Escribir un comentario

In this exciting study, Ahn and colleagues address two questions surrounding the Solanaceous coiled-coil NLR NbPtr1: how an autoactive receptor is held in check in the absence of pathogens, and how a single NLR comes to recognise six sequence-unrelated effectors from three bacterial genera. Through a genome-wide search and a transient autoactive cell death suppression screen, they identify a clade of small, single C-NOI-domain proteins (NbNOI-1, -4, -6, -8, -9) that suppress NbPtr1 autoactivity through direct interaction. Silencing these NbNOIs produces NbPtr1-dependent stunting and lethality, and CRISPR knockout of the nbnoi1,6 pair is rescued by loss of NbPtr1, establishing genetically that these proteins are bona fide negative regulators. The recognised effectors modify the NbNOIs (proteolytic cleavage by AvrRpt2 and RipE1, ADP-ribosylation by AvrRpm1) and disrupt the NbNOI-NbPtr1 association, providing a mechanistic link between negative regulation and effector sensing. Modeling with AlphaFold 3 combined with structure-guided mutagenesis pinpoints a conserved threonine (equivalent to AtRIN4 Thr166) docking into an ARC2 pocket formed by Glu431, Pro433 and Val477, and mutagenesis on both sides of the interface plus co-immunoprecipitation support a “ball-and-socket” interface.

We thoroughly enjoyed reading and discussing this manuscript. It tells a clear and compelling story, makes orthogonal use of genetics, biochemistry and structural prediction, and is for the most part nicely controlled. The comments below are offered in the spirit of being useful; we want to emphasise that the manuscript is already of very high quality, and we hope these suggestions help the authors strengthen it further.

General Comments

• The integration of genetics, biochemistry and AlphaFold 3-guided structural work into a single coherent narrative is a real strength, and the result is a compelling story that was nice to read.

• We found the evolutionary framing of decoy diversification especially appealing, as a lens on how non-orthologous NLRs (NbPtr1, RPM1 and RPS2) can converge on the same family of host proteins for surveillance and how one NLR can recognise multiple unrelated effectors.

• Guardee or decoy? The title and model frame the suppressor NbNOIs as decoys, that is, proteins whose primary role is recognition rather than an independent cellular function such as PTI regulation (the defining feature of RIN4 as a true guardee). However, this distinction is not directly tested, and we do not think a decoy assignment can yet be made. Although the fresh-weight measurements in the ptr1 background show no significant phenotype on silencing, which is consistent with a decoy interpretation, a non-significant result does not exclude an independent function, and the silenced plants do appear slightly smaller in the photos. A direct assay of basal immunity (ROS burst, MAPK activation, callose, or bacterial growth) with NOI knockout or knockdown in the absence of Ptr1 would test whether these NbNOIs regulate PTI and thereby distinguish guardee from decoy. Given that the decoy-diversification concept is a conceptual centrepiece, even a modest experiment here would substantially strengthen the framing; alternatively, the guardee/decoy language could be tempered.

• Structural and mechanistic dissection. The structure-guided portion toward the end of the paper is elegant and underpins much of the proposed activation mechanism, but the central interface model is of modest initial confidence. The authors do an excellent job of testing it experimentally; our comments under Figure 5 suggest ways to present this confidence transparently and to validate the interface more directly.

Figure 1 (NOI family identification and phylogeny)

• The Fig. 1A phylogeny is built with FastTree; using a model-based method such as IQ-TREE with branch support values would place the tree on firmer footing.

• It is not clear from the main text how the NOI domain was defined or how sequences were curated, and in particular whether the initial search, which filters on RIN4-specific features (the cleavage motif, RSM2, the Thr166-equivalent and a palmitoylation motif), may have excluded divergent NOI-domain proteins that lack these features. If so, this could bias both the family definition and the diversification narrative, and it would be worth stating the criteria and their potential effect explicitly.

• The readers would benefit from a stated basis for how NOI orthology was established (reciprocal best hits, synteny or phylogenetic orthology inference); if this was established in prior work, a pointer to that justification would suffice.

Figure 2 (silencing phenotypes; and associated supplements)

• Fig. 2C. The legend defines the values as NbPR1 expression relative to EV-VIGS Nb-1, but this could be made clearer in the panel itself, and error bars / a statistical comparison across the biological replicates would strengthen the very large fold-changes reported.

Figure 3 (effector-induced modifications of NbNOIs; and associated supplements)

• Fig. 3A. The VIGS background (NbNOI-1,4,6,8,9-VIGS) is given in the legend but could usefully be indicated in the panel itself. Two specificity controls would also be informative: the same experiment without NbPtr1 complementation, and the ptr1 background without NbNOI VIGS.

• Direct evidence for the full modification set, and a positive control. Cleavage (AvrRpt2, RipE1) and ADP-ribosylation (AvrRpm1) of the NbNOIs are demonstrated directly, which is excellent. The remaining modifications central to the model, acetylation by HopZ5 and AvrBsT and rhamnosylation by AvrB, are inferred from prior AtRIN4 work rather than shown on the NbNOIs here. The HopZ5 co-IP disruption (Fig. 4B) partly covers HopZ5, but we would suggest either providing direct biochemical evidence for these modifications on the NbNOIs or tempering the statements where they are currently presented as established. Including AtRIN4 (or an NbRIN4 ortholog) as a positive control in the cleavage and modification assays would also benchmark the readouts against a well-characterised substrate.

• Fig. 3C-E. The catalytic-dead effector mutants already serve as the unmodified reference and show that the NbNOIs accumulate well, which is the key control. An additional EV/GFP lane would nonetheless help anchor the unshifted band position.

• Detecting cleavage products. Because the AvrRpt2/RipE1 cleavage site sits close to the N-terminus, a larger N-terminal tag would let the cleaved N-terminal fragment be visualised directly, distinguishing cleavage from general degradation rather than relying on loss of full-length signal. Alternatively, C-terminal tagging might also allow the detection of a bandshift upon N-terminal cleavage.

• F11A rationale. The rationale for the NbNOI-1 F11A mutant (the conserved Phe of the cleavage motif) could be introduced more explicitly where it first appears (around line 228).

Figure 4 (NbPtr1-NbNOI interaction and NbPtr1 activation; and Fig. S9)

• In the co-IP (Fig. 4A), NbNOI-5 serves as the closely related non-suppressor control. NbNOI-7 is similarly close to the suppressor clade and also fails to suppress, so including it as a second interaction control would have further strengthened the specificity argument. We were wondering if there was a particular reason why this was not used?

• MADA / activation wording. Supplementary Fig. 9 already provides a helpful MADA-region alignment of NbPtr1 against ZAR1 and NRC4. It would be worth stating explicitly whether NbPtr1 scores as a MADA-type CNL (for example via running a MADA-HMM), and phrasing the activation statement around line 259 as oligomerisation into a ZAR1-like resistosome rather than the more generic “oligomerization.” Would also be super interesting to see what an NbPtr1 resistosome would look like when modeling CC-NBARC domains as pentamers with 50xOLA and 5xATP with AlphaFold 3.

Figure 5 (AlphaFold 3 model of the NbPtr1-NbNOI interface; and associated supplements)

• Reporting model confidence. We would encourage the authors to report the full confidence picture: Best practices for reporting AlphaFold models call for including PAE plots and pLDDT-coloured structures for the model shown (at minimum as a figure supplement). Also, the per-contact (interface) PAE for the specific Thr-EPV contacts highlighted in Fig. 5B, and the metrics across all seeds rather than the single best model should be reported. The Methods state that three seeds were used and that residues recurring “more than five times” were selected; it would help to state explicitly how many predictions were generated in total (the AlphaFold Server returns five models per seed, so presumably fifteen) and whether the interface was consistent across them, including the mean ipTM across the best models from each seed. Colouring the NbPtr1 domains (CC, NB, ARC1, ARC2, LRR) separately in Fig. 5A would also make the docking geometry much easier to interpret. The reported ipTM for the NbPtr1-NbNOI-8 model is 0.43, which is modest for a predicted interaction. Although the authors do an excellent job of experimentally testing this prediction, it would be good to be upfront about the fact that it is not initially of high confidence. It is actually an interesting observation that AF3 seems to not be confident about an interaction that is actually biologically relevant. Did the authors try repeating the modeling with more recent template databases? Along this same train of though, did the authors try modeling with AF2-Multimer? The MLA3-Pwl2 complex modeled in Gomez De La Cruz et al. (2026) yields extremely high confidence in AF2 Multimer (ipTM of around 0.8) but gives very low confidence when modeled with AF3 in roughly 9 out of 10 seeds. Might be worth giving it another go with AF2!

• Orthogonal validation of the interface, and resolving close non-suppressors. The current support for the model is essentially loss-of-function: mutating the conserved Thr, or the EPV pocket, abolishes suppression and interaction. This is consistent with the model but does not discriminate it from more general destabilisation. Two experiments would substantially raise confidence in this prediction. Ideally, a gain-of-function test: the model implies that interface residues, rather than the Thr alone, determine compatibility, since NbNOI-5 carries the conserved Thr but neither binds nor suppresses. Grafting suppressor-clade interface residues onto an incompatible NOI such as NbNOI-5 or NbNOI-7 to confer suppression of NbPtr1 would be a powerful validation. Alternatively, and perhaps more easily, reciprocal charge-swap or salt-bridge complementation across the Ptr1-NOI interface would be extremely compelling. Swapping the charge of each of the side chains involved in a salt bridge at the Ptr1-NOI interface should break the interaction, but swapping the charge of both side chains at the same time should restore it. We recently found some success with this approach in a recent work (See https://doi.org/10.64898/2026.05.18.725810 , particularly Figure 5). Relatedly, an alignment focused specifically on the NOI-Ptr1 interfaces across compatible and incompatible NOIs would let readers see how divergent these surfaces actually are; the manuscript states there is sequence variation but does not show it at the interface. We would also encourage extending the incompatible-NOI analysis beyond NbNOI-5 by including NbNOI-7 in the modelling (and, as noted under Figure 4, in co-IP). Importantly, the authors note that the NbNOI-5 model is not significantly different from those of the suppressors but do not show these data; making the NbNOI-5 (and ideally NbNOI-7) models available, and stating plainly whether AF3 can or cannot discriminate compatible from incompatible NOIs, would be valuable. If AF3 cannot distinguish them, that is itself an important observation about what can be done with AF3 in this system and is worth mentioning.

Discussion

• We enjoyed the coevolutionary model in Fig. 6. As noted under General Comments, a decoy assignment is not yet established; if the suppressor NbNOIs retain a guardee function, an alternative or additional driver of NOI diversification could be escape from effector targeting, rather than coevolution with the guard NLR alone. A sentence acknowledging this possibility would enrich the Discussion.

• The Du et al. citation on interfamily co-transfer of sensor and helper NLRs feels slightly disjointed; clarifying that the parallel is “transfer of an NLR together with its required partner protein” rather than a literal sensor-helper pairing would help.

Minor and presentation points

• Fig. S1 ordering. The 1xMYC-tagged NbNOI constructs are first described around line 215, but Fig. S1 (which uses them) is introduced earlier at line 142. Either introduce the constructs earlier or cross-reference Fig. S1 at line 215.

• Amino-acid nomenclature. One-letter and three-letter codes are used interchangeably (e.g. Thr166 vs T26A, Glu431/Pro433/Val477 vs EPV); consistency would improve readability.

• AlphaFold 3 is conventionally styled with a space (see Abramson et al. 2024).

• Table reference. The Methods (AlphaFold section) cite Table S8 for the seed values, but the table list indicates Table S9 holds the AF3 interaction-residue data; this looks like a mislabel.

• Reference numbering. The reference list is scrambled around entries 54-59: the Ge et al. 2026 helper-resistosome paper appears under number 57 between refs 54 and 56, no entry is numbered 55, and 57 is used twice. Worth checking against the citation manager, since the in-text citation numbers in this region may no longer resolve correctly.

• Supplementary files. Bundling the figure supplements as a single ZIP file would spare readers many individual downloads on bioRxiv.

Final Remarks

Despite the detailed comments above, we want to emphasise that this is already a high-quality and compelling study that makes an elegant conceptual contribution to how we think about indirect recognition and decoy evolution. We hope our comments are received in the constructive spirit in which they are intended and help strengthen an already excellent manuscript. We thank the authors for posting this work as a preprint, which accelerates dissemination and invites open discussion, and we enjoyed the accompanying thread on X/Bluesky by the first author, which communicated the findings clearly! We greatly enjoyed reading and discussing this work and look forward to seeing it published in its final form.

Conflicto de intereses

Los/as autores/as declaran que no tienen intereses en conflicto.

Uso de Inteligencia Artificial (IA)

Los/las autores/as declaran que no utilizaron IA generativa para concebir nuevas ideas para su revisión.