994; Lincoln et al., 1994; Ruberti et al., 1991; Carabelli et al., 1993; Schena and Davis, 1992; Quaedvlieg et al., 1995; Reiser et al., 1995; Lu et al., 1996). GL2 is most closely related to the HD-Zip group that contains a homeodomain located toward the N terminus followed by an amphipathic alpha helical domain (Kerstetter et al., 1994). The alpha helical region of GL2 is capable of interacting with itself in vitro (Di Cristina et al., 1996). C-terminal to the helical domain, GL2 encodes an additional 527 amino acid residues. This region does not contain any recognizable motifs; however, other genes have been identified that show significant similarity to the C-terminal domain of GL2. For example, Arabidopsis Thaliana Meristem Layer 1 (ATML1) and GL2 are 37% identical throughout their sequence and 60% identical in the homeodomain (Lu et al., 1996). GL2 also shares C-terminal sequence similarity to homeodomain genes from Phalenopsis (Nadeau et al., 1996), Helianthus (Valle et al., 1996), and to an uncharacterized Panicum sequence (Taniguchi et al., 1994).

To better understand the role of GL2 during trichome development, and the mechanisms by which the GL2 gene is regulated, we examined GL2 expression throughout leaf development. GL2 protein levels and GL2 promoter activity were analysed in developing leaves and trichomes. Variations in the level and location of GL2 expression were detected in the subepidermal cells and the trichomes of developing leaves. Analysis of the GL2 promoter indicates that two separate regions contribute to the observed leaf and trichome expression pattern. Evidence for the involvement of GL1 in the activation of the GL2 promoter in vivo is presented.

Plants were processed for GUS staining as described by Larkin et al. (1993). For sectioning of GUS-stained material, specifically staged plants were embedded in Tissue Tek OCT medium, and allowed to equilibrate for 4 hours at room temperature. Samples were cryo-sectioned in a Jung-Reichert Frigocut 2800 cryostat. Sections were transferred to poly-L-lysine coated slides and mounted using Aqueosmount for microscopy. For microscopic observations, samples were viewed using a Nikon SMU-Z stereoscope or a Nikon Diaphot-200 inverted microscope with DIC optics. Images were collected on color slides. Scanned images were labelled and compiled in figures using Adobe Photoshop (Adobe Systems Inc.) and Powerpoint^4.0 (Microsoft Corp.). For SEM analysis, soil grown plants were frozen in liquid nitrogen, mounted onto the cold stage of a Philips 500 scanning electron microscope and images were collected (Ahlstrand, 1996).

GUS activity was analysed quantitatively essentially as described by Gallagher (1992). Seedlings were germinated on MS plates with 10% sucrose and the cotyledons and leaves were separated from the hypocotyl at the four leaf stage. Plant samples were frozen in liquid nitrogen an homogenized with a plastic pestle in a microfuge tube. Additional grinding was done in the presence of 200 μl of GUS extraction buffer containing 50 mM NaPO₄, 1 mM EDTA, 0.1% sodium lauryl sarcosine, 0.1% Triton X-100, and 10 mM DTT. The homogenized sample was centrifuged for 10 minutes, and 5 μl of the supernatant was assayed for GUS activity in 195 μl extraction buffer containing 20% methanol and 1 mM 4-MUG (4-methyl-umbelliferyl glucuronide). The reaction was terminated by the addition of 800 μl of 0.2 M Na₂CO₃. MU (4-methyl-umbelliferone) flouresence was measured with a fluorometer (Hoefer Model TKO 100).

The gl1-1 null and ttg-1 alleles were used to analyze GL2::GUS expression. To quantitate GUS activity in gl1, ttg, gl1 ttg, 35S::R-GR ttg, and 35S::R-GR ttg gl1 lines a cross was made between gl1 GL2::GUS and ttg 35S::R-GR. F₁ progeny from that cross were allowed to self, creating F₂ segregating populations. For each genotype, at least 18 individual plants were analyzed for GUS activity. Plants that lacked significant GUS activity presumably did not contain the GL2::GUS reporter and were discarded from the analysis. F₂ seed was germinated on MS plates. ttg plants were scored based on the absence of anthocyanin production in the hypocotyl. Both TTG/n.d. and ttg plants were tranferred to MS plates containing 10 μM dexamethasone. 35S::R-GR is a dominant marker scored by a reduction of root hairs and the presence of increased trichomes in the presence of GL1. 35S::R-GR gl1 plants had reduced root hairs and did not initiate increased trichomes on the leaf surface. In order to isolate the ttg gl1 double mutant, F₂TTG/n.d. gl1 plants were identified as glabrous plants with normal seed coat muscilage. These F₂s were allowed to self, and F₃ seed was analyzed for segregation of ttg and the presence of the GL2::GUS transgene. One line was identified as homozygous for gl1 and GL2::GUS, and heterozygous for TTG. This line was used to quantitate GUS activity in the ttg gl1 and in gl1 TTG/n.d. plants. Similar results were obained with two other lines that were heterozygous for GL2::GUS and TTG and homozygous gl1. GL2::GUS quantitation of ttg GL1/n.d. was conducted on an F₃ line that was homozygous ttg, and segregating for GL2::GUS and gl1; heterozygosity at GL1 was confirmed by crossing 8 F₃ plants with a gl1-1 homozygote and scoring the F₁s for ttg and gl1.

The amino acid sequence of the c-myc epitope tag (MSEQKLISEEDL; Evan et al., 1985) was clone into the NcoI site at the 5′ end of the GL1 full length cDNA using a synthesized oligonucleotide. This new clone was moved as a cassette into an NcoI/EcoRI cleavage of the pSputk expression vector (Stratagene) downstream of the Sp6 promoter and the Kozak initiation site. This same epitope was cloned in a similar manner to the 5′ end of the GL1 myb domain cDNA (aa 1-126). These epitope fusions were cloned into the pSputk vector. The full length R cDNA (Ludwig et al., 1989) was also cloned into pSputk for in vitro expression.

In vitro transcriptions and translations were performed together using the TnT coupled reticulocyte lysate system (Promega). Translations were performed using [³⁵S]methionine and subsequently quantitated using TCA precipitation onto filter paper to ensure equal loading and molar ratios. Immunoprecipitations were performed with anti-c-myc antibody Ab-1 (Oncogene Science) and with protein A/G+ agarose beads (Oncogene Science) as suggested by the manufacturer. This involved clearing the lysate by incubation of the translation reactions with mouse normal anti-serum and beads for 1 hour on ice followed by low speed centrifugation. The cleared lysates were then incubated with the Ab-1 antibody at a 1:500 dilution and the beads at 4°C overnight with gentle shaking. The bead/antibody/protein complexes were then pelleted and washed 4 times with cold PBST buffer (1×PBS, pH 7.4, with 1% Triton X-100). Co-precipitations were treated identically except that the proteins were mixed at an equal molar ratio and allowed to sit at room temperature for 30 minutes prior to the clearing step.

Antibodies for immunocytochemistry were generated against a 21 amino acid peptide [SNGAHVQSIANLSKGQDRGNS] corresponding to residues 594-614 in the deduced amino acid sequence of GL2 (BioSynthesis Inc.). Antibodies were affinity purified using the peptide as an antigen. Arabidopsis seedlings were fixed overnight in buffered 4% paraformaldehyde at 4°C. Samples were dehydrated in an ethanol series, and infiltrated with Histoclear. Once equilibrated in 100% Histoclear, samples were infiltrated with Paraplast over a period of 4 days. Samples were mounted into blocks, and sectioned at 11 μm using a Reichert Jung microtome. Sections were dewaxed with Histoclear, and rehydrated in a graded ethanol series. Samples were detected using the Vectastain Elite Kit (Vector Laboratories) with minor modifications. Endogenous peroxidases were blocked in 0.3% hydrogen peroxide in methanol. After three washes in phosphate-buffered saline, samples were subjected to proteolysis with 20 μg/ml proteinase K (Sigma) for 20 minutes. Proteinase K digestion was stopped by incubation in 2 mg/ml glycine. Primary anti-GL2 antibodies were used at a 1:17,000 dilution along with 1:1,000 normal goat serum (Sigma). After incubation with a biotinylated goat secondary antibody and an avidin/biotin horseradish peroxidase complex, samples were detected with 10 mg/ml 3,3′-diaminobenzidine hydrochloride in 25 mM Tris-HCl, pH 7.5, 0.03% hydrogen peroxide, 0.04% nickel chloride. Staining was monitored under a stereomicroscope, and samples were dehydrated and mounted with Permount prior to microscopy.

A 2.1 kb HindIII/NheI fragment from the 5′-upstream region of the GL2 gene was cloned into pUC 118 using standard molecular techniques (Ausubel et al., 1995) to form pGL2pro. A HindIII/BamHI fragment containing the GL2 promoter was cloned as a transcriptional fusion to GUS using the pBI101.1 plasmid (Jefferson, 1987) to form pGL2::GUS. The ΔRI and ΔRV constructs was generated by digestion of pGL2pro with EcoRI and EcoRV, filling in the protruding ends with T4 DNA polymerase (NEB), and digestion with BamHI. These fragments were cloned into HindIII/BamHI digested pBI101.1 vector in which the HindIII overhang had been filled in with T4 DNA polymerase (NEB). All of the other deletion constructs shown in Fig. 5 were made by double digesting the pGL2pro plasmid with the enzymes indicated, filling in the protruding ends with T4 DNA polymerase (NEB), and religating the DNA molecule. The truncated GL2 promoter fragments were then cloned into pBI101.1 as HindIII/BamHI fragments. The GL1::GUS promoter corresponds to the GGE4 promoter GUS fusion described by Larkin et al. (1993).

A cross was made between 35S::GL1/35S::GL1 GL2::GUS /n.d. and 35S::R-GR/n.d.ttg/ttg plants. In addition, ΔRI::GUS and ΔXH::GUS were crossed to 35S::GL1/n.d. 35S::R-GR/n.d. ttg/ttg plants. The F₁ progeny were germinated on MS plates with 10 μM dexamethasone, putative 35S::GL1 35S::R-GR plants were transferred to soil. Seed was collected from at least five plants. The genotype of these individual plants was confirmed by segregation data from the F₂ generation, and by PCR amplification of genomic DNA with primers specific to 35S::GL1 and 35S::R-GR. F₂ progeny were analyzed histochemically for GUS activity after growth on MS plates containing 10 μM dexamethasone. For GL2::GUS, F₃ plants were scored as ttg, 35S::GL1 ttg, 35S::R-GR ttg, or 35S::GL1 35S::R-GR ttg. Floral explants from F₃ individuals were surface sterilized and incubated on MS plates with or without dexamethasone for 24 hours. Floral explants were then stained for GUS activity.

To establish a framework for discussing trichome development, trichome development will first be divided into several stages (Fig. 1). During the first stage, a committed cell begins to expand radially relative to its surrounding cells (stage 1; Fig. 1A), and the nucleus undergoes endoreduplication (Hülskamp et al., 1994). After the trichome precursor expands to two to three diameters greater than the surrounding cells, it emerges perpendicular to the epidermal plane and produces what will become the stalk of a trichome (stage 2; Fig. 1A). Next, the top of the growing cell develops secondary foci of cell expansion that will form the trichome branches (stage 3; Fig. 1A). Leaf trichomes typically produce two to four branches depending on the geographic race (Marks and Esch, 1994). After all branches have been initiated, the cell continues to expand via diffuse growth, such that it increases in diameter as well as in height. Initially, the growing branch tips are blunt (stage 4; Fig. 1A); but later, the tips become pointed (stage 5; Fig. 1B). After cell expansion ceases, the outer trichome surface develops numerous papillae (stage 6; Fig. 1B).

A 2.1 kb HindIII/NheI fragment from the 5′-untranslated region of the GL2 gene was cloned as a transcriptional fusion to the β-glucuronidase gene (GL2::GUS), and used to transform wild-type Arabidopsis plants. The activity of the GL2::GUS fusion throughout leaf development is shown in Fig. 2. In the young leaf primordia and in developing leaves up to approximately 400 μm in length, GL2::GUS activity was detected in developing trichomes and surrounding epidermal cells. In leaves longer than 500 μm, most of the non-trichome GL2::GUS activity was restricted to the basal regions of the leaf (Fig. 2A). GL2-dependent staining also was detected in the petiole and in the midvein of developing leaves. Transverse sections of GUS-stained leaves that were less than 400 μm in length showed that the promoter was active in all cell layers in the developing leaf (Fig. 2B). Once the developing leaf reached approximately 600 μm, GL2::GUS activity was detected in the basal regions of the leaf where trichomes continue to initiate (Fig. 1C), but GUS activity was only observed in trichomes in the apical two thirds of the leaf (Fig. 2C,D). Early stage trichomes (stages 1-3) forming at the base of older leaves (greater than 600 μm in length) often showed higher GL2::GUS activity relative to surrounding adjacent epidermal cells that do not enter the trichome pathway (Fig. 2E). However, on younger leaves, the level GL2::GUS activity in stage 2 and stage 3 trichomes relative to neighboring epidermal and subepidermal cells was often similar (Fig. 2F). GUS activity persisted during stages 2 and 3, but after stage 3, the level of GUS activity was variable and similarly staged trichomes often displayed differential GL2::GUS activity (compare trichomes A and B in Fig. 2G). As a trichome reached its final size, the level of activity appeared to increase (Fig. 2G, trichome B) and in mature trichomes, GL2::GUS activity remained stable for the lifetime of the cell (Fig. 2H). The data are consistent with the hypothesis that GL2 is expressed throughout trichome development.

To confirm the expression patterns of the GL2::GUS fusion observed in developing leaves, anti-GL2 antibodies were used to localize the endogenous GL2 protein. Wild-type and gl2-2 leaves were prepared for immunocytochemical staining using a polyclonal antibody directed against an epitope in the C terminus of GL2. gl2-2 served as a negative control because it contains a T-DNA insertion consisting of several tandemly linked T-DNA units in the coding sequence upstream from the deduced 20 amino acid GL2 peptide against which the anti-GL2 antibodies were generated (Rerie et al., 1994). The anti-GL2 staining results obtained were consistent and the results from a single experiment are shown in Fig. 3. Wild-type plants showed significantly greater staining than gl2-2 plants (compare Fig. 3A,B). Wild-type sections processed for immunocytochemistry without primary antibody or probed with pre-immune serum, as a control, did not display any significant staining (data not shown). Immunolocalization of GL2 protein in young wild-type leaves detected higher levels of GL2 protein in all cell layers compared to older leaves (compare Figs 3B and 3C). In the subepidermal cells of young leaves, the anti-GL2 signal was distributed throughout the cytoplasm with some nuclear staining (Fig. 3D); however, in trichomes at stage 4 (Fig. 3B,C) and at stage 3 (Fig. 3E) the signal was localized to the nucleus. Immunolocalization experiments with antibodies against GL1 did not reveal any significant cytoplasmic staining, and all GL1 signal was localized to the nucleus (D. Szymanski, unpublished results). Methylene blue staining of similar sections showed that the cytoplasm of developing trichomes was present and presumably accessible to the antibody reagent. The staining pattern of the anti-GL2 antibodies corroborate the results obtained using the GL2::GUS reporter construct. Furthermore, the partitioning of GL2 protein between the cytoplasm and nucleus revealed another potential level of control on GL2 activity.

Genetic analysis of the trichome developmental pathway showed that the TTG and GL1 genes are required for trichome initiation. Epistatic analysis placed GL2 downstream from GL1 and TTG (reviewed by Marks, 1997; Hülskamp et al., 1994), which suggests that GL2::GUS expression may require GL1 and/or TTG. In order to address this issue, the GL2::GUS construct was analyzed in gl1, ttg, and gl1 ttg backgrounds. For gl1 and ttg mutants, the GUS staining patterns were observed in at least 5 shoots removed from transformed callus tissue, and were confirmed in at least three independent transformants grown from seed, as well as in F<sub>_2> and F<sub>_3gl1, ttg and GL2::GUS were segregating. The transformant gl1 GL2::GUS D-1 was used as a parent in a cross with a ttg plant to analyze GL2::GUS in the double mutant. Results from representative plants are shown in Fig. 4. The leaf primordia and petiole staining observed in wild-type plants (Fig. 4A) was also observed in gl1 (Fig. 4B). However in gl1 there was a consistent reduction in GL2 staining in the marginal and apical domains of developing leaves compared to wild-type. In contrast, ttg plants lacked significant GL2::GUS staining in the petiole of fully expanded leaves, and staining was restricted to the margin of developing leaves (Fig. 4C). However, the diffuse leaf primordia staining in ttg was similar to gl1 and wild-type. In both gl1 and ttg, staining was not detected in any epidermal cells in mature leaves. This suggests that the GL2::GUS expression in fully expanded leaves of wild-type plants is dependent upon the presence of trichomes induced by the activities of GL1 and TTG; otherwise, a subset of epidermal cells on mature gl1 and ttg leaves would accumulate GUS. The detection of significant GL2::GUS in both gl1 and ttg suggested that either GL1 or TTG was sufficient for GL2 transcription in developing leaves. To address this possibility, the GL2::GUS reporter was crossed into a gl1 ttg background and analyzed for GUS staining. In the double mutant, GL2::GUS activity is faintly detected in the leaf primordia, and is restricted to the margins of developing leaves. In later leaves GL2::GUS activity was restricted to the margins of the leaf base and petiole (Fig. 4D). The GL2::GUS construct was also transformed into gl2 plants. It was found that GL2::GUS expression in gl2 shoots (Fig. 4E) and mature leaves (Fig. 4F) mirrored the expression in wild-type plants. Therefore GL2 does not function as an essential factor that autoregulates its expression in the shoot and in developing trichomes.

In order to quantify the relative levels of GL2 transcription, the total GUS activity of leaves and cotyledons at the four leaf stage was measured in vitro (Table 1). Plants that contained at least one wild-type copy of GL1 and TTG had a mean GUS activity of 2200 pmol MU/minute.mg. gl1 and ttg plants lacked trichomes and had reduced mean GUS activities of 280 and 250 pmol/minute.mg respectively. Consistent with the histochemcal staining, the gl1 ttg double mutant had a mean GUS activity of only 59 pmol MU/minute.mg. The ability of the maize R gene to complement ttg phenotypes was also reflected in its ability to rescue GL2::GUS transcription in the ttg background. 35S::R-GR gl1 ttg plants that displayed a mean GUS activity ²of 540 pmol MU/minute.mg.

Previous molecular complementation experiments have shown that the presence of the upstream domain between the EcoRI and EcoRV sites is required for full GL2 function in the leaf (Rerie et al., 1994). This result suggested that transcriptional control was a significant component of GL2 function in wild-type plants. To address this issue, several nested and internal deletions of the GL2 promoter were cloned as transcriptional fusions to GUS and analyzed for expression patterns in transformed plants. The physical map of the analyzed constructs and a summary of the observed expression patterns are shown in Fig. 5. Internal deletions from the HpaI site at position −2007 (relative to the first translation start codon) to the XbaI site at position −880 (construct 3) or to the MscI site at position −421 (construct 2) eliminated GL2::GUS trichome expression. However, faint staining in leaf primordia was detected with construct 3, suggesting that sequences between −880 and −421 may also play a role in regulating GL2 shoot expression. The results with constructs 2 and 3 indicate that the deleted region contains important elements for shoot and trichome expression. Plants containing constructs with external deletions down to the EcoRV site at position −1373 (construct 4) exhibited both non-trichome and trichome staining in the shoot. If the region between the EcoRV site and MscI of construct 4 are replaced with the region upstream of the EcoRV site (the region between the HindIII site at position−2149 and the EcoRV site; construct 6), shoot expression is maintained. This analysis indicates that sequences both upstream and downstream of the EcoRV site can mediate shoot and trichome expression. Both domains contained several sequence elements with similarity to well characterized binding sites for vertebrate (Biedenkapp et al., 1988) and plant (Grotewald et al. 1994; Sablowski et al., 1994) myb proteins.

Both the GL1 and TTG loci are required for trichome initiation, while GL2 is required for the earliest morphogenetic events of trichome growth. The presence of myb-class binding sites in the GL2 promoter provides additional circumstantial evidence for the involvement of GL1 in GL2 regulation. The spatial and temporal expression of the GL1 and GL2 genes also is consistent with the idea that GL1 regulates GL2 expression. A comparison of the staining patterns of representative GL1::GUS and GL2::GUS containing transgenic plants reveals the similarity between GL1 and GL2 expression profiles (Fig. 6). The GL1 promoter was active in non trichome cells at the base of developing leaves, and in developing trichomes. GL1::GUS staining was not detectable in older trichomes at the tip of the first leaf pair (Fig. 6A) or in the trichomes of mature leaves (data not shown). The GL2::GUS staining pattern in developing leaves is similar, except staining persists in mature trichomes (Fig. 6B). Transverse sections through mature GL1::GUS (Fig. 6C) and GL2::GUS (Fig. 6D) leaves confirmed a difference in transcription between GL1 and GL2 in mature trichomes. As indicated above, GL1 is not required for GL2 expression in the shoot, but the overlapping expression pattern of GL1 and GL2 suggests that GL1 could mediate GL2 expression during some stages of trichome development.

A basic model of the control of GL2 is one in which GL1 directly activates GL2 transcription. Previous experiments have shown that over expression of the GL1 gene using the 35S promoter results in a reduction in the number of leaf trichomes and the production of a few ectopic trichomes on the cotyledons (Larkin et al., 1994). However, over expression of the maize R gene with 35S::GL1 provided an additional activity such that there was a massive increase in the initiation of both ectopic and leaf trichomes (Larkin et al., 1994).

To examine the relationship between GL1, R and the GL2 promoter activity in vivo, plants that ectopically express GL1 and an inducible form of the maize R gene were analyzed for ectopic GL2 promoter activity. In Arabidopsis, the 35S::GL1 and 35S::R transgene combination causes seedling lethality (Larkin et al., 1994). However, 35S::GL1 plants that contain an inducible form of the R gene, that includes an N-terminal fusion to the vertebrate glucocorticoid receptor (35S::R-GR; Lloyd et al., 1994), are viable in the absence of the glucocorticoid analog dexamethasone. Plant viability is sensitive to TTG gene dosage; the most vigorous 35S::GL1, 35S::R-GR plants are homozygous for the ttg mutation. Inducible R activity has allowed us to monitor the effects of ectopic GL1 and R expression on the activity of the GL2::GUS transgene. Fig. 7 shows that in a ttg background (Fig. 7A) neither the 35S::GL1 (Fig. 7B) nor the 35S::R-GR (Fig. 7C) transgenes alone ectopically activate the GL2::GUS reporter in cotyledons and roots. Increased GL2::GUS staining in the midvein and margin of 35S::GL1 plants in wild-type or ttg/ttg background was often observed. Strong ectopic expression from the GL2 promoter was observed in five lines that ectopically expressed both GL1 and R-GR in the presence of dexamethasone (Fig. 7D). Induced plants that ectopically expressed GL1 and R-GR displayed widespread ectopic trichome initiation on their cotyledons, hypocotyls and leaves.

The ability of dexamethasone to induce ectopic expression of the GL2 promoter was tested using inflorescence explants from plants harboring the GL2::GUS reporter. Inflorescence explants from soil grown ttg/ttg, 35S::GL1; ttg/ttg, 35S::R-GR; ttg/ttg and 35S::GL1; 35S::R-GR; ttg/ttg plants were subcultured on MS plates with (induced) or without (uninduced) 1 μM dexamethasone. After 24 hours, the explants were analyzed for GUS activity (Fig. 7E-L). In wild-type plants, GL2::GUS is active in the outer cell layer of the seed coat and stained seeds are faintly visible through a cleared silique (Sattler and Marks, unpublished data). In the ttg background, wild-type GL2::GUS expression in developing seeds (data not shown) is absent (Fig. 7E,F), and ectopic expression of GL1 cannot rescue this activity (Fig. 7G,H). The 35S::R-GR construct can rescue GL2::GUS expression in developing seeds in a dexamethasone-independent manner, but does not give rise to ectopic GL2::GUS activation in the siliques (Fig. 7I,J). Strong dexamethasone-dependent activation in siliques was only observed in plants that ectopically expressed GL1 and R-GR in the presence of dexamethasone. 35S::GL1 35S::R-GR activation of GL2::GUS did not require adoption of the trichome cell fate, because differentiated epidermal cells of siliques ectopically expressed GL2::GUS. To rule out the possibility that the ectopic activation of the GL2 promoter in the 35S::GL1 and 35S::R-GR background represents a non-specific transcriptional activation, two additional GL2 reporter constructs were analyzed in that genetic background. The ΔXD construct, which lacks the GL2 promoter regions required for detectable trichome expression, was not ectopically activated by the 35S::GL1 35S::R-GR transgenes (Fig. 8A). The ΔRI construct, which contains both putative shoot transcription domains, was ectopically activated in the presence of ectopic GL1 and R-GR (Fig. 8B). These results indicate that GL1 and R require the region between the XbaI and EcoRI sites (Fig. 5) for the ectopic GL2 activation.

Because both GL1 and R are required to ectopically activate GL2, it is possible that GL1 and R function as a complex. This possibility has been examined through the use of in vitro co-immunoprecipitation experiments (Fig. 9). In order to provide a strong antigen, a construct containing the GL1 gene with a myc epitope as an N-terminal fusion was generated. This same epitope fusion was made to a truncated version of the GL1 protein that expressed only the N-terminal 122 amino acids, which contains the entire myb homologous domain. These two protein fusions, as well as the full length untagged R protein, were cloned into Sp6 expression vectors for the in vitro production of mRNA and subsequent reticulocyte translations in the presence of [³⁵S]methionine. Fig. 9A shows an autoradiograph of the SDS-PAGE gel of the translated myc-GL1 (lane 1), myc-GL1 myb (lane 2), unprogrammed lysate (lane 3), and R protein (lane 4). Two high molecular mass bands are observed in the R translation reactions; the lower molecular mass band may be due to the spurious initiation or termination of R translation reaction (lane 4). Following incubation of each translation reaction with the anti-myc antibodies, the antibody bound fraction was purified with sepharose G+/A beads. The resulting complexes were further separated by SDS-PAGE (Fig. 9B). Only the GL1 proteins expressing the myc epitope were precipitated (Fig 9B, lanes 1,2). It is important to note that while R is a myc homologous protein, it does not have any homology to the myc epitope tag and is not precipitated by the anti-myc antibodies (Fig. 9B, lane 3).

The translated proteins were incubated together prior to immunoprecipitation. Under these conditions it was found that R only immunoprecipitated in the presence of full-length GL1 (Fig. 9C, lane 1). The immunoprecipitated R protein corresponds to the high molecular mass form of R that is observed in Fig. 9A lane 4. The nature of the smaller products in Fig. 8A lane 4 are unknown; however, the fact that only the larger product was precipitated indicates that the interaction between GL1 and R is specific. In addition, since the myb domain of GL1 could not efficiently precipitate R it indicates that something in the carboxy terminus of GL1 is also necessary for a stable interaction and that the myc epitope itself is not the source of the interaction.

In Arabidopsis the TTG and GL1 genes are required for trichome initiation. GL2 function does not appear to be required for initiation since gl2 leaves have approximately normal numbers of trichomes. However gl2 trichome morphology is variable, and shows defects from stage 2 to stage 6. The expression of GL2 reflects a complex requirement for its activity during leaf development, and includes spatial and temporal regulation in developing leaves and within trichomes (Figs 2, 3). In leaf primordia, the GL2 gene is expressed in all cell layers. The apparently uniform level of expression in the epidermis of leaf primordia is low relative to that observed during some stages of trichome development. It is possible that this expression pattern reflects a permissive level of GL2 required for normal progression of the trichome morphogenetic program. The functional significance of the diffuse subepidermal primordia and petiole expression pattern is not known because the expression precedes trichome initiation and occurs in cell types that do not normally enter the trichome pathway. Because the number of cell layers and leaf morphology of developing gl2 leaves are not obviously different from wild-type, the non-trichome function of GL2 in developing leaves is unknown. GL2 protein appears to show cell-type specific subcellular localization in developing leaves, and is localized in the nucleus in developing trichomes (Fig. 2). The partitioning of GL2 within the cell may play a functional role in regulating GL2 activity.

Consistent with the early requirement for GL2 for accurate initial outgrowth of the trichome stalk, significant GL2::GUS activity was detected in developing trichomes. The apparent lack of this early GL2 activity in plants homozygous for strong mutant alleles results in a loss of trichome outgrowths. Some of the trichome outgrowths on gl2 plants expand somewhat normally, but have fewer branches and lack papillae. Therefore, the detection of GL2 protein from stages 2 through 5, and the expression of GL2::GUS during stage 6 is consistent with GL2 regulating multiple aspects of trichome morphogenesis.

The definition of the promoter elements that activate trichome gene expression is an important step in understanding the relationship between GL2 transcription and its regulation. Leaf primordia, leaf base, and trichome specific expression were not clearly separable activities and could be detected in two separate promoter domains flanking the EcoRV site at −1373 (Fig. 5). Either domain is sufficient for leaf and trichome expression. Inspection of the DNA sequences flanking the EcoRV site in the GL2 promoter revealed two sequence elements with similarity to vertebrate myb-class transcription factor binding sites. The orientation of both elements is inverted with respect to the directionality of the GL2 coding sequence. The more 5′ putative myb-binding site element (pMBS1: GACTAACGGTAAG) matches the consensus vertebrate myb-binding site in six of six positions (shown in bold; Biedenkapp et al., 1988). The proximal sequence element (pMBS2: TACTAACAGTATA) conforms to the vertebrate myb consensus in five of six positions (shown in bold). The sequence similarity between pMBS1 and pMBS2 extends beyond the myb consensus; they are identical in nine of ten positions. It is possible these binding sites are composite elements recognized by more than one factor and/or trichome-specific myb-binding sites that have additional sequence constraints outside of the myb-core sequence. The EcoRI to EcoRV domain also contains potentially significant sequences with identity in five of six positions to the core recognition sequence CC[T/A]ACC that is bound by the maize P gene (Grotewald et al., 1994) and myb305 from Antirrhinum (Sablowski et al., 1994). Previous molecular complementation experiments have shown that the presence of the upstream domain between the EcoRI and EcoRV sites is required for full GL2 function in the leaf (Rerie et al., 1994). Removal of this region resulted in trichomes with fewer branches. Therefore, the level of expression conferred by the two leaf expression domains appears to be important for normal trichome development, and the presence of multiple putative myb-binding sites reveals the possibility of complex combinatorial control of the GL2 promoter. The same two regions also mediate GL2::GUS expression in the outer seed coat (Sattler and Marks, unpublished data). The possibility that shared promoter elements control GL2 expression in seeds, shoots and trichomes will be addressed in future experiments.

There is extensive overlap in the shoot expression patterns of GL1 and GL2. Expression of both GL1 and GL2 is detected throughout young leaf primordia and during early trichome development. Mutations in GL1 are epistatic to those in GL2 and the GL2 promoter region contains potential GL1 binding sites. All of these findings suggest that GL1 may directly regulate some aspects of GL2 expression. It is also apparent that there are GL1-independent components to GL2 expression in the shoot. In mature trichomes GL2 transcription persists, while GL1 is apparently downregulated. It is possible that additional myb genes that are expressed in trichomes regulate GL2 (Li et al., 1996). In the absence of GL1 and TTG, mean GUS activity in the shoot was 59 pmol MU/minute.mg. In plants containing either TTG or GL1 mean GUS activity increased to approximately 250 pmol MU/minute.mg. This indicates either TTG or GL1 is sufficient to independently increase GL2 expression. The spatial pattern of GL2 expression in the leaf primordia in gl1 ttg plants is similar to gl1, ttg, and wild-type plants (Fig. 4); the major difference is in the relative level of expression and/or its persistence as the leaf develops (Table 1). Since this expression pattern does not persist in developing ttg gl1 leaves, the role of GL1 and TTG appears to be to stabilize and activate GL2 transcription in specific cells in the developing leaf. If the ttg-1 allele is a null, the reduced but detectable level of GL2 transcription in ttg gl1 mutants also indicates that an additional unknown factor(s) participates in GL2 transcription in developing leaves.

Ectopic GL2 expression cannot be induced by the ectopic expression of GL1 alone. In fact, 35S::GL1 transgenic plants express high levels of GL1 protein (D. Szymanski, unpublished results), yet show a reduction in trichome number (Larkin et al., 1994) and apparently normal GL2::GUS expression in the developing shoot and trichomes (Fig. 7). The maize R gene provides an additional activity such that plants that overexpress GL1 and R display widespread trichome initiation over much of their aerial surface and ectopic GL2::GUS transcription (Fig. 7). The presence of both ectopic GL1 and R bypasses the requirement for TTG and presumably the normal trichome initiation control mechanisms that operate via wild-type levels of GL1 and TTG. The detection of physical interaction between GL1 and R in co-immunoprecipitation experiments provides biochemical evidence that GL1 and R function as a complex. The interaction of GL1 with R is stable, specific, and occurs in the absence of DNA (Fig. 9). Therefore, it is possible that the association of GL1 and R, rather than the convergence of GL1 and R-dependent parallel pathways, is the mode of the activation of the GL2 promoter. The ability of this complex to constitute a unique biochemical activity is unknown, since we have not been able to identify a high affinity DNA-binding site for GL1, R, or the GL1-R complex on the GL2 promoter. Therefore it is not known if GL1 and R are part of a complex that directly or indirectly activates the GL2 promoter or if additional factors are required for a stable interaction with DNA. The fact that R can activate GL2 transcription in the absence of GL1 and TTG activity (Table 1) suggests that if R is part of a complex that regulates GL2, it is capable of interacting with additional components of the complex, perhaps the putative gl1-ttg-independent factor. Since R can compliment GL1-independent phenotypes of ttg plants it is expected that R interactions are not limited to GL1.

The relationship between the maize R gene and TTG is not clear. The dominant effects observed by overexpressing R have not been observed with any other plant bHLH containing gene. Furthermore, recent results indicate that TTG is not a R homolog, and its deduced amino acid sequence encodes WD 40 repeats (A. Walker and J. Gray, personal communication). The presence of R could bypass the need for TTG is several ways. For example, the function of TTG could be to regulate the expression or the activity of a myc factor. In this case, R could function independently of TTG. Alternatively, both GL1 and TTG may directly bind to the GL2 promoter, either alone or as part of a larger complex, but the interaction between GL1 and R somehow eliminates the requirement for TTG. The characterization of TTG and the identification of components of DNA-protein complexes that bind to the GL2 promoter should help to resolve this issue.

In conclusion, it has been found that both TTG and GL1 can influence GL2 expression in the shoot. Furthermore, the region of the GL2 promoter shown to be important for shoot expression is also needed for GL1-R mediated ectopic GL2 expression. However, many aspects of GL2 expression remain to be understood. What is the identity of the ttg-gl1-independent factor that gives rise to leaf primordia expression? What factors control the specific and persistent GL2 expression in mature trichomes? Finally, GL2 expression is unaltered in the roots and seeds of gl1 mutants. This indicates the presence of other factors that may participate in the regulation of GL2. The complexity of GL2 regulation and its expression pattern reflects its function throughout trichome development and in the control of cell fate in the root and seed. Further studies of the mechanisms of GL2 regulation will aid in the understanding of how complex and integrated gene networks control plant development.

We are grateful to Alan Lloyd (University of Texas) for the gift of the 35S::R-GR containing plants, and to Sue Wick for assistance in cryosectioning GUS-stained samples. We thank Daniel Klis for technical assistance. Thanks to Beth Kent, Scott Sattler and Pam Vanderweil for critical comments on the manuscript. The University of Minnesota Imaging Center provided excellent digital image technology support. This research was supported by NSF/IBN-9506192 and USDA 95-37304-2219.