PREreview of Unlocking expanded flagellin perception through rational receptor engineering

by Andrés Posbeyikian, Mauricio P. Contreras, Yordan Dolaptchiev, AmirAli Toghani, Emma Turley, Barbara Schrammel, and Hsuan Pai of the TSL Preprint Club

Published: December 2, 2024
DOI: 10.5281/zenodo.14262617
License: CC BY 4.0

General Remarks

The study titled 'Unlocking Expanded Flagellin Perception Through Rational Receptor Engineering' by Tianrun Li et al. investigates the engineering of the FLAGELLIN-SENSITIVE 2 (FLS2) receptor to broaden its recognition spectrum of bacterial flagellin-derived peptides. By employing a combination of diversity analyses, AlphaFold modeling, and amino acid property assessments, the authors identified key residues on the concave surface of FLS2 that interact with both the co-receptor and polymorphic flg22 residues. Subsequent synthetic biology approaches enabled the transfer of expanded recognition capabilities from Quercus variabilis FLS2 (QvFLS2) and Vitis riparia FLS2^XL to homologs with narrower ligand specificities, thereby enhancing the receptors’ abilities to detect diverse flagellin epitopes. Overall, the manuscript presents a coherent narrative, effectively detailing the methodologies and outcomes of the receptor engineering process.

The manuscript tells a straightforward and engaging story, making it easy to follow for readers. The experiments are well-structured, with clear figures that are thoughtfully arranged to support the narrative. Overall, the writing is polished and effectively communicates the study’s findings. Below, we include some suggestions that we believe could enhance the manuscript. We hope these are useful to the authors.

Figures

Figure 1 and Extended Data Figures 1-2

Including data on the transient overexpression of NbFLS2 (in the Nicotiana benthamiana FLS2 knockout background) would be beneficial, as it could address potential differences in FLS2 accumulation between wild-type N. benthamiana and CRISPR knockouts overexpressing other variants.
Alongside, Western blot analyses for a tagged NbFLS2 in Extended Data Figure 2c would confirm comparable expression levels, ensuring that observed phenotypic differences are due to receptor specificity rather than differences in protein accumulation.
Typo on text line 75 - there is no Extended Data Figure 1c.

Figure 2 and Extended Data Figure 4

The manuscript would benefit from a clearer delineation of the relative contributions of Repeat Conservation Mapping (RCM) and AlphaFold predictions when selecting amino acids for mutation. Currently, few mutations correspond to regions highlighted by RCM in Figure 2a. Reordering the narrative to instead present AlphaFold predictions before the RCM might enhance coherence. This applies particularly to Figures 2 and 3.
In Figure 2b, the Predicted Aligned Error (PAE) data for the AlphaFold-predicted complex is not shown; it would have been beneficial to include it for added clarity and depth of analysis. When conclusions are drawn about the relative positioning of residues based on AlphaFold models, it is essential to include PAE data to substantiate these claims and provide supporting evidence. This could also offer additional dimensions to the analysis, potentially clarifying which interfaces are more reliable, guiding the engineering efforts.
In Figure 2c, the representation of single-residue swaps is ambiguous, as the number of lines or their positioning on the individual LRR repeats does not correspond proportionally to the swapped residues. A more precise depiction (for example, a condensed version of Extended Data Figure 4b) would facilitate better understanding of the mutations' impact. These remarks also apply for Figure 3c.
In Figure 2e, the MAPK assay is very convincing, and in good agreement with the ROS assay data shown in Figure 2d. The fact that the ROS assay is normalized makes us wonder if the absolute intensity of responses triggered by the SynFcFLS2’s are lower or higher than those of QvFLS2 or FcFLS2.. For example, is this baseline consistently stronger or weaker when Pae is delivered? Could the data be normalised to the water control instead?
Have the authors conducted any AlphaFold modeling with the different peptides, and could this somehow explain the lack of recognition of the peptides Atu and Pvi in the synthetic receptors?

Have the authors considered painting the information from the RCM plot onto the structure? This may be a more intuitive way of visualizing the information.
In the Extended Data Figure 4b, could the authors clarify what the navy, yellow and white colour-coding means? This also applies to Extended Data Figure 5b.

Figure 3

The observation that transferring 30 residues from FLS2XL to VrFLS2 (= SynVrFLS2³⁰) does not disrupt VrFLS2's perception of Ddi, whereas SynVrFLS2²² (which shared more residues with VrFLS2 than SynVrFLS2³⁰) effectively loses this recognition, is intriguing. Similarly, none of the SynVrFLS2’s can perceive Eam, despite both ‘parent’ receptors triggering a response to this peptide in Figure 3d. A discussion exploring potential reasons for these unexpected outcomes would enrich the manuscript.

Figure 4

We personally felt that the PCA analysis in Figure 4 lacks a clear question and takeaway, making its relevance to the engineering efforts ambiguous. Thus, currently, the analysis of amino acid properties as a whole does not seem to contribute significantly to the overall findings of the paper and may be more appropriate as supplementary material. Clarifying whether this analysis is descriptive or predictive, and how it informs the engineering process, would strengthen the manuscript.

Figure 5

Incorporating PAE data and depositing models as supplementary materials would enhance transparency and facilitate follow-up analyses by readers. In previous modelling experiments, the co-receptor was included, predicting a tripartite complex between FLS2, peptide and co-receptor. Was the co-receptor excluded from the analysis conducted across Figure 5 for any specific reason? If included, how might these results be affected?

Figure 6

In the x-axis of Figure 6a, ‘LRR domains’ should be changed to ‘LRR’ or ‘LRR number’ if referring to individual repeats. This also applies to the figure legend.
Providing a full structural footprint rather than zoomed-in views would improve clarity. In subpanels d and e it could be made clearer which codons the asterisks indicating significance are pointing to.
It would be more informative to indicate the number of sequences and/or genomes used per taxa in either the main text or the methods section.

Discussion

Based on AlphaFold models of QvFLS2, this study identified 15 residues as being critical for binding the Rso_1 flg22 peptide. However, the experimental results in Figure 2 indicate that transferring these residues into the FcFLS2 background did not yield the anticipated recognition of Rso_1 flg22. This discrepancy highlights the limitations of structural modeling and underscores the need for empirical validation. A discussion addressing these limitations would provide valuable insights into the challenges of receptor engineering.
We noted that while some engineering attempts did not yield expected outcomes (see notes on Figure 3 above), these results were not explored further in the discussion. We personally think highlighting these findings could be useful to guide future engineering attempts.