Development of three different cell types is associated with the activity of a specific MYB transcription factor in the ventral petal of Antirrhinum majus flowers

Within a plant organ, there may be several different types of specialised cell. Even within a single tissue, such as epidermis, there are usually morphologically distinct cell types. The distribution and types of specialised cell depend on the specific function of particular organs. In the snapdragon, Antirrhinum majus, the zygomorphic flower is defined largely by the specialised structure of the corolla that is formed from five fused petals(two dorsal, two lateral and one ventral) that have distinct morphologies(Fig. 1A). The corolla is adapted for bee pollination. The lobes provide a brightly coloured visual signal to attract pollinators. The ventral petal, which is supported by the lateral petals, is bent at the hinge region between the tube and the lobe tissue and is reinforced to form a landing platform for bees. There are specialised conical cells on the adaxial surface (inner epidermis) of the petal lobes. The specialised shape of these cells optimises light capture by the pigments, causing the petal to appear deeper in colour, which increases the attractiveness of the flower to pollinating bees(Glover and Martin, 1998; Comba et al., 2000). Petals have other epidermal cell types, including multicellular hairs (trichomes). In Antirrhinum, trichomes are particularly numerous in the corolla tube,where they trap pollen from the surface of pollinators, act as nectar-guides for pollinators and produce scent to attract pollinators(Almeida et al., 1997; Kolosova et al., 2001). The processes regulating the differentiation of these specialised cell types constitute an important part of floral development.

In A. majus, the MIXTA gene regulates the differentiation of conical epidermal cells on the petal lobes. MIXTA encodes a MYB-related transcription factor that is necessary for the development of conical cells on the inner epidermis of the petals(Noda et al., 1994). MIXTA expression in Antirrhinum is limited to a narrow window of time relatively late in development of the petal epidermis and confined to adaxial petal epidermal cells(Glover et al., 1998). Ectopic expression of the MIXTA gene in tobacco results in the formation of two distinct cell types on leaves (Glover et al., 1998). All plants ectopically expressing MIXTAproduce outgrowths on their leaf epidermal cells, which resemble conical cells. Some lines also produce ectopic trichomes on their leaves and petals. These experiments show MIXTA to be necessary and sufficient for conical cell formation, but also indicate that two specialised cell forms,trichomes and conical cells, can be induced by MIXTA expression and therefore probably share at least part of a common developmental programme in tobacco. The same common developmental pathway is triggered by MIXTA in A. majus, as ectopic MIXTA expression also promotes formation of both cell types in its species of origin(Martin et al., 2002). Expression of MIXTA early relative to the commitment of cells to further rounds of division results in the formation of multicellular trichomes, while later expression gives rise predominantly to conical epidermal cells (Glover et al.,1998). In Antirrhinum, MIXTA is expressed in the petal epidermis only after cell division has ceased, explaining the observation that it is involved only in conical cell formation(Glover et al., 1998). mixta mutants develop floral trichomes normally, suggesting that other genes are responsible for the formation of trichomes in Antirrhinum.

Trichomes develop in specific regions of the petals of Antirrhinum, including the inner epidermis of the corolla tube and the outer epidermis of the petal lobes(Fig. 1A). Inside the mouth of the flower, in the throat of the corolla tube, a mass of trichomes forms that collects pollen from the undersides of pollinators. Beyond this mass,extending down the ventral/lateral petal border to the base of the anther filaments, are two strips of yellow trichomes that act as nectar guides for bees (Raman, 1990; Galego and Almeida, 2002). Analysis of graminifolia mutant flowers, in which the petals do not fuse to form the corolla tube, shows that the mass of trichomes in the throat of the tube forms largely on the ventral petal. The strips of yellow trichomes, which serve as nectar guides, are determined in part by the ventral petal and in part by the lateral petals (C.M., unpublished). The glandular heads of the trichomes are one of the sites of scent (methyl benzoate)production (Kolosova et al.,2001). Given that Antirrhinum flowers produce numerous multicellular glandular trichomes, but that MIXTA function is confined to the induction of conical cells on the inner petal epidermis, we investigated whether other MYB transcription factors related structurally to MIXTA might be involved in controlling trichome formation.

The AmMYBML1 (Antirrhinum majus MYB MIXTA LIKE 1) gene encodes a protein structurally very similar to MIXTA. Here, we describe the functional characterisation of AmMYBML1. Ectopic expression of the gene in tobacco shows that it can induce the production of both trichomes and conical cells on floral tissues. In Antirrhinum, the AmMYBML1 gene is expressed only in the ventral petal. Within the ventral petal its expression is localised to the conical epidermal cells of the hinge, the expanded mesophyll of the hinge and the inner (adaxial)epidermis, including trichomes, within the corolla tube. Floral expression is determined by the promoter of AmMYBML1. Loss of function of the DIVARICATA (DIV) gene, which specifies ventral petal identity (Galego and Almeida,2002), results in a large reduction in AmMYBML1transcript levels and loss of the three specialised cell types in which AmMYBML1 is highly expressed. Weak alleles of DEFICIENS(DEF), which encodes one of the B-function MADS-domain transcription factors required for petal identity, also result in reduced expression of AmMYBML1 in petals, and cause similar losses in the three specialised cell types as mutants of DIV. These data suggest that AmMYBML1 is involved in the differentiation of three distinct specialised cell forms and so contributes to petal form and function; this role is dictated, in large part, by the specific expression pattern of AmMYBML1, which is controlled by DIV in association with B-function MADS-domain proteins.

RNA in situ hybridisation was performed on sections of Antirrhinumbuds according to Jackson et al. (Jackson et al., 1991) and Bradley et al.(Bradley et al., 1993). An AmMYBML1 cDNA fragment (GenBank Accession Number AJ006292) and a MIXTA cDNA fragment, each cloned in pBluescript, were used as templates for riboprobe synthesis. The vector containing the Cyclin D3b gene from Antirrhinum has been described by Forbert et al.(Forbert et al., 1996).

A sense construct for ectopic expression of AmMYBML1 in tobacco was created by excising the whole AmMYBML1 cDNA from pCR2.1 using SnaBI and BamHI and inserting it into the HincII and BamHI sites of pJIT60(Guerineau and Mullineaux,1993). A cassette containing the cDNA driven by the double CaMV 35S promoter and with the CaMV terminator sequence was excised using SstI and BamHI sites, and inserted into the SstI and BglII sites of the pBin19 binary vector(Bevan, 1984).

The genomic sequences of AmMYBML1 were identified by screening a genomic library from A. majus in lambda EMBL4 (a gift from Hans Sommer, MPI Koln). Primers were designed to amplify a large region of promoter sequence (1.6 kb), and a selection of shorter promoter fragments, each with BamHI and XbaI restriction sites on the end. These PCR products were digested and ligated into a BamHI/XbaI-digested pGREEN-GUS binary vector(Hellens et al., 2000). The AmMYBML1 promoter sequence has been submitted to EMBL/GenBank(AY661653), along with the AmMYBML2 and AmMYBML3 cDNA sequences (AY821655 and AY661654, respectively).

Constructs were transformed into Agrobacterium tumefaciens LBA4404 using electroporation (Mattanovich et al.,1989). Tobacco leaf discs were transformed using the method of Horsch et al. (Horsch et al.,1985).

RNA was extracted and transcript levels were monitored by northern analysis, using the method of Martin et al.(Martin et al., 1985). Quantitative RT-PCR was undertaken using the method of Jin et al.(Jin et al., 2000) using oligonucleotide primers G3540 (5′-ATTTGGTGCTGAGGTTGAGA-3′) and G3543 (5′-ACAACTGACTCCAGCAAACG-3′) for the ubiquitincDNA, and F4266 (5′-CATTGTACGTACGTTCATCTTTAGTTAGCTTC-3′) and F4267(5′-GGGCGCGCCAGCTTCCATGACCATGTTCTT-3′) for the AmMYBML1cDNA. Twenty cycles were used to amplify the ubiquitin cDNA within the linear range and 15 cycles for AmMYBML1.

Tissue samples were immersed in GUS staining solution (40 mg/ml X-Gluc in 100 mM NaH₂PO₄, 10 mM Na₂EDTA, 0.5 mM K ferrocyanide, 0.5 mM K ferricyanide, 0.1% Triton X-100, pH 7) and incubated at 37°C for 3 hours. Samples were destained with 70% ethanol and examined using a light microscope.

Transformation of tobacco (Nicotiana tabacum cv Samsun)protoplasts was performed as described by Negrutiu et al.(Negrutiu et al., 1987). Plasmid DNA (10 μg) containing the AmMYBML1 promoter region (1571 bp) fused to the GUS gene was used for each transfection. Different combinations of 4 μg of plasmids containing cDNA sequences encoding DIV,DEF, GLO and DEF-VP16 under the control of the double CaMV35S promoter were used to test expression from the AmMYBML1 promoter. Varying amounts of the plasmid pJIT60 were used to ensure that equal total amounts of DNA were introduced for each transfection and to control for any possible effects of the CaMV35S promoter in titrating out transcription factors. After transfection, protoplast extracts were assayed for β-glucuronidase (GUS)activity according to Jefferson(Jefferson, 1987). All transfections were performed in duplicate, and GUS assays were performed in duplicate for each transfection. Similar results were obtained in at least two independent experiments for each effector combination. GUS activity was calculated as nmol methylumbelliferone per mg protein per minute.

Plant tissue was examined under a CamScan mark IV scanning electron microscope with a Hexland cryostage.

To identify MIXTA-like genes that might control trichome formation in Antirrhinum flowers, a cDNA library made from RNA from lobe tissue of the mixta mutant line was screened for clones encoding proteins structurally related to MIXTA, by probing with the full-length MIXTAcDNA (Glover et al., 1998) and washing at low stringency (3×SSC, 0.5% SDS, 65°C). Transcripts encoding two distinct R2R3 MYB proteins closely related to MIXTA were identified. Two full-length clones of AmMYBML1 and two full-length clones of AmMYBML2 were isolated from 100,000 pfu screened. During subsequent screening of a genomic library in bacteriophage lambda EMBL4 with the AmMYBML1 cDNA, a third, highly related, gene was identified(AmMYBML3). The predicted open reading frames in this genomic clone were identified and the sequence of the transcript was confirmed by cDNA isolation by RT-PCR (Frohman et al.,1988).

The amino acid sequences of the three MIXTA-like proteins, AmMYBML1,AmMYBML2 and AmMYBML3, were compared with the primary sequences of other R2R3 MYB proteins using CLUSTALW. The amino acid sequence in the two MYB domains of R2R3-MYBs is highly conserved, and phylogenetic analysis places AmMYBML1,AmMYBML2 and AmMYBML3 in subgroup 9 of the 24 subgroups that have been identified among the 125 Arabidopsis R2R3 MYB proteins(Stracke et al., 2001). Subgroup 9 contains the MIXTA-LIKE MYB genes, which are defined by both the conservation of their DNA-binding domains and by motifs present in the C-terminal domains (Kranz et al.,1998; Stracke et al.,2001). The functions of only two members of this group have been characterised. PhMYB1 from Petunia hybrida plays a similar role to MIXTA in conical petal cell development(Avila et al., 1993; van Houwelingen et al.,1998).

AmMYBML2 is most closely related to PhMYB1 from Petunia hybrida(Fig. 1B)(Avila et al., 1993). AmMYBML2 and AmMYBML3 are also closely related to AtMYB16 and AtMYB106. In fact,AmMYBML2 appears to be orthologous to PhMYB1 from our alignments, and MIXTA and AmMYBML1 appear to be products of more recently derived gene duplications of a subgroup 9 ancestral gene. We were interested in whether the developmental function of these related genes had radiated along with the changes in floral morphology that adapt members of the Scrophulariaceae to specific pollinators. Consequently, we examined the function of AmMYBML1 in Antirrhinum in greater detail.

Previous analysis had indicated that AmMYBML1 is expressed in petals, early in tissue development(Glover et al., 1998). Investigation of AmMYBML1 expression in other plant organs showed the AmMYBML1 transcript to be restricted to the floral tissues; no expression was detected in leaves or roots(Fig. 1C). Petal tissues express AmMYBML1 only during the early stages of development, with maximum transcript found in buds of length 0-5 mm. Analysis of expression in young, medium and old leaves indicated that AmMYBML1 is not expressed in leaves, irrespective of developmental stage.

In situ hybridisation of Antirrhinum flower buds showed that AmMYBML1 expression is limited to the ventral petal, and is absent from the other floral whorls and the lateral and dorsal petals(Fig. 2A,B). Probing equivalent sections with the cell cycle marker Cyclin D3B, showed that AmMYBML1 is expressed in the ventral petal while its cells are still competent for further division (Fig. 2B,C). In a double mutant of A. majus (cyc,dich)that has radially symmetrical flowers with only ventral petals(Luo et al., 1996) AmMYBML1 is expressed in all the petals, confirming that its expression is specific to the ventral petal(Fig. 2D).

Investigation of the localisation of AmMYBML1 transcript within ventral petal tissue at slightly later developmental stages identified three key regions of expression. AmMYBML1 was found to be expressed in the epidermal cells and in the trichomes which line the throat of the corolla tube and act to collect pollen from the undersides of pollinating bees(Fig. 2E). Transcript was also found in the adaxial epidermal cells overlying the hinge of the petal, which is the folded region of the ventral petal between tube and lobes(Fig. 2E,F). AmMYBML1expression was also detected in the mesophyll cells of the hinge region, but restricted to those cells on the adaxial side of this tissue(Fig. 2E,F). The expression of AmMYBML1 in mesophyll cells is in direct contrast to the expression pattern of MIXTA, which is restricted in its expression to the adaxial epidermal cells (Fig. 2G).

To test the biological function of AmMYBML1, and to compare it with MIXTA, a construct containing the AmMYBML1 cDNA under the control of the double CaMV35S promoter was used for transformation of tobacco. Expression of AmMYBML1 in regenerated tobacco plants was confirmed by RNA gel blot analysis. Eight independent lines with detectable levels of AmMYBML1 expression in leaves were identified and used for phenotypic characterisation (Fig. 3H). In the vegetative tissues, the phenotypes of the plants expressing AmMYBML1 were identical to those of wild-type tobacco plants. However, significant differences were observed in floral tissues. Scanning electron microscopy showed that conical cells and trichomes developed ectopically in the epidermis of several floral organs. The strongest phenotype was observed in carpel epidermal tissue, but the epidermal cells of the other floral organs, and the bracts, were also affected. Microscopic analysis of leaf and stem epidermal samples confirmed that, unlike MIXTA, AmMYBML1 was not able to regulate the production of conical cells or trichomes in vegetative tissues (Fig. 3A,B). However,bracts, sepals and stamens were found to develop conical cells in response to AmMYBML1 expression, while the petals and carpels were found to differentiate full multicellular trichomes, as well as ectopic conical cells(Fig. 3C-G). These data demonstrate that the AmMYBML1 protein can function in the same way as the MIXTA protein to direct the differentiation of specialised epidermal cell forms, both conical cells and trichomes. However, the activity of AmMYBML1 is restricted to the reproductive phase of plant development, suggesting that AmMYBML1 requires the presence of additional factors, which are restricted to floral tissues, to induce cellular differentiation. As tobacco provides an excellent model for testing gene function in Antirrhinum, as evidenced by the identical effects of ectopic expression of MIXTA in the two species (Martin et al.,2002), these data indicate that AmMYBML1 can promote both conical cell and trichome formation in Antirrhinum.

The detection of AmMYBML1 expression in the adaxial epidermis and adaxial mesophyll cells of the hinge region of the ventral petal(Fig. 2E,F) led us to investigate the possible function of the protein in this region of the Antirrhinum petal. Fig. 4A shows a freeze fracture section through the hinge of a mixta mutant petal, which has flat epidermal cells on the inner epidermis of the petal lobes. In the region of the hinge, however, a few conical cells (indicated by arrows in Fig. 4A) are formed in the mutant line, presumably induced by the expression of AmMYBML1 as the transposon insertion inactivates the MIXTA gene entirely (Noda et al.,1994). Conical cells in the hinge region were also observed in a second stable mixta mutant allele derived by EMS mutagenesis (C.M. and Z.S.-S., unpublished). These cells (indicated by arrows) can be seen in surface view in Fig. 4B.

Freeze fracture sections of the petal hinge indicated that the mesophyll cells in this region are tightly packed on the adaxial side(Fig. 4D). This contrasts with the loosely packed cells of the petal lobe mesophyll(Fig. 4E) and suggests a role for AmMYBML1 in the expansion of the adaxial hinge mesophyll cells. These cells contact each other through surface projections(Fig. 4F). Differential expansion across the mesophyll transect may result in the generation of the folds of the hinge structure that strengthen the landing platform and provide grips for pollinating bees. AmMYBML1 is indeed capable of inducing tissue thickening, because ectopic expression throughout the petal mesophyll in tobacco resulted in petals with much thicker mesophyll(Fig. 4H) than controls(Fig. 4G). Measurements of the thickness of equivalent regions of petal tissue gave a mean value of 251±119 μm for control tobacco flowers and 374±79 μm for flowers from the 35S-AmMYBML1 tobacco lines, values that were different with 95% confidence limits. Interestingly, flowers of tobacco plants expressing MIXTA under the control of the CaMV35S promoter(Glover et al., 1998) did not show such pronounced thickening of the petal mesophyll (mean value 330 μm)and Antirrhinum flowers ectopically expressing MIXTAretained the folds of tissue in the hinge region of the flower (see Fig. S1 in the supplementary material). These observations suggested that MIXTAis less effective than AmMYBML1 at promoting the expansion of petal mesophyll cells.

The AmMYBML1 promoter, 1610 bp upstream from the initiating ATG codon, was sequenced and analysed using databases of transcription factor binding sites. Fig. 5A shows the AmMYBML1 promoter with key regions highlighted. The putative TATA box is located between 171 and 162 bp upstream of the initiating ATG codon. Two motifs of interest were identified between –364 and –358 bp and between –1113 and –1109 bp upstream of the ATG. These contained putative I-box motifs, which is interesting because these sequences have been shown to be bound by proteins similar to DIVARICATA (DIV)(Rose et al., 1999; Galego and Almeida, 2002). DIV influences specifically the growth of ventral and lateral petals(Almeida et al., 1997; Galego and Almeida, 2002). In div mutant plants, each half of the ventral petal becomes a mirror image of the adjacent part of the lateral petal(Almeida et al., 1997),resulting in a loss or significant reduction of the hairs in the throat of the tube that collect pollen and in the size of the strips of hairs that pass down the corolla and mark the junction between ventral and lateral petals(Almeida et al., 1997). The DIV protein is a MYB-related transcription factor which has two MYB domains, which are separated in the primary sequence of the protein. Structurally, it is most similar to LeMYB1 from tomato (Rose et al., 1999). LeMYB1 binds the I-box sequence GATAAG, while StMYB1(a related protein from potato) binds to a similar sequence, GGATAAG, or to a core sequence, GGATA (Baranowskij et al.,1994; Rose et al.,1999). The putative I-box at –364 in the AmMYBML1promoter is AGATAAA. As A and G are both purines, and therefore functionally similar, the sequence AGATAAA may be similar enough to GGATAAG to be bound by similar proteins. Alternatively, between –1113 and –1109 bp the sequence GGATA, the core for StMYB1 binding, was found.

The promoter region of AmMYBML1 was used to produce four promoter-GUS fusion constructs of different sizes: 1571 bp, 1260 bp, 884 bp and 223 bp in length. Promoter activity was analysed by transfer of these reporter constructs into tobacco and histochemical assays of GUS activity in T1 plants. The 1610 bp and 1260 bp promoters drove GUS expression in outgrowing cells of the floral organs. This included all types of trichomes covering the floral organs, both glandular trichomes and conical petal cells(Fig. 5B,C). The reduced promoter (884 bp) and the short promoter (223 bp) conferred GUS expression in trichomes in all aerial organs (Fig. 5D,E) but also gave expression in the vascular tissue of the plant(Fig. 5D). No expression was apparent with any AmMYBML1 promoter-GUS constructs in root tissue of tobacco seedlings.

The expression pattern determined by the 223 bp promoter fragment showed that this sequence is all that is required for expression in trichomes. Consequently, this region is likely to include the important cis-acting elements regulating trichome-specific expression. Aside from the TATA box and one potential CAAT box, there are two GAMYB (MBSII)(Solano et al., 1995) binding motifs within this region.

The 1260 bp promoter was enough to limit the expression of AmMYBML1 to the trichomes of the flowers but the 884 bp promoter directed GUS expression in trichomes and vascular tissue in both floral and vegetative tissues. Thus, reducing the promoter length from 1260 bp to 884 bp resulted in GUS expression being directed outside the flower. These results suggest that negative regulatory elements silencing expression in vegetative tissues are located within the promoter of AmMYBML1 between 1260 bp and 884 bp upstream of the ATG. Within this region there are at least four potential MADS domain protein-binding sites (CArG boxes; Fig. 5A). One of the two putative I boxes is also located in this region, and may be involved in specifying ventral petal-specific expression in Antirrhinum.

Unfortunately, because tobacco flowers are radially symmetrical, we were unable to use expression in tobacco to identify sequences active in determining expression in the ventral petal only.

The observation that AmMYBML1 transcript is restricted to the ventral petal in wild-type Antirrhinum flowers(Fig. 2A) led us to investigate its expression in the div mutant. The div mutant lacks ventralising signals in its flowers, and all five petals are consequently dorsalised (Galego and Almeida,2002). The characteristic mass of trichomes in the throat of the tube in the wild-type flower is absent from this mutant(Fig. 6A-D). The mutant also fails to make the folds of tissue around the hinge region of the lower petals(Fig. 6A-D,F,G) and lacks conical cells in the region of the corolla epidermis just below the hinge(Fig. 6H). On the inner epidermis of the rest of the ventral petal lobe, conical cells develop normally (Fig. 6I,J),suggesting that MIXTA activity is unimpaired in div mutants(Fig. 6E). Because DIVencodes a MYB-related transcription factor responsible for determining ventral petal identity (Galego and Almeida,2002), the restriction of AmMYBML1 expression to the ventral petal suggested that it may be a target for transcriptional control by DIV, and that the phenotype of div mutants might reflect the loss of AmMYBML1 expression as a consequence of the loss of activity of its upstream regulator. Gel blots of RNA from petals of wild-type and divmutant flowers showed that AmMYBML1 was expressed in both lines, but the expression in wild-type petals was at least 10-fold higher than in the div mutant petals (Fig. 6E). MIXTA expression was unaffected by mutation of DIV (Fig. 6E).

Because AmMYBML1 expression is specific to petal tissue, it seemed likely that further regulation of AmMYBML1 expression might operate through proteins determining petal identity. In Antirrhinum, these include the B-function MADS-domain proteins DEFICIENS (DEF) and GLOBOSA (GLO). One problem in examining the effect of these genes on expression of petal-specific target genes is that null mutants of DEF and GLO produce no petals. Two weak alleles of DEF do produce petals, however: the temperature sensitive allele def101(Zachgo et al., 1995) and deficiens-chlorantha (def-chlor), an allele with patchy DEF expression in petals resulting from an alteration in the promoter of the gene (Sommer et al.,1990; Schwarz-Sommer et al.,1992). Quantitative RT-PCR analysis was used to estimate the expression of AmMYBML1 in the flowers of plants homozygous for the def101 (grown at the permissive temperature of 15°C) and def-chlor alleles compared with wild type. A reduced level of AmMYBML1 expression was found in both mutants, indicating that DEF activity influences the expression of AmMYBML1(Fig. 7E). This result is consistent with the presence of CArG motifs within the AmMYBML1promoter, as these are the recognition motifs of the B-function MADS-domain proteins (Schwarz-Sommer et al.,1992; West et al.,1998; Parenicova et al.,2003). Examination of the ventral petals of flowers of def101 (15°C) revealed them to have a reduced mass of trichomes in the throat of the tube (Fig. 7A,B), reduced folds of tissue on the hinge and flat epidermal cells in the region of the hinge (Fig. 7C,D), features predicted as a consequence of the reduction in AmMYBML1 expression.

Protoplast transfection assays were attempted with 35S:DIV,35S:DEF and 35S:GLO, and the AmMYBML1 promoter driving GUS as a reporter. Results from these results suggested that DIV and DEF plus GLO can bind to the AmMYBML1 promoter in plant cells. However, even in combination, their activity was not adequate to drive transcriptional activation of gene expression (see Fig. S2 in the supplementary material). Other interacting proteins, such as other MADS-domain proteins of the SEPALLATA and/or SQUAMOSA/AP1 classes, are probably required to provide a complex capable of transcriptional activation. Alternatively, because the transfection assays were performed with protoplasts derived from leaves (a technical constraint) and the full-length AmMYBML1 promoter is negatively regulated in vegetative tissues, the absence of transcriptional activation in response to DIV, DEF and GLO may be a consequence of over-riding transcriptional repression of the AmMYBML1 promoter in vegetative tissues.

Previous studies have demonstrated that the MIXTA protein is sufficient for the formation of both conical cells and trichomes in tobacco, but necessary only for the formation of conical petal cells in its species of origin, Antirrhinum majus. These observations led us to suspect that other proteins, similar in structure to MIXTA, might promote trichome formation in Antirrhinum flowers.

The AmMYBML1 gene encodes a protein with very strong structural similarity to MIXTA (Fig. 1B),such that both proteins belong to R2R3 MYB subgroup 9(Stracke et al., 2001). On the basis of this structural similarity, the AmMYBML1 protein was predicted to play a role in conical cell and trichome formation. This was confirmed by ectopic expression of AmMYBML1 in tobacco. AmMYBML1 induced the formation of both ectopic conical cells and trichomes on floral organs of tobacco. No phenotypic consequences of AmMYBML1 expression were observed in vegetative tissues, in contrast to the conical cells and trichomes induced on leaves in response to ectopic MIXTA expression(Glover et al., 1998). Therefore, AmMYBML1 can function only in floral tissues. This restriction of AmMYBML1 function is most probably attributable to the interaction of AmMYBML1 with other transcription factors with activity localised to floral tissues.

Petal and carpel tissue were most responsive to AmMYBML1expression. Tobacco petals are already covered with conical cells, but in the transgenic lines, ectopic trichomes were formed among the conical cells(Fig. 3D). The wild-type carpel epidermal surface was smooth, but in the transgenic lines ectopic conical cells and trichomes were produced (Fig. 3E-G). The excess trichomes on both organs were multicellular and competent to form glandular heads. These data confirmed that the AmMYBML1 protein, as predicted from its structural similarity to MIXTA, is able to induce the formation of both conical cells and trichomes but, unlike MIXTA,only in floral organs.

The endogenous role of a gene is dependent not only on the function of its encoded protein but also on its expression pattern. To investigate the role of AmMYBML1 in Antirrhinum, its expression pattern was determined using northern analysis and in situ hybridisation. Northern analysis indicated that expression of AmMYBML1 is restricted to floral tissues (Fig. 1C). From this evidence it can also be concluded that any role of AmMYBML1 in the differentiation of cells is restricted to the flower.

In situ hybridisation demonstrated that AmMYBML1 expression was restricted to the second whorl of the flower and, within the corolla, to the ventral of the five petals (Fig. 2A,B). Confirming this, in floral symmetry mutants where all five petals adopt ventral identity (dich, cyc) all five petals expressed AmMYBML1 (Fig. 2D). Several specialised cell types form on the ventral petal, including trichomes in the throat of the corolla tube, conical cells on the adaxial epidermis of the petal lobes and hinge region, and expanded adaxial mesophyll cells in the hinge region.

Analysis of the timing of AmMYBML1 expression relative to a predictive marker of cell division, Cyclin D3b(Forbert et al., 1996),indicated that AmMYBML1 is expressed in the ventral petal while cells are still dividing (Fig. 2A-C). Previous studies have shown that that the ability of MIXTA to direct the formation of conical cells or trichomes is linked to the competence of cells for further division when the gene is expressed(Glover et al., 1998). The observation that AmMYBML1 is expressed while cells are still competent for further division indicates that it may play a role in the formation of either trichomes or conical cells, or both.

In situ hybridisation studies identified AmMYBML1 expression in the trichomes of the ventral petal, especially those that form in a mass in the throat of the corolla tube and that collect pollen from pollinating bees(Fig. 2A, Fig. 6A,C). Expression in trichomes was confirmed by promoter::GUS analysis, which identified GUS staining in all floral trichomes (Fig. 5B). Deletions of the promoter expanded GUS expression to the trichomes of vegetative organs, suggesting that flower-specific expression is the result of negative regulation of AmMYBML1 expression(Fig. 5E). The activity of the AmMYBML1 promoter in trichomes was consistent with the observed ability of AmMYBML1 to induce trichome formation on the petals and carpels of tobacco, when ectopically expressed(Fig. 3D,G). Taken together with the early expression of AmMYBML1 relative to the progression of cell division (Fig. 2B,C) and the expression of AmMYBML1 in the trichomes of the ventral petal(Fig. 2E), these data suggest strongly that AmMYBML1 directs the formation of specialised trichomes in the ventral petal.

In situ hybridisation experiments also identified AmMYBML1transcript in cells of the inner epidermis of the hinge of the ventral petal(Fig. 2E,F). This epidermis develops conical cells, similar to those of the main expanse of the petal lobes. Observation of transcript in this region, in combination with the ability of AmMYBML1 to direct conical cell formation when ectopically expressed, led us to investigate further the role of this gene in the differentiation of these specialised cells. Previous studies had indicated the importance of MIXTA in the differentiation of petal conical cells in Antirrhinum and suggested that in mixta mutants, no conical cells are produced on the adaxial epidermis. Re-investigation of mixta mutant plants, which have flattened epidermal cells on the main expanse of their petal lobes, did reveal a role for AmMYBML1. The epidermis of the ventral petal in the region of the hinge still produced some conical cells in the mixta mutants(Fig. 4A,B), indicating that a gene other than MIXTA is sufficient for the development of conical cells in this restricted region of the corolla. In combination with the ability of AmMYBML1 to direct the formation of conical cells, the localisation of its transcript to this region and the fact that promoter analysis confirmed the expression of AmMYBML1 in conical cells of tobacco petals (Fig. 5C)strongly support the idea that AmMYBML1 is responsible for the differentiation of conical cells on the epidermis of the ventral petal hinge.

In situ hybridisation also identified AmMYBML1 expression in the adaxial mesophyll cells of the ventral petal hinge(Fig. 2E,F and Fig. 4C). Expression in mesophyll cells was not observed for MIXTA(Fig. 2G) and suggests a novel role for AmMYBML1. Scanning electron microscopy revealed that the mesophyll cells in this hinge region were expanded and tightly packed together, creating folds of tissue (Fig. 4D). Differential expansion of the mesophyll on the adaxial side may contribute to the overall form of the ventral petal which facilitates pollination. The folded region reinforces the corolla, which is important because the ventral petal functions as a landing platform for large pollinating bees. When AmMYBML1 was expressed in tobacco under the control of the CaMV35S promoter, petal lobes were considerably thicker than controls, and thickening was correlated with increased expansion of mesophyll cells (Fig. 4G,H). Therefore the combination of expression of AmMYBML1 in the hinge mesophyll and ability of AmMYBML1 to enhance mesophyll cell expansion in transgenic tobacco flowers leads us to conclude that AmMYBML1 is involved in expansion of the adaxial mesophyll cells of the ventral petal, helping to create the folds of the reinforced hinged landing platform. Interestingly, MIXTA was not as effective as AmMYBML1 at promoting mesophyll thickening in transgenic tobacco, suggesting a further functional distinction between the two proteins.

We have been unable to identify a mutant of AmMYBML1, despite extensive reverse genetic screens. However, examination of the divmutant of A. majus supported our interpretation of the function of AmMYBML1. DIV promotes AmMYBML1 expression significantly,although it does not affect expression of MIXTA. div mutants lack trichomes in the throat of their corolla tubes. This supports the idea that AmMYBML1 promotes trichome formation in the ventral petal tube directly. div mutants also lack conical cells in the region of the ventral petal epidermis around the hinge (Fig. 6H). As MIXTA expression is not affected by DIV, although DIV is expressed throughout the corolla in the early stages of development(Galego and Almeida, 2002),this observation supports a role for AmMYBML1 in promoting conical cell formation in this restricted region. Finally, div mutants lack the folds of tissue that normally form at the hinge, supporting the view that AmMYBML1 expression in the adaxial mesophyll in this region promotes the formation of the folds of petal tissue that reinforce the hinge and provide a robust landing platform for pollinating bees.

DIV determines ventral petal identity in A. majus(Almeida et al., 1997; Galego and Almeida, 2002) and promotes AmMYBML1 expression in the ventral petal. Our data suggest that DIV is an activator of AmMYBML1. Certainly, a significant number of the specific events in ventral petal development appear to be realised through DIV activation of AmMYBML1 expression.

The B-function floral homeotic genes, DEF and GLO, also contribute to the control of AmMYBML1 expression, as evidenced by the reduction in transcript levels of the gene in flowers from plants homozygous for weak def alleles (Fig. 7E), and by the reduction of those cellular features promoted by AmMYBML1 in the ventral petal of flowers of def101(Fig. 7A-D). Traditionally, DEF and GLO are defined as functioning in the determination of organ identity, and consequently it might be that their regulation of AmMYBML1 expression is indirect. However the inhibition of AmMYBML1 promoter activity by the combination of DEF and GLO in protoplast transfection assays (see Fig. S2 in the supplementary material) suggests that they may interact directly with the promoter. The absence of transcriptional activation of the AmMYBML1 promoter by DIV/DEF/GLO in any combination suggests that there are additional components in the transcriptional activation complex. Potential candidates are the SEPALLATA-like genes and SQUAMOSA encoding MADS box transcription factors known to form ternary complexes with DEF and GLO, and to be essential for their transcriptional regulatory activity(Egea-Cortines et al., 1999; Gutierrez-Cortines and Davies,2000).

Phylogenetic analysis suggests AmMYBML1 to be the product of a relatively recent duplication of an ancestral subgroup 9 gene. The specialised function of AmMYBML1 may have evolved at the same time as the zygomorphic floral structure of Antirrhinum and other Scrophulariacae. Its activity contributes to the adaptations of the Antirrhinum flower to specialised pollination by bees. Our data suggest that the divergence of developmental function between MIXTAand AmMYBML1 is largely the result of differences in the expression patterns of the two genes although there are also subtle differences in the biochemical functions of the proteins they encode. Gene duplication and specialisation of the AmMYBML1 expression pattern may have been the primary changes underpinning the evolution of the specialised cell types to provide some of the features of Antirrhinum flowers that adapt them to bee pollination. The AmMYBML1 protein directs the formation of three different specialised cell types in the ventral petal of Antirrhinum. The conical epidermal cells on the hinge epidermis may play a role in directing pollinators towards their nectar reward. The hinge structure created by the thickening and folding of the mesophyll acts as both a landing platform for bees and a deterrent to potential nectar robbers, as it requires an animal with the weight of a bumblebee to open the corolla at the hinge. The trichomes in the throat of the corolla tube, on the ventral petal, collect pollen from the surface of pollinating bees and redistribute it to the stigmatic surface for fertilisation. The coordination of development of these three different specialised cell forms results in the morphogenesis of a structurally complex petal, which is the function of this one transcription factor, AmMYBML1. The divergence of AmMYBML1 function from that of MIXTA and other R2R3 MYB genes of subgroup 9 in A. majus illustrates one of the ways specialised floral forms have been achieved in plants.

We thank Steve Mackay and Matthew Dorling for plant care, Ji Yang for help with sequence phylogeny, Hans Sommer for the Antirrhinum genomic library in lambda EMBL4, and Jorge Almeida for constructive discussions on the contents of the manuscript. This work was funded by a BBSRC grant to C.M. and a BBSRC Studentship to F.J.; C.M. is also supported by a BBSRC core strategic grant to J.I.C.