Avalilação PREreview de Dual recognition of structurally unrelated mildew effectors underlies the broad-spectrum resistance of Pm3e in wheat

General Assessment

The manuscript presents novel findings on Pm3d and Pm3e, demonstrating broad-spectrum resistance and dual recognition of unrelated effectors. The authors combine virulence profiling with 92 isolates (global collection), mapping in biparental populations to identify effectors, analysing haplotype variability of effector candidates, and performing NLR chimera analyses. The work provides mechanistic insight into the recognition specificity of Pm3d and Pm3e, involving both RALPH-like and AvrPm3e_1-like effectors.

A particularly interesting contribution is the suggestion that effector stability may explain why previous studies failed to detect these AVR candidates (e.g., BgtE-20069b being recognised by the Pm3e allele). To overcome this effector’s instability, the authors fused effector candidates with a highly soluble mRFP fluorophore. Another notable observation is the potential antagonism between Pm3d and Pm3e alleles when two NLRs are co-infiltrated, leading to interallelic suppression effects.

Overall, the study is strong, with clear evidence of effector variability present in virulent/avirulent isolates to overcome resistance conferred by Pm3 alleles. However, the lack of functional validation of the chimeras presented in this work—either through wheat transgenic plants or cell biology approaches such as microscopy, co-IP, or mass spectrometry—remains a limiting factor. The title’s claim of broad-spectrum resistance is also weakened by the lack of emphasis on the geographical origins of the isolates used.

Major Comments

1. Biological relevance and subcellular localisation

The BgtE-5754 effector localisation remains unclear. The authors showed that the signal peptide (S.P.) was poorly predicted and that removing the “predicted” signal peptide leads to a loss of HR. This is important, as most studies usually remove the S.P. before performing co-infiltration assays. The authors should clarify whether BgtE-5754 is assumed to act in the apoplast or cytoplasm, which could be “easily” predicted by EffectorP and might help draw some conclusions. Please include experimental or predictive evidence, or at least a discussion of the expected localisation. This is important to understand the biological relevance of the recognition.

2. Consistency in AlphaFold usage

There is inconsistency between the terms AlphaFold, AlphaFold2, and AlphaFold3 (e.g., Fig. 3F vs. line 231 vs. line 242). The manuscript should use a single consistent naming scheme or clarify why different AlphaFold versions were used at each step. There are major differences in how AlphaFold predicts structures from version 2 to version 3. Additionally, more methodological detail is required: seeds, recycles, templates, parameters, and the rationale behind model choice. The general impression is that the authors included a large amount of structural prediction work but did not use that information to draw major conclusions, so some of it feels a bit “lost.” While the authors present the predicted structures, they do not provide key confidence metrics such as the Predicted Aligned Error (PAE) plot or pTM/pLDDT values, which are necessary to evaluate the reliability of the model, domain orientations, and inter-domain confidence.

3. Missing data on protein stability and interactions

Throughout the manuscript, effector–NLR interactions are inferred mostly from HR responses. In some parts, the authors included western blots to check protein stability, but in other parts they did not. Several conclusions about critical amino acids would be strengthened by demonstrating:

• protein stability (e.g., western blots for all variants).

• whether the effector is still properly folded.

• protein–protein interaction assays (if feasible).

4. Need for structural analysis of chimeras

Given the emphasis on antagonism and synergistic effects among chimeric NLRs, AlphaFold predictions of the chimera structures would greatly support the claims. Although the authors provided evidence from co-infiltrations and F1 phenotyping from allelism tests, we would like to see Pm3d and Pm3e modelled with their corresponding two effectors to visualise how the structures compare. Additionally, some structural comparison with the chimeras that recognise both BgtE-5754 and RALPH effectors would help strengthen the conclusions.

5. Transgenic validation of chimeras

The manuscript claims that some chimeras restrict pathogen penetration and are promising candidates for stable resistance. However, this is based on HR/N. benthamiana assays and haustorial index tests. We strongly recommend including wheat transgenics for at least the key chimeras to truly validate agronomic potential. If this is not possible, the authors should provide more evidence for the haustorial index tests, e.g., some microscopy images that support the conclusions. Given the amount of supplementary data provided for the HR assays, it was surprising that no detailed data were shown for this critical part of the work.

6. Terminology: “engineering”

The manuscript uses “engineering,” but the described work is more classical domain swapping and chimera construction. This may not fit current meaning of “NLR engineering” (AI-guided modelling, rational design, integrated domains, de-novo binding etc.). Consider adjusting terminology.

8. Introduction lacking structural background

The introduction could briefly summarise basic NLR structural organisation to help contextualise the later structural and chimera experiments.

Specific Comments by Section

Textual Comments

• Line “92 Bgt isolates selected …” A short summary describing the geographical origins of these isolates would help, instead of making the reader check previous references. The authors mention that only a few isolates overcome Pm3d and Pm3e alleles and then use those to develop the biparental populations to map the effectors. It would be good to name these isolates in this section beforehand, so the reader becomes familiar with the key isolates early on. Also, include clear terms such as “resistant,” “susceptible,” and “leaf coverage,” possibly supported by an illustrative image.

• Line 231: Clarify whether AlphaFold or AlphaFold3 was used.

• Line 242: States AlphaFold2 was used for predicting 112 proteins. This contradicts Fig. 3F.

• Lines 263–303: For the co-infiltration assays of AvrPm3e_2, please clarify whether the SP was removed, as done earlier for other effectors. Consistency is important. We assume the SP for AvrPm3e_2 was robustly predicted and that removing it may have led to a loss of HR in co-infiltration assays, but this is not explained. Please comment.

• Line 404: Specify the chimera leading to the conclusion that “W657 alone is sufficient”. Likely Pm3_Ch2-HA, but please name it explicitly.

• Lines 408–410: Again, specify the exact chimera (likely Pm3_Ch13-HA).

• Line 513: “cluster 40” – possible typo?

• Lines 710, 732: Strange “squares” appear, likely formatting artifacts.

• Line 731: Indicate the Bgt conidiospore concentration used.

Figure-specific Comments

Figure 1

·       Clarify phenotyping statistics: number of leaves, replicates, independent experiments.

·       AlphaFold terminology inconsistent.

·       Pm3CS#19 etc. — the #19 identifier is unnecessary here; can be described in Methods.

Figure 3

·       Panel B: “structural model” is misleading; this is a schematic. Consider modifying the term.

·       Increase size of labels marking mutations.

·       Keep AlphaFold3 naming consistent across panels.

·       Panel F: show confidence scores and discuss reliability of AF3 predictions.

·       Panels G and I appear swapped.

·       Western blot: effector variants should include the appropriate tags consistently. They have not tested the tagged effectors back in co-infiltrations assays.

·       Supplementary Fig. 4 should include protein levels for all variants.

Figure 4

• If mRFP is used to stabilise effectors, also include mRFP alone as a control.

• Panel B: add amino acid numbering at the start/end of domain representations.

• Panel E seems dispensable.

Figure 6

• Panel B: bold the key chimeras (Pm3_Ch1-HA, Ch7-HA, Ch13-HA).

• Increase clarity of mutation labels; consider diagonal text or spacing to avoid overlap.

• Because Pm3d and Pm3e mutations (e.g., V1332) are close visually, increase bar spacing or change colouring (e.g., dashed lines).

• Panel C: the highlighted orange is difficult to see; looks more red.

• Panel D: yellow is hard to visualise; consider higher contrast colours. Increase image size slightly.

Competing interests

The authors declare that they have no competing interests.

Use of Artificial Intelligence (AI)

The authors declare that they did not use generative AI to come up with new ideas for their review.