A potato late blight pathogen effector interacts with ENTH-domain protein TOL9a and an activated helper NLR to suppress immunity
AVRcap1b binds the ENTH domain of the host trafficking protein NbTOL9a through one end and activated NbNRC2 through another, and needs both to suppress immunity, which makes it a bridge rather than an inhibitor.
AVRcap1b suppresses NRC-mediated cell death but does not touch resting NRC2, which left the mechanism unexplained. The crystal structure of AVRcap1b bound to the ENTH domain of NbTOL9a settles part of it. The effector adopts an L-shaped fold, unlike the linear PSR2, and the bend comes from a single module, LWY5, rather than from any loss of the conserved inter-repeat features that were thought to enforce linearity. WY1 and LWY2 grip the ENTH domain across roughly 800 A2. Mutating that interface costs both binding and suppression. But binding NbTOL9a is not enough: the P. ipomoeae ortholog binds it just as well and suppresses nothing. AVRcap1b engages sensor-activated NbNRC2 through a separate interface, and the P92E mutant retains that engagement while losing NbTOL9a. Two contacts, one bridge.
Key findings (10)
AVRcap1b binds NbTOL9a through its N-terminal ENTH domain specifically. Chimeric TOL proteins showed that swapping the NbTOL9a ENTH domain into the non-binding NbTOL6 confers AVRcap1b association, and the interaction was reconstituted in vitro with purified components.
Evidence in planta co-immunoprecipitation with NbTOL9a/NbTOL6 domain chimeras; analytical SEC (complex peak shift to 13.7 ml) confirmed orthogonally by SEC-MALS
Source Figures 1A-1C; figure S1
The crystal structure of AVRcap1b in complex with the NbTOL9a ENTH domain reveals an L-shaped fold, distinct from the linear stick-like architecture of PSR2, the only other structurally characterised RXLR-LWY effector. The model covers AVRcap1b residues 78-675 (most of WY1 and all of LWY2-LWY7 bar the last three residues) plus the full ENTH domain.
Evidence X-ray crystallography
Source Figure 2A; figure S2; table S1; movies S1 and S2; PDB 9RDC
The bend comes from LWY5, which introduces a 77 degree angle in the backbone. LWY5 has the highest pairwise RMSD against every other repeat (mean 6.08 A within AVRcap1b, 6.83 A against PSR2 repeats), while comparisons excluding LWY5 average below 4.83 A.
Evidence pairwise structural superposition of individual LWY modules
Source Figure 2B; figures S3, S4; table S2
The L-shape is not caused by loss of the features thought to enforce linearity. The Loop4-5 element and the four conserved hydrophobic pocket residues proposed to hold PSR2 straight are conserved throughout LWY2-LWY7 of AVRcap1b. Those features are therefore not sufficient to produce a linear arrangement.
Evidence multiple sequence alignment of AVRcap1b LWY domains against the PSR2 model
Source Figure S5
L-shaped effectors form a distinct family across the genus Phytophthora. A BLAST search returned 184 LWY effectors splitting into two clades, one of 152 sequences clustering with AVRcap1b. AlphaFold 3 models of representatives gave L-shaped topology for every AVRcap1b-clade member and stick-like topology for every PSR2-like sequence. The one exception was a truncated P. infestans ortholog (XP_002896947.1) lacking LWY6 and LWY7, which modelled as a short stick.
Evidence phylogenetic analysis of 184 LWY effectors; AlphaFold 3 modelling (AVRcap1b pTM 0.72, RMSD 1.05 A against crystal; PSR2 pTM 0.67, RMSD 1.10 A; clade 1c orthologs pTM >= 0.65)
Source Figures 3A and 3B; figure S6
The interface buries about 800 A2 on AVRcap1b, with the alpha1 and alpha2 helices of the ENTH domain inserting into a shallow concave pocket formed by WY1 and LWY2. Six AVRcap1b residues are implicated: R90, P92, G94, K98 from WY1 and S139, S143 from LWY2, contacting ENTH residues M5, R8, L14, I15 (alpha1) and D18, A20, M21, D24, D27, I28 (alpha2).
Evidence interface analysis of the co-crystal structure; the authors caution that resolution does not support accurate side-chain modelling
Source Figures 4A and 4B
AVRcap1b binding selectivity across the NbTOL family tracks polymorphism at these interface residues. AVRcap1b binds NbTOL9a strongly, NbTOL9b and NbTOL9c weakly, and NbTOL6 and NbTOL3 not at all, matching sequence differences at the key contact positions.
Evidence sequence comparison of ENTH domains against previously reported binding data
Source Figure 4C
Mutating the NbTOL9a-binding interface compromises both binding and suppression of NbNRC2-mediated cell death, so ENTH binding is required for full virulence activity.
Evidence co-immunoprecipitation of AVRcap1b variants; quantitative cell death scoring (0-7 scale) in nrc2/3/4 knockout N. benthamiana, three biological replicates, BestHR bootstrap analysis with 1000 iterations
Source Figures 5A-5C
NbTOL9a binding is necessary but not sufficient. Orthologs from the closely related clade 1c species P. mirabilis, P. ipomoeae and P. andina share 88.5-91% amino acid identity and preserve all six key ENTH-binding residues, yet only P. infestans AVRcap1b suppresses NbNRC2. The P. ipomoeae ortholog binds NbTOL9a as well as the P. infestans protein does.
Evidence cross-species ortholog suppression assays with matched expression; co-immunoprecipitation
Source Figures 6A-6C
AVRcap1b associates specifically with Rx/CP-activated NbNRC2, not with the resting state, through an interface distinct from the NbTOL9a one. AVRcap1bP92E fails to bind NbTOL9a but retains association with activated NbNRC2EEE, while the P. ipomoeae ortholog binds NbTOL9a but never NbNRC2 in either state. The two associations are not mutually exclusive, pointing to a ternary complex.
Evidence co-immunoprecipitation under resting and Rx/PVX-CP-activated conditions, repeated three times
Source Figure 7; figure S9
NLRs
| Name | Class | Role in this paper |
|---|---|---|
| NRC2 NbNRC2, NbNRC2EEE (MADA mutant) | helper subject | the suppressed helper; AVRcap1b binds its activated form only, through an interface separate from the NbTOL9a-binding one |
| Rx | sensor subject | activating sensor used throughout to switch NbNRC2 into its activated state |
| NRC3 NbNRC3 | helper cited-only | second known AVRcap1b target from prior work; the subject of the companion paper |
Effectors
| Name | Pathogen | Role |
|---|---|---|
| AVRcap1b | Phytophthora infestans | central subject; crystallised with the NbTOL9a ENTH domain and shown to require both contacts for immune suppression |
| AVRcap1b orthologs | Phytophthora ipomoeae, P. mirabilis, P. andina (clade 1c) | the specificity control that carries the argument: they bind NbTOL9a and preserve all six interface residues yet suppress nothing, separating binding from suppression |
| PSR2 | Phytophthora sojae | the only previously solved RXLR-LWY effector structure; the linear comparator against which the L-shape is defined (PDB 5GNC) |
| PVX coat protein | Potato virus X | activating trigger in the Rx/NRC2 system |
| SS15 | Globodera rostochiensis | contrasting mechanism discussed, not assayed: SS15 locks resting NRCs, AVRcap1b acts only on activated ones |
Structures
| Entry | State | Description |
|---|---|---|
| PDB 9RDC | 1:1 heterocomplex | AVRcap1b (residues 78-675) in complex with the NbTOL9a ENTH domain; L-shaped effector fold with a 77 degree bend at LWY5; interface buries ~800 A2 on AVRcap1b |
| PDB 5GNC (external reference) | monomer | P. sojae PSR2, linear RXLR-LWY effector used as the structural comparator |
| AlphaFold3 predicted | monomer | Representative LWY effectors from both clades across Phytophthora species, plus AVRcap1b and PSR2 as validation controls; AVRcap1b pTM 0.72 with RMSD 1.05 A to the crystal structure, PSR2 pTM 0.67 with RMSD 1.10 A |
Open questions
- How AVRcap1b association with NbTOL9a leads mechanistically to immune suppression is still unknown; the bridge model is inferred from two separate binding interfaces, not from a resolved ternary complex.
- Why the P. ipomoeae ortholog fails to suppress NbNRC2 despite binding NbTOL9a and preserving all six interface residues is unexplained. The determinant of NRC2 binding has not been mapped.
- Whether AVRcap1b acts on intermediate stages of NRC activation or on the mature hexameric resistosome is not resolved here; the companion paper addresses this directly.
- The activities of the wider L-shaped effector family are unknown. The NbTOL9a ENTH-binding interface is not conserved across the clade, so most members probably have different targets.
- Crystallographic resolution does not support accurate side-chain modelling, so the specific hydrogen bonds proposed at the interface are probabilistic rather than definitive.
- Whether AVRcap1b or other multi-repeat RXLR-LWY effectors adopt a linear form under other conditions remains open.
Connections in this corpus
| Paper | Why |
|---|---|
seager-2026-avrcap1b-nrc3 | Explicitly named as the companion study (reference 50), and the two must be read together. Madhuprakash establishes that AVRcap1b binds NbTOL9a and activated NbNRC2 through separate interfaces and needs both; Seager shows what the NRC-side contact does, stalling assembly at three protomers. The shared P92E result is the pivot: TOL9a-binding-deficient but still NRC-engaging in Madhuprakash, still partially suppressing in Seager. Shared authors: Seager, Contreras, Kamoun. |
guo-2026-ccg10-resistosome | Guo cites Madhuprakash et al. 2024 for NRC helper oligomerisation. Both papers depend on the resting-homodimer-to-hexamer model of NRC2 activation, Guo as a structural comparator and Madhuprakash as the state distinction that makes activation-dependent effector binding interpretable. |
toghani-2026-sensor-helper-interface | Both hinge on the conformational difference between resting and activated NRC2. Toghani defines the transient sensor-helper interface that produces the activated state; Madhuprakash shows an effector that recognises only that state. Same Rx/NRC2 system in both. |
toghani-2026-sni-undecamer | Shared use of AlphaFold 3 to survey structural variation across a protein family and then test predictions experimentally. Madhuprakash applies it to effector folds, Toghani to resistosome stoichiometry, with the same logic of prediction-then-validation. |
ibrahim-2026-nrg1-chloroplast | Ibrahim proposes organellar targeting as a possible route around effectors that suppress plasma-membrane NLR function; Madhuprakash characterises exactly such an effector and shows it co-opts a membrane trafficking pathway. Shared authors Yuen and Bozkurt. |
gomezdelacruz-2026-mla3-mimicry | Two contrasting effector-receptor logics. Gomez De La Cruz has an NLR evolving to mimic an effector's virulence target so it gets bound; Madhuprakash has an effector binding a helper NLR to shut it down. Both resolve the recognition surface structurally, and both use the interface to explain specificity between close relatives. |
Provenance and caveats
The authors state explicitly that crystallographic resolution does not permit accurate side-chain modelling of the complex, so the listed hydrogen-bond partners at the AVRcap1b-NbTOL9a interface are probable contacts, not confirmed ones. Any wiki statement naming specific residue pairs must carry that caveat.
Reference numbers marked [UNVERIFIED] were inferred from in-text citation context, not read from the reference list. Only reference 50 (Seager) is confirmed, because it is named in the Discussion text.
Sequence identity figures (88.5, 90, 91%) are stated in the text without being individually assigned to P. mirabilis, P. ipomoeae and P. andina in the extracted passage. [UNVERIFIED - do not attribute a specific percentage to a specific species without checking Figure 6.]
Sources
- Full citation: Madhuprakash J, Toghani A, Pai H, Harvey M, Bentham AR, Seager BA, Yuen ELH, De la Concepcion JC, Lawson DM, Stevenson CEM, Vergara-Cruces A, Derevnina L, Bozkurt TO, Banfield MJ, Kamoun S, Contreras MP (2026) A potato late blight pathogen effector interacts with ENTH-domain protein TOL9a and an activated helper NLR to suppress immunity. Sci Adv 12:eaea4500.
- DOI: 10.1126/sciadv.aea4500
- PDF on file:
papers/sciadv.aea4500.pdf - Extraction:
data/papers/madhuprakash-2026-avrcap1b-tol9a.json(schema v1.0) - References: 76 works, retrieved from OpenAlex
- Licence: CC BY