Receptor serine/threonine protein kinases in signalling: analysis of the erecta receptor-like kinase of Arabidopsis thaliana

Authors


Summary

The Arabidopsis ERECTA (ER) gene regulates elongation of above-ground organs. ER encodes a member of the leucine-rich repeats–receptor-like protein kinases (LRR–RLK) gene family, with the predicted protein containing a signal peptide, 20 leucine-rich repeats in the extracellular domain, a transmembrane domain, and a cytoplasmic serine/threonine protein kinase domain. The structural features of the predicted ER protein suggest its role in cell–cell signalling is through phosphorylating serine/threonine residues. Consistent with this hypothesis, in vitro protein kinase analysis indicates that ER is a functional serine/threonine protein kinase. Furthermore, a large-scale genetic screen was conducted to analyse new mutations in the erecta gene; 16 new er alleles were isolated, all of which were recessive. Here we present the identification of molecular lesions of seven alleles of er, which suggests the hypothesis that ERECTA might employ a mode of action distinct from other RLKs such as Xa21 or CLAVATA1, which function in disease resistance and developmental pathways, respectively.

Introduction

The coordinated development of a multicellular plant requires that cell division and expansion be regulated in order to generate a particular shape or form. Cell surface receptors, known as receptor-like protein kinases (RLK), have emerged as potential key regulators for such coordinated development. RLKs also function in diverse biological processes such as self-incompatibility or disease resistance (Kreis & Walker, 2000). Despite their wide biological roles, all RLKs are structurally similar, composed of an amino terminal signal peptide, a variable extracellular domain, a single transmembrane domain, and a cytoplasmic serine/threonine protein kinase domain (Lease et al., 1998). Mechanistically, RLKs are hypothesized to function in multistep pathways that lead to changes in gene regulation. However, little is known about downstream components or which genes are regulated by different RLKs. A greater understanding of how plants interpret environmental and internal cues will be gained by characterizing RLKs and their pathways.

To facilitate progress in understanding RLK function, Arabidopsis thaliana ERECTA (ER), was chosen as experimental material (Torii et al., 1996). The advantages of ER include a distinctive loss-of-function phenotype (Rédei, 1962), viable and fertile nullomorphic er alleles (Torii et al., 1996), and the benefits of the model system Arabidopsis. To gain insights into ER structure and function, we have carried out a genetic screen to analyse new mutations in the erecta gene.

Er dramatically alters organ shape, affecting the development of leaves, flowers, and fruits (Rédei, 1962; Bowman, 1993; Torii et al., 1996). The er phenotype is expressed throughout much of the plant’s life (Torii et al., 1996). While embryonic development of er mutants has not been analysed, starting at the seedling stage, er plants already have shorter petioles, and rounder cotyledons and juvenile leaves than wild-type (Bowman, 1993). All subsequent leaves of er mutants maintain this developmental pattern. Mirroring leaf development trends, er flowers are shorter and wider than wild-type and are borne on shorter pedicels (Torii et al., 1996). Similarly, inflorescence stems are both wider and shorter (Torii et al., 1996). Finally, er siliques are shorter, wider, and have a blunt tip (Redei, 1962; Bowman, 1993; Torii et al., 1996).

ER encodes a member of the RLK gene family, with the predicted protein containing a signal peptide, 20 leucine-rich repeats (LRR) in the extracellular domain, a transmembrane domain, and a cytoplasmic serine/threonine protein kinase domain (Torii et al., 1996). Sequence analysis of er alleles has suggested the importance of both extracellular and intracellular domains. The extracellular domain of er-103 has a methionine instead of an isoleucine at a LRR consensus position (Torii et al., 1996). The importance of the protein kinase domain for ER function was illustrated by the er-1 allele that possesses a missense mutation of isoleucine 750 to lysine, which is a conserved residue in the protein kinase domain of most RLKs reported. This allele is equivalent in phenotypic strength to er-105 or er-104 which have no mRNA signal on a Northern blot (Torii et al., 1996). Thus, er-1 is likely to be a null allele, suggesting that the protein kinase domain is critical for ER function.

The characterization of RLKs CLAVATA1 (CLV1) and BRASSINOSTEROID INSENSITIVEI1 (BRI1), which function in developmental pathways, and Xa21, which functions in disease resistance, provide a starting point for the analysis of the ERECTA RLK mechanism of action (Song et al., 1995; Clark et al., 1997; Li & Chory, 1997). The clv mutants have enlarged shoot and floral meristems, fasciated stems, extra flowers, and excess floral organs (Clark et al., 1993). CLV1 has been proposed to control the balance between proliferation and differentiation at shoot and floral meristems (Clark et al., 1993). The property of meristem stem cell identity is controlled by a regulatory loop in which CLV1 represses the WUSCHEL (WUS) gene and the WUS gene induces meristem identity and expression of CLV3, the ligand for CLV1 (vide infra) (Mayer et al., 1998; Brand et al., 2000; Schoof et al., 2000). CLV1 has been demonstrated to be an active protein kinase that autophosphorylates on multiple serines (Stone et al., 1998; Williams et al., 1997). Genetic evidence suggests that CLV1 functions as a multimer, based on dominant interference seen when some mutant alleles are heteroallelic with wild-type (Clark et al., 1995). Furthermore, size exclusion chromatography of immunoprecipitated CLV1 from plant extracts indicates that CLV1 is found as a disulphide-linked multimer in a molecular complex containing several other proteins (Trotochaud et al., 1999). CLV1 protein is found in complexes of two sizes in wild-type plants, 450 kDa and 185 kDa (Trotochaud et al., 1999). Analysis of the 185 kDa complex revealed that in this complex, CLV1 most likely exists as a disulphide-linked heteromultimer.

A candidate partner for CLV1 in the heteromultimer is the CLV2 protein. clv2 is the second of three mutants that express the clv phenotype. clv2 mutants have the characteristic club-shaped fruits of clv1 mutants, but they also express several traits absent from clv1 mutants suggesting other roles (Kayes & Clark, 1998). Notably, the fruits of clv2 mutants are valveless, and growth under short day conditions suppresses the clv2 phenotype. These phenotypes, which are not observed in clv1 mutants, were interpreted to indicate CLV2 function in both CLV1 and other developmental pathways.

CLV2 was cloned by T-DNA tagging and found to encode a protein predicted to have a signal peptide, extracellular domain consisting of leucine-rich repeats, and a small cytoplasmic domain (Jeong et al., 1999). Thus, CLV2 is similar to the predicted CLV1 protein, except that it lacks a protein kinase domain. Size exclusion chromatography analysis of CLV1 protein in extracts prepared from clv2 mutants shows an overall reduction in the quantity of CLV1 protein and a shift in the distribution of CLV1 from the wild-type high molecular weight complex, to the lower molecular weight CLV1 complex (Jeong et al., 1999).

A Rho GTPase-like protein is present in the CLV1450 kDa complex (Trotochaud et al., 1999). Small G proteins have become well established in metazoan signal transduction cascades, but their relevance to plant signal transduction cascades was unclear. As there are multiple Rho GTPase-like proteins in Arabidopsis, and the antibodies used to detect them react with at least several of these proteins, it is unclear which member of this class is in the CLV1 complex (Trotochaud et al., 1999). Furthermore, the relative importance and function of the Rho-GTPase-like protein in the complex remains to be determined.

Results from two complementary approaches suggest a role for a type 2C protein phosphatase, designated KAPP (Kinase Associated Protein Phosphatase) in CLV1 signalling. Overexpression of KAPP in a wild-type background caused a weak clv phenotype (Williams et al., 1997), and cosuppression-based loss of expression of KAPP in a clv1 background suppresses the clv phenotype (Stone et al., 1998). Finally, KAPP immunoprecipitates with CLV1 protein and is present in the 450 kDa CLV1 complex (Stone et al., 1998; Trotochaud et al., 1999).

Genetic analyses place CLV1 and a third gene, CLV3, in the same pathway because recessive hypomorphic alleles of clv1 become dominant when placed in a clv3 mutant background (Clark et al., 1995). Moreover, clv1 clv3 double mutants constructed with intermediate alleles express a phenotype no more severe than either a clv1 or clv3 single strong allele alone.

CLV3 was cloned by T-DNA and transposon tagging and is predicted to encode a small secreted polypeptide, a structure suggestive of a ligand (Fletcher et al., 1999). Further evidence that CLV3 functions as a ligand was obtained by analysis of the CLV1 protein complex from extracts made from strong clv3 mutant plants. In clv3 plants, the majority of the CLV1 protein exists as the 185 kDa complex (Trotochaud et al., 1999).

Recent evidence further supports the hypothesis that CLV1 and CLV3 form a receptor-ligand pair. Coimmunoprecipitation studies (Trotochaud et al., 2000) of cauliflower protein extracts followed by column chromatography show that a CLV3 antibody precipitates the 450 kDa CLV1 complex, but not the 185 kDa complex. Similarly, Western blots of complexes precipitated with a CLV1 antibody detect CLV3 in the 450 kDa complex, but not the 185 kDa complex. However, when the presence of CLV3 was assayed in extracts from clv1–10 plants, an allele of CLV1 that does not show kinase activity, it was only found in low molecular weight (c. 25 kDa) fractions. This result suggests that CLV1 kinase activity is required for the stable association of CLV3 with CLV1. When extracts of yeast expressing CLV1 and CLV2 are incubated with the 25 kDa CLV3 containing complex from plants binding is also observed, indicating that CLV3 binds to a CLV1/CLV2 complex. Because the CLV3 antibody cross-reacts with a 6-kDa protein under denaturing conditions these results support the hypothesis that CLV3 is part of a multicomponent complex that acts as the ligand for CLV1.

Taken together, the aforementioned data suggests a model in which binding of the CLV3 ligand to a CLV1-CLV2 disulphide-linked dimer results in a conformational change and autophosphorylation of CLV1. The activation of the CLV1 kinase then stabilizes CLV3 binding. Upon this event, other molecules such as KAPP, the Rho-GTPase-like protein, and other proteins bind to the activated receptor complex and cause the CLV1 signal to be transduced.

Another Arabidopsis thaliana mutant RLK is bri1. Several different screens for Arabidopsis dwarf mutants, insensitive to brassinosteroids (BR), have resulted in the identification the bri1 locus (Clouse et al., 1996; Li & Chory, 1997; Schumacher & Chory, 2000). The predicted BRI1 protein consists of a signal peptide, 25 leucine-rich repeats in the extracellular domain, a transmembrane domain, and a cytoplasmic serine/threonine protein kinase domain (Li & Chory, 1997). BRI1 contains a 70-amino acid island between the 21st and 22nd LRR.

Consistent with its role as a membrane-bound receptor, BRI1-green fluorescent protein fusions are localized to the plasma membrane (Friedrichsen et al., 2000). In addition, when wild-type, but not mutant BRI1 is expressed in human embryonic kidney cells, it autophosphorylates on serine and threonine. However, BR does not alter BRI1 kinase activity in these cells. This suggests additional plant factors are required for BRI1 signalling.

To determine if BR can directly activate BRI1, chimeric receptors containing the BRI1 extracellular domain were fused to the protein kinase domain of XA21 (He et al., 2000). XA21 is an LRR receptor kinase from rice that functions in disease resistance (Song et al., 1995). The BRI1:XA21 chimera was transformed into rice cells that were then treated with BR. To determine BR-dependent activity, a variety of disease resistance responses were assayed, including cell death, oxidative burst, and expression of defense response genes. Chimeras containing the BRI1 leucine-rich repeats and transmembrane domain responded to the BR application, but importantly chimeric receptors containing a mutation in the BRI1 extracellular domain or the XA21 protein kinase domain did not. These results support the hypothesis that the BRI1 extracellular domain senses BR.

Sequence analysis of bri1 alleles provides additional evidence for the importance of the extracellular LRR domain and the cytoplasmic protein kinase domain (Li & Chory, 1997; Noguchi et al., 1999; Friedrichsen et al., 2000). Mutations that result in strong phenotypes occur in both the protein kinase domain and the extracellular LRR domain, including the 70 amino acid island (Noguchi et al., 1999).

While previous studies with CLV1, BRI1, and Xa21 have provided models for RLK mechanisms, we still know little about ER biochemical activity or the mechanism of action in organ shape specification. In this report, we demonstrate that ER encodes a functional serine/threonine protein kinase. To gain some insights into the mechanism of action employed by ER we have collected a large number (total of 24) of erecta mutant alleles, and the phenotypic consequences of 16 er mutant alleles were assessed by morphometric analysis. Furthermore, we have identified the molecular lesions present in seven er alleles. The analysis of these mutants suggests the hypothesis that ER may employ a mode of action distinct from other RLKs such as Xa21 or CLV1.

Materials and Methods

Protein expression

pMal-ER was constructed by ligation of a 1270-bp EcoRI fragment encompassing amino acids 610–976 of ER with EcoRI digested pMalCRI vector (New England Biolabs, Beverly, MA, USA) and testing for the correct orientation by restriction digestion. This construct was transformed into E. coli strain PR745 for expression. 50 ml of LB broth containing 100 µg ml−1 ampicillin was inoculated with a single colony and grown for 12 h at 37°C. The 50 ml culture was added to 1 l of LB containing 100 µg ml−1 ampicillin and grown at 37°C until an OD600 of 0.6 was achieved, and then IPTG was added to a final concentration of 0.2 mM. The culture was then grown at 30°C for 6 h before harvesting cells by centrifugation. The maltose binding protein (MBP)-ER fusion protein was purified as described (Horn & Walker, 1994).

Phosphoaminoacid analysis

MBP-ER was autophosphorylated by incubation in protein kinase buffer for 1 h at room temperature. The reaction consisted of 0.25 µg µL−1 MBP-ER protein in 50 mM HEPES pH 7.4, 10 mM MgCl2, 10 mM MnCl2, 1 mM DTT, 10 µM [γ-32P]ATP (10 000 cpm pmol−1). The reaction was stopped by the addition of Laemelli loading buffer and heating at 70°C for 10 min. The sample was separated on a 10% SDS-PAGE gel at 200 V until the bromophenol blue had completely migrated off the gel. The gel was then fixed, stained, dried and exposed to X-ray film for one hour. Phosphoamino acid content was determined as described (Boyle et al., 1991).

Plant material and growth conditions

Columbia (Col) and Landsberg erecta (Ler) seeds were obtained from the Arabidopsis Biological Resource Center (ABRC) (Ohio State University, USA). The remainder of the erecta alleles were isolated either from 0.3% EMS Columbia gl1 nph1–1 lots (E. Liscum, unpublished), EMS mutagenized Columbia gl1 seeds (Lehle Seeds), or KA Feldmann’s T-DNA population (in the case of er-123, isolated from ABRC stock CS6502 pool 23). The er-2 allele generated by G. Rédei has been renamed er-106 and an additional er allele was isolated from a T-DNA mutagenized population (S. Sawa, unpublished) and named er-107. Thus, the 16 novel er alleles described in this report are named er-108 to er-123. Seeds were imbibed in 0.1% agarose for 3 d at 4°C and then sown on 8.9 cm square pots containing PRO-MIX potting soil (Premier, Allentown, PA, USA) at a density of 9 seeds per pot. Plants were grown at 23°C under continuous light (100 µmol m−2 s−1) and watered weekly with Miracle-Gro fertilizer. A minimum of 25 inflorescences and a minimum of 50 pedicels and fruits were measured per genotype.

Photography

The top 5 mm of the inflorescence was excised (7–10 d after flowering), superglued onto a piece of black velvet, placed on a dissecting scope stage, and photographed from above.

Sequencing of erecta alleles

Genomic DNA was prepared from rosette leaves of novel erecta alleles according to Edwards et al. (1991). To facilitate DNA sequencing, four overlapping genomic regions covering the erecta open reading frame (nt 1761–3497, nt 3255–4832, nt 4709–6203, and nt 5990–7423 of GenBank D83257) were amplified by PCR and sequenced directly using ABI 377 A automated sequencer (Applied Biosystems, Foster City, CA, USA). After removal of primer sequence, overlapping contigs were assembled by Sequencher™ 3.1.1 software (GeneCode Corp, Ann Arbor, MI, USA). The site of mutation was confirmed by sequencing both strands. Nucleotide positions refer to ER genomic DNA sequence (Genbank accession number D83257) unless otherwise noted.

For cDNA sequencing of er-116, total RNA was isolated from aerial portions of 4.5-wk-old plants with the Rneasy Midi kit (Qiagen, Valencia, CA, USA) according to manufacturer’s instructions. mRNA was purified from the total RNA using the PolyAttract mRNA isolation kit (Promega, Madison, WI, USA) according to manufacturer’s instructions. Two micrograms of mRNA was used to make first strand cDNA in the Riboclone cDNA Synthesis system (Promega). mRNA was primed with oligo(dT), according to manufacturer’s instructions, except that only first strand cDNA synthesis was performed. PCR was performed using ExTaq (Panvera, Madison, WI, USA), with either primers ERg1762 (5′ gtatatctaaaaacgcagtcg 3′) and ERg5757rc (5′tgacacggtgagtttagccaa 3′) to generate a 1.7-kb double-stranded cDNA encompassing the 5′ end of the ER cDNA (designated product #1) or primers Erg5022 (5′ cttgagtagaaatcatataact 3′) and ER26 (5′ tataggaagtctaagtgccag 3′) to generate a 1.7-kb double-stranded cDNA overlapping with product #1 and encompassing the 3′ end of the ER cDNA (designated product #2). PCR conditions were as follows: PanVera PCR buffer (proprietary composition except 2 mM MgCl2), 200 µM dNTPs, 0.2 µM forward and reverse primers in a 50-µL reaction. Hot starting was performed with 2.5 units of ExTaq, which were added after the reactions were heated to 96°C. The PCR cycle parameters consisted of 96°C for 5 min, followed by 36 cycles of 96°C for 15 s, 53°C for 30 s, and 72°C for 2 min. PCR products were then purified by low melting point agarose gel electrophoresis, phenol extraction, ethanol precipitation, and resuspended in 25 µL of dH2o. Both strands of the two overlapping products were determined by sequence analysis as follows: 2 µL of purified product #1 or product #2 were used in cycle sequencing reactions using ABI Prism dRhodamine Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer, Foster City, CA, USA), followed by ethanol precipitation. Sequences were determined by capillary electrophoresis. cDNA sequences were analysed using Seqman software (DNASTAR, Inc., Madison, WI, USA).

Sequencing of er-108 cDNA was performed as follows. Total RNA was isolated from inflorescence apices of er-108 plants by Rneasy Mini kit (Qiagen). The RT-PCR was performed using the ThermoScript RT-PCR system (Gibco BRL, Carlsbad, CA, USA) according to the manufacturer’s instruction. cDNA covering the er-108 mutation was amplified with the primers ERg5022 and ERg6182rc (5′ caatgatatacttctcacttag 3′) and the amplified fragment was directly sequenced using the same primers.

All new er alleles except er-108 have an A to T substitution at nt2800 (of the published ER genomic DNA sequence, Torii et al., 1996) within the fourth intron, which suggests there is a polymorphism in the parental plants that were subjected to mutagenesis. er-101 and er-102 were determined to contain the same lesion as er-1, with lysine substituted for isoleucine 750.

Results

Biochemical analysis of ER protein kinase activity

ER is predicted to encode a serine/threonine protein kinase (Torii et al., 1996). To test this hypothesis, we expressed the ER cytoplasmic domain as a translational fusion with MBP. MBP-ER was expressed at the expected molecular weight and possesses protein kinase activity (Fig. 1a,b). ER autophosphorylates on serine and threonine, but not on tyrosine residues (Fig. 1c). Furthermore, MBP-ER autophosphorylates on multiple residues as judged by analysis of tryptic digests separated by two-dimensional thin layer electrophoresis thin layer chromatography, as over 10 distinct spots were visible (data not shown). However, the exact number of phosphorylation sites is difficult to determine due to the possibility of incomplete digestion with trypsin. These results are consistent with the prediction that ER encodes an active serine/threonine protein kinase.

Figure 1.

Expression, activity, and phosphoaminoacid analysis of maltose binding protein (MBP)-ERECTA protein kinase. (a) MBP-ER and MBP stained with Coomassie blue (b) Protein kinase activity of MBP-ER (c) Two-dimensional thin layer electrophoresis of phosphoaminoacids. Circles mark the position of the migration of standards and +f sign indicates the origin. Dark spots near origin represent partially hydrolysed protein.

Genetic analysis of erecta alleles

To dissect the importance of the extracellular and protein kinase domains for ER function in vivo, we screened for new er mutant alleles. 100 000 M2 plants from 40 000 Col M1 EMS mutagenized seeds were screened and 60 putative mutants were identified that created an erecta-like inflorescence, with clustered flowers at the apex of the inflorescence. Thirty of the original 60 putative mutants were found to be heritable and fertile. Those mutants that are not allelic to er were designated elk mutants and will be described elsewhere. Sixteen mutants were determined to be allelic to er and were named er-108 to er-123. Mutants were determined to be allelic to er after analysis of F1 and F2 from reciprocal crosses with Ler or er-105. The F1 non-complementation was further confirmed by analysing F2 progeny and failing to see any wild-type plants segregate out. Chi-squared tests of an expected nine wild-type: seven erecta phenotype in the case of two unlinked recessive genes conferring an er phenotype showed a P-value of less than 0.001 in all cases. Dominance of er alleles to wild-type was determined by analysis of reciprocal F1 crosses to wild-type Col. All er alleles were recessive to wild-type (data not shown).

Phenotypic characterization of erecta alleles er mutants are characterized as having shorter inflorescences, pedicels, and fruits (Rédei, 1962; Torii et al., 1996). Furthermore, flowers are clustered at the top of the inflorescence and fruit tips are blunt-ended. The inflorescence apex of each er allele is shown in Fig. 2. The strength of each allele was determined by morphometric analyses in which we measured stem, pedicel, and fruit lengths (Fig. 3). Most er alleles identified are strong alleles, despite some variability, nearly equal in phenotypic strength to er-105, a null allele (Torii et al., 1996). However, one allele, er-116, stands out as weaker in strength. First of all, a less severe phenotype is obvious in floral apices of er-116, when compared with other strong alleles (Fig. 2). Additionally, er-116, has inflorescence stems that are about 80% as long as wild-type (Fig. 3a). In contrast, strong er alleles are about 40–50% as tall as wild-type. Similarly, pedicels of er-116 are 80% as long as wild-type, while pedicels of strong alleles are 40–50% as long as wild-type (Fig. 3b). Moreover, the fruit length of er-116 is 90% of the wild-type length, with strong er alleles approximately 80% as long (Fig. 3c). Finally, the fruit tips of er-116 mutants are only weakly blunted relative to strong alleles (data not shown). Taken together, these results show that er-116, while phenotypically distinct from wild-type, is far weaker than other isolated er alleles. Although a direct comparison of er-116 with er-103, a previously described weak allele, was not performed, it is likely that the er-103 phenotypic strength is intermediate between er-116 and er-105.

Figure 2.

A comparison of the floral apices of erecta alleles. (a) Col apex (b) er-105 apex (c) er-108 apex (d) er-109 apex (e) er-110 apex (f) er-111 apex (g) er-112 apex (h) er-113 apex (i) er-114 apex (j) er-115 apex (k) er-116 apex (l) er-117 apex (m) er-118 apex (n) er-119 apex (o) er-120 apex (p) er-121 apex (q) er-122 apex (r) er-123 apex. All photographs are at the same magnification, scale bar indicates 3 mm.

Figure 3.

Characterization of erecta mutant alleles comparing five week old plants. (a) inflorescence lengths (b) pedicel lengths (c) fruit lengths. Bars represent average, error bars indicate standard deviation.

Sequence analysis of erecta alleles

Determination of the molecular lesions present in these alleles might allow an inference into the function or mechanism of ER action. Among the 16 er alleles described in this report, seven alleles were sequenced. In each case, lesions were detected that were consistent with both strands of DNA. These results are summarized in Table 1 and Fig. 4.

Table 1. Molecular lesions in novel erecta alleles. † Based on ERECTA genomic DNA sequence Genbank accession number D83257
er allelePhenotypic strengthLesionGenomic positionPredicted change
er-101 strongT ⇒ A6565I750 ⇒ K
er-102 strongT ⇒ A6565I750 ⇒ K
er-108 strongG ⇒ A5649Frameshift and premature stop
er-111 strongG ⇒ A5759W562 ⇒ STOP
er-113 strongC ⇒ T3274R196 ⇒ STOP
er-114 strongG ⇒ A6807D831 ⇒ N
er-115 strongC ⇒ T3796Q283 ⇒ STOP
er-116 weakG ⇒ A6974In frame deletion of IMSK
er-117 strongG ⇒ A5203G489 ⇒ D
Figure 4.

Structure of the predicted ERECTA (ER) protein and location of genetic lesions. Hatched boxes represent the signal sequence and transmembrane domain, dotted boxes represent leucine-rich repeats–receptors (LRRs), and the horizontal lined box represents the protein kinase domain. The predicted consequences of each lesion are shown. Z indicates a nonsense mutation. Sequence data for er-1 and er-103 are from Torii et al. (1996), and the remaining data are from this report.

Two alleles, er-108 and er-116 have lesions mapping to conserved splice site residues at either the donor or acceptor ends of introns, respectively. er-116 was of particular interest because of the weaker nature of its phenotype. er-116 was determined to have a mutation in the conserved splicing acceptor sequence AG at the 3′ end of the 26th intron. Northern blot analysis showed that steady-state levels of ER mRNA were expressed at the same levels in er-116 as in wild-type (data not shown). To determine the consequence of this splicing site lesion for the mature ER mRNA, we isolated and sequenced an er-116 cDNA. The er-116 cDNA was shown to have a 12-bp deletion, missing nucleotides 2575–2586 cDNA (+1 being the first nucleotide of the initiation ATG of the ER cDNA). This deletion is predicted to remove four amino acids, IMSK, and then restore the proper open reading frame. Interestingly, the mutation of the normal splicing acceptor site AG allowed for a cryptic splice acceptor site at the end of the lysine codon (AAG) to define a slightly larger than normal intron, creating a 12-bp in frame deletion in the cDNA. The four missing amino acids belong to subdomain X of the protein kinase domain. The crystal structure of human protein kinase A α catalytic subunit (PKAα) has been solved (Zheng et al., 1993), and subdomain X of PKAα corresponds to an alpha helix (alpha helix G) on the lower lobe near the carboxy terminus of the protein kinase domain.

A second allele, er-108, has a lesion affecting the conserved GT at the splicing donor site of the 22nd intron, with a guanine transition to adenine. Isolation and sequencing of an er cDNA from er-108 revealed that a cryptic splicing donor site 4 bp downstream of the normal GT donor site and the normal 3′ splice acceptor site are utilized resulting in the addition of 5 bp. This cDNA change causes a frameshift, predicted to result in the addition of 14 novel amino acids and a premature stop. Thus, the ER protein expressed in er-108 is predicted to have a full-length extracellular domain but lack the transmembrane and protein kinase domains. However, the level of ER mRNA or ER protein expression has not been examined in this allele.

Two of the sequenced alleles, er-114 and er-117, are predicted to be missense alleles. er-114 results in the replacement of a conserved aspartic acid with asparagine, in subdomain IX of the protein kinase domain. This is a nearly invariant residue among all protein kinases, and this substitution likely abolishes protein kinase activity.

er-117 was determined to have G to A transition at nucleotide 5203, predicted to cause a missense mutation replacing glycine 489 with aspartic acid. This mutation changes a conserved residue in the eighteenth leucine-rich repeat (LRR) of ER. ER has a consensus LRR pattern of LG-L-L-L-L-N-L-G-IP-, with the mutated residue of er-117 highlighted in bold. (Torii et al., 1996; Li & Chory, 1997). When the LRRs of ER, BRI1, or CLV1 are aligned, every LRR contains a glycine at this same position (Torii et al., 1996; Clark et al., 1997; Li & Chory, 1997). Replacement of a glycine, which has a small uncharged side-chain, to aspartic acid, which has a larger, negatively charged side-chain, is a dramatic change for a residue which is highly conserved in the LRRs of several RLKs.

The remaining three sequenced alleles, er-111, er-113, and er-115 each contain predicted nonsense mutations. In the case of er-111, the predicted ER protein would be truncated after the paired cysteines flanking the carboxy end of the extracellular LRRs. er-113 would result in a truncation in LRR six, while er-115 would result in a truncation in LRR 10. In addition to the nonsense mutation at nucleotide 3796 in er-115, there is also a silent adenine to guanine transition at nucleotide 5906. All three alleles are predicted to produce secreted forms of the extracellular domain of ER because they still have an amino terminal signal peptide, but lack the transmembrane domain.

Discussion

Verification of serine/threonine protein kinase activity of ER

ER was predicted to encode a serine/threonine protein kinase, yet there was no biochemical evidence supporting this claim (Torii et al., 1996). We tested this hypothesis by expressing a recombinant MBP-ER protein and analysing phosphoamino acid content. These experiments support previous activity predictions and support a model in which ER autophosphorylates on multiple serine and threonine residues.

Identification and verification of erecta alleles

Sixteen mutants were identified that were shown to be allelic with er by reciprocal complementation tests. All mutants were found to be recessive to wild-type. However, we cannot exclude the possibility of some alleles having weak semidominance as heterozygotes. Interestingly, the segregation of er-116 in F2 crosses with er-105 was roughly three weak erecta phenotype: one strong erecta phenotype (data not shown). Thus, er-116 is completely dominant to er-105. Because an intermediate phenotype was not observed, which might have occurred in the expected 50% of the F2 with the genotype er-105er-116, it suggests that the level of ER signal in er-116 can tolerate a 50% reduction without reaching a rate-limiting threshold and impacting phenotype. Since er-114 is likely to express an inactive protein kinase, but behaves in a recessive manner, this suggests that a 50% level of er-114 does not interfere with the normal ER function. This may be the case if ER acts as a homodimer that undergoes transphosphorylation. However, the possibility that er-114 has reduced expression cannot be excluded.

Significance of phenotypic variation of er alleles

Among the 16 alleles recovered, 15 were strong, while one is weak in strength. The recovery of a weaker allele indicates that smaller degrees of activity are possible. Combined with previously identified alleles, there are 24 er alleles in total. Because the ER promoter, exons, and introns encompass about 8 kilobases of DNA, if mutations in ER occurred randomly there would be about one mutation every 350 bp. This could potentially allow for a dissection of ER activity into discrete domains with different biological functions. For example, there could hypothetically be a domain controlling fruit shape and a separate domain controlling pedicel growth which would have been evidenced by an er allele only expressing a subset of the er phenotypes. However, phenotypes affected by er are not genetically separable in er alleles. That is, there were no alleles that affected one er trait without coordinately affecting other er traits to the same degree. This suggests that if there is any branching or divergence of ER signal to control various aspects of plant development, it must occur downstream of ER.

Implications of er lesions for ER function

There is paradoxical evidence regarding the importance of the protein kinase domain for RLK function. While many mutations have been found that map to the protein kinase domains of RLKs and produce strong mutant phenotypes, there are two interesting cases that deviate from this expectation (Torii, 2000). clv1–6 contains a frameshift mutation, which is predicted to replace the protein kinase domain with 16 novel amino acids followed by a premature stop codon (Clark et al., 1997). Biochemical studies support this prediction, because antibodies raised against the carboxy terminus of CLV1 react with protein of the expected CLV1 molecular weight in extracts from wild-type plants, but no signal is evident in extracts prepared from the clv1–6 background (Trotochaud et al., 1999). clv1–6 expresses a weak phenotype (Clark et al., 1997).

A second piece of data that questions the importance of the protein kinase domain in RLK functioning comes from a natural variant of the Xa21 gene from rice. Xa21 is a LRR RLK which confers resistance to Xanthamonas oryzae pv oryzae (Song et al., 1995). The Xa21 gene belongs to a multigene family containing seven members on rice chromosome 11. One of the members of this groups, Xa21D, has a retrotransposon called Retrofit inserted close to the carboxy end of the extracellular LRR-containing domain (Wang et al., 1998). Thus, Xa21D is predicted to result in a truncated version of the Xa21 protein containing the signal peptide and LRRs, but no transmembrane domain or protein kinase domain. Xa21D mRNA expression has been demonstrated and transgenic rice carrying a construct expressing the predicted Xa21D protein has an intermediate resistance to the pathogen (Wang et al., 1998). Furthermore, Xa21D transgenic plants maintain the same race-resistance conferred by the normal Xa21 gene. Taken together, these results suggest that the protein kinase domain may be partially dispensable for functioning. However, to have a completely normal response, the protein kinase domain is necessary. To explain these results, we can hypothesize that these RLKs may interact with additional components that, under some circumstances, can make the protein kinase domain of CLV1 or Xa21 partly redundant.

The isolation of erecta alleles er-108 and er-111 allowed for a genetic test of this phenomenon for ER. er-108 and er-111 are both predicted to encode a secreted, truncated forms of ER, as they possess a signal peptide, all 20 LRRs but neither transmembrane nor protein kinase domains. The truncations are brought about through different kinds of mutations: er-111 contains a nonsense mutation that should truncate ER, while er-108 contains a splicing donor site mutation that may also produce a truncated protein (see Results section). The expected result if ER functions in an analogous manner to Xa21 or CLV1 was that er-108 and er-111 would express weak or intermediate phenotypes. Instead, er-108 and er-111 have a strong phenotype, essentially equivalent to the null allele er-105, suggesting the hypothesis that ER must have some mechanistic differences from CLV1 or Xa21 such that the protein kinase domain of ER is absolutely essential for its function. At this moment, it is not clear whether the truncated ERECTA proteins are expressed at a comparable level to the wild-type proteins, or whether the truncation leads to mislocalization. It should be noted that clv1–6 possesses a transmembrane domain, while Xa21D does not. Detection of the truncated gene products as well as analysis of their subcellular localization, once specific antibodies against the LRR domain of ERECTA are available, will provide a clear answer as to the ERECTA mode of action. Further, the hypothesis that the protein kinase domain of ER is essential for ER function could also be tested by a reverse genetics approach expressing a truncated ERECTA gene in the null allele er-105.

Why does er-116 express a weak phenotype? er-116 possesses a 12-bp in-frame deletion predicted to remove four amino acids from subdomain X of the protein kinase domain. This subdomain consists of an alpha helix whose function is unknown (Hanks & Hunter, 1995). The alignment of several RLK protein kinase domains proposed by Li & Chory (1997) shows that among several RLKs, the first residue of IMSK is conserved as a hydrophobic residue, but the remaining three residues are very different. In fact, HAESA (RLK5) contains a deletion of the last three residues corresponding to IMSK, according to this published alignment. Furthermore, many protein kinases contain large insertions between subdomains X and XI (Hanks & Hunter, 1995). If we model ER as being similar to PKAα (Zheng et al., 1993), this deletion is predicted to shorten alpha helix G by approximately one helical turn. However, it is unclear what effect this would have upon ER to result in a weak phenotype. Potential consequences could be a decrease in enzyme activity, a loss of a potential phosphorylation site (IMSK), a reduction in substrate binding, or decreased interactions with downstream effectors.

Two missense mutants, er-117 and er-114, reemphasize the importance of the extracellular domain and protein kinase domain, respectively. er-117 contains a missense mutation changing glycine to aspartic acid. This mutation changes a highly conserved glycine, found in every LRR repeat of ER, BRI1, and CLV1. Thus, er-117 is likely to disrupt the structure of the extracellular domain and perturb ligand binding or dimerization. er-114 substitutes asparagine for a conserved aspartic acid residue in subdomain IX of the protein kinase domain, resulting in a strong mutant phenotype. This residue is nearly invariant among protein kinases, although sometimes it is substituted with an alanine or threonine. Structural/functional analysis of this residue in the tyrosine protein kinase Fer showed that mutating this residue to arginine abolished protein kinase activity (Cole et al., 1999). Furthermore, they proposed that deleterious substitutions at this position affect protein kinase activity because of a shift in the backbone of the catalytic loop.

Together, the biochemical analysis of ER activity and characterization of erecta mutants provide further support for the importance of both extracellular and protein kinase domains for ER function, and implicates a potential mode of action of ER. Sequence analysis of other er alleles may also give insight as to which residues of the ER protein are critical for ER signalling. This work, combined with additional structure/function analyses may provide insights into conserved and distinct features of plant RLK signalling.

Acknowledgements

KAL was supported by a predoctoral fellowship from the University of Missouri Maize Biology Training Grant and the University of Missouri Molecular Biology Program.

NYL was supported by the Washington NASA Space Grant Undergraduate Fellowship. These studies were supported by a National Science Foundation grant (MCB 9809884) and the University of Missouri Food for the 21st Century Program grant to JCW, and in part by a University of Washington Royalty Research Fund (R2499) to KUT.

Ancillary