Molecular mimicry of a pathogen virulence target by a plant immune receptor
The barley receptor MLA3 detects the blast effector Pwl2 by copying the binding surface of the host protein Pwl2 normally targets, which is the opposite trick to an integrated decoy.
Integrated domains explain how some NLRs gain new recognition: the receptor swallows the effector's target whole and keeps it as a decoy. MLA3 does something else. Its LRR C terminus binds the Magnaporthe oryzae effector Pwl2 through a surface that overlaps the one Pwl2 uses on its real host target, the heavy-metal-associated protein HIPP43, without sharing any sequence similarity or fold with it. Reciprocal residue exchanges between MLA3 and OsHIPP43 transfer binding both ways. Three residues unique to MLA3 among more than thirty Mla alleles carry the specificity, and a single D926K substitution is enough to give the non-recogniser MLA23 the ability. Moving that interface into rye SR50 produced a receptor that still recognises AvrSr50 and now also Pwl2, resistant to both stem rust and blast in transgenic barley.
Key findings (10)
MLA3 confers Pwl2-dependent resistance to Magnaporthe oryzae. The near-isogenic barley line S02, carrying Mla3, was susceptible to a pwl2 mutant isolate but resistant to the wild-type isolate. S13, which expresses both Mla3 and Mla23, behaved identically, and resistance was attributed to MLA3 alone since MLA23 neither triggers cell death with Pwl2 nor associates with it.
Evidence spot inoculation with M. oryzae KEN54-20, a pwl2 mutant (M61) and a PWL2-complemented mutant; transcriptome analysis; cell death and co-immunoprecipitation assays
Source Figures 1A-1D; table S1
Recognition maps to the C terminus of the LRR domain. MLA23 differs from MLA3 by only 19 amino acid polymorphisms at the LRR C terminus (98.6% identity), and MLA3 binds Pwl2 through its last 85 residues, with direct binding confirmed independently.
Evidence protein alignment; fragment binding assays in N. benthamiana; yeast two-hybrid
Source Figures 1B and 1E; figure S2
MLA3 mimics the binding interface of HIPP43, the Pwl2 host virulence target. AlphaFold2 modelling of the MLA3 LRR domain with Pwl2 (pLDDT 84.4, pTM 0.859) gives an interface that overlaps the experimentally determined Pwl2-OsHIPP43 interface (PDB 8R7A), with both partners engaging Pwl2 in related orientations despite no sequence similarity.
Evidence AlphaFold2 modelling compared against the experimental Pwl2-OsHIPP43 crystal structure
Source Figures 2A-2C; figures S3-S6
The mimicry is of interface, not of domain. The HMA domain of OsHIPP43 is four antiparallel beta strands and two alpha helices; the MLA3 C terminus is three parallel beta strands and one alpha helix, corresponding to the last three LRR repeats. So the Pwl2-binding region of MLA3 is not a misannotated integrated HMA domain.
Evidence secondary structure comparison of the two binding regions
Source Figure S5
MLA3 and OsHIPP43 compete for Pwl2, and their interfaces are interchangeable. Co-expressing OsHIPP43 weakened MLA3/Pwl2 cell death, while the Arabidopsis thioredoxin AtTRXh5 control did not. Reciprocal exchange of Pwl2-binding residues preserved Pwl2 association when MLA3 residues were put into OsHIPP43, and maintained effector recognition when OsHIPP43 residues were put into MLA3.
Evidence competition cell death assays in N. benthamiana; reciprocal residue swaps in both proteins
Source Figures 2D-2G; figures S7A, S8
Three residues unique to MLA3 among the Mla allelic series carry the specificity: Lys926, Val931 and Tyr932. Only 3 of 12 predicted Pwl2-contacting residues are exclusive to MLA3 across more than 30 alleles sharing over 90% identity. Lys926 sits in a negatively charged pocket of Pwl2; Val931 and Tyr932 correspond to OsHIPP43 Val56 and Thr57, both hydrogen bonding to Pwl2 Arg44, whereas other alleles carry His931 and usually Pro932 whose bulky side chains clash with Arg44.
Evidence codon-based alignment of Mla alleles; structural modelling and interface analysis
Source Figures 3A, 3C, 3D; figures S9-S11
MLA3 was the only Mla allele predicted by AlphaFold2 to bind Pwl2 with model confidence above 0.75, averaged over three independent prediction rounds.
Evidence AlphaFold2 modelling of the LRR domain of all known Mla alleles with Pwl2, scored as (0.8 x ipTM) + (0.2 x pTM), three replicates
Source Figure 3B; figure S8
A single substitution transfers recognition between alleles. The MLA23 chimera carrying the MLA3 Swap1 region gained Pwl2 response and the reciprocal MLA3 chimera lost it. Narrowing further, the S1a region alone sufficed, and the single D926K substitution in MLA23 was enough to confer Pwl2 recognition, while K926D in MLA3 abolished it. Of all 19 possible substitutions at position 926 in MLA3, only K926R retained recognition, showing a positive charge is required. S1b contributes but is not sufficient alone. All chimeras and mutants accumulated in planta.
Evidence reciprocal domain swaps and saturating single-residue substitution; cell death assays with three biological replicates and at least six technical replicates each; immunoblotting
Source Figures 3E-3G; figures S12-S16
Transplanting the interface into rye SR50 creates a dual-specificity receptor. SR50 shares 77% identity with MLA3, and its AvrSr50 specificity maps to the ascending lateral chain of the LRR up to position 870, distinct from the C-terminal Pwl2 site. Both the SR50-3Cterm and the narrower SR50-3BI chimeras gained Pwl2 recognition, and SR50-3BI retained AvrSr50 recognition. Gain of recognition correlated with gain of Pwl2 binding in planta.
Evidence chimeric receptor construction; cell death assays; co-immunoprecipitation; AlphaFold2 modelling of SR50-3BI with both effectors
Source Figures 4A-4F; figures S17-S19
The engineered receptor confers resistance to two pathogens in planta. Transgenic barley lines expressing SR50-3BI under the MLA promoter were resistant to a Pgt isolate carrying AvrSr50, comparable to wild-type SR50, and both chimeras conferred resistance to M. oryzae carrying Pwl2. Infections were repeated at least twice across independent transgenic lines.
Evidence transgenic barley lines challenged with Puccinia graminis f. sp. tritici isolates F2B126 (AvrSr50) and 07WA140-151_QCMJC (avrSr50), and with M. oryzae KEN54-20 and a pwl2 mutant
Source Figures 4G and 4H; figures S19-S21
NLRs
| Name | Class | Role in this paper |
|---|---|---|
| MLA3 Mildew locus a 3, Mla3, RGH1 allele | singleton subject | central subject; the receptor shown to mimic HIPP43 and thereby bind Pwl2 directly |
| MLA23 Mla23 | singleton subject | the closest non-recognising allele; the D926K substitution alone gives it Pwl2 recognition, making it the key specificity control |
| MLA allelic series MLA1, MLA6, MLA28, Mla family | singleton comparator | the allelic background against which MLA3's three unique interface residues are defined |
| SR50 Sr50, SR50-3Cterm, SR50-3BI | singleton subject | the bioengineering chassis; receives the MLA3 Pwl2 binding interface and becomes dual-specificity |
Effectors
| Name | Pathogen | Role |
|---|---|---|
| Pwl2 | Magnaporthe oryzae | the recognised effector; a MAX-fold effector whose host-target binding surface MLA3 mimics |
| AvrSr50 | Puccinia graminis f. sp. tritici | the native SR50 effector; its recognition is retained in the dual-specificity chimera, proving the two interfaces are independent |
| AVRa3 | Blumeria graminis f. sp. hordei | the Bgh effector recognised by MLA3, whose identity remains unknown. Noted as unlikely to be MAX-fold, since the Bgh repertoire is not predicted to contain MAX effectors |
Structures
| Entry | State | Description |
|---|---|---|
| AlphaFold2 predicted | 1:1 heterocomplex | MLA3 LRR domain in complex with Pwl2; best-ranked model pLDDT 84.4, pTM 0.859 |
| PDB 8R7A (external reference) | 1:1 heterocomplex | Experimental structure of Pwl2 in complex with its rice host target OsHIPP43; the comparator that establishes the mimicry |
| AlphaFold2 predicted | 1:1 heterocomplex | LRR domains of all known MLA alleles in complex with Pwl2, three replicates each; only MLA3 exceeded a confidence score of 0.75 |
| AlphaFold2 predicted | 1:1 heterocomplex | SR50-3BI LRR domain modelled separately with AvrSr50 and with Pwl2, showing two spatially distinct binding interfaces on one LRR |
Open questions
- The identity of AVRa3, the Blumeria graminis f. sp. hordei effector recognised by MLA3, is unknown. Since the Bgh repertoire is not predicted to contain MAX-fold effectors, MLA3 probably recognises it through a different interface, but this is untested.
- How widely interface mimicry occurs across NLRs is unknown; the paper raises it as a question rather than answering it.
- Whether other NLRs conferring multiple disease resistance do so through multiple LRR binding interfaces is proposed but not demonstrated beyond the engineered case.
- There is no experimental structure of MLA3 with Pwl2. The mimicry claim rests on an AlphaFold2 model compared to an experimental structure of the other complex, supported by reciprocal mutagenesis.
- Whether Pwl2 binding at the MLA3 C terminus sterically interferes with the nucleotide-binding domain to promote activation, as proposed, is a modelling inference.
- Whether the engineered SR50-3BI retains durable resistance in the field, or imposes a fitness cost, is not addressed.
Connections in this corpus
| Paper | Why |
|---|---|
madhuprakash-2026-avrcap1b-tol9a | Two opposite readings of the same structural logic. Here an NLR evolves to mimic an effector's virulence target so that it gets bound; there an effector binds a host trafficking protein and a helper NLR to shut immunity down. Both resolve the decisive surface structurally and both use interface residues to explain specificity between close relatives, MLA3 against MLA23 and P. infestans AVRcap1b against its P. ipomoeae ortholog. |
guo-2026-ccg10-resistosome | Both are cereal NLR papers using structure to explain a receptor property, and they share the MLA system as a reference point: Guo cites the MLA13-AVRa13 heterodimer as evidence for alternative activation modes, which is the same allelic family worked on here. Guo also uses Sr35, a wheat CC-NLR closely comparable to the rye SR50 used here as a bioengineering chassis. |
toghani-2026-sensor-helper-interface | Both take a predicted interface, validate it by reciprocal exchange between two proteins, and then use it to engineer new specificity. Toghani reciprocally charge-swaps a salt bridge to rescue sensor-helper signalling and expands lettuce helper compatibility; this paper reciprocally exchanges interface residues between MLA3 and OsHIPP43 and expands SR50 effector recognition. Same experimental grammar applied to receptor-receptor and receptor-effector contacts. |
gao-2026-vnt1-splicing | The two non-Kamoun papers of the corpus, and both concern how NLR recognition is unlocked rather than how networks are wired. Gao finds a prodomain that must be spliced out before Rpi-vnt1.1 can assemble; this paper finds an interface that must be present before MLA3 can bind. Both also engineer or exploit the controlling element. |
seager-2026-avrcap1b-nrc3 | Both dissect a protein-protein interface by mutagenesis and then confirm it by a gain-of-function swap between orthologs: Seager transplants LWY7 from P. infestans AVRcap1b into the non-suppressing P. ipomoeae ortholog, this paper transplants the Pwl2 interface from MLA3 into SR50. The reciprocal gain-of-function is the shared standard of proof. |
ibrahim-2026-nrg1-chloroplast | Both use AlphaFold modelling to explain a functional property from structure and then test it by targeted substitution, Ibrahim on membrane targeting through the extended fourth helix, this paper on effector binding through three C-terminal LRR residues. Worth contrasting on confidence: this paper's MLA3-Pwl2 model scores pTM 0.859 against Ibrahim's ipTM 0.43 for the NRG1 resistosome. |
Provenance and caveats
Not a Kamoun-lab paper. Kamoun and lab members (Sugihara, Toghani, Posbeyikian, Ludke, Contreras) are thanked in the acknowledgements for discussions only. The About page should be clear that inclusion in this corpus reflects topical scope, not authorship.
There is no experimental structure of MLA3 with Pwl2. The mimicry claim rests on an AlphaFold2 model (pLDDT 84.4, pTM 0.859) compared against the experimental Pwl2-OsHIPP43 structure (PDB 8R7A), backed by reciprocal mutagenesis in both directions. The mutagenesis is what carries the claim; the model alone would not.
This is the only paper in the corpus using AlphaFold2 rather than AlphaFold3. Confidence metrics are therefore not directly comparable with the AlphaFold3 scores reported elsewhere, and the composite score used here, (0.8 x ipTM) + (0.2 x pTM), differs from the raw ipTM values quoted in the other papers. Any cross-paper comparison of model confidence must say so.
Residue numbering differs across MLA alleles: MLA3 Lys926 corresponds to MLA1 Glu926 and MLA23 Asp926, while the Val931/Tyr932 pair corresponds to MLA28 His931/Arg932, MLA6 His929/Leu930 and MLA23 His930/Pro931. Do not present these as different positions.
The DOI recorded for reference 27 (AlphaFold2, Jumper et al. 2021) was inferred from the standard citation rather than read from the PDF, which prints no DOIs. Verify at Stage 3.
Sources
- Full citation: Gomez De La Cruz D, Ingram T, Zdrzalek R, Taylor J, Wawryk-Khamdavong A, Bachowska K, Banfield MJ, Talbot NJ, Moscou MJ (2026) Molecular mimicry of a pathogen virulence target by a plant immune receptor. Science 392:1050-1055.
- DOI: 10.1126/science.aef9946
- PDF on file:
papers/science.aef9946.pdf - Extraction:
data/papers/gomezdelacruz-2026-mla3-mimicry.json(schema v1.0) - References: 46 works, retrieved from OpenAlex
- Licence: [UNVERIFIED] - not stated on PDF