A novel Arabidopsis CHITIN ELICITOR RECEPTOR KINASE 1 (CERK1) mutant with enhanced pathogen-induced cell death and altered receptor processing



  • Plants detect pathogens by sensing microbe-associated molecular patterns (MAMPs) through pattern recognition receptors. Pattern recognition receptor complexes also have roles in cell death control, but the underlying mechanisms are poorly understood. Here, we report isolation of cerk1-4, a novel mutant allele of the Arabidopsis chitin receptor CERK1 with enhanced defense responses.
  • We identified cerk1-4 in a forward genetic screen with barley powdery mildew and consequently characterized it by pathogen assays, mutant crosses and analysis of defense pathways. CERK1 and CERK1-4 proteins were analyzed biochemically.
  • The cerk1-4 mutation causes an amino acid exchange in the CERK1 ectodomain. Mutant plants maintain chitin signaling capacity but exhibit hyper-inducible salicylic acid concentrations and deregulated cell death upon pathogen challenge. In contrast to chitin signaling, the cerk1-4 phenotype does not require kinase activity and is conferred by the N-terminal part of the receptor. CERK1 undergoes ectodomain shedding, a well-known process in animal cell surface proteins. Wild-type plants contain the full-length CERK1 receptor protein as well as a soluble form of the CERK1 ectodomain, whereas cerk1-4 plants lack the N-terminal shedding product.
  • Our work suggests that CERK1 may have a chitin-independent role in cell death control and is the first report of ectodomain shedding in plants.


Plants recognize potentially harmful microbes via conserved signatures referred to as pathogen- or microbe-associated molecular patterns (PAMPs/MAMPs). Several corresponding pattern recognition receptors (PRRs) have been identified and were shown to be either receptor-like kinases (RLKs) or receptor-like proteins (RLPs). RLKs are plasma membrane-localized transmembrane proteins with an N-terminal, extracellular ligand binding domain (ectodomain) and a C-terminal intracellular protein kinase domain that functions as a signal transduction module. RLPs share the same domain organization but lack the kinase domain (Newman et al., 2013). Well-studied examples of PRRs perceiving peptide MAMPs are the Arabidopsis leucine-rich repeat (LRR) RLKs FLAGELLIN SENSITIVE 2 (FLS2) and EF-TU RECEPTOR (EFR), which recognize bacterial flagellin and elongation factor TU, respectively (Gómez-Gómez & Boller, 2000; Zipfel et al., 2006). Similarly, the lysin motif (LysM) RLK CERK1 is the Arabidopsis receptor for the fungal cell wall polysaccharide chitin (Miya et al., 2007; Wan et al., 2008; Petutschnig et al., 2010; Liu et al., 2012) and is also required for chitin sensing in rice (Oryza sativa) (Shimizu et al., 2010). Moreover, CERK1 was shown to play a role in the perception of the bacterial MAMP peptidoglycan in Arabidopsis (Willmann et al., 2011) and is targeted by the Pseudomonas syringae effector AvrPtoB (Gimenez-Ibanez et al., 2009; Zeng et al., 2012).

PRRs do not relay defense signals on their own, but are part of larger signaling complexes at the plasma membrane. RLKs and RLPs may bind to MAMPs cooperatively or associate with co-receptors for full activation of signal transduction (Monaghan & Zipfel, 2012). The contribution of PRR complexes to MAMP-triggered immunity is well established (Newman et al., 2013; Liebrand et al., 2014). However, their role extends beyond the recognition of microbe-derived signals, as several components have been described to modulate cell death. One prominent example is BRI1-ASSOCIATED RECEPTOR KINASE 1 (BAK1), the co-receptor of the PRRs FLS2 and EFR as well as the brassinosteroid receptor BRI1 (Kim et al., 2013). bak1 knock-out mutants show increased cell death formation upon infection with several pathogens (Kemmerling et al., 2007), suggesting that BAK1 has an additional function as a cell death regulator. Indeed, double mutants of BAK1 and its partially redundant homolog BAK1-LIKE 1 (BKK1) die shortly after germination as a result of massive spontaneous cell death (He et al., 2007). Knock-out mutants of BIR1, a BAK1-interacting LRR-RLK, resemble bak1 bkk1 double mutants, displaying very strong cell death early in development (Gao et al., 2009). Plants with reduced concentrations of a related BAK1 interactor, BAK1-INTERACTING RECEPTOR-LIKE KINASE 2 (BIR2), show increased pathogen-inducible cell death, similar to knock-outs of BAK1 (Halter et al., 2014). The cell death phenotype of bir1 is dependent on another LRR-RLK, SUPPRESSOR OF BIR1 1 (SOBIR1) (Gao et al., 2009), which is a co-receptor for several RLP-type PRRs in Arabidopsis and tomato (Solanum lycopersicum) (Liebrand et al., 2013; Zhang et al., 2013, 2014). Taken together, these studies provide clear evidence that PRR complexes play an important role in cell death regulation, but the underlying mechanisms appear to be intricate and are poorly understood.

In animals, controlled proteolysis is a well-known general mechanism regulating the function of transmembrane proteins. Proteolytic cleavage of the extracellular portion of transmembrane proteins at or near the cell surface is referred to as ectodomain shedding. It can be either constitutive or stimulus-induced and has been demonstrated for a large number of animal proteins, including cell adhesion molecules, immunomodulators, growth factors, receptors and enzymes (Arribas & Borroto, 2002; Hayashida et al., 2010). Prominent examples are epidermal growth factor receptors (EGFRs) and their ligands. Both EGFRs and EGFR ligands are synthesized as membrane-anchored precursors and their soluble ectodomains are released by proteolysis (Higashiyama et al., 2011). EGFR family receptors are receptor tyrosine kinases and structurally similar to plant RLKs. Ectodomain shedding often primes substrates for a second proteolytic cleavage within the transmembrane domain (regulated intramembrane proteolysis (RIP)), leading to the release of a soluble intracellular fragment. The cytosolic domain may then act as a signaling messenger within the cell (Lal & Caplan, 2011).

Here we report the isolation of a novel Arabidopsis chitin receptor mutant, cerk1-4, which carries an amino acid exchange in the ectodomain of CERK1. cerk1-4 plants show enhanced salicylic acid (SA)-dependent cell death formation and resistance to powdery mildews. The phenotype of cerk1-4 mutants is independent of chitin signaling, does not require CERK1 kinase activity and can be conferred by an N-terminal fragment comprising only the ecto- and transmembrane domain of CERK1-4. We also show that wild-type CERK1 undergoes ectodomain shedding and that cerk1-4 lacks the extracellular shedding product. Our findings suggest that CERK1 may have a yet unidentified role in cell death control that is mediated by its ectodomain and is independent of chitin perception.

Materials and Methods

Arabidopsis mutant screen and map-based cloning of cerk1-4

Arabidopsis thaliana (L.) Heynh. Columbia (Col-3 gl1) seeds were mutagenized with ethyl methanesulfonate and resulting M2 plants were inoculated with the barley powdery mildew Blumeria graminis (DC.) Speer f.sp. hordei (Bgh). Plants displaying macroscopic lesions were selected and the phenotype confirmed in the next generation. The mutant was crossed with Landsberg erecta (Ler) and mapping was performed as described previously (Lipka et al., 2005). Simple sequence length polymorphism (SSLP) marker analysis located cerk1-4 in a target interval bordered by markers nga162 and ciw11 on chromosome 3 (Lukowitz et al., 2000). To fine-map cerk1-4, 977 F2 plants were screened for individuals with recombination break points between these markers. Twenty-four recombinants were recovered and their genotypes at the cerk1-4 locus were deduced by scoring F3 progeny for their interaction phenotype with Bgh. Use of additional cleaved amplified polymorphic sequence (CAPS) and SSLP markers (Supporting Information Table S1a) inferred from the CEREON database of Col and Ler polymorphisms (Jander et al., 2002) allowed delimitation to an interval of c. 33 kb (bordered by markers MIL23C and MIL23F on BAC MIL23) harboring 11 predicted open reading frames (ORFs) (At3g21550–At3g21650). Sequencing of cerk1-4-derived cDNAs and genomic sequences revealed a single nucleotide mutation in position 370 of the coding region of At3g21630 (CERK1), whereas no sequence polymorphisms were found for the other genes.

Pathogen infection assays and microscopy

Inoculation with powdery mildews was performed as described previously (Lipka et al., 2005). Plants were 4–5 wk old at the time of inoculation and pictures were taken at 5–7 d post inoculation (dpi) for Bgh and 10–14 dpi for Golovinomyces orontii (Castagne) V.P. Heluta infections. Inoculation and scoring with Alternaria brassicicola (Schwein.) Wiltshire were performed according to Kemmerling et al. (2007). Trypan blue staining was performed as described in a previous report (Lipka et al., 2005). Confocal microscopy was performed on a Leica TCS SP5 system (Leica Mikrosysteme Vertrieb GmbH, Wetzlar, Germany) equipped with an Argon ion laser and a HyD hybrid detector. For plasmolysis, leaf discs were mounted in 0.5 M NaCl before microscopy.

Mutant analyses

The cerk1-4 mutant was crossed with the mutants PENETRATION 2 (pen2-1) (Lipka et al., 2005), SALICYLIC ACID INDUCTION DEFICIENT 2 (sid2-2) (Wildermuth et al., 2001), PHYTOALEXIN DEFICIENT 4 (pad4-1) (Glazebrook et al., 1996) and ENHANCED DISEASE SUSCEPTIBILITY 1 (eds1-2) (Aarts et al., 1998), as well as with a transgenic line expressing the salicylate hydroxylase NahG (Lawton et al., 1995). The pen2-1, pad4-1 and eds1-2 mutations and the NahG gene were detected as described previously (Nishimura et al., 2003; Lipka et al., 2005). The sid2-2 mutation was genotyped with primers sid2-2 F and sid2-2 R and cerk1-4 was genotyped by amplification with primers UL154 and UL166, followed by digestion with XapI (Table S1b).

Quantitative RT-PCR

RNA extraction was performed with the RNeasy Plant Mini Kit (Qiagen) or the innuPREP Plant RNA Kit (Analytik Jena AG, Jena, Germany) according to the manufacturer's instructions. cDNA was synthesized with RevertAid H Minus M-MuLV Reverse Transcriptase (Fermentas/Thermo Fisher Scientific, Schwerte, Germany). qPCR was performed with a CFX96 Real-Time PCR System (BioRad) using SsoFast EvaGreen Supermix (BioRad) as recommended by the manufacturer. The amplification protocol consisted of 30 s of initial denaturation at 95°C, followed by 45 cycles of 95°C for 5 s and 55°C for 10 s. Melting curves were recorded to ensure single product amplification. For each experiment, dilution series of pooled cDNA were run under the same conditions to calculate primer efficiencies. PATHOGENESIS-RELATED GENE 1 (PR1) was amplified using primers EP227 and EP228 and PR1 expression levels were normalized to ACTIN (amplified with primers EP223 and EP224; Table S1c).

Plant protein work

Protein extraction from Arabidopsis leaves, microsomal fractionation, chitin pull-downs, Mitogen-activated protein kinase (MAPK) assays, western blotting and protein staining were performed as described previously (Petutschnig et al., 2010). For apoplastic washes, leaves of 6-wk-old plants (grown under short-day conditions) were harvested and vacuum-infiltrated with water in a desiccator. They were gently dried using a salad spinner. The leaves were then wrapped in aluminum foil and placed in centrifuge buckets fitted with perforated plastic beakers. Apoplastic wash fluids were collected by centrifugation at 700 g for 5 min at room temperature. They were concentrated using 10 000 MWCO PES Vivaspin or Vivacell protein concentrators (Sartorius AG, Göttingen, Germany). Protein extraction with phenol was performed according to Meyer et al. (1988). For SDS extraction, frozen, pulverized plant materials were directly mixed with SDS loading buffer (50 mM TRIS, pH 6.8, 100 mM DTT, 2% SDS, 10% glycerol, and 0.025% bromophenol blue) and immediately boiled at 95°C for 10 min. Debris was removed by centrifugation at 10 000 g for 10 min. For mass spectrometry analysis, microsomal and soluble protein extracts from leaves of 6-wk-old Col-3 gl1 plants were prepared as described in a previous report (Petutschnig et al., 2010). Chitin enrichment of full-length CERK1 from microsomal fractions was performed according to the same protocol. Soluble protein extracts from Arabidopsis typically contain chitinases and other chitin-binding glycosyl hydrolases at relatively high concentrations. In SDS PAGE, these enzymes migrate at a size similar to that of the free CERK1 ectodomain and therefore would interfere with its detection. Thus, they were precipitated from the soluble fraction with 50% ammonium sulfate. After 60 min of equilibration at room temperature, precipitated proteins were pelleted by centrifugation at 17 000 g for 15 min. The supernatant, which contained the free CERK1 ectodomain, was used for binding to chitin magnetic beads as described previously (Petutschnig et al., 2010).

Mass spectrometry

In-gel tryptic digestion of proteins was carried out according to Shevchenko et al. (1996). Nano LC was performed on a RSLCnano Ultimate 3000 system (Thermo Scientific): peptides in the 1–6-μl sample solution were trapped and washed with 0.05% trifluoroacetic acid on an Acclaim® PepMap 100 column (75 μm × 2 cm; C18; 3 μm; 100 Ǻ; Thermo Scientific) at a flow rate of 4 μl min−1 for 12 min. Peptide elution and further analytical separation by reverse phase chromatography were performed on the online-coupled Acclaim® PepMap RSLC column (75 μm × 15 cm; C18; 3 μm; 100 Ǻ; Thermo Scientific) running a gradient from 96% solvent A (0.1% formic acid) and 4% solvent B (80% acetonitrile and 0.1% formic acid) to 50% solvent B within 25 min at a flow rate of 300 nl min−1. The eluting peptides were ionized by nano-electrospray (nESI) using the Nanospray Flex Ion Source (Thermo Scientific) at 2.4 kV and continuously transferred into the Orbitrap Velos Pro mass spectrometer (Thermo Scientific). Full scans within m/z of 300–1850 were recorded by the Orbitrap-FT analyzer at a resolution of 30 000 with parallel data-dependent top 10 MS2-fragmentation in the LTQ Velos Pro linear ion trap. LCMS method programming and data acquisition were performed with the software XCalibur 2.2 (Thermo Fisher). MS/MS2 data processing for protein analysis and identification was performed with the Proteome Discoverer 1.3 software (Thermo Scientific) using the Sequest and Mascot search engines and the TAIR10 protein database (Lamesch et al., 2012).

Information on plant growth conditions, SA measurement, constructs and cloning, northern blotting, structural images and statistical analysis may be found in the online version of this article (Methods S1).


A novel mutant with deregulated pathogen-induced cell death

To identify novel components that contribute to Arabidopsis nonhost resistance against nonadapted powdery mildew fungi, we performed a forward genetic screen for mutations that lead to altered interaction phenotypes with barley powdery mildew Bgh. Wild-type Arabidopsis plants do not show any macroscopic disease symptoms upon inoculation with Bgh (Lipka et al., 2008). We identified a dominant ethyl methanesulfonate (EMS)-induced mutant that developed necrotic lesions and leaf chlorosis 5–7 d after Bgh challenge (Fig. 1a). We later identified the gene affected as CERK1 (see below; Figs 2, S4) and therefore named the mutant cerk1-4. Three days after inoculation with Bgh, clusters of dead cells were visible in cerk1-4, which were absent in the background line Col-3 gl1 or noninoculated plants (Fig. 1b). Microscopic analysis of the cerk1-4 mutant revealed a deregulated cell death response at Bgh interaction sites. While Col-3 gl1 control plants showed restricted epidermal cell death at sites of fungal penetration, cell death frequently spread to surrounding epidermis and subtending mesophyll cells in the mutant (Fig. 1c,d). The pen2-1 mutant allows higher penetration rates of Bgh than wild-type Arabidopsis plants (Lipka et al., 2005). Bgh-induced spreading cell death was drastically increased when cerk1-4 was crossed with pen2-1 (Fig. 1d), confirming that cell death formation in cerk1-4 depends on the invasion success of the fungus. The exaggerated cell death reaction was also mounted in response to compatible powdery mildews such as Golovinomyces orontii, rendering cerk1-4 more resistant than wild type (Fig. S1a). Conversely, in interaction assays with the necrotrophic fungus Alternaria brassicicola, the mutant showed increased disease symptoms (Fig. S1b,c). Unchallenged cerk1-4 plants were indistinguishable from wild type up to c. 5 wk of age, but upon prolonged growth under short-day conditions they showed spontaneous cell death on older leaves. Late in development new leaves appeared small and crinkly, leading to an overall reduced rosette size (Fig. S2a).

Figure 1.

Upon Blumeria graminis f.sp. hordei (Bgh) attack, Arabidopsis thaliana cerk1-4 plants show enhanced cell death reactions and hyper-accumulation of salicylic acid (SA). Col-3 gl1 and cerk1-4 plants were inoculated with Bgh. Control plants were grown under the same conditions, but not inoculated. (a) Macroscopic symptoms on inoculated and control plants 5 d post inoculation (dpi). (b) Leaves of Bgh-inoculated (3 dpi) and control plants were stained with trypan blue to visualize cell death. (c) Microscopic images of Bgh interaction sites in Col-3 gl1 and cerk1-4 (3 dpi). Dead cells and fungal structures were stained with trypan blue. Spores causing cell death are marked with red arrows. Bars, 100 μm. (d) Deregulated cell death in cerk1-4 is triggered in response to fungal penetration. Col-3 gl1, cerk1-4, pen2-1 and pen2-1 cerk1-4 plants were inoculated with Bgh, and samples taken 3 dpi for trypan blue staining. For each of the indicated genotypes, interaction sites were analyzed microscopically and classified as follows: no cell death (papilla or haustorium formation; light gray), cell death restricted to the epidermal layer (medium gray) and spreading cell death in epidermal and mesophyll layers (dark gray). The data are presented as mean ± SD of four biological replicates with 100 interaction sites each. Within each interaction class, bars with different letters are significantly different according to Tukey's test (< 0.0001). (e) Total SA concentrations. Data are presented as mean of three replicates ± SD. Significant differences between cerk1-4 and Col-3 gl1: *, < 0.0001. Whole rosettes were harvested 3 and 5 d after Bgh inoculation, respectively. Fifteen plants were pooled for each sample. (f) PR1 expression was analyzed by quantitative RT-PCR and values were normalized to the expression of ACTIN. The results are presented as averages of three technical replicates ± SD. Significant differences between cerk1-4 and Col-3 gl1: *, < 0.0001. Whole rosettes were harvested after 3 and 5 d, respectively, and 15 plants were pooled for each sample. Experiment (d) was performed twice with similar results, experiment (e) was performed three times and experiments (a–c, f) were performed more than four times with similar results.

Figure 2.

The cerk1-4 mutation affects a conserved leucine residue in the second lysin motif (LysM) of the Arabidopsis thaliana chitin receptor CERK1. (a) Domain organization of CERK1. The position of the cerk1-4 mutation (L124→F) is indicated by an arrow. The epitope recognized by the CERK1 antibody used in this study is designated by the letter ‘α’. SP, signal peptide; TM, transmembrane domain. CERK1 derivatives described in this study are indicated by lines underneath the CERK1 organigram: FL, full-length CERK1; ED, CERK1 ectodomain; ID, CERK1 intracellular domain. In parentheses, their expected molecular masses are given. *, approximate molecular mass assuming full N-glycolsylation. (b) Protein structure of the CERK1 ectodomain (Liu et al., 2012). The central domain containing LysM2 is colored green. Leucine124, which is mutated in CERK1-4, is shown as a sphere model in red. Amino acids involved in interaction with chitin (Liu et al., 2012) are shown as blue spheres. (c) Alignment of the central LysM domain of all LysM receptor-like kinases (RLKs) (LYKs) and LysM receptor-like proteins (RLPs) (LYMs) from Arabidopsis, as well as selected LysM-RLKs and proteins from rice, Lotus japonicus and Medicago truncatula. The ligands they bind and/or the pathways they function in are indicated on the right. The highly conserved CxC motifs are shown in green. The LysM, as determined for AtCERK1 (Shimizu et al., 2010) and rice (Oryza sativa) CHITIN-ELICITOR BINDING PROTEIN (CEBiP) (OsCEBiP) (Hayafune et al., 2014), is indicated by a black box. AtCERK1 amino acids involved in interaction with chitin (Liu et al., 2012) are highlighted in blue. Isoleucine122 has been determined as a crucial residue for chitin binding in OsCEBiP (Hayafune et al., 2014). OsCEBiP isoleucine122 and positionally equivalent isoleucine or leucine residues in other proteins are highlighted in yellow. In CERK1-4, leucine124 is mutated to phenylalanine. CERK1 leucine124 and positionally equivalent leucines in other proteins are highlighted in red.

Deregulated cell death in cerk1-4 is salicylic acid dependent

SA promotes cell death and resistance to biotrophic pathogens. As the cerk1-4 mutant is affected in both processes, we analyzed its SA concentrations. In untreated 5-wk-old plants, cerk1-4 and Col-3 gl1 had comparable total SA concentrations. Upon Bgh treatment, SA accumulated in both genotypes; however, SA concentations were considerably more elevated in cerk1-4 (Fig. 1e). This was also reflected by significantly stronger induction of the SA marker gene PR1 (Fig. 1f). SA concentrations increase in Arabidopsis plants over time. Similar to Bgh-induced samples, older cerk1-4 plants accumulated SA to much higher concentrations than the Col-3 gl1 background line and also showed increased expression of PR1 (Fig. S2b,c).

To test if cell death deregulation in cerk1-4 is the consequence of elevated SA concentrations, we crossed cerk1-4 with mutant or transgenic lines impaired in the SA pathway. Indeed, a mutation in the SA biosynthesis gene sid2 strongly reduced Bgh-induced cell death in cerk1-4 and introduction of the salicylate hydroxylase NahG completely abolished the phenotype. Similarly, mutations in eds1 and pad4, two lipase-like genes involved in SA signaling, fully suppressed the cerk1-4 phenotype (Fig. S3). These findings strongly suggest that the formation of the cerk1-4 phenotype requires SA. Low levels of Bgh-induced spreading cell death in sid2-2 crosses with cerk1-4 may be explained by the fact that sid2-2 plants are not completely SA deficient (Wildermuth et al., 2001). As reported previously (Lipka et al., 2005), single knock-out mutants of the SA pathway are not sufficient to break nonhost resistance and thus did not support macroscopically detectable growth of Bgh. The same holds true for the cerk1-2 mutant.

cerk1-4 harbors a single amino acid exchange in the second LysM of CERK1

Map-based cloning and sequencing of candidate genes revealed the gene affected in the cell death mutant to be the LysM-RLK CERK1 (Fig. S4; Miya et al., 2007; Wan et al., 2008). The mutant harbors a point mutation, causing a single amino acid exchange (L124→F) in the second LysM of the CERK1 ectodomain (Fig. 2a). In Arabidopsis CERK1, the second LysM domain mediates chitin binding via a number of surface-exposed residues (Liu et al., 2012; Fig. 2b). Leucine124, the amino acid mutated in cerk1-4, is located on the opposite face of LysM2, pointing toward the center of the CERK1 ectodomain (Fig. 2b). Interestingly, this leucine can be found in in the central LysM domain of all Arabidopsis LysM-RLKs and LysM-RLPs and is also highly conserved in LysM receptors from other plant species (Fig. 2c), indicating that it fulfils an important structural function.

The cerk1-4 phenotype is independent of chitin signaling

As the cerk1-4 mutation is located in the ectodomain which mediates chitin binding (Petutschnig et al., 2010; Liu et al., 2012), we tested the chitin signaling capacity of the CERK1-4 protein. Upon chitin binding, CERK1 is phosphorylated, which is visible as a mobility band shift in SDS PAGE (Petutschnig et al., 2010). Western blotting showed that chitin treatment of cerk1-4 plants led to phosphorylation of the full-length receptor (Fig. 3a). CERK1-4 is also able to initiate downstream signaling, as no significant changes in chitin-induced generation of reactive oxygen species (ROS) or activation of MAP kinases were observed in the cerk1-4 mutant (Fig. 3b,c). In line with these findings, CERK1-4-GFP showed normal localization at the plasma membrane (Fig. S5). Together, these results suggest that the overall chitin response in cerk1-4 plants is not different from that in the wild type.

Figure 3.

Arabidopsis thaliana cerk1-4 mutants retain chitin-signaling capacity and the cerk1-4 mutant phenotype is independent of CERK1 kinase activity. (a) Col-3 gl1 and cerk1-4 plants were infiltrated with 100 μg ml−1 chitin and incubated for 20 min. The mobility shift of CERK1 indicating phosphorylation was detected by western blotting using the CERK1-specific antibody. cerk1-2 was included to demonstrate specificity of observed bands. The lower panel shows Coomassie brilliant blue (CBB)-stained protein. (b) Leaf discs of Col-3 gl1, cerk1-4, Col-0 and cerk1-2 plants were treated with 100 μg ml−1 chitin and reactive oxygen species (ROS) generation was monitored using a luminol-based assay. Chemiluminescence at 12 min after treatment (peak of chitin-treated samples) is presented as the mean of eight samples ± SEM (control, dark gray bars; chitin, light gray bars). RLU, relative light units. Significant differences between mutant and background: *, < 0.001. (c) In vitro grown Col-3 gl1, cerk1-4, Col-0 and cerk1-2 seedlings were treated with 100 μg ml−1 chitin for 12 min. Immunocomplex MAPK assays were performed with specific MAP KINASE 4 (MPK4) and MAP KINASE 6 (MPK6) antibodies. Upper panel, autoradiograph of phosphorylated substrate myelin basic protein (MBP; 18.5 kDa); lower panel, western blots detecting MPK4 (43 kDa) or MPK6 (45 kDa). CBB, Coomassie brilliant blue stained protein (loading control). (d) cerk1-2 knockout mutants were transformed with constructs containing CERK1, cerk1-4, cerk1 loss-of-function (LOF) (K350→N, enzymatically inactive) or cerk1-4 LOF double mutant cDNA driven by the CERK1 promoter. For each construct, three independent transgenic lines were analyzed. Col-3 gl1, cerk1-4, Col-0 and cerk1-2 served as controls. Upper panels, western blots of chitin-binding proteins probed with the CERK1 specific antibody; lower panel, CBB-stained membranes. (e) The transgenic lines and controls described in (d) were inoculated with Blumeria graminis f.sp. hordei (Bgh) and pictures were taken 5 d post inoculation (dpi).

To rule out the possibility that the cerk1-4 phenotype is caused by locally restricted or transient changes in chitin signaling that might be easily missed, we generated a double mutant version of CERK1 containing the cerk1-4 mutation as well as an amino acid exchange in the ATP binding site (K350→N). The latter is a loss-of-function (LOF) mutation that renders CERK1 enzymatically inactive and unable to restore chitin sensitivity in cerk1-2 knock-out plants (Petutschnig et al., 2010). We transformed cerk1-2 plants with the cerk1-4 LOF double mutant construct, as well as constructs containing cerk1-4 or cerk1 LOF single mutant or wild-type CERK1 cDNA. The expression of CERK1 or cerk1 LOF had no effect on Bgh-induced cell death (Fig. 3d,e). By contrast, cerk1-2 plants transgenically expressing cerk1-4 cDNA looked like cerk1-4 mutants, confirming the amino acid exchange in cerk1-4 as the cause of the cell death phenotype (Fig. 3d,e). Interestingly, cerk1-4 LOF also conferred the cerk1-4 phenotype to the respective transgenic lines (Fig. 3d,e), even though it lacks the capacity to mediate chitin signaling. These data corroborate the hypothesis that the cerk1-4 phenotype is independent of chitin signal transduction. They also demonstrate that, in contrast to chitin perception (Petutschnig et al., 2010), the deregulated cell death phenotype of cerk1-4 is independent of CERK1 kinase activity.

CERK1 is proteolytically processed and cerk1-4 mutants lack an N-terminal cleavage product

Western blotting with a specific antibody (Gimenez-Ibanez et al., 2009; Petutschnig et al., 2010) revealed that, in addition to full-length CERK1 at c. 75 kDa, wild-type extracts also contain a signal c. 33 kDa (Fig. 4a). As the antibody recognizes an epitope near the N-terminus of CERK1 (compare Fig. 2a), the 33-kDa signal represents an N-terminal fragment of CERK1. Consistent with this notion, both the full-length CERK1 band and the 33-kDa signal could be pulled down with chitin beads, indicating that they have chitin-binding activity. As expected, the cerk1-2 knock-out mutant (Miya et al., 2007) did not show any signal for full-length CERK1 and also lacked the 33-kDa band (Fig. 4a). On occasion, extremely low levels of 33-kDa signal were observed in cerk1-2, which is possible because the cerk1-2 T-DNA insertion is located near the 3′-end of the CERK1 gene and upstream transcripts can be detected (Miya et al., 2007). Interestingly, the cerk1-4 mutant contained the full-length CERK1 protein, but the 33-kDa band was absent (Fig. 4a). The cerk1-4 mutant shows enhanced cell death reactions and hyper-accumulation of salicylic acid upon pathogen attack and during senescence. Thus, we investigated the CERK1 band pattern under these conditions. In wild-type plants, the 33-kDa signal accumulated both after Bgh inoculation and at old age (Fig. 4b,c). However, the 33-kDa band was never detected in cerk1-4 mutants.

Figure 4.

Arabidopsis thaliana wild-type CERK1 is subject to ectodomain shedding, releasing a soluble CERK1 ectodomain into the extracellular space, whereas cerk1-4 mutants lack the shed ectodomain. CERK1 derivatives observed: FL, full-length CERK1; ED, free CERK1 ectodomain; ID, CERK1 intracellular domain (compare Fig. 2a). CBB, Coomassie brilliant blue-stained membrane (loading control). (a) Anti-CERK1 western blots of cerk1-4, cerk1-2 and their respective background ecotypes Col-3 gl1 and Col-0. Total protein extracts (left) and chitin pull-downs (cpd; right) were probed with the CERK1 lysin motif 1 (LysM1)-specific antibody. (b) Total protein extracts were prepared from 5- and 8-wk-old cerk1-4 and Col-3 gl1 plants and used for a chitin pull-down. Chitin-binding proteins were analyzed by western blotting with the CERK1-specific antibody. The blot shows three biological replicates (a–c) for each genotype at age 5 and 8 wk. Each replicate is a pool of 10 plants. (c) Col-3 gl1 and cerk1-4 plants were inoculated with Bgh. Control plants were grown under the same conditions, but not inoculated. Samples were collected 3 and 5 d after inoculation (dpi) and protein extracts were prepared. Western blots of chitin-binding proteins were probed with the specific CERK1 antibody. Each sample is a pool of 15 complete plants. (d) Protein extracts of cerk1-4, Col-3 gl1, cerk1-2 and Col-0 leaves were separated into microsomal and soluble protein fractions. Chitin pull-downs were performed with all samples before analysis by western blotting with the CERK1-specific antibody. (e) Apoplastic wash fluid (AWF) was collected from Col-0 leaves and part of the AWF was concentrated c. 50 fold (conc. AWF). Total extracts were prepared from leaves as a control (total). All sample types were analyzed by western blotting directly (dir) and after chitin pull-down (cpd), using the specific CERK1 antibody. Control blots were performed with an antibody against MPK4. The CERK1 ectodomain fragment could be detected in AWF after chitin pull-down and in concentrated AWF without further enrichment. MPK4 was easily detectable in total extracts, but signals were very weak in AWF, even when concentrated 50-fold, indicating low cytosolic contamination of the apoplastic protein preparations. (f) cerk1-2 knock-out plants transgenically expressing CERK1-GFP or cerk1-4-GFP under the CERK1 promoter were analysed by western blotting. Total proteins were extracted by direct addition of SDS loading buffer to frozen, powdered plant materials. The resulting samples were used in a western blot with a GFP antibody. For both CERK1-GFP and cerk1-4-GFP, two independent transgenic lines (a, b) are shown, Col-0, cerk1-4 and cerk1-2 served as controls. All experiments were performed at least three times with similar results.

The wild-type-specific 33-kDa CERK1 signal was also observed when plant material was extracted directly in SDS loading buffer or the strong denaturing agent phenol to abolish any enzymatic activity after harvesting (Fig. S6a). These experiments demonstrate that the 33-kDa band is not a result of in vitro proteolysis during extraction, but a form of CERK1 that exists in planta. Northern blots provided no evidence for CERK1 splicing variants and no difference in CERK1 transcripts were observed between cerk1-4 and wild type (Fig. S6b). Also, transformation of cerk1-2 plants with cerk1-4 cDNA constructs conferred the cerk1-4 phenotype excluding aberrant splicing as the cause (see above; Fig. 3d,e). Altogether, these findings suggest a co- or post-translational cleavage mechanism to produce the 33-kDa CERK1 derivative.

Wild-type but not cerk1-4 plants accumulate soluble CERK1 ectodomain

Analysis of microsomal preparations confirmed that full-length CERK1 is a membrane-bound protein. By contrast, the 33-kDa CERK1 cleavage product appeared in the soluble protein fraction, indicating that it lacks the transmembrane domain (Fig. 4d). This hypothesis is supported by the fact that the 33-kDa signal was also found in apoplastic wash fluids of leaves (Fig. 4e). To confirm the identity and specificity of the 33-kDa signal, we used the chitin-binding activity of CERK1 for its enrichment and subsequent analysis by mass spectrometry. Protein extracts from Arabidopsis leaves were separated into membrane and soluble fractions. Chitin-binding glycosyl hydrolases that are present at high levels in the soluble fraction were removed by ammonium sulfate precipitation. Chitin-binding proteins were then enriched from both soluble and membrane fractions using chitin magnetic beads and were analyzed by LC-MS/MS. As expected, membrane preparations contained peptides corresponding to both the CERK1 ectodomain and kinase domain. By contrast, only peptides of the CERK1 ectodomain were found in the soluble samples (Fig. S6c, Table S2a,b). The mass spectrometry data corroborate the production of the soluble CERK1 ectodomain in addition to the full-length membrane-bound CERK1 receptor. Furthermore, they suggest that cleavage occurs near or within the transmembrane domain.

In this experiment, we also detected two further chitin-binding LysM-RLKs, LYSM-CONTAINING RECEPTOR-LIKE KINASE 4 and 5 (LYK4 and LYK5), as well as the Glycosylphosphatidylinositol (GPI)-anchored LysM-RLP LYSIN MOTIF DOMAIN-CONTAINING GLYCOSYLPHOSPHATIDYLINOSITOL-ANCHORED PROTEIN 2 (LYM2), confirming previous findings (Petutschnig et al., 2010; Wan et al., 2012). Interestingly, peptides corresponding to the ectodomain of LYK4 were also found in the soluble fraction, suggesting that LYK4 may undergo ectodomain shedding like CERK1 (Table S3a). Moreover, we detected a soluble form of LYM2 that might be generated by cleavage of its GPI anchor (Table S3b). LYK5 was only found in the microsomal fraction, possibly as a consequence of its low overall abundance in the analyzed samples (Table S3c).

The cleavage of CERK1 near or at the transmembrane domain probably gives rise to an intracellular CERK1 fragment. To detect the postulated C-terminal CERK1 cleavage product in planta, we made several attempts at raising antibodies against the CERK1 kinase domain or C-terminus. As none of the antibodies showed sufficient affinity, we generated transgenic lines expressing C-terminally GFP-tagged, functional versions of CERK1 and CERK1-4 in the cerk1-2 background (Fig. S5). Western blotting with a GFP antibody indeed detected signals that corresponded in size to the tagged intracellular domain (c. 65 kDa; compare Fig. 2a), in both CERK1-GFP and CERK1-4-GFP (Fig. 4f). The presence of C-terminal fragments with both proteins may suggest that receptor cleavage occurs normally in cerk1-4, but the released CERK1-4 ectodomain does not accumulate. While in direct SDS extracts the C-terminal cleavage products of CERK1 and CERK1-4-GFP were readily detectable (Fig. 4f), they were only faintly visible when extracted under nondenaturing conditions. This is similar to reports from animal systems, where cleaved intracellular domains of receptor kinases are known to be notoriously difficult to detect because of their metabolically labile nature (Carpenter & Liao, 2009). Because of this, we were unfortunately unable to investigate whether the intracellular CERK1 fragment is soluble or membrane bound. However, confocal microscopy of CERK1-GFP and CERK1-4-GFP showed only plasma membrane localization (Fig. S5), suggesting that C-terminal fragments are (mainly) membrane associated.

The CERK1-4 ectodomain and transmembrane domain are sufficient to deregulate cell death containment

Experiments with an enzymatically inactive CERK1-4 variant showed that the manifestation of the cerk1-4 phenotype is independent of chitin signaling and does not require CERK1 kinase activity. Plant genomes contain large numbers of RLKs with inactive kinase domains which probably transduce their signals via phosphorylation-independent mechanisms (Castells & Casacuberta, 2007). This could also be the case for a potential CERK1 function in cell death signaling. Alternatively, the cerk1-4 phenotype may be mediated by the CERK1-4 ectodomain alone whereas the kinase domain may not be needed at all. To address this question, we expressed the CERK1-4 ectodomain (amino acids 1–232; compare Fig. 2a) in the cerk1-2 knock-out background from the CERK1 promoter. None of the resulting transgenic plants showed a cerk1-4-like phenotype (Fig. S7a). However, the transgenic CERK1-4 fragment could not robustly be detected in western blots (Fig. S7b), which made a straight-forward interpretation difficult. To circumvent this problem, we generated constructs for N-terminal CERK1-4 fragments that included the ectodomain as well as the transmembrane domain (amino acids 1–253) or transmembrane and juxtamembrane domain (amino acids 1–321; compare Fig. 2a). The resulting transgenic proteins were readily detectable (Fig. 5a) and cerk1-2 plants expressing either one of these CERK1-4 fragments from the CERK1 promoter strongly resembled cerk1-4 mutants (Fig. 5b). These findings suggest that not the kinase domain, but the N-terminal part of CERK1-4 mediates the formation of the cerk1-4 phenotype and that the ectodomain plus transmembrane domain are sufficient.

Figure 5.

N-terminal CERK1-4 fragments containing the transmembrane domain confer the cerk1-4 phenotype to Arabidopsis thaliana plants independently of full-length CERK1. Transgenic plants were generated in the cerk1-2 background, expressing N-terminal CERK1-4 fragments from the CERK1 promoter: CERK1-4 ectodomain plus transmembrane domain (amino acids 1–253) and CERK1-4 ectodomain, transmembrane domain and intracellular juxtamembrane domain (amino acids 1–321). (a) Total protein extracts were prepared from four different transgenic plants per construct as well as controls (Col-0, cerk1-2, Col-3 gl1 and cerk1-4) and pull-downs with chitin magnetic beads were performed. The chitin-bound proteins were analyzed by western blotting using the CERK1-specific antibody. FL, full-length CERK1; ED, CERK1 ectodomain fragment. (b) The transgenic lines and controls were inoculated with Blumeria graminis f.sp. hordei (Bgh) and pictures were taken after 7 d.

Targeted generation of a second CERK1 ectodomain mutant allele confers a cerk1-4-like cell death phenotype

Recently, the molecular structure of the CERK1 ectodomain was elucidated (Liu et al., 2012). The three LysMs are tightly packed against each other, resulting in a globular overall structure. The cerk1-4 mutation is located in a loop that connects the two α-helices of LysM2 and points toward the center of the ectodomain (Fig. 6a). We speculated that the L→F exchange might interfere with the tight packing of the three LysMs and reasoned that this might make the CERK1 ectodomain more accessible for proteases or change the interaction with other proteins, potentially leading to increased turnover. In the CERK1 ectodomain structure model, the first β-strand of LysM1 is directly facing the cerk1-4 mutation site. We exchanged two amino acids on β1 that are close to the cerk1-4 site (Fig. 6a) for amino acids with similar properties but bulkier side chains (A34 S35→LT). The resulting mutant gene variant, which we named cerk1-5, was transformed into cerk1-2. CERK1 and cerk1-4 constructs were transformed as controls. In western blot analyses, transgenically expressed CERK1 showed a protein pattern like endogenous CERK1, containing the full-length receptor as well as a clear band for the cleaved ectodomain (Fig. 6b). By contrast, CERK1-4 and CERK1-5 proteins showed a drastically reduced abundance of the free ectodomain band (Fig. 6b). Very weak signals at the position of the ectodomain band probably stem from the cerk1-2 background. However, because of the high expression levels of the respective lines, the presence of low levels of ectodomain shed from the transgenic proteins cannot be excluded. Importantly, CERK1-5 not only matched CERK1-4 in protein band patterns, but also induced a cerk1-4-like cell death phenotype (Fig. 6b).

Figure 6.

Arabidopsis thaliana cerk1-2 plants transgenically expressing cerk1-4 or cerk1-5 show a cerk1-4-like protein band pattern and cerk1-4 phenotype. (a) Position of cerk1-4 and cerk1-5 mutations within the CERK1 ectodomain. The cerk1-4 (L124) and cerk1-5 (A34 S35) mutation sites are shown as red and orange spheres, respectively. Disulfide bridges are shown as yellow sticks. (b) cerk1-2 knock-out mutants were transformed with constructs expressing the CERK1, cerk1-4 or cerk1-5 cDNA under the CERK1 promoter. Total protein extracts were prepared from six independent transgenic lines per construct and controls (Col-0, cerk1-2 and cerk1-4). Chitin-binding proteins were enriched on chitin magnetic beads and analyzed by western blotting with the specific CERK1 antibody (upper panel). The transgenic lines and controls were inoculated with Blumeria graminis f.sp. hordei (Bgh) and pictures were taken 7 d post inoculation (dpi; lower panels). FL, full-length CERK1; ED, CERK1 ectodomain fragment.


In this work we identified a dominant Arabidopsis mutant with an enhanced SA-dependent cell death response to the nonadapted powdery mildew Bgh. The mutant also showed increased resistance to adapted powdery mildews as well as hypersusceptibility to a necrotrophic fungus. The causative mutation was identified to be a single amino acid exchange in the ectodomain of the chitin receptor CERK1. A recent survey of dominant mutations in Arabidopsis (Meinke, 2013) lists five mutants with enhanced SA-dependent defense responses and resistance to biotrophic pathogens. Among them is a mutant of the ankyrin-repeat protein ACCELERATED CELL DEATH 6 (ACD6) (Lu et al., 2003), which has been described as a modulator of defense responses in naturally occurring Arabidopsis accessions (Todesco et al., 2010). The other mutants comprise two Toll/Interleukin-1 Receptor (TIR)-Nucleotide Binding Site (NBS)-LRR proteins (Shirano et al., 2002; Zhang et al., 2003), an RLK (Bi et al., 2010), and an LRR-RLP (Zhang et al., 2010). The two TIR-NBS-LRRs are probably functional R-proteins and the mutant phenotypes are probably caused by constitutive protein activity (Shirano et al., 2002; Zhang et al., 2003). The functions of the RLK and RLP are not known, but based on their domain organization they are probably receptors and could act as PRRs like CERK1. Together with the data from this study, these findings suggest that dominant mutations causing increased pathogen resistance reveal components involved in immune responses and/or signaling.

We demonstrated that the cell death phenotype of the cerk1-4 mutant is completely independent of chitin signaling. In contrast to chitin perception, manifestation of the cerk1-4 phenotype does not require CERK1 kinase activity and can be conferred by the N-terminal part of the receptor. This raises the possibility that CERK1 fulfils two independent functions. Knock-out of several immune receptors causes enhanced cell death responses, including the PRR co-receptor BAK1, its closest homolog BKK1 and the two BAK1 interacting kinases BIR1 and BIR2 (He et al., 2007; Kemmerling et al., 2007; Gao et al., 2009; Halter et al., 2014). The pathways by which these receptor kinases regulate cell death are still a mystery. However, a recently discovered point mutant of BAK1 demonstrated that the role of BAK1 in PAMP perception can be uncoupled from cell death control and therefore the two pathways are probably mediated by distinct molecular mechanisms (Roux et al., 2011; Schwessinger et al., 2011). Therefore, fulfilling two or more mechanistically independent functions might be a common theme for RLKs involved in immune signaling.

In this work we also showed that wild-type plants contain free, soluble CERK1 ectodomain in addition to the full-length receptor. Thus this study represents the first report of ectodomain shedding of a plant cell surface protein. cerk1-4 mutants lack the soluble CERK1 ectodomain derivative. Analyses of plants expressing C-terminally tagged versions of CERK and CERK1-4 demonstrated the presence of C-terminal fragments. These findings may suggest that CERK1-4 is cleaved normally, but the released ectodomain fails to accumulate. However, because of difficulties in further characterization of the C-terminal CERK1/CERK1-4 fragments we do not currently know whether the observed N- and C-terminal derivatives arise through the same cleavage event or mechanism, and consequently the data need to be interpreted cautiously. However, the hypothesis that the stability of the cleaved ectodomain is affected in cerk1-4 mutants is supported by the fact that the mutation is located in the second LysM domain of CERK1 and not in close proximity to the cleavage site, which was shown to be within either the transmembrane domain (TM) or the eight amino acids upstream. In fact, the cerk1-4 mutation site is right at the center of the CERK1 ectodomain structure (Liu et al., 2012). As CERK1-4 shows normal chitin binding and cerk1-4 plants can perceive chitin, the mutation probably does not affect this function of LysM2. Because the cerk1-4 mutation results in an exchange for a bulkier amino acid (L–F), it may disturb the tight packing of – and consequently the interaction between – the three LysMs of CERK1. Indeed, a targeted mutation at a site in LysM1 (cerk1-5), which is also at the center of the CERK1 ectodomain, conferred a cerk1-4-like cell death phenotype, providing further evidence for this theory. While LysM2 has been shown to mediate chitin binding (Liu et al., 2012), the functions of LysM1 and LysM3 are not known. Conceivable roles are binding of yet unidentified ligands or protein–protein interaction and complex formation. The cerk1-4 mutation may perturb one of these roles, while leaving the chitin signalling capacity unaffected.

Because the cerk1-4 phenotype does not require the kinase domain, it is tempting to speculate that the CERK1 ectodomain could potentially act like an RLP and transduce a signal other than chitin via interaction with an unknown RLK. Notably, Arabidopsis contains three LysM-RLPs (LYM1–3) that have been indicated to mediate peptidoglycan recognition via CERK1 (Willmann et al., 2011) or regulate cell-to-cell connectivity (Faulkner et al., 2013). They are GPI-anchored proteins (Borner et al., 2003) and may therefore be released into the apoplast like the CERK1 ectodomain.

An alternative explanation for cell death phenotypes associated with PRR complexes is receptor guarding. Plant pathogens have evolved effector molecules which they transfer into the host to subvert its immunity, and several of them have been shown to target PRRs (Deslandes & Rivas, 2012; Dou & Zhou, 2012). Thus, it would be advantageous for plants to monitor the integrity of receptor complexes involved in immunity. It has been proposed that complexes containing BAK1 are guarded by a surveillance system which detects alterations in their composition or activity (Gao et al., 2009; Halter et al., 2014). Similarly, plants may sense the altered characteristics or stability of CERK1-4 or its interaction with complex partners and trigger cell death. One conceivable scenario would be interference of the CERK1-4 receptor with the function of known cell death-modulating RLKs such as BAK1, BIR1 or BIR2. However, as CERK1 was demonstrated to be targeted by an effector for degradation (Gimenez-Ibanez et al., 2009) it seems more likely that the integrity of CERK1 itself is monitored. CERK1 is involved in the perception of at least two PAMPs (chitin and peptidoglycan) of fungal as well as bacterial origin, and thus it would be particularly favorable for plants to guard CERK1. Interestingly, Symbiosis Receptor-Like Kinase SYMRK, a Lotus japonicus LRR-RLK required for symbiosis with rhizobia and arbuscular mycorrhiza, was recently reported to be cleaved within the ectodomain to release an N-terminal subdomain. This was shown to enhance complex formation with Nod factor receptor 5 (NFR5) and to promote turnover of the remaining SYMRK receptor (Antolin-Llovera et al., 2014), demonstrating that receptor processing can affect protein–protein interactions and stability. The fact that cerk1-4 plants do not accumulate the soluble CERK1 ectodomain may suggest that breakdown products of the shed ectodomain cause the observed cell death deregulation. Such CERK1 fragments may be generated at low levels by proteases in wild-type plants during pathogen attack or senescence and may function as mobile apoplastic elicitor molecules that allow signaling to neighboring cells. Indeed, in animal systems ectodomain shedding has a well-established role in the generation of free extracellular signaling molecules (Higashiyama et al., 2011; Weber & Saftig, 2012).

Another important function of ectodomain shedding in animals is priming of substrates for the release of intracellular fragments, thereby facilitating signaling to the nucleus (Lal & Caplan, 2011). Ligand-induced release of the kinase domain has been proposed for the rice LRR-RLK Xa21 (Park & Ronald, 2012); however, the role of the Xa21 ligand is currently unclear (Lee et al., 2009, 2013). CERK1 ectodomain shedding does not require a chitin stimulus, suggesting that this process does not play a role in chitin signaling into the cell. Accumulation of the CERK1 ectodomain is constitutive, but is enhanced upon pathogen attack and during senescence. This would be consistent with a role of CERK1 ectodomain shedding in cell death control. A definitive answer to these questions may be given in the future by generation of shedding-deficient CERK1 variants that lack both the extracellular and the intracellular cleavage products.


We thank Dr Marcel Wiermer for critical reading of the manuscript and Dr Cyril Zipfel for helpful discussions. We also thank Sabine Wolfarth, Katharina Dworak and Sarah Gardner for technical assistance. This work was supported by the Gatsby Charitable Foundation and DFG Priority Programme 1212 ‘Microbial Reprogramming of Plant Cell Development (Plant-Micro)’.