The Arabidopsis ACR4 gene plays a role in cell layer organisation during ovule integument and sepal margin development

Plant meristems are composed of organised layers (files or plates) of cells arranged parallel to the `outside' of the meristem. Each layer undergoes cell divisions in a defined plane or planes pushing cells to the periphery of the meristem where they are either incorporated into new meristems or become differentiated. In Arabidopsis the shoot apical meristem (SAM) has two outer tunica layers, the L1 and the L2, which undergo regulated divisions in the anticlinal plane. The inner cell layer or corpus is designated L3 and undergoes both anticlinal and periclinal divisions. As organ primordia arise on meristem flanks, changes in the regulation of cell division patterns occur. In dicotyledon leaf primordia, the epidermal cell layer is exclusively L1-derived and L1-derived cells continue to divide largely anticlinally until late in development. In contrast the L2 layer undergoes both anticlinal and periclinal divisions to contribute the leaf mesophyll, while the L3 contributes to both leaf mesophyll and the vasculature(Stewart and Burk, 1970). The contributions of meristematic cell layers to organ primordia vary. Whilst the Arabidopsis leaf is usually formed from L1-, L2- and L3-derived cells, petal primordia have been shown to contain cells of only L1 and L2 origin and ovule integuments are entirely L1 derived(Jenik and Irish, 2000). Integument cells undergo carefully regulated divisions, mainly in the anticlinal plane, so that the completed organ is a tubular plate of cells only 2-3 cells thick and effectively entirely epidermal(Schneitz et al., 1995; Robinson-Beers et al.,1992).

Experiments and observations in many plant species have shown that the developmental behaviour of cells in meristems and developing organs is largely dictated by their position rather than by lineage. Thus if the progeny of cells from one layer invade another layer during development, the displaced cells differentiate according to their new position(Stewart and Derman, 1975; van den Berg et al., 1995; Kidner et al., 2000). For this developmental plasticity to be achieved, cells must constantly receive and interpret information from their neighbours. Our understanding of how plant cell layers communicate is currently limited to a few specific examples. In Arabidopsis roots, an inside to outside movement of transcription factors (notably the SCARECROW (SCR) protein) is required for normal differentiation of ground cell layers(Nakajima and Benfey, 2002). In contrast, inter layer communication in shoot meristems appears to require the interaction of a diffusible ligand with a cell-autonomous receptor kinase complex (Fletcher et al.,1999). A similar interaction is invoked in the development of maize leaves and endosperm, where the receptor kinase-encoding CRINKLY4 (CR4) and the calpain-encoding DEFECTIVE KERNEL 1 (DEK1) genes are required for specification and maintenance of`outer' cell layer identity during endosperm and leaf development(Becraft et al., 1996; Becraft et al., 2002; Lid et al., 2002). The maize EXTRA CELL LAYERS 1 (XCL1)gene seems to be involved in pathways regulating division behaviour in L1 cells during organ formation. The Xcl1 mutant provides intriguing evidence that cell identity can be uncoupled from positional cues at least late in development. (Kessler et al.,2002).

In a search to identify genes involved in inter-cell layer communication in Arabidopsis, a study of ACR4, an Arabidopsis CR4homologue, was carried out. ACR4 was found to be required for normal cell organisation during ovule integument development and the formation of sepal margins. Both these tissues are formed exclusively from plates of L1 cells arranged back to back. By isolating the functional ACR4promoter, ACR4 was shown to be expressed in L1 cells in all apical meristems and young organ primordia, including those of the developing ovule integuments. In addition, ACR4 is expressed in an intriguing pattern in root meristems. The kinase activity of ACR4 was demonstrated and,using fusion proteins expressed under the ACR4 promoter, ACR4 protein localisation was visualised in vivo in the plasma membranes of L1-derived cells. The wide expression pattern of ACR4 compared to its associated mutant phenotype may be a result of functional redundancy with other related proteins or functionally related pathways. Taken together, the data presented indicate a role for ACR4 in the cellular signalling pathways required for correct cell organisation in ovule integuments and sepal boundaries, and may provide important clues as to the types of signalling involved in cell layer maintenance and specification in the wider context of plant development.

The ACR4 open reading frame (ORF) was PCR amplified from Arabidopsis thaliana genomic DNA ecotype Columbia (Col0) with CR5(5′-TGGTACCTTTGAAAAGAATGAGAATGTTCG) and 5′-GAGCTCAGAAATTATGATGCAAGAACAAGC. The ACR4 promoter was amplified with 5′-TGTCGACATAGTCAAGAAATGGCCTTTCC and 5′-TTCTAGACAAAGTCAACACACACGCTT. Products were cloned into pGEMT-easy(Promega) (pL92 and pL93 respectively). Probes (antisense and sense) for in situ hybridisation were made by linearising pL92 with NcoI or SalI, respectively, and transcribing with Sp6 or T7 RNA polymerase,respectively. In situ hybridisations were carried out using a standard protocol (Jackson, 1991). For promoter expression analysis, the GAL4::VP16-encoding sequence and terminator were isolated from an enhancer trap vector(http://www.plantsci.cam.ac.uk/Haseloff/Home.html)(Haseloff, 1999) and transferred to the binary vector pSPTV20(Becker et al., 1992). The ACR4 promoter was inserted upstream of the GAL4::VP16 coding sequence(pL143). The ACR4 promoter was placed upstream of H2B::YFP, by cloning the H2B::YFP-coding sequence from pBI121(Boisnard-Lorig et al., 2001)into the binary vector pBIBHyg (pMD4)(Becker, 1990). The ACR4 promoter was inserted upstream of H2B::YFP (pMD6). For deletion-1026 an XhoI-XbaI fragment from pL93 was cloned into pMD4(pL227). For deletions -857 and -405, L93 was fully digested with XbaI and partially digested with HindIII. Appropriate fragments were cloned into MD4 (pL226 and pL225 respectively). To place the ACR4 promoter upstream of mGFP6, an mGFP6-encoding fragment was cloned from pBSmGFP6 to pBIBHyg and the ACR4 promoter was placed upstream (pL228). The mGFP6 variant is identical to mGFP5(Haseloff, 1999) with two amino acid changes; F64-L and S65-T (J. Haseloff, personal communication). For protein localisation studies the full-length ACR4 ORF was amplified with CR5 and 5′-GAGCTCGAGAAATTATGATGCAAGAACAAG, and mGFP6 was amplified from pBSmGFP6 with 5′-CTCGAGAATGAGTAAAGGAGAAGAAC and 5′-TCTAGTGTTTGTATAGTTCATCCATG so as to remove the ER retention signal. Green fluorescent protein (GFP) was cloned downstream of ACR4 and the fusion protein-encoding fragment was then cloned into pBIBHyg. The ACR4 promoter was then added (pMD11). For complementation studies the ACR4 ORF was cloned into pBIBHyg. The ACR4 promoter was added (pMD5). Plant transformations were carried out using Agrobacterium GV3101 (Koncz and Schell, 1986) and a floral dipping technique(Clough and Bent, 1998). Fluorescence studies were carried out using an Olympus Fluoview confocal microscope.

To express recombinant GST fusion proteins in bacteria, the ACR4kinase domain was amplified using 5′-AGGATCCGTCCGGATCTTGATGAG and 5′-GAGCTCGAGTTTCCCATTAGCTGTGC, and cloned as an in-frame fusion with GST coding sequences in pGEX-3x (Amersham Pharmacia Biotech). Protein expression and purification using GST-sepharose (Amersham Pharmacia Biotech) was carried out according to the manufacturer's guidelines. Site directed mutagenesis was carried out using the QuikChange site-directed mutagenesis kit (Stratagene)with primer 5′-GGAACCACTGTTGCAGTGATGAGAGCGATAATGTC and its reverse complement. GST fusion proteins were assayed for kinase activity by incubation in 30 μl (final volume) with 20 mM Tris (pH 7.5), 100 mM NaCl, 12 mM MgCl₂ with 10 μCi of [γ-³²P]ATP for 1 hour at room temperature. Samples were boiled in loading buffer and analysed by SDS-PAGE. Coomassie Blue-stained gels were dried and exposed to film.

To isolate acr4-1 the Wisconsin collection was screened with oligos 5′-TGCCATCTCAGTACTTCATGACTCTCTCT and 5′-CTCTCTGCCTCTTTGTTACTTTCCTGCCT as described previously(Krysan et al., 1999). The mutants acr4-2, acr4-3, acr4-4 were identified on the Syngenta website (Sessions et al.,2002). To estimate insertion number, probes against the GUS marker gene or BAR selection gene were made by amplifying the GUS ORF with primers 5′-GTGGGAAAGCGCGTTACAAGAAAGC and 5′-CACCATTGGCCACCACCTGCCAGTC or the BAR ORF with 5′-CGTACCGAGCCGCAGGAAC and 5′-ATCTCGGTGACGGGCAGGAC. For histological analysis, tissue was submerged overnight in 84 mM Pipes (pH 6.8) solution containing 4% acrolein, 1.5%glutaraldehyde 1% paraformaldehyde and 0.5% Tween 20. Tissue was rinsed several times in 100 mM Pipes and dehydrated using an ethanol series. JB4 resin was infiltrated into the tissue over a period of 2 weeks before embedding. 4.5 μm sections were stained in Toluidine Blue and visualised using a Leica standard light microscope. For creation of the ATML1marker line, the ATML1 ORF was amplified by reverse transcription PCR and cloned into pGEMT-easy using oligos ATML1A and ATML1B(Abe et al., 2001). GFP was amplified using 5′-AGCTAGCATGAGTAAAGGAGAAGAAC and 5′-AGCTAGCGTGTTTGTATAGTTCATC, and cloned pGEM-9z (Promega). The ATML1 ORF was fused downstream of GFP and the fused construction was cloned downstream of the pAS99 HindIII insert [containing the full ATML1 promoter (Sessions et al.,1999)] in pBIBHyg (pL178).

Roots were incubated for 2 hours in 100 μM brefeldin A (BFA) (B7651,Sigma-Aldrich). The working BFA solution was made by diluting a 10 mM DMSO stock 1:100 in water. Control roots were incubated for the same period of time in a 1:100 dilution of DMSO in water.

Similarity searches were carried out using the maize CR4(Becraft et al., 1996) protein against the annotated Arabidopsis genome. Five genes encoding predicted products showing sequence and structural similarity to CR4 were identified. One predicted protein, encoded by ACR4(Tanaka et al., 2002), was considerably more similar to CR4 than the other sequences identified, both within the extracellular domain and the kinase domain. RNA in situ hybridisations were carried out to determine the distribution of ACR4transcripts in developing Arabidopsis tissues(Fig. 1). Embryonic ACR4 expression was first observed at the eight-cell stage,throughout the eight cells of the embryo proper(Fig. 1A-D) and then became restricted to the outer cell layer (protoderm) of the developing embryo soon after the dermatogen stage. Expression was maintained at high levels in all protoderm cells until the early torpedo stage, when it diminished in non-meristematic cells. Cells of the embryonic root and shoot meristems continued to express ACR4 at high levels until embryo maturity. No ACR4 mRNA could be detected in the developing endosperm at any stage. Post-germination, ACR4 transcripts were detected in the L1 cell layers of seedling apical meristems, inflorescence meristems(Fig. 1F), floral meristems and young leaf and floral organ primordia but decreased rapidly in older organs before cell expansion had initiated. Expression was also detected in ovule primordia, where it was initially limited to external cell layers as in other organs, and then detected in integument primordia. At maturity, expression in the ovule was most strongly maintained in the internal layer of the inner integument, the endothelium, although it was detectable throughout the integuments. In main and lateral root primordia, results were unclear although expression was observed in the outer (epidermal) cell layer of young roots in some transverse sections, and diminished as roots expanded. Transcript distribution at the root tip appeared strong in root-cap cells near the quiescent centre. In summary, ACR4 transcripts were detected in all meristematic tissues tested and were, with the exception of roots,specifically localised to outer cell layers.

Because ACR4 RNA expression levels were low, a two-component transactivation approach was used for promoter analysis. A 1.9 kb genomic fragment finishing at the presumptive ATG of the ACR4 gene was placed upstream of a sequence encoding the chimaeric GAL4::VP16 transcriptional activator (Haseloff, 1999). Homozygous single-insertion transformants were crossed to plants containing a HISTONE 2B::YFP protein fusion encoding gene under control of a 35S minimal promoter and the GAL4-UAS (Boisnard-Lorig et al., 2001). In the immediate products of these crosses,nuclear-localised YFP was detected in embryos as early as 48 hours after pollination. Embryonic pACR4-driven marker gene expression was protoderm localised, mirroring exactly ACR4 mRNA distribution(Fig. 1E). The observation that ACR4 is not expressed in the developing endosperm was confirmed. Post-germination expression patterns correlated with in situ hybridisation results in root, vegetative, inflorescence(Fig. 1G) and floral meristems as well as in leaf and floral organ primordia(Fig. 1H). In ovules all integument cells showed marker expression although expression was stronger in the ovule epidermis, the `outer' layer of the inner integument, and the endothelium (Fig. 1I). H2B::YFP placed directly under control of the 1.9 kb ACR4promoter gave expression that was identical to, but weaker than trans-activated marker expression, confirming that the transactivation system amplified promoter activity without distorting expression patterns.

In the roots of plants transactivating H2B::YFP, marker expression was observed in the quiescent centre (QC) central cells, columella initials and cells below the QC, the lateral root cap (LRC) and the initial cells destined to give rise to the root epidermal cell file and the LRC(Fig. 1J). However, expression was not observed in epidermal cells until the point where they emerged from under the LRC (Fig. 1K). This transition was sharp, with cells initiating expression as soon as they started to lose contact with the LRC. Expression in the root epidermis was maintained into the elongation zone, where it diminished. In more distal positions on the root, initiating lateral root primordia could be identified on the basis of their expression of H2B::YFP. Expression initiated in lateral root primordia at the four- to eight-cell stage, usually in a double file of cells (not shown). Expression in lateral roots resembled that observed in apical root meristems. The expression pattern of ACR4 in roots differed from that in apical regions, firstly, in that a population of meristematic L1 cells(epidermal cell file under LRC) did not express ACR4, and secondly in that populations of `internal' cells (QC, and lateral root primordium initials) expressed ACR4.

In contrast to in situ hybridisation results, H2B::YFP remained visible in developing organs until relatively late in development. To investigate this phenomenon, a sequence encoding a cytoplasmically localised version of mGFP6 was placed under the control of the 1.9 kb ACR4 promoter. Lines expressing this construction showed expression in the same meristematic zones observed for lines expressing H2B::YFP, although fluorescent protein `leaked'from outer cell layers into internal cell layers, especially in young embryos and floral/inflorescence meristems. GFP expression was not maintained in mature organs indicating that in some tissues H2B::YFP may persist in nuclei after gene expression has terminated.

To determine the extent of the functional ACR4 promoter, the 1.9 kb full-length promoter was reduced distally from -1849 (where -1 is the base before the ATG) to give a -1026, a -857 and a -405 deletion. These fragments were placed directly upstream of the H2B::YFP reporter gene previously described, and transformed into plants. Their ability to drive L1-specific expression was assessed in young roots, developing seeds and inflorescence meristems, and compared to that of the full-length promoter.Δ-1026 and Δ-857 both gave expression patterns identical to that shown by the full-length promoter in roots, embryonic and meristematic tissues(verified in 20 independent transformants). Δ-405 gave no detectable H2B::YFP expression (40 independent transformants screened). Thus all sequences required for normal ACR4 expression were located in the first 857 bases of the promoter.

To gain material for functional analysis of ACR4, collections of T-DNA insertion lines were screened. One insertional mutant in ACR4was identified in the Wisconsin population(Krysan et al., 1999) and shown to be heterozygous for a double (back to back) T-DNA between bases 1066 and 1100 of the ORF. This allele was designated acr4-1. Three mutant lines were uncovered in the Syngenta collection(Sessions et al., 2002): the acr4-2 allele contained a T-DNA insertion at base 249 of the ACR4 ORF, acr4-3 contained an insertion 570 bp downstream of the ACR4 ORF and acr4-4 housed two insertions in the ACR4 promoter, one 1.6 kb and one 810 bp upstream of the start of transcription. PCR and subsequent Southern blot analysis confirmed that the progeny of heterozygous acr4-1, -2, -3 and -4 plants segregated wild-type, heterozygous and homozygous individuals in a 1:2:1 ratio. Southern blot analysis also showed that the acr4-1 and acr4-2 and backgrounds contained no other T-DNA insertions than those at the ACR4 locus, but that both the acr4-3 and acr4-4 backgrounds contained multiple independently segregating TDNAs. The positions of the insertions in acr4-1 and acr4-2would be predicted to give strong mutant alleles and were therefore of particular interest for functional studies.

Segregating populations carrying acr4-1, acr4-2, acr4-3 and acr4-4 were analysed to identify potential mutant phenotypes associated with disruption of the ACR4 gene. No differences in gross plant morphology between homozygous mutants and wild-type plants were noted in any of the four populations. However, all acr4-1 and acr4-2homozygotes showed abnormalities in both the shape and texture of developing seeds. Instead of being elliptical and smooth, the developing seeds were rounded and rough in appearance. In addition, seeds were heterogeneous in their development compared to wild type, and siliques contained unfertilised ovules and aborted seeds at a rate of 40-85%(Fig. 2A,B). The developmental stage of seed abortion varied from just after pollination to just prior to maturity. When selfed heterozygous plants were analysed, no seed abnormalities were found, indicating that the phenotypes described were due to the maternal genotype. No seed defects were observed in the siliques of homozygous acr4-3 and acr4-4 plants.

To confirm that seed morphology and abortion phenotypes were entirely under maternal control, flowers from homozygous acr4-2 plants were emasculated and pollinated either with self pollen, or pollen from heterozygous or wild-type siblings. Control flowers from heterozygous and wild-type siblings were either self pollinated or cross pollinated with pollen from the homozygous plant. Siliques from crosses onto heterozygous or wild-type plants were full of morphologically normal seed, independent of the genotype of the male parent (5 crosses of each). Self-pollinated siliques from homozygous plants were only 15-60% full, and contained seeds exhibiting the mutant phenotypes previously described. Siliques from crosses of wild-type or heterozygous pollen to a homozygous female presented identical phenotypes to self-pollinated homozygotes (10 crosses of each). In all cases mature seed germinated successfully and segregated homozygous, heterozygous or wild-type seedlings in the proportions expected, confirming that the embryo sac genotype plays no role in the seed phenotype observed.

To understand the developmental basis of the observed seed phenotype, ovule morphology in mutant plants was analysed. Mutant ovules displayed phenotypes of varying severity (Fig. 3B-D). All ovules showed epidermal irregularities, including abnormal cell size and shape, callus-like outgrowths, and occasional inappropriate cell types such as stomata. Ovules sometimes fused together(Fig. 3D). In most (>90%) of mutant ovules the abaxial zone of the integuments failed to elongate sufficiently to give the curvature seen in wild-type ovules. In some cases the embryo sac/nucellus protruded from the shortened integuments(Fig. 4H,J). In addition to disruption in ovule epidermal organisation, lack of organisation of integument cell layers was observed, with some ovules showing loss of cell layers, and others showing sporadic over-proliferation of integument cells. A varying proportion (20-50%) of ovules lacked a recognisable embryo sac(Fig. 3C,D). In extreme cases the endothelium was absent or reduced to a few disorganised cells. In other cases the endothelium cells enclosed differentiated/divided cells, or an empty space. In 30-50% of mutant ovules the egg apparatus (synergids, egg cell and polar nucleus) could be distinguished (Fig. 3B).

In order to ascertain at what stage ovule developmental defects first occurred, scanning electron microscopy (SEM) of developing ovules was undertaken. Wild-type development was as previously described (Schnietz et al., 1995; Robinson-Beers et al.,1992). Ovule primordia arose as bulges along the placenta, and developed into finger-like protrusions(Fig. 4A). Subsequently the inner and outer integuments initiated as two ring-shaped growths encircling the megasporocyte-containing ovule tip (nucellus), with the inner integument initiating just before the outer integument(Fig. 4C). Both integuments then elongated as sleeves of cells engulfing the nucellus, with the outer-integument growing faster than, and eventually overgrowing the inner integument (Fig. 4E,G). In acr4 mutant ovules, development was normal until the point of integument initiation (Fig. 4B). However, instead of initiating as smooth ring-like bulges,the integuments of mutant ovules initiated unevenly, with some cell files bulging out, and others remaining flat. In many cases more than two sets of bulging cells could be seen in the proximodistal axis, and integuments did not initiate as coherent rings, suggesting that the points of integument initiation were not well defined (Fig. 4D). After initiation, mutant integuments appeared thicker than wild-type, and their more rounded cells gave developing ovules a rough texture(Fig. 4F). Integuments grew more slowly in mutant than in wild-type plants, and the leading edge of the integument, instead of being smooth, appeared disorganised. At maturity, even in the most `normal' mutant ovules, integuments failed to fully enclose the nucellus (compare Fig. 4G with 4H). In some cases integument elongation either of one(Fig. 4I) or both integuments was severely compromised (Fig. 4J). Abnormal protruding cells were often observed on the surface of mutant ovules (Fig. 4H)

Defects observed in ovules were maintained in developing seeds when fertilisation had been possible. In particular, the texture of the seed coat was abnormal, with outgrowths observed, particularly in retarded seeds. A lack of proximodistal elongation of the mutant embryo sac after fertilisation caused the mutant endosperm to develop in a reduced volume giving seeds a round rather than elliptical shape (Fig. 2D). Although defects in embryo organisation were not observed,seeds with more severe defects in integument organisation were also retarded in embryo and endosperm development. Histological analysis supported the hypothesis that these seeds were those observed to abort.

To study the epidermal abnormalities observed in developing seeds, SEM analysis of mature mutant seeds was carried out. Although seed coat abnormalities were observed, particularly at the funiculus abscission scar and at the micropylar region, the majority of seed coat cells had a similar structure to those observed in wild-type seeds(Fig. 2C,D). Because homozygous seeds still differentiated appropriate epidermal cell types, and even in ovules, mis-specification of cell types (for example the presence of stomata)involved epidermal-specific identities, the expression of an L1 marker in mutant ovules was investigated. Homozygous acr4-2 and acr4-1plants were crossed to marker lines expressing an N-terminal GFP::ATML1 fusion protein (unpublished results) under the ATML1 promoter(Sessions et al., 1999). These lines expressed nuclear localised fusion protein in the L1-specific pattern previously reported for ATML1 expression in embryos and meristems(Lu et al., 1996; Sessions et al., 1999). ATML1 fusion protein expression was observed in the outer cell layer and endothelium of mature ovules in wild-type plants, with weak expression occasionally observed in the inner cell layer of the inner integument. In acr4mutant ovules ATML1 expression was similar to or more widespread than in wild type. In excrescences on the ovule surface, both protruding callus-like cells and underlying cells showed expression. Strong expression was sporadically seen in cells situated between the ovule epidermis and the endothelium. In several cases, the egg sac space was filled with expressing cells. This analysis suggests that although mutant ovule integument cells showed abnormalities in organisation, they did not loose their L1 identity.

Because acr4 mutants showed abnormalities in ovule integuments,sepal margins, which have a similar structure (appressed layers of L1 cells)were examined in more detail. Although no major defects in sepal morphology were observed in acr4 mutants, it was noted that the cells at sepal boundaries appeared less well organised than in wild-type plants, giving a somewhat ragged appearance (Fig. 4K,L). In general the border region was thicker (contained more cells) in the abaxial/adaxial dimension than in wild type, suggesting that outgrowth of sepal margins could be affected. Mutant margin cells were irregularly shaped and showed abnormal `lumpy' areas and regions devoid of the cuticular decoration seen in wild-type cells. No defects at the margins of leaves or petals could be discerned.

Although two independent mutant alleles in two different backgrounds both gave identical phenotypes, a further confirmation that the observed phenotype was due to loss of ACR4 function was obtained by genetic complementation of acr4-2. Homozygous mutants were crossed to hygromycin-resistant transformants carrying a full-length ACR4promoter driving the ACR4 ORF. Four F₂ families corresponding to four independent transformants were selected on hygromycin and PCR-genotyped for homo- or heterozygosity of acr4-2. The phenotypes of homozygous plants were compared with those of heterozygous and wild-type plants in each case. For two families homozygosity of acr4-2 plants was verified by Southern blot. For all four families full phenotypic complementation was apparent in immature and mature seeds of homozygous mutant plants, confirming that the observed mutant phenotypes were due to loss of ACR4 function.

To establish whether ACR4 protein encodes a functional kinase, as predicted from its sequence, a GST fusion protein construct was engineered to express the ACR4 kinase domain in bacteria. A 61 kDa protein encoding the GST-kinase was expressed and purified (Fig. 5). To act as a control in kinase assays, Lys 540 (a crucial amino acid in the kinase activation loop) was mutated to methionine. GST-kinase and GST-kinase-null proteins were subjected to in vitro kinase assays. The kinase domain showed phosphorylation that was absent in the kinase-null variant(Fig. 5). Incubation of the kinase domain with GST protein alone did not result in phosphorylation of GST,indicating that the kinase domain could autophosphorylate inter or intramolecularly in vitro (results not shown).

Structural predictions indicated a plasma membrane localisation for ACR4. To test this prediction the entire ACR4 ORF was fused at the C terminus in frame with GFP, placed under control of the complete ACR4promoter and introduced into plants. In order to test whether the fusion protein was being correctly localised, plants from two different expressing lines were crossed to homozygous acr4-2 mutants, and F₂plants were genotyped for the acr4-2 allele. Full complementation of the acr4-2 mutant phenotype was observed for one line, and partial complementation for the other line tested. Partially complementing plants showed reduced seed death, and a more normal seed shape, although seed texture was still abnormal. Expression of fusion proteins was detected in regions where H2B::YFP reporter expression had previously been observed(Fig. 6), and was identical in wild-type and in complemented homozygous mutant plants. Cellular localisation of fusion proteins varied from tissue to tissue. In some cells, for example those on the surface of ovules, most fluorescence appeared to be associated with plasma membranes (Fig. 6A). In the L1 cells of embryos, inflorescence and floral meristems and roots, plasma membrane localisation was observed, but fluorescence also localised to multiple small intensely staining bodies within cells (Fig. 6B,C,D,G). These bodies did not co-localise with red-fluorescing chloroplasts, but were the same size or smaller. To confirm that fluorescent protein was localised to plasma membranes rather than cell walls, roots were treated with 0.8 M mannitol to induce plasmolysis. Under these conditions fluorescence was pulled away from cell-cell boundaries, indicating that fluorescent proteins were indeed associated with the plasma membrane, rather than the cell wall(Fig. 6F). In order to further address ACR4::GFP localisation, the effect of brefeldin A (BFA) on protein localisation in roots was examined. BFA targets and inhibits the action of proteins involved in vesicle formation, thereby inhibiting vesicle trafficking within cellular membrane compartments and to and from the plasma membrane(Nebenführ et al., 2002). After treatment with BFA, ACR4::GFP localisation was compromised(Fig. 6G,H). The relative intensity of plasma membrane-associated fluorescence decreased, and instead of multiple small cytoplasmic bodies, one or two large fluorescent bodies were observed in each cell. An identical phenomenon has been observed using immunolocalisation of the auxin efflux carrier PIN1 in BFA-treated roots(Geldner et al., 2001; Geldner et al., 2003). The described result of BFA treatment on ACR4::GFP localisation supports the hypothesis that ACR4 is usually exported to the plasma membrane via the ER and Golgi, and that this export, or possibly some form of recycling, is inhibited by BFA. It seems likely that the cytoplasmic bodies observed in cells not treated with BFA correspond to elements of the endomembrane system, such as excretory vesicles or endosomes.

In all tissues studied, fusion protein was present in plasma membranes adjacent to both anticlinal and periclinal cell walls, although the degree of localisation adjacent to periclinal cell walls was variable. In root meristems(QC and root cap initials) localisation was observed uniformly in both anticlinal and periclinal plasma membranes(Fig. 6E). In cells situated on the surface of the plant, the amount of protein visible in plasma membranes adjacent to the outer periclinal cell wall appeared lower than that on anticlinal and inner periclinal cell plasma membranes(Fig. 6D,I). This phenomenon was particularly noticeable in the outer cells of ovule outer integuments where all cells expressed fusion protein, although this could in part be due to the additive signal from two appressed internal membranes(Fig. 6I).

Despite the wide ranging expression pattern observed for ACR4,probable null mutants only show defects in two tissues; ovule integuments and sepal boundaries. Characterisation of mutants in several genes affecting integument development including INNER NO OUTER, SHORT INTEGUMENTS 1,SHORT INTEGUMENTS 2, BELL, AINTEGUMENTA, ABERRANT TESTA SHAPE and NOZZLE, has shown that integuments play an important role in female gametophyte development and maturation(Reiser and Fischer, 1993; Villanueva et al., 1999; Robinson-Beers et al., 1992; Broadhvest et al., 1999; Baker et al., 1997; Schneitz et al., 1998;Balasubramanian et al., 2002). In particular the presence of an intact endothelial cell layer is crucial, possibly because nutritionally and developmentally important substances are channelled to the gametophyte through this specialised cell layer (Kapil and Tiwari, 1978). We observed no defects in ovule development until integument initiation, when megaspores usually initiate meiosis, suggesting that the observed lack of a female gametophyte in some mature acr4ovules may be due to degradation or de-differentiation rather than to lack of initiation of gametophyte development.

Many acr4 mutant ovules are never fertilised because of severe morphological abnormalities, but of the ones that are, those with more severe organisational defects abort as developing seeds. Abortion is independent of zygotic genotype and is, moreover not due to developmental defects in embryo and endosperm development, although both tissues are retarded at the time of seed death. Retardation and abortion probably occur because defective seeds provide insufficient maternal support, in terms of nutrients, for embryo sac development. Similar retardation and death of embryo/endosperm was observed when reduced expression of the genes FBP7 and FBP11 led to developmental abnormalities and degeneration in the endothelium and seed coat of Petunia (Colombo et al.,1997). The total lack of zygotically derived embryo development defects and the observation of seed coat abnormalities in our study contradicts results obtained using antisense experiments to reduce ACR4 expression (Tanaka et al.,2002).

ACR4, as a membrane-localised receptor-like kinase, probably acts by perceiving extracellular ligands. Several genes encoding possible ligands, or ligand processing molecules for CR4 and related proteins have been proposed. These include the subtilase encoded by the ABNORMAL LEAF SHAPE 1(ALE1) gene (Tanaka et al.,2001). During embryo and endosperm development, signals from surrounding tissues (as could be provided by the action of genes such as ALE1) might be important in signalling required for `outside' cell layer specification. However, it seems more likely that in organ primordia, as has been shown in root cell layer differentiation, an `inside to outside'signalling process is involved in regulating cell layer behaviour, combined with a role for signals from neighbouring cells in the same cell layer(Nakajima and Benfey, 2002). Our observation that ACR4 protein is localised on `internal' plasma membranes of `outside' cells supports the hypothesis that ACR4 may perceive signals from underlying cells and/or same-layer neighbours. If this is the case, the restriction of the acr4 phenotype to ovule integuments and sepal margins could be attributable to the fact that these tissues are unique in the Arabidopsis plant, in being composed of two appressed layers of L1 cells. If normal L1 behaviour (i.e. anticlinal divisions giving rise to a monolayer of L1 cells) were dependent on perception of positional information both from underlying cells, and from same-layer neighbours, then a loss in signalling between same-layer neighbours could be compensated for by signals from underlying cells in most tissues. However, in the case of ovule integuments and sepal margins, positional information would be effectively limited to that exchanged between same-layer neighbours. The cells in these organs would thus be particularly sensitive to disruption of this signalling pathway, which would be expected to lead to a loss of cellular organisation and thus abnormalities in organ outgrowth, similar to the phenotype observed in acr4 mutants.

The restricted mutant phenotype of ACR4 compared to maize CR4 mutants is surprising since ACR4 appears to be unique in Arabidopsis in its degree of similarity to maize CR4. Unlike studies of cr4 in maize, we find no evidence for a loss of epidermal identity in acr4 mutants, but rather solely a loss of cell organisation. The cell disorganisation observed in acr4 ovule integuments and sepal margins is, however, reminiscent of aspects of the epidermal defects observed in the leaves of maize cr4 mutants. Notably, both phenotypes involve deregulation of the planes of division, and organisation of populations of L1 cells. Striking differences in expression also exist between ACR4 and CR4. In maize, CR4 is expressed in the aleurone cell layer and one of the major phenotypes associated with cr4 mutants is a defect in aleurone differentiation(Becraft et al., 1996). ACR4 shows no endosperm expression, although it is arguable whether Arabidopsis can be considered to differentiate a structure analogous to the cereal aleurone layer (Berger,1999). In addition, unlike ACR4, CR4 appears to be expressed throughout apical meristems, without restriction to the L1 layer until late in leaf development (Becraft et al., 1996), and no CR4 expression has been reported in maize root tissue.

Functional redundancy between ACR4 and four other Arabidopsis genes showing weaker similarity to CR4 cannot be ruled out as an explanation for some of the differences in phenotypic severity between cr4 and acr4 mutants. The two most closely related genes encode proteins lacking a conserved kinase catalytic domain required for kinase activity (domain 8) (Hanks et al.,1998). ACR4 encodes a functional kinase, and kinase activity is probably required for at least some of its functions. However,kinase-inactive receptors can retain partial function, possibly by interaction with other unrelated kinases. The kinase-null clv1-6 allele, which causes part of the kinase domain of the CLAVATA1 protein to be deleted, causes only a weak mutant phenotype. The mutant protein thus retains functions that are independent of its ability to auto/transphosphorylate itself and other proteins (Torii and Clark,2000). Of the two less similar genes, one encodes a protein closely related to tobacco CRK1, which has recently be implicated in cytokinin responses (Schafer et al., 2002). The other shares many more residues with CRK1 than with ACR4 and CR4, especially in the extracellular domain adjacent to the trans-plasma membrane domain, where ACR4 and CR4encode putative TNFR-like repeats.

An alternative explanation for the weak acr4 phenotype could be that although several independent mechanisms regulate L1 behaviour in both Arabidopsis and maize, mechanistic differences in organ primordium development in monocotyledonous and dicotyledonous species have led to less functional overlap in maize than in Arabidopsis. Considering the relatively large numbers of genes expressed in L1 cell layers from early in development in both species, this possibility seems realistic, and will be investigated using ongoing mutagenesis and double mutant analysis approaches in the near future.

We would like to thank Dr Jim Haseloff (University of Cambridge, UK), Dr Corinne Boisnard-Lorig and Dr Frederic Berger (Ecole Normale Superieure, Lyon,France) for providing pBI121, pBSmGFP6 and the GAL4::VP16 coding sequence. We are most grateful to Dr Allen Sessions (Syngenta, Torrey Mesa Research Institute, San Diego) for providing pAS99, to Dr Chris Jeffree for help with SEM, Kathryn Degnan for technical assistance, Dr Jane Langdale, Dr Andrew Hudson, Dr Justin Goodrich and three referees for critical comments on the manuscript, Dr John Golz for help with in situ hybridisations and resin embedding and Dr Thomas Guebitz for help with phylogenetic analysis. We acknowledge the work of the NASC, ABRC, Syngenta and the University of Wisconsin for creating and providing seed stocks. G.I. is supported by a Royal Society University Research Fellowship, and M.G. by a BBSRC studentship. The project also benefited from a JREI grant (BBSRC/Olympus optical).