Escribir un comentario

In the preprint “De novo design of protein nanoparticles with integrated functional motifs” by Haas et al., the authors develop a generalizable method for designing novel protein complexes that integrate a specific functional motif (e.g. scaffolding antigens during vaccine delivery) within their structure. Using this method, they design several symmetric nanoparticles and characterize the resultings structures in depth. Their rigorous characterization revealed that their designs have unique folds and distinct backbones, demonstrating that their method generates novel and highly diverse structures. Furthermore, they observed a variety of hydrophobic and polar interactions across the interfaces, further speaking to the diverse structures their method was capable of generating. Finally, the authors test novel nanoparticles tailored to a specific application. They address whether they can build a nanoparticle scaffold for the influenza virus hemagglutinin tri-head antigen and successfully identify a nanoparticle that evoked a robust immune response in mice. 

Strengths:

We find it compelling that the authors successfully build a scaffold that retains functional motifs throughout its surface which is a nontrivial task. The authors have shown impressive control over geometric assemblies. Their claim of broad applicability within the area of de novo protein design is convincing and it does seem that this method is largely generalizable. Though there are likely to be limitations on the overall size of these nanoparticles and constraints on the sequences that could be adopted. We appreciate the robust characterization of the resulting nanoparticles at such a high level of detail. 

Main concerns:

1. We ask that the authors make the “hit” criteria explicit and quantify outcomes across the entire screening funnel (not just the selected examples. The screening counts are well detailed but it remained unclear to us which metric qualified as “on-target” (SEC profile? DLS monodispersity? nsEM 2D classes? 3D reconstructions?). This part could benefit from a supplemental figure or a flowchart defining each decision point and providing per-construct outcomes (expression/solubility, yield, SEC behavior, DLS PDI, nsEM call: on-target/off-target/mixed/disassembled. The high-resolution structures are beautiful, but it is difficult to judge whether they are representative of the full set of assemblies that appear “on-target” by lower-resolution assays. Add quantitative summaries across a broader set (perhaps summary of AF2 confidence for all tested constructs or report fraction of constructs with detectable alternative assemblies). This will help underscore the generalizability aspect of this workflow. We are also curious as to what criteria were used for “visual inspection” for downselecting outputs for experimental characterization.

2. Reconcile the wording around reported “success rates,” especially the icosahedral rate. The authors report “>10% for icosahedral assemblies” which should be reconciled with the explicit screening numbers in the results, and it would be useful to know whether the denominator excludes non-expressing constructs. We encourage the authors to explicitly state the “% of expressed and SEC-purified constructs,” or “% of all genes synthesized”. 

3. The authors note aggregation shoulders in SEC, disassembly in nsEM for some architectures, and off-target species in a substantial fraction of samples. If possible, relating these outcomes to interface hydrophobicity, buried unsatisfied polar atoms, and predicted shape complementarity would increase confidence for robustness.

4. The immunogenicity data shows robust responses across positive immunogens but not clearly superior functional activity compared with controls. In particular, we are confused as to why responses to HA-Foldon are similar to TH-Cage and HA-Cage. We are also confused as to why I3-326-MI15-FL yields a higher response than I3-326-MI15-TH. 

5. At the end of Figure 3 and their subsection involving the structure determination of the de novo protein nanoparticles the authors extensively discuss the interactions occurring at the interfaces of the designed assemblies. They highlight how different nanoparticles have variation in polar and hydrophobic interactions and then state that these observations motivate further analysis of the design models chosen for experimental characterization. It was not clear why the observed variability drove further analysis of the design models and whether their observations were a feature or a bug of their methodology. We interpreted these findings to mean that the variation in interactions spoke to their ability to develop varied and unique structures that utilized different approaches to achieve the same end. Explaining the broader impact or value of having variability in interactions across their different generated structures would aid in the interpretability of the findings. 

Minor concerns:

• In the subsection titled “Designing tailored protein nanoparticles for specific applications”, the authors state that the design of nanoparticle scaffolds for displaying antigens is a difficult task. This is because certain vaccine targets have symmetry mismatches or geometric constraints that would benefit from a more specifically tailored nanoparticle scaffold. However, it is unclear if their chosen antigen - the HA trihead antigen fulfills any of these difficult to design for criterion. A sentence of rationale would help readability. 

• The authors selected one-component nanoparticles with point group symmetry as demanding targets for their ML-based approach. It is unclear why symmetry is so important in this specific context. Their sentence in the discussion about B-cell response to symmetric compounds would be well suited to the introduction to motivate their design choices. 

• Mainly a curiosity, you note that icosahedral designs favor more hydrophobic interfaces while octahedral designs skew less hydrophobic. Are these differences a fundamental property of the distinct point-group symmetries?

• What is the justification for only using female BALB/cAnNHsd mice in the final figure? It has been shown that there are sex differences in immune response, specifically that female mice exhibit more pronounced immune responses (Sex as a Biological Variable, NIH) (Klein & Flanagan, Sex differences in immune responses, Nat Rev Immunol, 2016). 

• Were off-target states observed in raw micrographs or do any emerge during classification?

• Terminology consistency. Define “building block” vs “subunit” vs “protomer” clearly and use consistently; this will help readers follow when the text refers to the diffused oligomeric component versus the final designed assembly.

• We are curious about the motivation behind designing protomers that were either 150 or 200 residues in length. Were these specific lengths chosen arbitrarily?

• The authors attribute the selective secretion of I3-326-MI15-TH to the “idiosyncratic” nature of nanoparticle secretion. Some of the determinants of scaffold–antigen compatibility would be nice to state here. 

• It would be useful to know the expression levels / purification yields / stability (thermal melting temps) of the nanoparticles. At what scale can these nanoparticles be produced and how stable are they?

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.