Novel receptor-like kinase ALE2 controls shoot development by specifying epidermis in Arabidopsis

In multicellular organisms, including in animals and higher plants, the epidermis differentiates during embryogenesis and supports the further development and survival of adult organisms. In animals, cells ingress from the surface into the interior of the embryo during gastrulation to generate outer (ectoderm) and inner (mesoderm and endoderm) fields, which serve as germ layers. Subsequently, an alternative cell-fate specification, either into neural or epidermal tissue, takes place in the ectoderm. Specification of the epidermis requires the activities of: the bone morphogenetic protein (BMP)pathway, involving extracellular regulators such as subtilisin-like proteases;BMPs cleaved by the proteases; and receptor kinases(Cui et al., 1998; Stern, 2005).

In higher plants, specification of outer (protoderm) and inner (provascular and ground) tissues takes place during early embryogenesis in a way that is apparently different from the mechanisms involved in animal systems; the tissue specification in higher plants does not involve cell ingression(Esau, 1977; Jürgens and Mayer, 1994; Ito et al., 2002). Instead, in both monocotyledonous and dicotyledonous plants, cells at the outermost part of the early embryo, via currently unknown mechanism(s), acquire distinct characteristics, including the formation of epidermal cuticle and of a regulated cell-division plane perpendicular to the surface of the embryo(anticlinal division), which enables self-maintenance of the protodermal cell-layer that generates cuticle. During post-embryonic development,epidermal cells of most aerial organs are derived from the protoderm of the shoot apical meristem (SAM; L1 layer)(Satina et al., 1940).

In Arabidopsis thaliana, the first indication of the morphological differentiation of the protoderm is visible in the embryo at the 16-cell stage, when outer and inner cells are generated(Jürgens and Mayer,1994). Outermost cells of early embryos express a specific set of genes, which include those for homeodomain transcription factors(ATML1 and PDF2) (Lu et al., 1996; Abe et al.,2003), for fatty-acid metabolism (FDH)(Yephremov et al., 1999), for a putative extracellular protein (PDF1)(Abe et al., 1999) and for a receptor-like protein kinase (RLK; ACR4)(Tanaka et al., 2002; Gifford et al., 2003). The homeodomain proteins ATML1 and PDF2 are essential for the differentiation of the epidermis, the expression of PDF1 and ACR4 in seedlings(Abe et al., 2003), and for maintaining ATML1 promoter activity(Takada and Jürgens,2007). ATML1 and PDF2 proteins bind to a cis-regulatory element, the L1 box, which is found in protoderm-specific genes such as PDF1 and FDH, as well as in the genes for ATML1 and PDF2 themselves (Abe et al., 2001; Abe et al., 2003). Thus, it seems possible that the expression of protoderm-specific genes involves positive-feedback regulation by ATML1 and PDF2. Our understanding of the regulatory mechanisms that control the activity and/or expression of ATML1 and PDF2 is, however, very limited.

We have shown that the abnormal leaf shape 1 (ale1)mutation in Arabidopsis results in impaired cuticle formation, in the adhesion of endosperm and embryo, and in the fusion of cotyledons and leaves(Tanaka et al., 2001). The ALE1 gene encodes a member of the subtilisin-like serine protease family and is preferentially expressed within the endosperm(Tanaka et al., 2001). These observations imply that a signal from the endosperm is required for the formation of cuticle around the embryo. Mutations in the ACR4 gene,which encodes a RLK, affect the organization of cell layers in ovules and at sepal margins, as well as affecting cuticle formation in leaves and ovules(Gifford et al., 2003; Watanabe et al., 2004). Thus,signaling systems that include the subtilisin-like serine proteases and RLKs have been shown to be involved in epidermal differentiation in both animal and plant embryos, yet our understanding of this signaling pathway in plants is still very limited.

We isolated a novel Arabidopsis mutant, designated abnormal leaf shape 2 (ale2), which has epidermal defects similar to those of ale1 and/or acr4 mutant plants. We describe here the molecular cloning of the ALE2 gene and show that it encodes a novel RLK. Our genetic analysis revealed that ALE2, ALE1 and ACR4 play collectively essential roles in protoderm specification,and in the formation of the primordia of cotyledons, during embryogenesis. Furthermore, our results suggest that ALE2 and ACR4 act in close harmony whereas ALE1 functions in a different manner to ensure robust facilitation of the differentiation of plant epidermis.

Arabidopsis mutant lines ale1-1, ale1-2, acr4-1 and acr4-5 have been described previously(Tanaka et al., 2001; Watanabe et al., 2004). The ale2-1 mutant was isolated from a T-DNA-tagged population with a Columbia-6 background(http://www.arabidopsis.org/abrc/jack.jsp). Lines were backcrossed at least three times with a wild-type strain [Columbia(Col-0), Landsberg erecta (Ler) or Wassilewskija (Ws)], as indicated in the text and Figures. The ale2-2 mutant was isolated by reverse-genetic screening of a T-DNA-tagged population, generated at the University of Wisconsin (Krysan et al.,1996), with following gene specific primers: UWBC-ALE2#1,5′-GAGGAAGCTATATGGTTACACCTTTGGTT-3′; and UWBC-ALE2#4,5′-ACCGGAAATCTTTACCCAGATGAACCGAA-3′. Plants were grown as described elsewhere (Tanaka et al.,2004b).

To assess the permeability to water-soluble molecules of the leaf surface,we immersed plantlets in an aqueous solution of toluidine blue (TB test), as described by Tanaka et al. (Tanaka et al.,2004b). For quantification of the amount of TB bound to plant tissues, the aerial parts of each plant were washed and ground thoroughly in a microtube that contained 200 μl of buffer [200 mM Tris-HCl (pH 8.0), 250 mM NaCl, 25 mM EDTA]. Next, 400 μl of ethanol was added, with vortex mixing,and plant debris was pelleted by centrifugation. The supernatant was examined spectrophotometrocally and the amount of TB was determined by the absorbance at 630 nm (A₆₃₀). The major peak of absorbance due to plant material (A₄₃₅) was used for normalization. Relative levels of TB were calculated as the ratio of A₆₃₀:A₄₃₅. When it was necessary to determine the genotype of a particular plant (e.g. to identify a segregating ale2 mutation), DNA was extracted from the pellet. Transmission electron microscopy was performed as described previously(Watanabe et al., 2004).

An F₂ mutant population was generated by genetic crossing of ale2-1 (isolated in ecotype Col) and Ler. Rough mapping of ALE2 was performed with CAPS and SSLP markers, available from TAIR(http://www.arabidopsis.org). CAPS markers between PHYB and ER were designed on the basis of information from the CEREON collection of polymorphisms(http://www.arabidopsis.org/Cereon/index.jsp). Meiotic-recombination breakpoints were generated near ALE2 by screening 142 ale2 mutants in the F₂ generation for recombinants between the ale2-1 mutation and flanking markers. The interval of recombination breakpoints closest to the ale2 mutation was narrowed down to a 55 kb region that was located on one BAC (F11A3). Overlapping DNA fragments covering this region were amplified from ale2-1 genomic DNA by PCR and DNA sequences were determined with automated sequencers (ABI3100; Applied Biosystems). Reverse transcriptase(RT)-PCR was performed as described previously(Watanabe et al., 2004). ALE2 cDNA, covering the entire coding region, was amplified with primers ALE2RT-5′ (5′-GAGGCTTGGTGTCTCCGTTATTGACTAAT-3′) and ALE2RT-3′ (5′-CCTCCTTTTCCCTTTTCTTC-3′) and cloned into pBluescript SK (-) (Stratagene), and its nucleotide sequence was determined. For the complementation experiment, a fragment of genomic DNA covering the entire coding region of ALE2 (nucleotides -2179 to 4389 relative to the initiation codon) was cloned into the binary vector pGreen0029(Hellens et al., 2000)(http://www.pgreen.ac.uk/)and introduced into ale2-1/+ plants by Agrobacterium-mediated transformation.

An ALE2 cDNA fragment encoding the protein kinase domain (residues 324-619) was PCR amplified and cloned into the GST fusion vector pGEX4T-3(Amersham Pharmacia Biotech). To generate kinase-inactive ALE2, a cDNA fragment corresponding to residues 289-619 was PCR amplified, a point mutation at the 377th codon (AAA to TGG) was introduced with mutagenic primers, and this fragment was then cloned into the pGEX4T-3 vector. An ACR4 cDNA fragment encoding the protein kinase domain(Watanabe et al., 2004) was cloned into the NusA fusion vector pET50 (Novagen). Recombinant proteins (1μg each) and myelin basic protein (MBP; 10 μg) was incubated in 20 μl of kinase buffer (Sasabe et al.,2006) containing 50 μM ATP and 10 μCi of[γ-³²P] ATP. The reaction products were then separated by SDS-PAGE on polyacrylamide gels.

For observations of cell morphology, we prepared semi-thin sections (500 nm thickness) as described previously(Watanabe et al., 2004). Morphology of ovules and embryos were examined as described by Tanaka et al.(Tanaka et al., 2004a).

In situ hybridization was performed essentially as described previously(Tanaka et al., 2001). Template DNA for transcription in vitro was amplified by RT-PCR and cloned into the EcoRV site of pBluescript SK (-). For ATML1, we used primers ATML1-5′D (5′-AGCCTAAGACCAAGTCCGAT-3′) and ATML1-3′U (5′-CCAGTAGTAGTAACCACTTCAAGA-3′) for RT-PCR, and we cloned the product of PCR into the vector to generate pNU562. For FDH, we used primers FDH-5′D(5′-CGGTGAACTTGAAGTACGTGA-3′) and FDH-3′U(5′-ACACGTGTCTTCTCGAAGAGTT-3′), generating plasmid pNU823. For PDF1, we used primers PDF1-5′D(5′-CGTAAGGTTTGAGGATGCCA-3′) and PDF1-3′U(5′-TCCAAGCAAGCCCCATATCA-3′) to generate pNU563. We generated an antisense probe for ATML1 by linearizing pNU562 with HindIII, with subsequent synthesis of RNA by T3 RNA polymerase. We generated an antisense probe for FDH by linearizing pNU823(FDH) with ClaI, with subsequent transcription by T3 RNA polymerase. For the PDF1 probe, we digested pNU563 with SmaI and transcribed the probe with T7 RNA polymerase. For fusion of the PDF1 promoter (pPDF1) with a gene for green fluorescent protein (GFP), a 1595 bp fragment (nucleotides -1602 to -8 relative to the initiation codon) containing the cis-regulatory sequence of the PDF1 gene (Abe et al.,1999) was amplified by PCR and cloned upstream of sGFP(S65T) (Chiu et al.,1996) in pGreen. Two independent transgenic lines (ecotype Ler) were crossed with ale1-1; ale2-1/+ plants and the fluorescence due to GFP was examined under a fluorescence stereomicroscope.

The ale2 mutation was identified by genetically screening for mutations associated with irregular leaf morphology. Homozygous ale2plants produced irregularly malformed leaves with eventual leaf fusion, as shown in Fig. 1C. In addition,the ale2 mutant leaves exhibited increased permeability to toluidine blue (TB) compared with wild-type leaves, suggesting that defects in cuticle on the plant surface were present (Fig. 1D). These epidermal and morphological defects were associated with a recessive mutation.

At the reproductive stage, ale2 plants generated malformed and fused floral organs (Fig. 1E). In addition, ale2 plants were sterile. Reciprocal crosses between ale2 and wild-type plants indicated that sterility was mainly due to defects in female reproductive tissue(s), although male fertility was also reduced. Therefore, we maintained ale2 mutants as the self-progeny of heterozygous plants or via genetic crosses with pollen from homozygous parents. In order to identify the primary defect responsible for the female sterility of ale2 plants, we examined the morphology of female gametophytes (ovules). Ovules from wild-type parents were oval with a smooth surface (Fig. 1F), reflecting the organized development of flattened layers of epidermal cells. The inner cells - including the adaxial outer integumental cells, inner integumental cells and endothelial cells - are derived from the protoderm (L1 layer)(Jenik and Irish, 2000) and form organized layers (Fig. 1H)that cover the female gametes (embryo sacs; Fig. 1H, asterisk). Scanning electron microscopy revealed that adjacent ovules generated from ale2parents had fused to one another. Moreover, their surfaces were rough and disorganized as a consequence of the swollen outer cells(Fig. 1G). Histological analysis confirmed that organization in the integument and endothelium had been disturbed by the ale2 mutation, which had induced irregularly orientated cross walls and the deformation of cells(Fig. 1I). In addition, mature ovules from ale2 plants contained degenerated embryo sacs(Fig. 1I), suggesting that the sterility of ale2 plants was due, at least in part, to the failure of ovules to develop. These observations suggested that the ALE2 gene is involved in the regulation of cell morphology, of the plane of cell division and of cuticle formation in epidermis-related tissues.

Using a map-based approach, we narrowed down the region that contained the ale2 mutation to a 55 kb region, as shown in Fig. 2A. We determined the nucleotide sequence of the entire 55 kb region from the ale2-1 mutant and compared this sequence with that of the wild type (Col-0). The only difference was the insertion of a single nucleotide, and this insertion was located within a putative gene (At2g20300). Subsequent amplification by reverse transcriptase (RT)-PCR, determination of the cDNA sequence, and comparison of the cDNA and the corresponding genomic DNA sequences indicated that the At2g20300 gene consists of eight exons, with an open reading frame of 2232 base pairs, which is capable of encoding a protein of 744 amino acids(Fig. 2B,C). To confirm that the single insertion was responsible for the ale2 phenotype, we introduced a fragment of genomic DNA that covered the At2g20300 gene into ale2-1 plants. Plants with the introduced transgene in addition to the homozygous ale2-1 mutation had a normal phenotype(Fig. 1E). In addition, an independently isolated ale2-2 mutant that harbored T-DNA within the fifth exon of the At2g20300 gene (Fig. 2B,C) had morphological defects essentially identical to those of ale2-1 plants (data not shown). We concluded that the At2g20300 gene was equivalent to the ALE2 gene.

The predicted sequence of the ALE2 protein contains a hydrophobic sequence at the N-terminus and a hydrophobic sequence in the middle region(Fig. 2C). ALE2 also includes an amino acid sequence in the C-terminal region that is strongly conserved in protein serine/threonine kinases (Fig. 2C). These structural features are typical of those of RLKs, and it is likely that the N-terminal region between the two hydrophobic regions serves as the extracellular region. When we used the predicted amino acid sequence of the putative extracellular domain to search available databases,several protein sequences predicted either from cDNA and/or genomic DNA sequences were found to contain sequences significantly homologous to that of ALE2. These sequences were rice XP473126 and XP468388, and Arabidopsis At5g56885. All these sequences had a cluster of basic amino acid residues (5-12 residues in each member), followed by sequences containing four cysteine residues (Fig. 2D), which are somewhat similar to that of the processing sites in animal insulin-related peptides. Sequences of these four proteins exhibited significant similarity between the basic region and putative transmembrane domain (Fig. 2D).

To examine whether ALE2 is a functional protein kinase, a recombinant protein in which glutathione S-transferase (GST) was fused to the putative cytoplasmic region that covered the kinase domain (KD) of the wild-type ALE2 protein (GST:ALE2KD-WT) was synthesized in E. colicells. As a negative control, a mutant fusion protein in which the lysine residue at position 377 in the potential ATP-binding domain was replaced with a tryptophan residue (GST:ALE2KD-KW) was also synthesized. In vitro phosphorylation experiments revealed that the wild-type fusion protein,GST:ALE2KD-WT, had activity to phosphorylate itself and myelin basic protein(MBP), whereas GST:ALE2KD-KW did not (Fig. 3A).

To gain further insight into the molecular nature of ALE2, we examined whether ALE2 and ACR4, another RLK that is involved in epidermal differentiation (Gifford et al.,2003; Watanabe et al.,2004), could phosphorylate one another. A recombinant protein that contained the putative cytoplasmic region of ACR4 with the protein-kinase domain was synthesized in E. coli cells as a histidine- and NusA-tagged fusion (His-Nus:ACR4KD-WT), and was purified. A mutant fusion protein in which the lysine residue in the putative ATP-binding site of ACR4 was replaced with a tryptophan residue (His-Nus:ACR4KD-KW) was also synthesized (Fig. 3B). As shown in Fig. 3C, His-Nus:ACR4KD-WT and GST:ALE2KD-WT proteins phosphorylated the kinase-inactive recombinant proteins GST:ALE2KD-KW and His-Nus:ACR4KD-KW, respectively. When ALE2KD-WT and ACR4KD-WT were used for the reaction, levels of phosphorylation of both proteins significantly increased compared with reactions with either type of kinase-negative proteins (Fig. 3C,D). Thus, ALE2 and ACR4 appear to have the potential to phosphorylate one another.

The potential biochemical interaction, as well as the similarity between the phenotypes of ale2 and acr4 mutants, prompted us to perform phenotypic analysis of an ale2 acr4 double mutant. We found that ale2 acr4 plants were viable and able to produce sterile flowers, as was the ale2 single mutant(Fig. 4A,B). With respect to the number of seeds per silique and the morphology of ovules, the effects of the acr4 mutation were weaker than those of ale2, whereas plants that were double homozygous for the ale2 and acr4mutations had a phenotype similar to that of the ale2 single mutant(Fig. 4A-C). We next examined the phenotypes of plants that were homozygous for one mutation and heterozygous for the other (i.e. ale2/+; acr4 and ale2; acr4/+). On the homozygous ale2 mutant background, the effect of a heterozygous acr4 mutation was unclear(Fig. 4A,B). By contrast, on the acr4 homozygous mutant background, the effects of the ale2 mutation were semi-dominant in terms of reduced fertility(Fig. 4B, lanes 7-9).

We generated plants that were homozygous for the ale1 and heterozygous for the ale2 mutation to examine the genetic interaction between ale1 and ale2. Dried seeds obtained from the parental plants included deformed (Fig. 5Aa) and shrunken (Fig. 5Ab) seeds [95 (17.8%) and 48 (9.0%) out of 535 seeds,respectively], which accounted for approximately 25% of all seeds and these phenotypes were rarely observed in the wild type or in single mutants. Shrunken seeds did not germinate, even on solid medium; however, seeds with aberrant morphology germinated, generating seedlings with various degrees of abnormality (Fig. 5Ac-Ae); only a single cotyledon-like structure was present in the most-severely affected cases (Fig. 5Ae). In severely affected seedlings, leaves were formed only after long-term culture or as deformed structures (data not shown). As expected, these seedlings were homozygous for both the ale1 and ale2 mutations (data not shown).

Next, we examined the permeability of the leaf surface to TB in single- and double-mutant plants. Wild-type leaves did not allow the dye to permeate their surface (Fig. 5B). Leaves from single mutants were typically stained in a patchy pattern, which suggested the partial loss of cuticle on the surface, as well as the presence of a back-up pathway that was still active and continued to promote cuticle formation(Fig. 5B). Remarkably, the surface of ale1 ale2 double mutants incorporated significantly more TB (Fig. 5B,C). By contrast,the ale2 acr4 double-mutant seedlings exhibited a range of epidermal surface defects that were only slightly stronger than, but mainly overlapped those of, the ale2 single mutant(Fig. 5C). Ultrastructural analysis revealed an enhanced defect in cuticle deposition in ale1 ale2 double mutants (Fig. 5D). Loss of electrondense cuticle in ale1 ale2 double mutants was often associated with the appearance of electron-dense granules in the cytoplasm (Fig. 5D,asterisks).

Because ale1 ale2 double mutants had impaired epidermal function and impaired organ formation, we postulated that the double mutants might be defective in the differentiation of epidermal precursor cells, which may be important for the formation of leafy organs. To test this hypothesis, we examined the morphology of cells in sections of shoot apices. In sections of wild-type SAM, the L1 layer consists of cells that are relatively regular and rectangular in profile, resulting in the generation of the smooth surface of the SAM (Fig. 5E). The L1 layer in ale1 or ale2 single-mutant plants was mainly a single layer of cells (Fig. 5E),whereas, in the shoot apex of ale1 ale2 double mutants, the outermost cells did not form an organized layer and each cell was swollen, with a rounded appearance (Fig. 5E).

We followed the course of embryogenesis in the wild type and the mutants,focusing on the morphology of protodermal cells and the development of cotyledons. As we reported previously, the ale1 single mutant was indistinguishable from wild type in terms of the morphology of the embryos before the heart stage (Tanaka et al.,2001) (Fig. 6A,B). In the case of the ale2 single mutant, we examined embryos obtained from heterozygous parents, because homozygous parents were sterile. We observed a change in the morphology of protodermal cells in approximately 2%of embryos at the globular stage; in these embryos, the protodermal cells were slightly swollen (Fig. 6C,D,E). At the triangular and the heart stages, we observed embryos with a rough surface at frequencies of approximately 13 and 26%, respectively(Fig. 6C,E). In spite of the malformed protodermal cells, ale2 embryos formed normal cotyledons(Fig. 6C). Morphological defects in embryos were obvious from the globular stage in the case of embryos obtained from ale1; ale2/+ parents(Fig. 6D). At the triangular stage and the heart stage, significant fractions of embryos (approximately 13 and 16%, respectively) had very deformed protodermal cells, whose heights were obviously greater than that of anticlinal cell walls of the protoderm(Fig. 6D,E). As the normal-looking siblings developed to the heart and torpedo stages, the deformation of cotyledons became more obvious, with the occasional loss of bulging of either one or both primordia(Fig. 6D). The frequencies of severe defects in the morphology of protodermal cells were below 25% at all developmental stages (Fig. 6E),suggesting that these phenotypes were present in double-homozygous embryos. However, when mildly affected embryos were taken into account, the percentage of embryos with swollen protodermal cells was more than 25%(Fig. 6E; especially obvious at the triangular stage), perhaps because of a dosage effect of the ALE2gene in the absence of activity of the ALE1 gene.

The severe functional and morphological defects in the outermost cells in ale1 ale2 double mutants prompted us to examine whether the mutant embryos had properly specified protoderm. We performed in situ hybridization experiments using probes for various genes that are preferentially expressed in the protoderm. Using probes for transcripts of the PDF1, FDH and ATML1 genes, we detected clear signals in the protodermal cells of developing wild-type embryos from the 16-cell through to the torpedo stages(Fig. 7A,E,G,I, and data not shown).

In each single mutant, we detected PDF1 transcripts in all protodermal cells of heart-stage embryos(Fig. 7B,C). However, in cells in the outermost layer of ale1 ale2 embryos, the signal was often below the limit of detection, although strong signals were detected in some outermost cells (Fig. 7D). We obtained similar results after hybridization with FDH and ATML1 probes (Fig. 7E-H). Moreover, when ale1 ale2 double mutants failed to form one or both cotyledonary primordia, no expression of FDH and ATML1 genes was detectable in the outermost cells in the apical region from which cotyledonary primordia would otherwise have developed(Fig. 7F,H). These results suggest that ALE2 is required for the specification of protodermal cell fate, for the transcription of protoderm-specific genes and/or for the accumulation of the transcripts of protoderm-specific genes in the absence of ALE1 activity.

We previously found that acr4 ale1 double mutants had very severe defects in epidermal surface function and in seedling morphology as compared with each single mutant (Watanabe et al.,2004). To determine whether ACR4, like ALE2, is responsible for the establishment and/or maintenance of the identity of protoderm in the absence of ALE1 activity, we examined the expression of protoderm-specific genes in ale1 acr4 embryos. We failed to detect any hybridization signals due to the ATML1 and FDH probes in the outermost cell layer of the apical region, which included possible sites for the formation of cotyledonary primordia in ale1 acr4 embryos(Fig. 7I-L). Such abnormal patterns of transcription of the ATML1 and FDH genes in ale1 acr4 double mutants were similar to those, described above, in ale1 ale2 embryos.

To gain some insight into the possible mode of regulation of protoderm-specific gene expression by ALE1 and ALE2, we used a GFP reporter gene, driven by the PDF1 promoter(pPDF1::GFP), and examined its expression in wild-type and ale1 ale2 double-mutant seedlings. In wild-type seedlings, strong and uniform signals were observed on the surfaces of young leaves(Fig. 7M,N). In ale1 ale2 seedlings, which had deformed leaves, fluorescence was observed in a patchy pattern on the surfaces of the leaves(Fig. 7O,P).

We have described here the isolation and initial characterization of the ale2 mutant, a novel mutant with defects in morphology and in the functions of epidermal cells. We have shown that the ALE2 gene encodes a previously uncharacterized RLK, and our results indicate that ALE2 plays an essential role in regulating the differentiation of the protoderm and epidermis in both embryogenesis and postembryonic development.

The At2g20300 gene, which turned out to be equivalent to the ALE2gene in this study, was originally defined as the gene for a receptor-like cytoplasmic kinase (RLCK), which lacks predicted extracellular domain. Our experiments by RT-PCR and a cDNA sequence from another source (GenBank AY091071) indicated that the ALE2 gene encodes a protein with a structure typical of RLKs, resembling the majority of members of the Arabidopsis RLK gene family (Shiu and Bleecker, 2003). Thus, ALE2 might function as cell-surface receptor. The putative extracellular-domain sequence of the predicted ALE2 protein is similar to that of members of the extensin-like RLK subfamily(Shiu and Bleecker, 2003) (36%and 18% identity to that of the At5g56885 and At4g02010 gene products,respectively). The biological functions of other members in this family are unknown and, thus, ALE2 is unique, at present, insofar as it has a known biological function. Our results suggest that ALE2 might function in the perception and transmission of a signal at the cell surface that is required for the appropriate differentiation and morphogenesis of epidermal cells.

We reported previously that the ale1 and acr4 mutations have a synergistic effect on epidermal-surface function, as well as in the formation of cotyledons (Watanabe et al.,2004). These genes encode a putative subtilisin-like serine protease and a RLK, respectively, which both might potentially be involved in intercellular signaling (Tanaka et al.,2001; Tanaka et al.,2002). Because individual mutations seemed to abolish the functions of the encoded proteins, it seems likely that ALE1 and ACR4 can promote the formation of epidermis independently. Our present experiments revealed an effect of the ale2 mutation that was qualitatively reminiscent of that of acr4, although the phenotype of acr4was weaker than that of ale2(Gifford et al., 2003; Watanabe et al., 2004) (Figs 1, 4). There seemed to be a strict requirement for ALE2 activity in the absence of ACR4, because ale2was associated with a semi-dominant phenotype on the acr4 mutant background (Fig. 4B). In addition, epidermis-related defects in the ale2 acr4 double mutants were reminiscent of those in the ale2 single mutant (Figs 4, 5). Taken together, our results suggest the presence of a single or closely overlapping pathway mediated by ACR4 and ALE2 (Fig. 8). Because ALE2 and ACR4 were able to phosphorylate each other in vitro, a possible model for the mode of their action is that they function as an enzyme-substrate pair in plants, thereby functioning in a single signaling pathway that may promote epidermal specification. In this scenario, their expression patterns should overlap with each other. ACR4 is expressed in epidermis-related tissues of various organs (Tanaka et al.,2002; Gifford et al.,2003). ALE2 transcripts appear to be evenly distributed throughout the embryo proper during early embryogenesis, and are preferentially detected both in the outermost cells and in the inner cells of cotyledonary primordia at the early heart stage (see Fig. S1 in the supplementary material). Thus, the pattern of ALE2 expression is overlapped to some extent with that of ACR4. The molecular interaction of these RLKs must, however, be examined by further experiments. Meanwhile, our results support a putative scenario wherein ALE1 functions in a manner that is somehow related to, but is distinct from, the mode of action of ALE2 and ACR4 (Fig. 8). A reasonable hypothesis would be that protoderm specification involves an intertissue communication (i.e. a signal from endosperm that might be generated by ALE1) and an extracellular signaling molecule regulating ALE2 and/or ACR4 activity. Molecular identification and characterization of such molecules would provide further insight into intercellular communication regulating protoderm specification in plants.

A striking feature of the accumulation of transcripts of protoderm-specific genes in ale1 ale2 and ale1 acr4 double mutants was the patchy pattern of such accumulation, with signals as strong as those in wild-type cells in some cells, and below the level of detection in other cells, at the outermost margins of embryos(Fig. 7). A similar patchiness of expression of the pPDF1::GFP reporter gene was also observed in leaves (Fig. 7P). These observations suggest that ALE1 and ALE2 might not be absolutely required for the transcription of protoderm-specific genes. Rather, they might be involved in the organized expression of these genes during development. Abe et al.(Abe et al., 2003) proposed that the expression of ATML1 and PDF2 might involve self-activating feedback regulation via the L1 box in their promoter regions. Recently, it has been shown that epidermis-specific expression of ATML1 is controlled by several regulatory sequences in its promoter(Takada and Jürgens,2007). It is tempting to speculate that ALE1, ALE2 and ACR4 impinge on the activity of such regulatory elements, allowing the uniform expression of protoderm-specific genes. It is increasingly clear that many genes required for the formation of cuticle are expressed in an epidermis-specific manner (reviewed by Tanaka and Machida, 2006; Kurdyukov et al., 2006) and that cuticle is essential for the prevention of the fusion of aerial organs(Tanaka and Machida, 2006). Remarkably, it has been shown that the function of FDH, expression of which was seriously affected in ale1 ale2 and ale1 acr4double mutants, is essential for proper leaf surface function [see Lolle et al. (Lolle et al., 1998) and references therein] (Tanaka et al.,2004b). Therefore, it is likely that the severe cuticular defects and organ fusions in ale1 ale2 and ale1 acr4 double mutants are due, at least in part, to the impaired expression of protoderm-specific genes, including FDH.

In addition, ale1 ale2 and ale1 acr4 double mutants were defective in the formation of cotyledons, as is the atml1 pdf2 double mutant (Abe et al., 2003). Because ATML1 and PDF2 are specifically expressed in protodermal cells and are required for the expression of protoderm-specific genes, it is possible that appropriate gene expression in the protoderm is a prerequisite for the formation of cotyledons. Our results suggest that the failure of cotyledon formation in ale1 ale2 and in ale1 acr4double-mutant embryos was associated with a loss of detectable levels of expression of protoderm-specific genes. Another gene that plays a role in the regulation of protoderm-specific gene expression is AtDEK1 (also known as DEK1), which encodes a protein with a calpain-protease domain. The product of this gene is essential for the expression of reporter genes that are driven by the promoters of protoderm-specific genes, such as ATML1 and ACR4 (Johnson et al., 2005). In the atdek1 mutant, development is arrested at the globular stage, without the formation of cotyledons(Johnson et al., 2005). Thus,currently available circumstantial evidence supports the hypothesis that properly differentiated protoderm is somehow essential for the initiation and/or continuous growth of the cotyledons. A further genetic link between the functional differentiation of the epidermis and cotyledon formation has been provided by the analysis of the gurke (gk; also known as acc1) mutant: strong gk mutant alleles are associated with the failed formation of cotyledons(Torres-Ruiz et al., 1996; Baud et al., 2004; Kajiwara et al., 2004); and leaky gk mutations allow the formation of plantlets but result in the adhesion of aerial organs (Torres-Ruiz et al., 1996) (H. Tanaka and Y.M., unpublished).

In addition to the defect in cotyledon formation, the organization of cell layers in the SAM was severely perturbed in ale1 ale2 double mutants(Fig. 4E). Such mutants often ceased development after germination, although some eventually generated several small leaf-like organs (Fig. 5A, Fig. 7O). Such morphological and growth defects were associated with the appearance of swollen spherical cells in the outermost layer of the SAM and with a disorganized pattern of expression of the pPDF1::GFP. These observations support the hypothesis that proper differentiation of the epidermal cell layer is required for the functions of the SAM and for the subsequent development of leaves. The plant hormone auxin plays a crucial role in the generation of aerial organs (Okada et al., 1991; Reinhardt et al., 2000). The patterns of expression and the subcellular localization of proteins involved in auxin transport indicate that auxin is transported towards the incipient primordia of leafy organs through the protoderm and epidermis (Benková et al., 2003; Reinhardt et al.,2003) (reviewed by Tanaka et al., 2006). Thus, the relationships between the differentiation of the protoderm, the regulation of polar auxin transport, and organ formation might be an interesting topic for future research.

We thank Jirí Friml for enabling H. Tanaka to complete this work in his laboratory. We also thank Yasuo Niwa, the John Innes Centre and the Arabidopsis Biological Resource Center for providing materials. The work was supported in part by a Grant-in-Aid for Scientific Research on Priority Areas (no. 14036216) to Y.M., and by a grant for Core Research in Evolutional Science and Technology (CREST) to C.M. from the Japanese Ministry of Education, Science, Culture, Sports and Technology. H. Tanaka was supported by a grant from CREST and was a fellow of the Japan Society for the Promotion of Science (JSPS).