Sensor-helper communication in NLR networks

In an NLR network, many sensor receptors detect pathogens and a few helper receptors execute the response. Eight papers here bear on how the signal passes between them, and the answer that emerges is that it passes by contact, briefly, and then the contact is gone.

Why the interaction was hard to see

Co-immunoprecipitation cannot resolve it. Rx gives interaction signals with NbNRC2 and NbNRC4 comparable to those with the unrelated NbZAR1, so the assay cannot separate specific from non-specific binding. That has been the position for roughly a decade (toghani-2026-sensor-helper-interface).

The same absence shows up from the other direction. No Rx sensor was captured in any purified NRC3 complex, by blue native gel or by in planta purification, and the helper assembles its resistosome without it (seager-2026-avrcap1b-nrc3). Two methods, one conclusion: the engagement is transient.

The interface

AlphaFold 3 predicts the complex that the pulldowns could not hold. Ten independent seeds converge on one geometry, mean pTM 0.788 and mean ipTM 0.759, in which the sensor's NB domain grips both the NB and LRR domains of the helper. Together those two interfaces contribute 185 contacts and bury 2026 A2; a third, LRR-LRR interface is smaller and turns out to be dispensable (toghani-2026-sensor-helper-interface).

The prediction is tested rather than admired. Fifteen charged interface residues were charge-reversed and eight abolished helper activation. The decisive experiment is reciprocal: D224K in the sensor is dead, K284E in the helper is dead, and the two together restore cell death. Control predictions against a receptor outside the network never exceed ipTM 0.3, so the method is not producing confident complexes indiscriminately (toghani-2026-sensor-helper-interface).

No experimental structure of any sensor-helper complex exists, here or elsewhere. The interface rests on prediction plus mutagenesis, and the mutagenesis is what carries it (toghani-2026-sensor-helper-interface).

Specificity is a small number of residues

The interface residues split in two. Four are required for activation of NbNRC2, NbNRC3 and NbNRC4 alike; others are helper-specific, so one substitution can keep signalling through one helper while losing another (toghani-2026-sensor-helper-interface).

That granularity matters because helper dependency is otherwise a clean phylogenetic property. In lettuce, group 1 sensors signal only through LsNRC0, group 3 only through Ast-LsNRC1, and group 2 through either — a boundary that selection analysis independently marks, with no positively selected sites in the helpers or in group 2 (pai-2026-lettuce-nrc). Two substitutions in the NB interface move a group 1 sensor across it (toghani-2026-sensor-helper-interface).

What the helper does next

Sensor binding is predicted to prime rather than to assemble. The modelled sensor-bound helper conformation sits between the resting homodimer and the active hexamer in terms of NB-HD1 rotation (toghani-2026-sensor-helper-interface). An effector that recognises only the activated helper, never the resting one, is consistent with a distinct intermediate existing (madhuprakash-2026-avrcap1b-tol9a).

Helpers differ in what else they need. NRC helpers oligomerise without partner proteins; NRG1 requires the EDS1-SAG101 module (seager-2026-avrcap1b-nrc3, ibrahim-2026-nrg1-chloroplast). Whether that distinction or compartment choice is the deeper split between helper classes is open.

One awkward observation sits against all of this: in legumes, TIR-NLRs are predominantly root-expressed while the CCR-NLR helpers they signal through are predominantly leaf-expressed (marques-2026-legume-tissue). A dependency in which sensor and helper are in different organs is not obviously a dependency.