The leucine-rich repeats in allelic barley MLA immune receptors define specificity towards sequence-unrelated powdery mildew avirulence effectors with a predicted common RNase-like fold

Nucleotide-binding domain leucine-rich repeat-containing receptors (NLRs) in plants can detect avirulence (AVR) effectors of pathogenic microbes. The Mildew locus a (Mla) NLR gene has been shown to confer resistance against diverse fungal pathogens in cereal crops. In barley, Mla has undergone allelic diversification in the host population and confers isolate-specific immunity against the powdery mildew-causing fungal pathogen Blumeria graminis forma specialis hordei (Bgh). We previously isolated the Bgh effectors AVRA1, AVRA7, AVRA9, AVRA13, and allelic AVRA10/AVRA22, which are recognized by matching MLA1, MLA7, MLA9, MLA13, MLA10 and MLA22, respectively. Here, we extend our knowledge of the Bgh effector repertoire by isolating the AVRA6 effector, which belongs to the family of catalytically inactive RNase-Like Proteins expressed in Haustoria (RALPHs). Using structural prediction, we also identified RNase-like folds in AVRA1, AVRA7, AVRA10/AVRA22, and AVRA13, suggesting that allelic MLA recognition specificities could detect structurally related avirulence effectors. To better understand the mechanism underlying the recognition of effectors by MLAs, we deployed chimeric MLA1 and MLA6, as well as chimeric MLA10 and MLA22 receptors in plant co-expression assays, which showed that the recognition specificity for AVRA1 and AVRA6 as well as allelic AVRA10 and AVRA22 is largely determined by the receptors’ C-terminal leucine-rich repeats (LRRs). The design of avirulence effector hybrids allowed us to identify four specific AVRA10 and five specific AVRA22 aa residues that are necessary to confer MLA10- and MLA22-specific recognition, respectively. This suggests that the MLA LRR mediates isolate-specific recognition of structurally related AVRA effectors. Thus, functional diversification of multi-allelic MLA receptors may be driven by a common structural effector scaffold, which could be facilitated by proliferation of the RALPH effector family in the pathogen genome.


Introduction
Plants have evolved sophisticated innate immune systems to protect themselves against colonization by pathogenic microbes [1,2]. At the population level, a host-adapted pathogenic species is comprised of numerous isolates/races with distinctive genetic blueprints which determine their infection phenotypes on individual accessions (genotypes) of a plant host population. In isolate-specific resistance, individual host accessions often mount a hypersensitive immune response against a subset of pathogenic isolates [3,4]. Isolate-specific resistance is mediated by genetic interactions between plant host resistance (R) genes and matching pathogen avirulence (AVR) effector genes (gene-for-gene model) [5]. Plant R genes often encode intracellular nucleotide-binding domain leucine-rich repeat-containing receptors (NLRs) [6]. These receptors have a characteristic modular domain architecture, consisting of a variable N-terminal Coiled-Coil (CC), Toll-Interleukin (TIR) domain or HeLo domain (named after the fungal HET-S and LOPB proteins), a central NB-ARC (nucleotide-binding adaptor shared by APAF-1, certain R gene products, and CED-4) domain, and C-terminal leucine-rich repeats (LRRs) [7,8]. The LRRs often constitute a determinant for specific pathogen recognition. NLRs can detect AVRs by direct interaction [9][10][11], a receptor-integrated decoy [12], or indirectly detecting effector-mediated alterations of a host target [13]. Upon AVR recognition by NLRs, a localized host cell death is typically, but not invariably, associated with receptor-mediated immunity.
CSEPs, the isolated AVR A effectors also lack the residues critical for catalytic activity [30,31]. By taking advantage of previously engineered hybrid receptors of MLA1 and MLA6 we confirm that the molecular basis of isolate-specific disease resistance against Bgh isolates A6 and K1 lies in the specific recognition of AVR A6 and AVR A1 effectors by six and 12 C-terminal LRRs of MLA1 and MLA6 receptors, respectively. We find that the LRRs of allelic MLA10 and MLA22 are largely sufficient for specific perception of allelic AVR A10 and AVR A22 effector proteins. Co-expression of hybrid effectors generated from allelic AVR A10 and AVR A22 with MLA10 and MLA22 receptors revealed that multiple effector residues participate in AVR A recognition specificities. Our findings imply a model in which co-evolution of the barley Mla-Bgh AVR a pathogen interaction is driven MLA sequence diversification upon detection of a common structural effector scaffold. This co-evolutionary process may have contributed to the proliferation and sequence diversification of RALPH effectors in the powdery mildew genome.

TWAS identifies BLGH_00709 (CSEP0254) as an AVR a6 candidate
For the isolation of AVR a6 , we examined the gene-wise association of Bgh transcriptomes with the published infection phenotypes of 27 Bgh isolates on Mla6 barley lines [11,28]. We used the previously described in planta fungal transcripts of the 27 Bgh isolates and their published infection phenotypes on barley Mla6 near-isogenic lines (NILs) of the barley cultivars (cv.) Pallas and Manchuria [11]. In short, we integrated high-confidence non-synonymous variants over each annotated Bgh gene and considered absence of a transcript as a missing genotype to obtain gene-wise genotypes. We then tested the gene-wise genotypes for association with the observed avirulence phenotypes using Fisher's exact test [11]. Association mapping identified a number of genes encoding CSEPs. The csep encoding genes with the lowest p-values in this analysis were BLGH_00709 (CSEP0254; gene-wise: p = 7,40E-07) and BLGH_00697 (p = 7.25E-07), suggesting them to be top-ranking candidates for AVR a6 (Fig 1A, S1 and S2 Tables). To first determine which candidate is recognized by MLA6 in barley, we transiently co-expressed the AVR a candidate gene and Mla6 receptor gene in cv. Golden Promise barley protoplasts. Co-expression of matching AVR a -Mla pairs in this system triggers a reduction in luciferase (LUC) reporter gene activity as a proxy for cell death [33]. Co-expression of BLGH_00709 and Mla6 led to a significant reduction of relative LUC activity in comparison to the empty vector (EV) control (98% reduction), while co-expression of BLGH_00697 with Mla6 did not reduce LUC activity (S1 Fig). This suggests that BLGH_00709 is recognized by Mla6 in barley, and we therefore subjected BLGH_00709 to further analysis as the top AVR a6 candidate.
First, we analyzed the gene architecture of the AVR a6 candidate gene in the available genomes of the Bgh isolates DH14 and RACE1 (both avirulent on Mla6 lines). The annotated near chromosome-level reference genome of the DH14 Bgh isolate harbors two more identical copies of BLGH_00709. These copies are annotated as BLGH_00708 and BLGH_07091 ( Fig  1B). Additionally, DH14 harbors another CSEP0254 paralog, called BLGH_07092. In comparison to BLGH_00709, BLGH_07092 carries a frameshift mutation that predicts an altered sequence from aa 64 onwards in the BLGH_07092 encoded protein (Fig 1D). BLGH_00708 and BLGH_00709 are located in close proximity to each other in a head-to-tail orientation next to the cyclin B1 gene on scaffold 16 of the DH14 genome, while BLGH_07091 resides with BLGH_07092 on scaffold 309 in a head-to-head orientation. In the genome of the Mla6 avirulent RACE1 isolate, three identical copies of BLGH_00709 can be found on tig00005311: BLGHR1_15960 (syntenic position to BLGH_00709) is located next to the cyclin B1 gene and on the same scaffold BLGHR1_15970 (syntenic position to BLGH_07091) and BLGHR1_15971  a6 . The x axis represents the Bgh DH14 genes, sorted by Bgh gene ID; the y axis shows −lg of p-values for all genes with at least one nonsynonymous coding SNP, indels as well as presence or absence of transcripts. CSEPs with a p < 0.018 (dotted line) are depicted by arrowheads. The candidate AVR a6 gene copies BLGH_00709 (CSEP0254) and BLGH_07091 are designated in the plot with bright green arrowheads. BLGH_07092 is designated by a dark green arrowhead. The other candidates BLGH_00697 (CSEP0058) and BLGH_00700 are depicted with a dark red and a bright red arrowhead, respectively. (B) Schematic illustration of the chromosomal regions harboring the AVR a6 candidate BLGH_00709 and its paralogues and family members with corresponding gene IDs in the genomes of Bgh isolates DH14, RACE1, and K1. All CSEPs are depicted by arrows. (C) Phylogeny of CSEP family 8 containing AVR A6 , which can be divided into clade 1 (BLGH_00709, BLGH_07092, BLGH_00700, BLGH_00698, BLGH_00697) and clade 2 (BLGH_05882, BLGH_05875, BLGH_05881), based on the protein sequences excluding the signal peptide and using BLGH_05397 as an outgroup. (D) Protein sequence alignment of AVR A6 and CSEP family 8 members including their respective signal peptides. https://doi.org/10.1371/journal.ppat.1009223.g001

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold (syntenic position to BLGH_07092) reside in a head-to-head orientation. Further analysis of genes surrounding the AVR a6 locus in DH14 and RACE1 suggests major genomic rearrangements: a gene cluster containing a tRNA nucleotide transferase and the DNA repair protein RadR is present in inverted orientation in scaffold 16 and tig00005311, and the BLGH_00700 family member cannot be found on tig0005283 of the RACE1 genome ( Fig 1B). For simplicity, from here on we will refer to the three sequence-identical paralogues BLGH_00709, BLGH_00708, and BLGH_07091 as AVR a6 . AVR A6 is part of the CSEP family 8, which contains six additional members and can be subdivided into two clades: AVR A6 , BLGH_07092, BLGH_00698 (CSEP0333), BLGH_00700, and BLGH_00697 (CSEP0058) belong to clade 1, whereas BLGH_05881 (CSEP0151), BLGH_05875 (CSEP0147), and BLGH_05882 (CSEP0148) belong to clade 2 ( Fig 1C). The clade 1 CSEP family 8 members share a 48.3-60.4% sequence identity with AVR A6 , while the clade 2 family members share a 32.9-39.7% sequence identity with the avirulence effector candidate. As BLGH_07092 is most likely a nonfunctional copy of AVR a6 , and its expression is lower when compared to AVR a6 in every Bgh isolate (S2 Fig), we did not subject it to further analysis.
Analysis of transcriptomic data revealed that all isolates avirulent on the cv. Manchuria and cv. Pallas Mla6 NILs express AVR a6 , which encodes a 115-aa-long protein with a predicted 24-aa-long signal peptide (SP) (S3 and S4 Figs). AVR a6 possesses one intron, which is spliced out in all transcripts of avirulent isolates. However, we identified transcripts of AVR a6 carrying this intron in all Bgh isolates virulent on Mla6 NILs. The transcription of the AVR a6 intron may be facilitated by two different mechanisms (S4 Fig): The AVR a6 transcript variant expressed in the virulent isolates CC66 and CC148 exhibits a T270C mutation in the intron branch point consensus sequence, suggesting that intron retention may be caused by inefficient or nonexistent U2 spliceosome recognition (S4B Fig). If this is the case, the intron retention leads to a premature stop codon and a truncated protein, which is only 79 aa-long including the signal peptide. This deduced truncated protein variant, which we named AVR A6 -V1, would also harbor A48S, A75T, and R79S amino acid substitutions. The isolates K1, K2, K3, B103, S15, S16, S22, and S25 are virulent on Mla6 lines and contain a splice donor site mutation (S3 and S4B Figs) in the transcripts that maps to the AVR a6 gene of the reference genome. Genome analysis of isolate K1 [11] confirmed this splice-site mutation (Fig 1B). The predicted intron retention in the transcript of the AVR a6 variant expressed by the K1, K2, K3, B103, S15, S16, S22, and S25 isolates leads to a premature stop codon as well as to a predicted 79-aa-long truncated protein (S4B Fig). In addition, the encoded protein exhibits two amino acid substitutions when compared to AVR A6 : A75T and R79I and we named this variant AVR A6 -V2. Furthermore, the number of transcripts from virulent Bgh isolates that map to AVR a6 is approximately four-fold lower than the AVR a6 transcripts in the avirulent DH14 isolate (S2 Fig), which could be a consequence of nonsense-mediated mRNA decay. In conclusion, virulence of Bgh isolates on barley NILs harboring Mla6 is likely conferred by SNPs in splice sites that may lead to intron retention in the respective genes. This is associated with reduced levels of the transcripts that map to the AVR a6 gene in Bgh isolates virulent on Mla6 lines.

Transient co-expression assays provide evidence for specific recognition of AVR a6 by Mla6
To determine if the candidate AVR A6 is specifically recognized by MLA6, we first co-expressed AVR a6 and Mla6 in a transient barley protoplast system containing leaf mesophyll cells [33]. Co-expression of matching AVR a6 or AVR a1 with Mla6 or Mla1, respectively, triggered significant reductions in LUC activity, when compared to reference samples where the effector gene has been exchanged to an EV (98% and 85% reductions, respectively, Fig 2A). Co-expression Golden Promise protoplasts were transfected with pIPKb002 vectors containing cDNAs of Mla6 or Mla1 and either an empty vector (EV), AVR a6 , AVR a6 -V2 , AVR a1, AVR a1-V1, or AVR a1 -V2 variants lacking their respective signal peptides together with a pUBI:Luciferase construct. The LUC activity relative to the EV sample was measured as a proxy for cell death 16 h post transfection. Box plot diagrams show median of the relative LUC activity of six independent transfections, which are represented by dots, while the box shows the interquartile range. Significant differences between samples were analyzed using non-parametric Kruskal-Wallis (KW) analysis followed by a Dunn's test. Calculated KW p-values are as follows: Mla6: p = 0.007146; Mla1: p = 0.0007392. Samples labeled with identical letters did not differ significantly (p < 0.05) in the Dunn's test for the corresponding Mla variant. (B) cDNAs of clade 1 AVR a6 family members BLGH_00698, BLGH_00697, BLGH_00700, AVR a6 , and AVR a1 variants were expressed without their respective signal peptides and stop codons and with a C-terminal mYFP fusion under the control of a 35S promotor in N. benthamiana. The effectors were co-expressed with Mla1 and Mla6 cDNAs fused Cterminally with a 4xmyc tag under the control of a 35S promotor. Cell death was scored five days post infiltration and of Mla6 with a DNA sequence that encodes a truncated version of AVR a6 (likely encoded by AVR a6 -V2) or co-expression of AVR a6 with Mla1 did not lead to significantly reduced LUC activity compared to samples co-expressing AVR a6 with Mla6, confirming the specificity of the recognition (Fig 2A).
To examine whether AVR A6 is recognized by MLA6 in a heterologous expression system without the presence of other barley proteins, we co-expressed the C-terminally mYFP-tagged effector fusion protein with the C-terminally 4xmyc-tagged receptor in N. benthamiana. Unlike the essentially complete cell death observed in the barley protoplast system (as evidenced by the very low levels of LUC activity), co-expression of AVR a6 -mYFP and Mla6-4xmyc triggered a cell death of varying confluence in Agrobacterium tumefaciens infiltrated tissue in independent N. benthamiana leaves compared to the cell death observed when coexpressing AVR a1 and Mla1 (Figs 2B and S14). Co-expression of AVR a6 -mYFP with Mla1-4xmyc did not elicit cell death, confirming the specific recognition of candidate AVR A6 by MLA6 but not MLA1. No cell death was observed when Mla6-4xmyc was co-expressed with a DNA sequence that encodes a truncated version of AVR a6 -mYFP (here named AVR A6 -V2-mYFP), even though both AVR A and MLA proteins are detectable in N. benthamiana leaf extracts (Figs 2B, 2C and 5C). Furthermore, co-expression of the clade 1 CSEP family 8 member BLGH_00698-mYFP, which shares the highest sequence similarity with AVR a6 , with Mla6 did not lead to cell death, even though BLGH_00698-mYFP is detectable in N. benthamiana leaf extracts (Fig 2B and 2C). Detection of BLGH_00700 and BLGH_00697 proteins was possible only after enrichment with a GFP-Trap, suggesting that their protein stability is lower than those of AVR A6 and BLGH_00698 in this system. A faster-migrating protein band for BLGH_00697-mYFP and a double band visible after blotting for BLGH_00700-mYFP suggest that these proteins may either be cleaved post-translationally by proteases in heterologous N. benthamiana or that they are not stable in the plant extraction buffer (Fig 2C). Taken together, co-expression of AVR A6 with MLA6 in both homologous and heterologous plant expression systems triggers a significant and specific cell death response, indicating specific effector recognition by the matching MLA immune receptor.

AVR A effectors have low sequence similarity, but show predicted structural homology to RNases
We subjected AVR A6 to a phylogenetic analysis including all annotated CSEPs in B. graminis formae speciales poae, lolium, avenae, tritici (isolate 96224), hordei DH14, secalis (isolate S1459), triticale (isolate T1-20), and dactylidis, but were unable to detect significant polypeptide sequence relatedness to other known Bgh AVR A effector proteins or to the so far isolated wheat powdery mildew avirulence effectors, AVRPM2, AVRPM3 A2/F2 , AVRPM3 B2/C2 , and AVRPM3 D3 (Fig 3A). However, we noted that all the avirulence proteins isolated from barley and wheat powdery mildews belong to CSEPs with a length of approximately 80 to 130 amino acids when neglecting their respective signal peptides. The same is true for CSEP0064, which was shown to form a RNase-like protein structure. To determine potential structural similarity between AVR A6 and known Bgh effectors, we subjected AVR A6 to structural prediction using IntFOLD version 5.0 [32]. AVR A6 exhibited high predicted structural similarity to the RNase-Figures show a representative of at least 15 co-transformations. (C) Protein levels of AVR A1 -mYFP, AVR A6 -mYFP, AVR A6 -V2-mYFP, BLGH_00698-mYFP, BLGH_00697-mYFP and BLGH_00700-mYFP. Samples for total protein extraction were harvested two days post infiltration. mYFP fusion proteins were enriched by an GFP-Trap. Proteins were separated using 10% or 12% polyacrylamide gels and proteins were detected using α-GFP and α-myc western blotting (WB). IP = immunoprecipitation. CBB = Coomassie Brilliant Blue. https://doi.org/10.1371/journal.ppat.1009223.g002

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold like fold observed for the X-ray structure, suggesting that AVR A6 possesses a ribonuclease-like fold similar to AVR A13 and AVR A7 (S5 Fig). This prompted us to also reanalyze all isolated AVR A effectors using the with IntFOLD version 5.0. We found that AVR A1 , AVR A9 , and AVR A10 /AVR A22 are also predicted to harbor a central α-helix directly facing three to four βsheets with a topology characteristic of ribonucleases ( Fig 3B). This is reminiscent of structural predictions for Bgt effectors AVRPM2 and AVRPM3 D3 , which also suggested structural similarities to ribonucleases (Fig 3, [22,[34][35][36]). Although the relationship to ribonucleases is less clear for Bgt AVRPM3 A2/F2 and AVRPM3 B2/C2 , these two effectors were predicted to exhibit a central α-helix and two to four ß-sheets like AVRPM3 D3 ( [22], Fig 3B). We examined AVR a and AvrPm gene models for the presence of an intron, which is thought to be characteristic for RALPH effector-encoding genes [29,31] and identified one intron at position +163 to +201 bp after the end of the signal peptide (S6A and S7A Figs). The relative intron position in the structural predictions is in an unstructured loop between the first and the second ß-sheet, except in AVR A13 , AVRPM3 A2/F2 , and AVRPM3 B2/C2 (Fig 3B). For the latter three effectors, the intron is positioned in a predicted unstructured loop C-terminal to the second ß-sheet (Figs 3B and S6). No common sequence motifs are detectable in the highly diverse intron sequences (S7B Fig). Thus, it remains unclear whether all known Blumeria avirulence effectors are descendants of one common "ur-RALPH" ancestor [29]. Irrespective of this, our analysis suggests that isolated Bgh and Bgt avirulence effectors are structurally related to fungal ribonucleases.

RNase-like AVR A effectors do not show ribonuclease activity
Using an RNase activity assay, we tested whether AVR A proteins are truly catalytically inactive as suggested for RNase-like Bgh effectors previously [29,30]. First, we expressed N-terminally GST-tagged AVR A6 , AVR A10 , and AVR A13 in Escherichia coli and purified them by GST affinity chromatography. We then cleaved the GST tag and applied size-exclusion chromatography (S8A Fig). Successful protein expression and purification of the AVR A effector proteins was tested by SDS-PAGE (S8B Fig). We then incubated AVR A6 , AVR A10 , and AVR A13 effectors with denatured HvRNA or native rRNA to test for ribonuclease activity [37]. Using RNA gel electrophoresis, we observed a degradation of RNA when co-incubated with a commercially available T1 RNase, which has the same function as the Fusarium F1 RNase. We did not observe RNA degradation when incubating HvRNA or rRNA with the AVR A effectors (Fig 4A  and 4B). These results were independently validated with AVR A6 , AVR A10 , and AVR A13 effector proteins that were produced in eukaryotic insect cells, followed by affinity chromatographic purification (S9A and S9B Fig). Taken together, the data indicate that AVR A effectors have no RNase activity, which is consistent with the previous prediction that ascribed pseudoenzymatic function to the RALPHs [29,30].

The C-terminal leucine-rich repeats of MLA1 and MLA6 receptors account for specific discrimination of structurally homologous AVR A1 and AVR A6 effectors in planta
A previous study showed that most of the residues under positive selection in allelic MLA resistance specificities in barley populations are located in the LRR region [19]. Using singlecell expression of MLA chimeras in barley leaf epidermal cells, C-terminal LRR regions of Mla1 and Mla6 were shown to encode determinants for isolate-specific immunity in barley to Bgh isolates K1 (carrying AVR a1 ) and A6 (carrying AVR a6 ) [38]. Here, we tested if the LRRs determine isolate-specific immunity by specifically recognizing AVR A effectors in barley. Therefore, we made use of the previously constructed intron-containing DNAs of the chimeric receptors M16666, M11166, M61111, and M66111 ( [38] (protein sequence shown in S10 Fig) and co-expressed them with matching AVR a6 , AVR a6 -V2, AVR a1 , and AVR a1 -V1 cDNAs lacking the signal peptide (SP) in the pIPKb002 vector under the control of a strong maize ubiquitin promoter in barley protoplasts. Upon co-expression of AVR a6 with M16666 or M11166 we detected a significant 79% or 92% reduction in relative LUC activity when compared to the EV samples, respectively ( Fig 5A). This reduction was not observed when M16666 or M11166 were co-expressed with AVR a6 -V2, AVR a1 , or AVR a1 -V1, respectively, suggesting that both chimeric receptors specifically recognize their matching effector ( Fig 5A). These findings suggest that the last six C-terminal leucine-rich repeats of a total of 15 deduced LRRs in MLA6 account for the recognition specificity of AVR A6 in barley. Furthermore, we discovered that co-expression of AVR a1 with M61111 or M66111 triggered a significant and specific 92% reduction of relative LUC activity in barley protoplasts indicative of a cell death response compared to the EV sample ( Fig 5A). This suggests that, out of 15 predicted LRRs in the MLA1 receptor, the 12 C-terminal ones contribute to the specific recognition of AVR A1 . Together, these findings corroborate previous experiments that show a significant growth reduction of Bgh isolate A6 when barley cells express M16666 and M11166, or growth reduction of Bgh isolate K1, when barley cells express M61111 or M66111 [38]. While our data does not exclude that other MLA domains contribute to the association with the Bgh AVR A effector proteins, we conclude that the MLA LRR regions confer Bgh recognition specificities to of the different Mla alleles in barley. To test for ribonuclease activity, heterologous AVR A6 , AVR A10 , and AVR A13 proteins, purified upon expression of the respective genes in E. coli, or T1 RNase were co-incubated with (A) denatured HvRNA and (B) native rabbit rRNA always. All samples were separated on non-denaturing 2% agarose gels (top panels) and analysed on a Bioanalyzer (lower panels) to check for RNA degradation.

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold To determine if these results can be independently validated in a heterologous system, we co-expressed C-terminally 4xmyc-tagged M16666, M11166, M61111, and M66111 and C-terminally mYFP-tagged AVR a6 and AVR a1 variants in N. benthamiana. We observed a strong cell death response upon co-expression of M11166-4xmyc with AVR a6 -mYFP, suggesting that the last six C-terminal LRR repeats of MLA6 are sufficient for recognition of AVR a6 even in the absence of further barley-specific host proteins ( Fig 5B). We detected weak recognition of AVR A1 by M61111 under UV light at 302 nm in N. benthamiana (indicative of accumulation of autofluorescent compounds in dying plant cells; S11 and S14 Figs), whereas the specific recognition of AVR A6 by M16666 and AVR A1 by M66111 seen in barley protoplasts was completely lost, suggesting that other barley-specific protein(s) might be necessary for the functionality of these chimeric receptors in the heterologous expression system.

The LRRs of the MLA10 and MLA22 receptors specifically recognize allelic AVR A10 and AVR A22
To test if the LRRs of MLA NLRs are necessary to specifically recognize not only sequence unrelated, but also sequence related allelic AVR A effectors, we designed two chimeric receptor genes encoding the MLA10 CC-NB fused with the MLA22 LRRs (MLA10LRR22) and the MLA22 CC-NB domain fused with the MLA10 LRRs (MLA22LRR10) (S12 Fig). Subsequently, Mla10-4xmyc, Mla22-4xmyc, Mla10Lrr22-4xmyc or Mla22Lrr10-4xmyc were co-expressed with AVR a10 -mYFP, AVR a10 -V/AVR a22 -V-mYFP, or AVR a22 -mYFP in heterologous N. benthamiana. Co-expression of Mla10-4xmyc or Mla22Lrr10-4xmyc with AVR a10 -mYFP led to cell death in N. benthamiana leaves, but no cell death was observed when these Mla NLRs were co-expressed with an empty vector (EV), AVR a10 -V / AVR a22 -V-mYFP, or AVR a22 -mYFP ( Fig  6A). Additionally, expression of Mla22-4xmyc and Mla10Lrr22-4xmyc led to cell death when co-expressed with AVR a22 -mYFP, but not when these proteins were co-expressed with AVR a10 -mYFP, AVR a10 -V/AVR a22 -V-mYFP, or an EV control ( Fig 6A). The MLA10LRR22 and MLA22LRR10 receptor chimeras were detectable in N. benthamiana leaf extracts at levels comparable with the MLA10 and MLA22 receptors ( Fig 6B). To determine whether these results were reproducible in the homologous barley protoplast system, we expressed Mla10Lrr22 or Mla22Lrr10 together with AVR a10 and AVR a22 in leaf protoplasts and measured reduction of LUC activity as a proxy for cell death. In comparison to the EV reference sample, co-expression of AVR a22 with Mla10Lrr22 and AVR a10 with Mla22Lrr10 lead to an average 80% and 40% reduction of relative LUC activity, respectively, and this was not the case when AVR a10 was co-expressed with Mla10Lrr22 or when AVR a22 was co-expressed with Mla22Lrr10 ( Fig 6C). Taken together, the results suggest that the 58 amino acid differences between the LRRs of Mla10 and Mla22 are major determinants of respective recognition specificities for the allelic AVR A10 and AVR A22 effectors.
An association between MLA10 and AVR A10 was previously shown to be detectable in plant extracts and in yeast [11]. Using a previously established split-LUC complementation assay, we therefore tested, whether the MLA22LRR10 hybrid receptor also specifically interacts with the AVR A10 effector when co-expressed in planta. We generated constructs expressing AVR a variants fused C-terminally to the N-terminal part of the LUC reporter (AVR a -nLUC); days post infiltration and separated using a 10% polyacrylamide gels and detected using α-myc western blotting, CBB = Coomassie Brilliant Blue. https://doi.org/10.1371/journal.ppat.1009223.g005

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold and Mla variants fused C-terminally to the C-terminal part of the LUC reporter (Mla-cLUC) [11]. We then performed A. tumefaciens-mediated transformation of N. benthamiana leaves to express AVR a10 -nLUC or AVR a10 -V/AVR a22 -V-nLUC (not recognized AVR a10 variant) together with either Mla10-cLUC, Mla22Lrr10-cLUC or Mla10Lrr22-cLUC. Forty hours post infiltration, we determined LUC activity as a proxy for AVR A /MLA association, as described previously [11]. LUC activity was significantly higher in samples that co-expressed AVR a10 -nLUC with Mla10-cLUC or Mla22Lrr10-cLUC, when compared to samples where AVR a10 -

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold Nicotiana benthamiana leaves were transformed transiently with vectors containing cDNAs of Mla10-cLUC, Mla10Lrr22-cLUC or Mla22Lrr10-cLUC together with vectors containing cDNAs of AVR a10 -nLUC or AVR a10 -V/ AVR a22 -V nLUC lacking signal peptides (SPs), under the control of a 35S promoter. LUC activity was determined forty hours after transfection. The experiment was performed on at least three independent days with two to four replicates (independent set of plants) each day. Significant differences between AVR a10 -nLUC or AVR a10 -V/ AVR a22 -V nLUC were analyzed using one-way Kruskal-Wallis (KW) analysis. Calculated KW p-values are as follows: Mla10: p = 0.0001491; Mla22Lrr10: p = 0.009035; Mla10Lrr22: p = 0.8079. Samples labeled with different letters differed significantly (p < 0.05). (B) Protein levels of MLA10-cLUC, MLA22LRR10-cLUC and MLA10LRR22-cLUC in N. benthamiana leaf extracts. Proteins were separated on a 8% SDS-PAGE gel and a detected using anti-LUC western blot (WB). (C) Yeast was co-transformed with cDNAs of N-terminal LexABD-fused MLA and N-terminal B42AD-fused AVR A variants. Growth on media lacking Leucine indicates association of respective proteins fused to AD (activation nLUC was exchanged with its virulent variant AVR a10 -V/AVR a22 -V. This was not the case when the AVR a10 -nLUC variants were co-expressed with Mla10Lrr22-cLUC (Fig 7A and 7B). Our data suggests a reduced association of AVR A10 with the MLA22LRR10 hybrid receptor compared to wild-type MLA10, which is in agreement with differences in cell death scores of N. benthamiana leaves co-expressing AVR a10 -mYFP with Mla10-4xmyc and Mla22Lrr10-4xmyc (S14D Fig).
We also tested whether AVR A10 can interact with MLA22LRR10 in the absence of other plant proteins in yeast. We co-expressed LexABD-Mla10Lrr22 or LexABD-Mla22Lrr10 under the control of a constitutive ADH1 promoter with B42AD-AVR a10 , B42AD-AVR a10 -V/AVR a22 -V, or B42AD-AVR a22 under the control of a galactose (GAL1)-inducible promoter. Coexpressing B42AD-AVR a10 with LexABD-Mla22Lrr10 or with LexABD-Mla10 in a yeast twohybrid assay (Y2H) led to yeast growth on leucine-deprived media ( Fig 7C). Little growth was detectable when LexABD-Mla22Lrr10 or LexABD-Mla10 was co-expressed with B42AD-AVR a10 -V/AVR a22 -V, while no growth was detected when co-expressing B42AD-AVR a22 (Fig 7C), even though all effector and receptor fusion proteins were detectable in yeast extracts ( Fig 7D). Taken together, our findings suggest that the MLA10 LRR domain is responsible for specific recognition of AVR A10 and that this is dependent on effector-receptor association. However, we were unable to detect an interaction of LexABD-Mla22 or Lex-ABD-Mla10Lrr22 with B42AD-AVR a22 in this Y2H assays (Fig 7C).
Multiple residues in AVR A10 and AVR A22 are responsible for differential recognition specificities of MLA10 and MLA22 It has been proposed that direct fungal effector-plant NLR receptor interactions are mediated by cumulative binding of multiple effector aa residues to the surface of its corresponding NLR receptor [27,39]. We aimed to resolve which of the 11 amino acid residues that are polymorphic between AVR A10 and AVR A22 alleles (excluding the SP) are responsible for the specific recognition by the cognate MLA10 and MLA22 receptors. On the basis of AVR A10 secondary structural predictions, we divided AVR A10 /AVR A22 proteins into three equally long parts: an N-terminal (residues 22-54 aa, comprising ß1-ß2 sheets and the α1-helix, which included the two amino acid substitutions D45G, D53E), a central (55-86aa; comprising the ß3-ß4 sheet and cluster of most amino acid differences Q55H, D58N, G59D, Q61P, H64Y, and the residue F77Y), and a C-terminal part (87-118 aa; including the ß5-ß6 sheets and three amino acid differences V93L, W96L, I111N) ( Fig 8A). These individual regions were exchanged between AVR A10 /AVR A22 effector peptides and we then tested the interactions of the six resulting chimeric AVR A10 /AVR A22 effector constructs (called chimera11, chimera12, chimera13, chi-mera14, chimera15 and chimera16) with MLA10 and MLA22, in N. benthamiana as described above. All chimeric proteins were detectable after GFP-Trap enrichment except for chimer-a16-mYFP, which was not consistently detectable ( Fig 8C). Co-expression of chimera14-mYFP with Mla22-4xmyc led to cell death (Figs 8B, S13 and S14). These findings suggest that the Cterminal polymorphic residues V93, W96, and I111 are not responsible for the specific recognition by MLA22.
While in chimera14-mYFP, the three C-terminal residues of AVR A22 are exchanged for the respective amino acids found at these positions in AVR A10 , in chimera12-mYFP the three Cterminal residues of AVR A10 are exchanged for AVR A22 -specific amino acid residues (Fig 8A). domain) and BD (Binding domain). (D) Protein levels of LexABD-MLA and B42AD-AVR A variants. Proteins were precipitated using an ammonium-acetate buffer and dissolved in a urea-SDS sample buffer before separation on a 10% or 12% polyacrylamide gel and detection by either α-LexA or α-HA WB. https://doi.org/10.1371/journal.ppat.1009223.g007

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold Even though both effector chimeras were stable in planta, chimera12-mYFP was not recognized by MLA10-4xmyc (Figs 8B, 8C and S13 and S14). This suggests that single, double, or triple amino acid mutations at aa positions 93, 96, or 111 within the C-terminus of AVR A10 can lead to a loss of MLA10-specific recognition. We sought to determine which amino acid residues at the C-terminus, if exchanged to the respective AVR A22 residues, would result in abrogation of MLA10-specific recognition. To this end, we introduced L93V, L96W, or N111I single amino acid substitutions in chimera12-mYFP. The resulting chimera28-mYFP (L93V), chimer-a29-mYFP (L96W), and chimera30-mYFP (N111I) constructs were co-expressed with Mla10-4xmyc and with Mla22-4xmyc. Co-expression of chimera29-mYFP with Mla10-4xmyc led to cell death, while co-expression of chimera28-mYFP or chimera30-mYFP with Mla10-4xmyc did not (Figs 8B-8D and S13 and S14), suggesting that the tryptophan at position 96 in AVR A10 is important for MLA10-specific recognition. However, we cannot exclude that differences in protein

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold stability between mYFP fused chimera 29 and 28, or 30, accounts for the lack of cell death in samples co-expressing chimera28-mYFP and chimera30-mYFP together with Mla10-4xmyc.
Even though chimera14-mYFP and chimera15-mYFP differ by only two N-terminal amino acids at positions 45 and 53 and both proteins are detectable, co-expression of chimera15--mYFP with Mla22-4xmyc did not trigger cell death (Figs 8B, 8C and S13 and S14). This suggests that residues at one or both of these N-terminal positions are necessary for the specific recognition by MLA22. To test this hypothesis, we introduced single D45G or D53E substitutions into chimera15-mYFP (Fig 8A). The resulting chimera21-mYFP (D53E) and chimera22--mYFP (D45G) constructs were co-expressed with Mla10-4xmyc or Mla22-4xmyc. Coexpression of chimera22-mYFP but not chimera21-mYFP with Mla22-4xmyc led to cell death, while no cell death was observed upon co-expression of chimera21-mYFP with Mla22-4xmyc, suggesting that the glycine at position 45 but not the glutamic acid at position 53 is essential for MLA22-specific recognition (Fig 8A, 8B and 8E).
To test if the N-terminal and central parts of the AVR a22 effector are sufficient to trigger cell death when co-expressed with Mla22, we constructed a deletion construct (Δ93-118) of chi-mera22, which we termed chimera23 (Fig 8A). Co-expression of chimera23-mYFP with Mla10-4xmyc or Mla22-4xmyc did not trigger cell death, although the chimera23-mYFP protein was seemingly more stable than AVR A10 and all other chimeric proteins (Figs 8B-8E and S13 and S14). Even though the C-terminal residues L93, L96, and N111 are not specifically recognized by MLA22, our findings suggest that the C-terminal region of the AVR A22 effector potentially stabilizes the conformation of the N-terminal and central regions, which are necessary for MLA22-specified recognition. In summary, the N-terminal glycine at position 45 and the C-terminal tryptophan at position 96 are important for MLA22-and MLA10-specific recognition, respectively.
To independently verify the data, we also co-expressed a selection of AVR a10 /AVR a22 chimeras (chimera12, chimera14, chimera21, chimera22, chimera24, chimera26 and chimera29) together with Mla10 or Mla22 in protoplasts of barley cv. Golden Promise and determined cell viability by LUC activity, as described above. LUC activity was approximately 50% lower in samples co-expressing Mla10 with AVR a10 when compared to samples that co-express Mla10 with AVR a22, and this is in agreement with previously published data [11] (Fig 9A). We detected intermediate LUC activity when co-expressing Mla10 together with chimera26 or chi-mera29, but this reduced LUC activity did not differ significantly from the sample co-expressing Mla10 and AVR a10 (Fig 9A). This was not the case for samples co-expressing Mla10 together with chimera12, chimera14, chimera21, chimera22, or chimera24 (Fig 9A).
LUC activity was on average 80% lower in samples co-expressing Mla22 with AVR a22 when compared to samples that co-express Mla22 with AVR a10 , and this is again in agreement with published data [11]. LUC activity of samples co-expressing Mla22 together with chimera14, chimera22, chimera24 and chimera26 was not statistically different from the activity observed when co-expressing Mla22 and AVR a22 (Fig 9B). This was not the case for samples co-expressing Mla22 together with chimera12, chimera21 or chimera29. Notably, we detected an intermediate relative LUC activity when co-expressing Mla22 together with chimera26 (Fig 9B). Similarly, cell death scores of N. benthamiana leaves co-expressing Mla22-4xmyc together with chimera26-mYFP were also lower than when co-expressing Mla22-4xmyc together with AVR a22 -mYFP (Figs 8 and S14). We thus conclude that the barley protoplast cell death data (Fig 9) overall recapitulate the MLA10 and MLA22 specificities towards AVR A chimeric constructs observed in the heterologous N. benthamiana system (Fig 8).

Fig 9. Co-expression of a selection of AVR A10 /AVR A22 chimeras with MLA10 and MLA22 in barley protoplasts.
Barley protoplasts of cv. Golden Promise were co-transfected with a LUC reporter assay and pIPKb002 vectors containing cDNAs of AVR a10 , AVR a22 , chimera12, chimera14, chimera21, chimera22, chimera24, chimera26 or chimera29 without signal peptide or with an empty vector (EV) together with either (A) Mla10 or (B) Mla22 in pipkb002. Transfections were performed at least six times independently. Significant differences between samples were analyzed using non-parametric Kruskal-Wallis (KW) analysis followed by the Dunn's test. Calculated KW p-value are as follows: Mla10: p = 1.439e-07, Mla22: p = 5.374e-11. Samples labeled with different letters differed significantly (p < 0.05) in the Dunn's test.

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold In summary, our findings suggest that for triggering MLA10-specific cell death, the four residues D53, H64, F77, and W96 of AVR A10 cannot be exchanged to the residues found in AVR A22 . In turn, to trigger MLA22-specific cell death, the five amino acid residues G45, H55, N58, D59, and P61 of AVR A22 cannot be exchanged to the residues found in AVR A22. Furthermore, deletion of the C-terminal third of the AVR A10 and AVR A22 effectors leads to loss of avirulence function (Fig 8).
To determine if MLA-mediated cell death initiated by recognition of the AVR A effector also correlates with receptor-effector association in plant extracts, we again applied the split-LUC complementation assay. We transiently expressed Mla10-cLUC together with AVR a10 -nLUC, chimera26-nLUC, chimera29-nLUC and as a control, chimera22-nLUC in N. benthamiana leaves, followed by LUC measurements at 40 hours post infiltration of leaves with the A.

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold tumefaciens carrying constructs of interest. LUC activity of samples co-expressing Mla10-cLUC with AVR a10 -nLUC, chimera26-nLUC and chimera29-nLUC was significantly higher than the LUC activity observed in the samples expressing Mla10-cLUC together with chimer-a22-nLUC (Fig 10A and 10B). Chimera26 and chimera29 but not chimera22, can trigger MLA10-mediated cell death in co-expression assays (Figs 8 and 9), and as such, we conclude that the recognition specificities mediated by MLA10 towards the AVR A chimeric variants 26, 29 and 22 correlate with receptor-effector association. We again only observed intermediate levels of LUC activity in samples co-expressing Mla10-cLUC together with chimera26-nLUC or chimera29-nLUC (Fig 10A). This is in agreement with the quantitative cell death assay in barley protoplasts (Fig 9A), and suggests that when compared to AVR a10, these constructs are impaired in their ability to activate and associate with MLA10 in planta and this may be associated with levels of protein expression (Fig 10B).

Two amino acids of AVR A10 that cannot be exchanged to respective AVR A22 residues are located in a predicted positively charged area that corresponds to the catalytic cleft of the fungal F1 RNase
Microscale thermophoresis assays suggested that CSEP0064 has some affinity to total RNA but its X-ray structure suggests that it lacks residues required for RNA hydrolysis [30]. Our data also suggests that the putative RNase-like fold of AVR A effectors is not associated with RNase activity (Fig 4). Here, we examined the location of the AVR A10 and AVR A22 residues that are required for specific recognition by MLA10 and MLA22, and if the corresponding residues in the F1 RNase are required for RNA binding and hydrolysis. To do this, we superimposed AVR A10 and AVR A22 predicted structures on the structure of Fusarium moniliformis F1 RNase. Residues Y38, Y42, and Y45 of the F1 RNase are involved in binding the ribose and phosphate of 2' GMP and the respective amino acids found in AVR A10 (F51, F54, and H57) are identical to those of AVR A22, and as such, do not account for specific MLA recognition ( Fig  11A-11C). The R77 residue in the F1 RNase, also forms a contact with the phosphate in 2' GMP. The corresponding residue can also be found in AVR A10 and AVR A22 (residue R81, Fig  11A-11C) but not in the CSEP0064 structure or the predicted structures of any other AVR A effector isolated so far. The residues W96 and H64 of AVR A10 are L96 and Y64 in AVR A22 and have dissimilar properties to the corresponding residues (H92 and E58, respectively) in the F1 RNase. Electrostatic potential prediction using Adaptive Poisson-Boltzmann Solver (APBS) of surfaces suggests that H64, R77, and W96 in AVR A10 and L64, R77, and L96 in AVR A22 belong to a positively charged cleft (Fig 11B-11D). In AVR A10 , these residues are predicted to be part of the positively charged surface patch. For recognition by MLA10, H64 and W96 can indeed not be exchanged to the respective residues found in AVR A22 (Fig 11B). In contrast, the AVR A22 residues that cannot be exchanged to the respective AVR A10 residues without losing MLA22 avirulence activity, can be found in a negatively charged surface patch away from the positively charged area, presumably required for MLA10 recognition (Fig 11D). These results suggest that the AVR A residues, which confer specific recognition by MLA10 and MLA22 receptors, are located in distinct predicted surface patches of the avirulence proteins encoded by allelic AVR a10 and AVR a22 .

Identification of AVR A6
Long-read DNA sequencing and high-quality genome assembly of the DH14 Bgh isolate recently recovered 30 Mb of previously unassembled repetitive regions of the Blumeria

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold genome, which linked together multiple non-contiguous series of genomic sequences, including scaffold 16 [40]. This may explain why we failed to identify BLGH_00708, BLGH_00709 and BLGH_07091 as AVR A6 in our previous studies [11,28] and emphasizes the importance of highquality Blumeria genome assemblies for identification of novel AVR A effectors. We found three AVR a6 paralogues in the RACE1 genome, and in line with this observation, three non-identical paralogues of AVR a7 have been described in RACE1 [11], indicating that these AVR a effectors were duplicated [11]. Similarly, BLGH_07092 is likely an AVR a6 copy with a frameshift mutation. This AVR a duplication could facilitate the gain of new virulence functions. The location of amino acid residues in AVR A10 and AVR A22 that determine MLA10 and MLA22 recognition specificities. (A and C) show structural superimposition of the crystal structure of Fusarium moniliformis RNase F1 (yellow) and IntFOLD version 5.0 structural predictions of AVR A10 or AVR A22 (grey). Depicted in green is the F1 RNase ligand 2'-guanosine monophosphate (2' GMP); residues of the F1 RNase catalytic triad and corresponding AVR A residues are depicted in red; residues of the F1 RNase RNA binding pocket and corresponding AVR A residues are shown in blue. The residues of AVR A10 and AVR A22 required for specific MLA10 and MLA22 recognition as determined in Figs 8-10 are framed with a purple rectangle. (B and D) Predicted electrostatic surface potential of the AVR A10 and AVR A22 effector surface calculated using Adaptive Poisson Boltzmann Solver (APBS) [67]. The residues of AVR A10 and AVR A22 required for specific MLA10 and MLA22 recognition as determined in

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold Unlike other virulent variants of Bgh AVR a effectors that carry SNPs in the coding region or transposon insertions, or are transcriptionally silent [11,28], here we report that virulence of Bgh isolates on Mla6 NILs is possibly caused by splice site mutations in the intron of transcripts from isolates virulent on Mla6 lines that map to the AVR a6 gene. The intron retention observed in most of these transcripts is potentially caused by either a splice branch point mutation or a splice donor site mutation and leads to premature stop codons (S3 and S4 Figs).
Even though the encoded truncated proteins are detectable in heterologous N. benthamiana (Fig 2C), RNA-Seq analysis of virulent Bgh isolates on barley leaves suggests that these mutations result in significantly reduced transcript levels. The latter is possibly a consequence of nonsense-mediated mRNA decay (NMD) [41]. This observation, together with a high virulence frequency on Mla6 NILs in European Bgh populations [42], suggests that the loss of the AVR a6 effector gene does not have a detrimental impact on Bgh virulence. Future studies might clarify whether AVR a6 virulence function can be compensated by AVR a6 family members or more distantly related effectors.

A predicted common ribonuclease-like fold among Blumeria AVR A, AVRPM3 and AVRPM2 effectors
AVR A6 was predicted to be structurally similar to CSEP0064 (Fig 3). Upon re-analysis of all previously reported Bgh AVR A effectors using the version 5.0 of the IntFOLD structural prediction server, we report here evidence supporting a common structural fold amongst the so far isolated AVR A effectors despite the lack of relatedness between their DNA and protein sequences (Fig 3). Moreover, AVR A effectors appear to share this structural fold at least with Bgt AVRPM2 and AVRPM3 D3 effectors that are recognized by wheat NLRs PM2 and PM3, respectively (Fig 3; [ 22,35,36]). In addition, protein alignments and in silico tertiary structure modelling suggest that Bgt AVRPM3 A2/F2 and AVRPM3 B2/C2 also have a central α-helix that faces three to four β-sheets [22]. Even though AVR A effectors seem to have a common predicted fold similar to RNases, the residues critical for catalytic activity are lacking in the predicted AVR A structures and we did not detect ribonuclease activity when AVR A effectors were co-incubated with RNA ( Fig 4A).
Given that multi-allelic barley MLA, multi-allelic wheat PM3, and wheat PM2 NLRs are sequence-unrelated and are encoded on non-syntenic chromosomal locations, it is possible that these immune receptors arose by convergent evolution to detect distinct members of the structurally related superfamily of powdery mildew RALPH effectors.
We speculate on two evolutionary scenarios that might explain the diversity of extant RNase-like effectors in Bgh and Bgt. In a common descent scenario, all RALPH effectors have diversified from an "ur-RALPH", which was present in the last common ancestor prior to host specialization of grass powdery mildews [29]. For instance, striking sequence conservation of AvrPm2 among Blumeria ff spp secalis, tritici and hordei [36] and the presence of orthologous candidate effector gene families among specialized forms of grass powdery mildews support a common descent model [43]. The relatively small number of RALPH effectors in dicot-infecting powdery mildew species such as Erysiphe pisi [44] and in the early-diverged Parauncinula polyspora [45] suggests that RALPHs could have evolved 80-90 million years ago [46]. However, the variable intron location in Blumeria AVR genes (S6 and S7 Figs) questions this hypothesis. In an alternative scenario, the predicted structural relatedness of known AVRPM and AVR A effectors could be the product of convergent evolution from different fungal RNases in the order Erysiphales after the differentiation of formae speciales. In summary, our data is reminiscent of findings indicating the existence of sequence unrelated but structurally related MAX effectors (Magnaporthe Avrs and ToxB-like) that account for 10% of the M. oryzae effector repertoire in this Ascomycete pathogen [47]. Another example for sequence-diversified but structurally-related effectors are oomycete RXLR effectors, such as ATR1 and ATR13 in Hyaloperonospora arabidopsidis [48,49], and PexRD2 and AVR3a11 in Phytophthora infestans [50].

The LRRs of allelic MLA receptors determine recognition of matching AVR A effectors
Earlier work with hybrids built from MLA1 and MLA6 receptors demonstrated that the MLA LRR is a determinant of Bgh isolate-specific recognition [38]. In this previous study it could not be clarified whether the respective AVR A effectors are the only fungal components that determine isolate-specific recognition by the MLA LRR domains. We show here that the LRR of four MLA receptors (MLA1, MLA6, MLA10, and MLA22) is responsible for specific recognition of matching AVR A effectors. We found that for the recognition of AVR A6, the C-terminal six LRR repeats of MLA6 cannot be exchanged to the ones found in MLA1, while the 12 C-terminal LRR repeats of MLA1 cannot be exchanged to those of MLA6 for recognition of AVR A1 . Similarly, four C-terminal LRR repeats of the flax allelic L5 and L6 receptors are necessary for recognition of matching AvrL567 effectors [27]. Most sites of positive selection among MLA resistance specificities cluster on the predicted concave site of the C-terminal LRRs [19,20]. Thus, it is possible that these LRR residues are contact sites for specific recognition and association with structurally related AVR A effectors. We confirmed this assumption for the LRRs of MLA10 by observing an association of MLA22LRR10 with AVR A10 when the proteins were co-expressed in plants or in yeast (Fig 6C).
Unexpectedly, whereas M11166, M16666, M61111, and M66111 chimeras clearly recognize the corresponding AVR A effectors in barley protoplasts (Fig 5A), the latter three hybrid receptors are non-functional in heterologous N. benthamiana. The resistance function of several barley Mla recognition specificities, including Mla6, is genetically dependent on HvRar1, HvSgt1, and HvHsp90 [38,51,52], which form a chaperone/co-chaperone complex in which HvHSP90 directly interacts with the LRR of MLA [52,53]. Thus, the hybrid MLA1/MLA6 receptors might be dependent on an additional barley protein for full resistance function. Interestingly, the function of M11166 is known to be fully independent of HvSGT1 and HvRAR1 in barley and we have shown here that this is the sole MLA6/MLA1 chimeric receptor functional in N. benthamiana [38,51].

Multiple polymorphic AVR A10 /AVR A22 residues influence recognition by MLA10 and MLA22
AVR a10 and AVR a22 effector alleles are maintained as a balanced polymorphism in a worldwide collection of Bgh isolates [11], which implies an important virulence function for the AVR a10 /AVR a22 gene, supported by its membership in the Blumeria core effectorome [40]. As the effector alleles with only 11 polymorphic amino acids likely adopt an identical protein structure, we aimed here to pinpoint polymorphic residues in AVR A10 /AVR A22 recognized by MLA10 and MLA22, respectively. Some effector chimeras, including chimera11 or chimera13 containing only two polymorphic residues compared to the respective WT avirulence effectors, escaped recognition by MLA10 and MLA22, even though these are stable proteins in N. benthamiana. This is consistent with the observation that other naturally occurring virulent AVR A effector variants can escape recognition by only one or two amino acid substitutions in the respective AVR A polypeptides [11,28]. Similarly, one amino acid exchange in Bgt AVRPM3 A2/F2 leads to a loss of pathogen recognition [54]. We identified four residues in AVR A10 that cannot be exchanged to the ones found in AVR A22 without losing recognition by MLA10. Four different residues in AVR A22 cannot be exchanged to the ones found in AVR A10 without losing recognition by MLA22. Of note, chimera26 can be recognized by MLA10 and MLA22 (Figs 8 and 9), further suggesting that differential regions in AVR A10 and AVR A22 are recognized by MLA10 and MLA22, respectively. Our findings are consistent with the identification of multiple residues, spread along the Bgt AVRPM3 B2/C2 effector polypeptide, that cannot be exchanged for specific detection by wheat PM3b or PM3c NLRs [22]. Multiple, additive contact points of the AvrL567-A and -D flax rust fungus alleles are recognized by the flax receptors L5 and L6, respectively [27]. In summary, our findings suggest that multiple residues on the AVR A effector surface determine the specific recognition by MLA receptors, and this may influence the functional diversification process of these receptors.
Previous studies suggested that RALPH effectors are pseudoenzymes that cannot cleave RNA due to the absence of a catalytic amino acid triad present in the fungal F1 RNase that are needed for enzymatic RNA catalysis [30]. These catalytic residues are E58, R77, and H92 [55]. We found that only one residue involved in RNA catalysis (R81) is conserved in the deduced AVR A10 /AVR A22 RNase-like effectors. Notably, four amino acids in AVR A10 cannot be exchanged to the ones found in AVR A22 (D53, F77, H64, and W96), and these are located close to a predicted positively charged area on the effector's surface. In the F1 RNase, the corresponding area forms the catalytic cleft. Residues required for MLA22 recognition are found in a negatively charged surface patch away from the negatively charged area presumably recognized by MLA10 (Fig 11A and 11B). This data is underlined by the recognition of chimera26 through MLA10 and MLA22, as chimera26 carries both of the described recognition patches. Recently, a few RALPH effectors were found in E. pisi, which infects a dicotyledonous host, and structural predictions showed that residues for RNA catalysis are partially conserved and are located within a positively charged binding cleft [44]. This is in agreement with our findings for Bgh AVR A10 and AVR A22 but contrasts with the absence of any catalytic triad residue as well as a positively charged binding cleft in the structure of Bgh CSEP0064 [30]. If these residues, which are located in the predicted positively charged cleft, and are recognized by MLA10, are also involved in potential RNA binding of AVR A10 and AVR A22 remains to be determined.

Plant material
The barley cultivar Golden Promise was grown at 19˚C, 70% humidity and under a 16 h photoperiod. N. benthamiana plants were grown and maintained under standard greenhouse conditions.

Association analysis
RNA-seq read alignment, variant calling, and association analysis were performed as described in Saur et al., 2019 [11].

PLOS PATHOGENS
LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold

Transient gene expression assays in barley protoplasts
The isolation and transfection of barley protoplasts was performed as described in Saur et al., 2019 [33]. In short, cDNAs of the AVR a s were co-expressed with cDNAs of Mla10, Mla22, Mla10Lrr22, and Mla22Lrr10 using the pIPKb002 vector with a strong ubiquitin promoter or with intron-containing DNA of chimeras M16666, M11166, M61111, or M66111 in a pUBI-NOS-vector (described in Shen et al., 2003 [38]) in barley cv. Golden Promise protoplasts. Protoplast solution (300 μl of 3.5 x 10 5 cells/ml) was transfected with 4.5 μg of LUC reporter construct, 10 μg of Mla plasmid, and 6.5 μg of the respective AVR a effector or an empty vector (EV). The protoplasts were incubated for 16 h at 21˚C in a plant growth chamber and then harvested by centrifugation at 1,000 x g. Subsequently, the supernatant was removed, and protoplasts were lysed by addition of 180 μl of cell culture lysis reagent (Promega, E1531). The LUC activity of samples was measured in a luminometer (Centro, LB960) using a 96-well plate in which 50 μl of protoplast lysate were mixed with 50 μl of the LUC substrate (Promega, E1501). The relative LUC units (RLU) were calculated by setting the absolute value of the EV sample to 1.

Transient gene expression assays in Nicotiana benthamiana
Expression constructs for AVR A and MLA and respective chimeras were always freshly transformed into Agrobacterium tumefaciens strains GV3101::pm90 and GV3101::pm90RK and selected on LB media containing the respective antibiotic resistance. Single colonies were inoculated into liquid LB medium and grown overnight at 28˚C with agitation at 220 rpm to a maximal OD 600 = 1.5. Agrobacteria were centrifuged at 2500 x g for 15 min and the pellet was resuspended in infiltration buffer (10 mM MES, pH 5.6, 10 mM MgCl 2 , and 200 μM acetosyringone) to an OD 600 of 1 to 1.2. The suspensions of Agrobacteria were incubated at 28˚C with shaking 150 rpm for at least 2 h. Leaves of four-week-old N. benthamiana plants were infiltrated with a 1:1 mix of bacteria carrying AVR a constructs or Mla constructs. The cell death score was assessed at four days post infiltration. Leaf tissue was harvested two days post infiltration for western blot analysis and 40 hours for split-LUC assays.

Split-luciferase complementation assay
The assay was performed as described in Saur et al., 2019 [11].

Plant protein extraction and immunoprecipitation for detection of fusion proteins
N. benthamiana leaf material was frozen in liquid nitrogen and ground to a fine powder using a Retsch bead beater.
For the detection of AVR A -mYFP and MLA-4xmyc proteins, 300 mg of ground leaf tissue were dissolved in 2 mL of ice cold extraction buffer (150 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, 10% (v/v) glycerol, 10 mM DTT, 2% (v/v) plant protease inhibitor cocktail (Sigma), 1 mM NaF, 1 mM Na 3 VO 4 , 1 mM PMSF, and 0.5% (v/v) IGEPAL). Extracts were centrifuged twice for 16 min at 16,000 x g at 4˚C. For the detection of MLA-4xmyc proteins, the extracts were diluted 4:1 with 4 x SDS loading buffer for SDS-PAGE. Samples were heated for 5 min at 95˚C. For the detection of AVR A -mYFP, the proteins were concentrated using GFP-trap-MA (Chromotek) beads. Beforehand, the beads were incubated in equilibration buffer (150 mM Tris-HCl, pH 7.5, 150 mM NaCl, 10 mM EDTA, pH 7.5, 10% Glycerol, 1.5% (w/v) BSA) for 1 hour at 4˚C with slow rotation. The protein extracts were incubated with the equilibrated beads for 4 h at 4˚C with slow rotation. Subsequently, the beads were washed five times with cold wash buffer at 4˚C. The conjugated proteins were stripped off the beads by boiling the samples in 30 μl 4 x Laemmli sample buffer at 95˚C for 10 min.

Protein expression and purification from Escherichia coli
AVR A6 (25-115) AVR A10 , and AVR A13  were expressed in E. coli as fusion proteins with N-terminal GST tags. The expression plasmids pGEX6p-1 (GE Healthcare) were transformed into the E. coli strain BL21 (DE3) (Novagen) by heat shock and grown at 37˚C in Luria-Bertani broth to an OD 600 of 0.6. Isopropyl-β-D-thiogalactoside (IPTG, Sigma) was added to induce protein expression at 18˚C for a further 12 h. The cells were harvested by centrifugation at 6,000 x g for 10 min at 4˚C and resuspended in resuspension buffer (25 mM TRIS pH 8, 150 mM NaCl). Cell suspensions were lysed by sonification. Cell debris was removed by centrifugation at 30,000g for 2 h. The soluble fractions were collected and allowed to flow through GST resin (GE Healthcare). After washing with two column volumes of the same buffer used for resuspension, another 2 ml of buffer and 10 μl of PreScission protease (GE Healthcare) were added to the column followed by overnight incubation to cleave off the AVR A proteins from the GST resin. The cleaved AVR A proteins were then eluted and further purified by size-exclusion chromatography using a Superdex 200 10/30 gel filtration column (GE Healthcare).

Protein expression and purification from insect cells
AVR A6 (25-115) AVR A10 , and AVR A13  were expressed in insect cells as fusion proteins with N-terminal GST tags. The expression plasmids pFASTBAC1 (Invitrogen) were transformed into the E. coli strain DH10Bac (Invitrogen) by heat shock. Successful transformation was validated by blue-white selection and bacmids of positive colonies were subsequently isolated with a DNA Mini Kit (QIAGEN). Sf21 insect cells (Invitrogen) were transfected with sequence verified bacmids by CellfectinII (Thermo Fisher). After five days incubation at 28˚C, recombinant baculovirus P0 were harvested and used to amplify P1 virus for another three days. Insect cells were infected at concentration of 2.0 x 10 6 -2.5 x 10 6 cells/ ml with P1 virus for 60 h. Insect cells were harvested and re-suspended in resuspension buffer (25 mM TRIS pH 8, 150 mM NaCl,15mM imidazole) followed by sonification lysis. Cell debris were removed by centrifugation at 30,000g for 2 h. The soluble fractions were collected and allowed to flow through a GST affinity trap. After washing with two column volumes of the same buffer used for resuspension, another 2 ml of buffer and 10 μl of PreScission protease (GE Healthcare) were added to the column, followed by overnight incubation to cleave off the AVR A proteins from the GST resin. The cleaved AVRA proteins were then eluted and further purified by size-exclusion chromatography using a Superdex 200 10/30 gel filtration column (GE Healthcare). The proteins were tested for RNase activity with the same method applied for E. coli purified proteins.

RNase activity assays
Leaf material from three-week-old barley cv. Golden Promise plants was harvested to extract total RNA using the RNeasy Plant Mini Kit (QIAGEN). The remaining genomic DNA was removed by treating RNA with TURBO DNase enzyme (Ambion). Purified AVR A effectors from E. coli were incubated with denatured total barley RNA. Then, 30 μl reaction mixtures (1 μg RNA, 1 μM protein in 15 mM Tris-HCl (pH 8.0), 15 mM NaCl, 50 mM KCl, and 2.5 mM EDTA) were incubated at 25˚C for 90 min. RNase F1 (Sigma) was included as a positive control. For analysis by the Bioanalyzer 2100 (Agilent Technologies, USA) 10 μl of sample were used.
To test the consumption of native rabbit rRNA, purified AVR A effectors were incubated with 20 μl of rRNA from rabbit reticulocyte lysate (Promega) following the method of Kao et al. 2001 [37]. 30 μl reaction mixtures (20 μl rabbit reticulocyte lysate, 1μM purified AVR A effectors in 15 mM Tris-HCl pH 8.0, 15 mM NaCl, 50 mM KCl, 2.5 mM EDTA) were incubated at 25˚C. After 60 or 30 min, the reaction was terminated by adding 20 μl phenol/chloroform and was vortexed for 30 seconds. Samples were sedimented at 14,000 rpm for 15 min and 30 μl of the aqueous layer was removed and mixed with 6 μl electrophoresis loading buffer. For analysis by the Bioanalyzer 2100 (Agilent Technologies, USA) 10 μl of sample were used.

Yeast two-hybrid assays
Mla variants were cloned into the pLexA-GW vector [64] for expression with an N-terminal LexA activation domain under the control of a constitutive ADH1 promoter (BD-MLA). The AVR a variants were cloned into pB42AD-GW [64] for expression with an N-terminal B42 activation domain followed by the HA tag under the control of an inducible GAL1 promoter (AD-AVR a ). Using the lithium acetate method [65], Mla bait constructs and AVR a prey constructs were co-transformed into the yeast strain EGY4.8 p8op-lacZ and successful transformants were selected by colony growth on SD-UHW/Glu (4% (w/v) Glucose, 0.139% (w/v) yeast synthetic drop-out medium pH 5.8 without uracil, histidine, tryptophan, 0.67% (w/v) BD Difco yeast nitrogen base, and 2% (w/v) Bacto Agar). Yeast transformants were grown to OD 600 1 in liquid SD-UHW/Glu before harvesting cells for serial dilution on SD-UHW/Gal/ Raf media (SD-UHW without glucose but with 2% (w/v) Galactose 1% (w/v) Raffinose, with (-UHW) or without Leucine (-UHWL)) and incubated for 14 days at 30˚C.

Yeast protein extraction
For protein extraction, 10 ml of co-transformed yeast strains were grown to an OD 600 of 1 in SD-UHW/Gal/Raf liquid medium at 30˚C with shaking at 200 rpm. The proteins were precipitated using the ammonium acetate method (modified from Karginov and Agaphonov et al., 2016 [66]). In short, cells were harvested by centrifugation at 700 xg for 5 min. The pellets were resuspended in 200 μl NH 4 -acetate buffer (1 M NH 4 (CH 3 COO), 150 mM NaCl, 30 mM Tris-HCl, pH 7.5, 10 mM PMSF, 5 mM EDTA, and one tablet of Protease Inhibitor Cocktail (Roche). The yeast suspension was transferred into BeadBug-prefilled tubes with 0.5-mm silica glass beads (Sigma) and ground in a Precellys homogenizer (two times at 6,200 rpm for 30 sec, break: 15 sec). Afterwards, the DNA was sheared using a Diogenode Bioruptur ultrasonic water bath (twice for 30 sec at high power, break: 90 sec). The suspension without the beads was transferred into a Protein LoBind tube (Eppendorf). The glass beads were washed three times with 250 μl NH 4 -acetate buffer. The washes were combined with the suspension and incubated for 1.5 h on ice. Precipitated proteins were harvested by centrifugation (16,000 x g for 10 min). Precipitates were washed with 1 ml 1 M NaCl and the pellet was resuspended with 200 μl Urea-SDS sample buffer (50 mM Tris-HCl, pH 6.8, 2% SDS, 8 M Urea, 1% ß-mercaptoethanol, 2 mM EDTA, 5% glycerol, and 0.004% bromophenol blue) at room temperature. Resuspension in urea-SDS buffer and omission of the boiling step is essential for detection of LexA-MLA fusion proteins. For western blotting, 10-15 μl of the sample were loaded on 8% or 12% SDS page gels, blotted onto PVDF membranes and probed with either anti-HA (Merck, clone 3F10, RRID:AB_390914) or anti-LexA (Santa Cruz, Biotechnology, sc7544, RRID:AB_627883) primary antibodies, followed by incubation with secondary anti-rat (Santa Cruz Biotechnology, sc2065, RRID:AB_631756) or anti-mouse IgG-HRP antibodies (Santa Cruz Biotechnology, sc2005, RRID:AB_631736) for the detection of AVR A or MLA proteins, respectively. HA and LexA fusion proteins were detected by HRP activity on SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Fisher 34095) using a Gel Doc XR and gel documentation system (Bio-Rad). (B) Schematic illustration of the genomic, transcriptomic, and deduced protein sequences of the three AVR a6 effector variants: AVR a6 , AVR a6 -V1, and AVR a6 -V2. Depicted is the mutation in the consensus sequence of the branch point in AVR a6 -V1 transcripts of Bgh isolates CC66 and CC148 and the splice donor site mutation in AVR a6 -V2, which can be found in transcripts and the genome of Bgh isolate K1. This splice site mutation likely lead to an intron retention, which is supported by RNA-seq reads. Intron retention introduces an early stop codon leading to anticipated truncation of the AVR A6 -V1 and AVR A6 -V2 proteins. showing absorbance at 280 nm (y-axis) against the retention volume (ml) (x-axis) and the respective fraction above (A34 and A36 for AVR A6 , A31 and A33 for AVR A10 ; A32 and A34 for AVR A13 ), which was used for further RNase activity assays. (B) Stain-free SDS-PAGE (Bio-rad) showing the fractions of purified AVR A6 , AVR A10 , and AVR A13 proteins used for further RNase activity assays with a white arrow. AVR A protein fractions were separated on a PLOS PATHOGENS LRR domain of MLAs define specificity towards avirulence effectors with a predicted common fold 12% polyacrylamide gel and visualized by the ChemiDoc MP Imaging System (170-8280). (TIF) S9 Fig. Protein quality control of AVR A6 , AVR A10 , and AVR A13 . (A) Size exclusion chromatogram of AVR A6 , AVR A10 , and AVR A13 purified from insect cells. Fractions eluted at 18.5, 17, 17.5ml of AVR A6 , AVR A10 , and AVR A13 proteins were verified by SDS-PAGE (indicated by red, green and blue arrows) and used for further RNase activity assays. (B) After size exclusion, insect cell-purified AVR A6 , AVR A10 , and AVR A13 proteins or T1 RNase were incubated with denatured HvRNA. All samples were separated on non-denaturing 2% agarose gels and analyzed on a Bioanalyzer to determine for RNA degradation. (TIF) S10 Fig. Amino acid sequence alignment of M61111, M66111, M16666, and M11166. Colored boxes depict different domains of the receptors: blue = CC-domain, green = NB-ARC domain, red = LRR as defined previously in [19]. Grey boxes depict individual LRRs. Significance of cell death scores was calculated by Fisher's exact test and an asterisk depicts p < 0.05: (B) MLA6, AVR A6 -1: 8.68xe -13 ; MLA1, AVR A1 : 2.6xe -11 (C) M11166, AVR A6 -1: 3.98xe -12 ; (D) MLA10, AVR A10 : 1.25xe -10 ; MLA22LRR10, AVR A10 : 3.07xe -07 ; MLA22, AVR A22 : 7.74xe -08 ; MLA10LRR22, AVR A22 : 5.8xe -07 (E) MLA10, AVR A10 : 3.07xe -29 ; MLA10, chimera26: 1.54xe -13 ; MLA10, chimera29: 3.73xe -12 ; MLA22, AVR A22 : 2.77xe -49 ; MLA22, chimera14: 6.68xe -28 ; MLA22, chimera22: 6.39xe -14 ; MLA22, chi-mera24: 1.5xe -11 , MLA22; chimera26: 9.59xe -09 . (TIF) S1 Table. Top-ranking AVR a6 candidates in the transcriptome-wide association study (TWAS) determined by gene-wise calling. � � The table columns show the new and former BLGH_ID, the description, the scaffold localization and the p-value of the top-ranking candidates for the gene-wise association of Bgh transcriptomes with infection phenotypes on Mla6 near-isogenic lines (NILs). Color codes depict top-ranking AVR a6 candidates and are consistent with the color code used in Fig 1A: bright green: CSEP0254 paralogues, dark green: BLGH_07092, dark red: BLGH_00697, bright red: BLGH_00700. (TIF) S2 Table. Top-ranking AVR a6 candidates in the transcriptome-wide association study (TWAS) determined by variant-wise calling. � � The table columns depict the scaffold localization, the effect that the mutation has on the reference gene (non-synonymous mutation, gained stop codon), the codon change and the respective aa exchange, a gene description, the CSEP_ID and the p-value of the top-ranking candidates for the variant-wise association of Bgh transcriptomes with infection phenotypes on Mla6 near-isogenic lines (NILs). Color codes designate top-ranking AVR a6 candidates and are consistent with the color code used in Fig 1A: bright green: CSEP0254 paralogues, dark green: BLGH_07092 and bright red: BLGH_00700. (TIF) S3 Table. Primers used in this study. (TIF)