For the TIGS and transient over-expression experiments 7-day-old seedlings of the Bgh-susceptible spring barley cv. Maythorpe were used. Stable transgenic barley plants of cv. Golden Promise were generated as described (Hensel et al., 2008). Transient over-expression in wheat was done by using 7-day-old seedlings of the Bgt-susceptible cv. Kanzler. Bombarded leaf segments or transgenic plants were inoculated with Swiss Bgt field isolate FAL 92315, or Swiss Bgh field isolate CH4.8 throughout the study.

Initial searches of the NCBI¹ and RIKEN EST² databases, using HvLEMK1 as the query sequence identified the best match to be the full-length EST TPLB0008D17, with a nucleotide similarity of 94%. Using this sequence, primers were designed using the software CLC DNA workbench, targeting a region upstream of the predicted start codon and downstream of the predicted stop codon: TaLEMK1_FL1_f1, TaLEMK1_FL1_r1, TaLEMK1_FL1_f2, TaLEMK1_FL1_r2 (Supplementary Table S1). The cDNA synthesis was performed using SuperScript III (Invitrogen) and the specific primer TaLEMK1_FL1_r2 with a mixture of RNAs from the wheat cultivar Renan. Renan was inoculated with the adapted (Bgt) and non-adapted (Bgh) pathogens of powdery mildew to ensure high levels of expression of TaLEMK1 transcripts. RNA was extracted 24 h after inoculation. Long range PCR was performed using the “Expand Long Template PCR system” from Roche Co. and following the manufacturer’s recommendation for cDNA templates between 0.5 and 9 kb. DNA fragments of the desired length were gel purified and cloned into pGEMT-easy. Colonies where tested for TaLEMK1 inserts by colony PCR. Plasmids with expected insert size where verified by sequencing. For allele comparison between wheat genotypes, seedlings (14 days after sowing) of the winter wheat varieties Arran, Brock, Cadenza, Pastiche, Vault, and Zebedee were inoculated with Bgh (isolate CH4.8), and RNA was extracted 24 h after inoculation. The Illumina, RNA-seq pair-end reads from each of the six varieties were assembled into contigs using a compiled hexaploid wheat reference sequence (TGAC, Norwich, UK). Transcripts were assembled using this reference sequence. TaLEMK1.1 (clone 7) and TaLEMK1.2 (clone 12) were used as query sequences in BlastN searches against the assembled transcripts of the six winter wheat varieties.

Transient-induced gene silencing constructs were generated and transferred by particle bombardment into leaf epidermal cells of 7-day-old barley seedlings as described (Douchkov et al., 2005). Leaf segments were inoculated 3 days after the bombardment with Bgh at a density of 180–200 conidia mm^-2. Transformed GUS-stained epidermal cells as well as haustoria-containing transformed (susceptible) cells were counted 48 h after inoculation, and TIGS effects on the susceptibility index (SI) were statistically analyzed as described in (Spies et al., 2012). For transient (over)expression, the HvLEMK1-containing BAC clone HvMRXALLhA0027N11 was bombarded into leaf segments of barley cv. Maythorpe or wheat cv. Kanzler, followed by challenge inoculation with the corresponding adapted pathogen Bgh or Bgt at a density of 180–200 conidia mm^-2 4 h after the bombardment and microscopic assessment of SI 48 h after inoculation. For the verification of HvLEMK1 transgene effects, a 17.4 Kb HvLEMK1-containing BAC sub-clone (GenBank Acc. KR610392) was excised as StuI/SphI fragment and inserted into SmaI/SphI sites of transient expression vector pIPKTA09. In addition, barley and wheat cDNAs were excised and inserted as NotI fragment into to the multiple cloning site of pIPKTA09 (Zimmermann et al., 2006). The resulting sequence-verified constructs were bombarded into barley or wheat as described for BAC clones. Relative SI to empty BAC clone or to empty pIPKTA09 was calculated, log(2) transformed and tested for statistically significant deviation from the hypothetical control value “0” by a 1-sample t-test, 2-tailed.

Barley cv. Golden Promise was transformed with pIPKb009_HvLEMK1 (Himmelbach et al., 2007) as described (Hensel et al., 2008). Phenotypic evaluation of Bgh and Bgt interactions was done microscopically on second, detached leaves of 12–14 day-old T1 plants placed on phytoagar plates (23,2 cm × 23,2 cm) and inoculated at a spore density of 30–40 conidia mm^-2. Leaf segments inoculated with either Bgh or Bgt were incubated strictly separated from each other in order to prevent cross-contamination. Golden Promise azygous T1 segregants served as internal negative controls. Inoculated leaf segments were incubated for 48 h (Bgh) or 7 days (Bgt) followed by staining with Coomassie brilliant blue R 250 (Schweizer et al., 1993). The number of growing Bgt colonies/leaf area was counted manually under a standard light microscope at 100× magnification. In case of Bgh, colonies were counted using the HyphArea software (Baum et al., 2011). Because of variability of residual Bgt susceptibility of the azygous control plants between the different inoculation experiments, we normalized the number of Bgt colonies to the average number of Bgt colonies on the azygous control in the corresponding experiment. Statistics: For Bgt susceptibility, non-normally distributed values from transgenic- or azygous control plants (a combined pool of null-allelic segregant plants from all T1 families) were subjected to Mann–Whitney test (2-tailed) for significant differences from the wildtype control. Data were obtained in three independent inoculation experiments. In the case of Bgh infection, data were normally distributed and tested by non-paired t-test against azygous control plants (2-tailed). Azygous control plants were used here as control because they were significantly more susceptible than wildtype plants, for unknown reasons. Data were obtained in two independent inoculation experiments.

To determine HvLEMK1 transcript amounts in transgenic plants, total RNA was isolated from the 4th leaf using the RNeasy Plant Mini Kit with on-column DNase digestion (Qiagen, Hilden, Germany). Two micrograms of total RNA from 11 to 18 transgenic T1 progeny plants per primary transformant, or from a total of 32 azygous segregants, were reverse-transcribed using iScript^TM cDNA Synthesis Kit (Bio-Rad Laboratories, Inc.). Transcripts of HvLEMK1 were quantified in triplicates by using TaqMan probes in a reaction volume of 10 μL (Maxima Probe qPCR Mastermix;Thermo Fisher Scientific, Waltham, MA, USA). Amplification and detection of fluorescent signal was performed in three technical replicates per cDNA sample on a 7900 HT Fast Real-Time PCR system (Life Technologies/Applied Biosystems, Darmstadt, Germany). To determine transcript amounts of proposed non-host-resistance marker genes, plants were grown in 22.5 cm × 18 cm pots in a plant incubator (Panasonic, Hamburg, Germany) at 20°C constant temperature, 60% relative humidity and 16 h illumination (intensity level 5) provided by daylight fluorescent tubes. Twelve-day-old plants were inoculated with Bgt spores (20–30 conidia/mm²), and second leaves were collected at respective time points for RNA isolation. Primary leaves were used for the hygromycin assay to identify azygous plants. RNA isolation, cDNA synthesis and qRT-PCR were done as mentioned earlier except a BRYT Green^® dye (Promega Corporation, USA) based system was used instead of TaqMan probes. Primer sequences for all PCR reactions are provided in Supplementary Table S1. Thermal cycling conditions consisted of initial denaturation at 95°C for 10 min followed by 30 cycles of (95°C/15 s, 55°C/40 s, 72°C/35 s). Ubiquitin conjugating enzyme 2 (HvUBC, Acc. AY220735) was used as internal normalization standard. To quantify the transcript levels in each sample, a standard curve for each gene with a serial dilution series was made from pooled RNA samples with three technical replicates each. Transcript quantities were determined using the SDS.2.4 software (Life Technologies GmbH, Darmstadt, Germany). For both HvLEMK1- and PTI-marker transcript quantification, two independent biological replicates were analyzed using plants grown on different dates.

For subcellular localization, the full-length sequence of HvLEMK1 was N- and C-terminally fused in-frame to yellow fluorescent protein (YFP) gene in pIPKTA48 and pIPKTA49 vectors, respectively, using Gateway^® cloning technology (Thermo Fischer Scientific, New York, NY, USA) (Supplementary Figure S1; Supplementary Table S1). Resulting YFP-fusion constructs were transiently expressed in 7-day-old leaf segments of barley cv. Golden Promise by particle bombardment and examined after 24 h of incubation without B. graminis inoculation using confocal laser scanning microscopy. The plasma-membrane marker aquaporin (AtPIP2A) of Arabidopsis thaliana (plasmid pm-rk CD3-1007), or the endoplasmic-reticulum (ER) marker Wall-associated kinase 2 (AtWAK2; plasmid ER-rk CD3-959), both fused to mCherry as described in Nelson et al. (2007), were mixed in an equimolar concentration and co-bombarded for co-localization experiments. For plasmolysis of epidermal cells, leaf segments were floated on 15% glycerol for 5–10 min, immediately prior to microscopy.

Seven-day-old barley plants of cv. Vada or wheat plants of cv. Renan were inoculated with Bgh or Bgt, and the abaxial epidermis of inoculated primary leaves or from non-inoculated control leaves was peeled at 6–74 h after inoculation, as described (Zellerhoff et al., 2010; Spies et al., 2012). Total, quality-controlled RNA was hybridized to Agilent Gene Expression 44K microarrays (design ID: 020599 for barley and 022297 for wheat; Agilent) as described (Chen X. et al., 2011). Single-channel array processing was utilized followed by data normalization with default parameters, and significant transcript-regulation events were determined by using GeneSpring GX (v11.5.1) software (Agilent Technologies, Waldbronn, Germany). Transcripts were assumed to be significantly regulated if p-values corrected for false-positive rate (FDR, Benjamini-Hochberg method) were smaller than 0.05 and if regulation factors between inoculated and corresponding control samples harvested in parallel exceeded 2.0. All quantile-normalized signal intensities of the analyzed candidate genes are shown in Supplementary Table S1, and the raw data from the corresponding array slides were deposited at ArrayExpress [Accession Nr. E-MTAB-2916 for barley and E-MTAB-3803 for wheat]. Array procedures followed MIAME guidelines throughout (Brazma et al., 2001).

For the analysis of PAMP-induced TaLEMK1 expression, wheat leaf segments were placed into 2 ml tubes submerged in H₂O and pre-infiltrated by vacuum three times for 45 s. Leaf segments were left to recover in the growth cabinet for 16 h to avoid gene expression response to water infiltration. After this time, water was replaced by fresh water (control) or 1 mg/ml chitin (Yaizu Suisankagaku Industry Co., Ltd). Samples were drained, snap-frozen in liquid N₂ at different time points and stored at -80°C. RNA was extracted using Qiagen RNAeasy Plant kit followed by removal of genomic DNA by DNase Turbo DNA-Free (Ambion). For quantitative real-time PCR (qPCR) analysis, first-strand cDNA was synthesized from 1 μg of total RNA using the SuperScript III First Strand Synthesis System. Quantitative RT-PCR (RT-qPCR) analysis of TaLEMK1, TaCMPG1-like and TaPub23-like was conducted as described in Tufan et al. (2009) using gene-specific primers (Supplementary Table S1).

Domain delineation, LRR repeat delineation, sequence analysis, and molecular modeling of the three domains of HvLEMK1 protein core were performed as described in Sela et al. (2012) and Slootweg et al. (2013). For model refinement an iterative procedure consisting in global simulated annealing with harmonic restraints on backbone atoms found in definite secondary structure states, followed by model quality assessment with MetaMQAP (Pawlowski et al., 2013). In order to optimize the HvLEMK1-LRR scaffold the Joint Fragment Remote Homology Modeling procedure (Sela et al., 2012; Slootweg et al., 2013) had to be used. The resulting 3D model was brought within 2.8 Å RMSD by repeated rounds of local remodeling from an optimal polypeptide path and a GDT_TS score of 60 according to MetaMQAP.

Sequence propensity analysis resulted in the following template selection for the three globular domains of (a) HvLEMK1-LRR: the LRR domain from FLS2 (PDB code: 4MN8) with sequence identity = 31%, confidence = 99; (b) HvLEMK1-malectin: the malectin from Xenopus laevis (PDB code: 2KR2) with sequence identity = 22%, confidence = 100; and (c) HvLEMK1-kinase: the kinase domains from BAK1 (PDB code: 3TL8) with sequence identity = 45%, confidence = 100 and from IRAK1 (PDB code: 2NRU) with sequence identity = 40%, confidence = 100. In addition the Joint Fragment Remote Homology Modeling procedure (Sela et al., 2012; Slootweg et al., 2013) was used to optimize the scaffold for Rnr8-LRR domain generation.

The Rnr8 gene encodes for an LRR/malectin-domain RLK designated HvLEMK1 that contains also a transmembrane and a C-terminal full-length protein-kinase domain (Figure 1A). Transient silencing of HvLEMK1 was reported to weaken non-host resistance of barley to the wheat powdery mildew fungus Bgt, and therefore we searched for the presence of an orthologous protein in wheat, which is the natural host for Bgt (Douchkov et al., 2014). By PCR we identified two genes, TaLEMK1_1 (Acc. KX529076) and TaLEMK1_2 (Acc. KX529077), differing by five predicted amino acids in the LRR domain (Figure 1B; Supplementary Figure S2). Comparison of the full length sequences of TaLEMK1 to the wheat chromosomes on URGI database³ revealed that the gene is located on the group 5 chromosomes, in syntenic position to the barley orthologous gene (Douchkov et al., 2014). Although exhibiting 94% nucleotide identify to HvLEMK1, evolutionary separation of the genus Hordeum and Triticum approximately 12 MY ago led to the accumulation of a total of 31 non-conservative amino-acid exchanges between the wheat and barley orthologs, which corresponds to approximately 3% of the protein (Supplementary Figure S2). Extending the search for close homologs of HvLEMK1 in more distantly related plants revealed the existence of HvLEMK1-like proteins in all analyzed species, and the conservation of the intron/exon structure suggested the existence of true orthologs (Figure 1C; Supplementary Figure S3). TaLEMK1_1 and TaLEMK1_2 were most closely related to genes on hexaploid wheat chromosome 5A and 5D, respectively, suggesting that they represent homeologs from the A and the D genomes of wheat. Because some of the non-Triticum plant species are not infected by any known powdery mildew fungus, HvLEMK1 might function in sensing stress-related endogenous signals or highly conserved PAMPs present in different pathogens, or act as co-receptor in RLK complexes.

The putative ligand-binding LRR domain of HvLEMK1 has a homogeneous repeat length of 24. In order to put differences between the orthologous barley and wheat proteins into a structural context, a three-dimensional (3D) model was generated starting from the LRR domain of FLS2, a RLK protein from Arabidopsis, which contains repeats of the same length as HvLEMK1 (Figure 1D). Variable residues between HvLEMK1-LRR and TaLEMK1_2-LRR were mapped onto the 3D structure. The overall set of variable amino acids is shown here in red and magenta. Of these only those shown in magenta are mutations that might alter locally the structure and/or the interaction potential of the LRR domain. All these variable sites are clustered in the C-terminal region of the LRR domain pointing to a possible interaction surface with pathogen- or plant interaction partners. More specifically this cluster is located on the backside of repeats 8, 9, and 10 (Figure 1E). The malectin domain presents a conserved ordered beta core and a more unstructured N-terminal subdomain which hosts the lectinic site involved in glycan recognition of the malectin prototype protein in animals (Supplementary Figure S4A) (Chen Y. et al., 2011). Due to low sequence identity between HvLEMK1-malectin and the template protein, the 3D structure is still a rough model, with a current RMSD of 4 Å and a GDT_TS of ∼50. Finally, the HvLEMK1-kinase represented in Supplementary Figure S4B shows high sequence similarity with both plant RLK Bak1 and with human IRAK1, not only in the secondary structure patterns but also in the enzymatic active-site region. The model was brought within less than 3 Å RMSD from an optimal polypeptide path with an overall quality GDT_TS score of 58.

The encoded HvLEMK1 protein contains a transmembrane domain between malectin- and kinase domains, suggesting plasma-membrane localization, as has been shown for other RLKs. To verify this prediction, HvLEMK1 cDNA was fused with its 3′-end to YFP and bombarded into barley epidermal cells. The resulting C-terminal fusion protein was tested for co-localization with an aquaporin:mCherry fusion protein that was used as plasma membrane marker, and with a AtWAK2:mCherry fusion protein used as marker for localization to the endoplasmic reticulum (ER) (Figure 2). This experiment revealed co-localization of HvLEMK1 with the plasma membrane marker in turgescent- and in plasmolyzed cells, and also in Hechtian strands (Figures 2a–h). In addition, we could co-localize a fraction of the protein with the ER marker (Figures 2i–l). It is known that plasma membrane-localized RLKs such as the malectin-domain protein Ferronia are retained in the ER, through which they are delivered to the cell periphery, if folding and quality-control chaperones in the ER are limiting or missing (Li C. et al., 2016b). Therefore ER-localization of part of the HvLEMK1:YFP protein pool may reflect overloading of the secretory system by transient over-expression under the control of the strong CaMV 35S promoter. However, because malectin is known to be ER-localized (Chen Y. et al., 2011) we cannot exclude a dual role of HvLEMK1 in both compartments. Finally, another fraction of HvLEMK1:YFP appeared to be localized to the nucleus, without a ring as seen with the mCherry ER marker (inset in Figure 2k) This unexpected result therefore awaits further functional examination.

Although RLKs are expected to exhibit a basal level of expression in order to fulfill their function in signal-perception, we often found their transcripts to be up-regulated upon pathogen attack of barley (Douchkov et al., 2014). We therefore interrogated two transcript-profiling datasets of leaf epidermal peels from barley and wheat challenged with Bgh and Bgt, which corresponds to a reciprocal set of host- and non-host interactions, by using the Agilent 44K Gene Expression Arrays of the two species. As shown in Figure 3, both HvLEMK1 and TaLEMK1 were up-regulated upon powdery-mildew attack. Interestingly, Bgh triggered maximum transcript levels at 6 h after inoculation in both plant species whereas the maximum response to Bgt attack was relatively delayed by 6 h. Therefore the different induction kinetics appears to be determined by the behavior of the respective pathogen (race) such as the kinetics of PAMP release, rather than by the host-vs. non-host status of barley or wheat.

We stably transformed barley with an RNAi construct against HvLEMK1 and identified 24 primary transgenic plants carrying intact versions of the silencing cassette. These were tested in the T1 generation for an altered interaction phenotype with Bgt (Supplementary Figure S5). T1 families with enhanced colony- or conidiospore formation of the non-adapted fungus in the first experiment were selected for subsequent tests in two more inoculation experiments, and for the quantification of target-transcript levels (Figures 4A,B). The susceptibility phenotype to Bgt could be confirmed in a statistically significant manner for four T1 families, and these also showed significant reduction of HvLEMK1 transcript levels. On the other hand, we observed no difference of quantitative host resistance against Bgh in any of these four families (Figure 4C) suggesting no effect of HvLEMK1 in quantitative host resistance. The microscopic inspection of Bgt development 7 days after inoculation (d.a.i.) of HvLEMK1-silenced leaves revealed the presence of colonies that, however, remained smaller than those of Bgh from parallel inoculations and were not visible to the naked eye (Figure 4D). Importantly, conidial chain formation of Bgt (highlighted by red arrows) was found in all four selected RNAi T1 families but never in the rare micro-colonies that were formed by Bgt on wildtype or azygous control plants (Figures 4D,E). We conclude that silencing of HvLEMK1 resulted in a partial breakdown of non-host resistance to Bgt. Because the HvLEMK1-silenced plants showed no visible growth abnormalities or stress symptoms (Supplementary Figure S6) we assume that the HvLEMK1 gene does not have essential developmental or housekeeping functions.

We characterized transcriptional induction of five proposed marker genes of non-host resistance in the RNAi transgenic families. These genes were selected because their transcripts were more strongly induced at 6–12 h after inoculation in non-host resistant- compared to host susceptible interactions of barley and wheat with B. graminis. By contrast, they showed no differential induction between barley lines carrying the Mla6 major R-gene or being mutated at the locus (mla6) (Supplementary Figure S9; Supplementary Table S2). As shown in Figure 5, three of the five candidate markers were less induced in HvLEMK1-silenced T1 populations at 12 h after Bgt-inoculation, compared to azygous segregant plants. At 24 h after inoculation the difference between azygous and transgenic plants had disappeared.

We next tested the option that heterologous expression of the barley ortholog in wheat and vice versa might confer non-host-like resistance across species borders due to, for example, reduced functional suppression by one or several effectors from the corresponding non-adapted powdery mildew pathogen (Deslandes and Rivas, 2012; Feng et al., 2012; Mentlak et al., 2012). Bombardment of wheat-leaf segments with either a BAC clone carrying the HvLEMK1 gene, the coding part of the gene under the control of the CaMV 35S promoter, or a full-coding sequence cDNA clone significantly enhanced resistance, as compared to a control BAC clone containing no annotated genes or to the empty over-expression vector pIPKTA9 (Figure 6). Transient over-expression of the two closely related homoeo-alleles in wheat revealed a protective effect of TaLEMK1_2 comparable to the barley gene, whereas TaLEMK1_1 showed no significant effect. One possible explanation to this observation is that the TaLEMK1_2 but not TaLEMK1_1 protein is free from functional suppression by a hypothetical Bgt-encoded effector. Attempts to transiently silence TaLEMK1_1 and TaLEMK1_2 for further clarifying their function in quantitative host resistance to Bgt and non-host resistance to Bgh failed because the bombarded wheat-leaf segments built up a strong resistance to Bgt within the minimum required incubation period of 2 days after bombardment that was required for effective TIGS. This bombardment-induced resistance, possibly due to mechanical stress, resulted in an average haustorial index of 0.006 from five independent experiments using the empty vector control. Stable transgenic RNAi lines of wheat would therefore be required to address the role of TaLEMK1 genes in non-host resistance. In summary, at least the TaLEMK1_2 protein appears to quantitatively contribute to host resistance in wheat while its role in non-host resistance is currently unclear. In the light of the enhanced resistance caused by over-expression of endogenous TaLEMK1_2 in wheat, the partial protection observed after transient expression of the heterologous HvLEMK1 transgene might also be due to its over-expression rather than transgenomic functional complementation, even when bombarded as a BAC clone under the control of its own promoter.