Avalilação PREreview de High–resolution cryo-EM structures of small protein–ligand complexes near the theoretical size limit

This exciting study from Park et al. represents a significant milestone for the cryo-EM community that we have eagerly anticipated since the authors first shared their preliminary findings on Twitter in March 2023. The work demonstrates that single-particle cryo-EM (SPA) can resolve protein structures near the theoretical lower molecular weight limit, and importantly challenges the assumption that commonly used structural fiducials are universally required for reconstruction of small proteins and protein-ligand complexes. The authors used single particle cryo-EM analysis to determine the structures of maltose-bound MBP at a reported 2.32 Å resolution and onvansertib-bound PLK1 at 3.04 Å. These notable and intriguing achievements serve to push the perceived limits of single-particle cryo-EM forward, and we are excited about the prospects posed by this study.

The presented study beautifully demonstrates how traditional cryo-EM hardware and software can be used on well-prepared samples with careful analysis to extend the limits of SPA cryo-EM towards increasingly smaller targets. This of course carries important implications for future drug development. To ensure the cryo-EM community can fully benefit from these important advances and build upon this foundational work, we offer the following constructive suggestions to strengthen the manuscript before publication:

Major points:

1. The greatest novelty of this study and its potential benefit for the cryo-EM field lies in our ability to reproduce the presented methods. Thus, to maximize the impact of this work for our community, the authors should provide more detailed and comprehensive description of their methods. All relevant sample preparation, data collection, and processing steps should be highlighted, including any non-default processing parameters used for achieving the MBP reconstruction. Understanding which processing steps contributed most significantly to resolution improvements would be invaluable for researchers attempting to apply these methods to their own small protein targets.

2. Despite their best efforts, the reported reconstruction of PLK1 suffers from strong preferred orientation (as evidenced in Fig. S4C), meaning that the resolution of the reconstruction is not isotropic, and thus the global and local resolutions of the structures by FSC are over-estimated. The presented data shown in the manuscript is not of sufficient quality to support confident model building of side-chains nor ligands (contrary to the authors' statement: line 100). We encourage the authors to employ established methods to overcome preferred orientation, such as using support layers (graphene), incorporating secondary detergents (LMNG, DM, OG, etc.). Given the low molecular weight of the PLK1, tilted data collection is unlikely to yield usable data due to increased ice thickness, but it's worth trying. Addressing the preferred orientation problem to present a directionally isotropic reconstruction would substantially strengthen the conclusions regarding PLK1 structure determination.

3. Notably, the MBP comparison involves a potentially unsuitable x-ray structure of an MBP-fusion construct (8IIZ), for which the authors cannot exclude conformational changes induced by the construct mismatch rather than differences attributable to methodology. We suggest performing this comparison using a construct of MBP only (e.g., 1ANF) and including an X-ray structure of apo-MBP (e.g., 1JW4) to better visualize maltose-induced changes and provide a more appropriate evaluation.

4. Most importantly, while the authors have successfully resolved MBP to high resolution (a remarkable achievement!) the manuscript would benefit from clearly articulating a reproducible and broadly applicable strategy for resolving small proteins to high resolution. Given the community's enthusiasm for implementing these approaches, we encourage the authors to explicitly state what differentiates their successful attempts from earlier efforts by providing detailed comparisons of instrumentation (including the specific effects of their 300 keV TEM with Falcon 4 and energy filter), data collection settings (including usage of low defocus: -0.4 to -1.8 µm, along with statistics of which particles from defocus bins contributed to the final reconstruction, helping us to better understand how low contrast conditions can lead to higher quality outcomes), and sample preparation protocols (including their choice of grids and vitrification procedures).

5. Access to the curated stack of movies would be beneficial for the community and software developers, enabling the development of novel processing routines specifically tailored for small proteins and advancing the field's collective capabilities. The datasets should be made available on EMPIAR.

To reiterate, we are excited about these important advancements and look forward to seeing how this work will inspire and enable future breakthroughs in structural biology of small proteins and drug targets.

Minor points:

1. Abstract: Rephrase to emphasize that the structure determination of small proteins does not necessarily require structural fiducials to aid alignment.

2. Motivation for the study requires clarification: Current drawbacks in drug development using x-ray crystallography should be stated more clearly, including the forced symmetry for assemblies and multiple ligand binding conformations, artifacts from crystal contacts, or the absence of conformational dynamics.

3. Fig. 1H and Fig. 2H, the legend for the color key (C-alpha RMSD) should be added to improve clarity. The authors should use RMSD over RMS (figure legend). A small area in Fig. 1H shows higher RMSD values. Also please add the atomic model of ligands to show where this region is positioned. What is the relevance of this high-RMSD region in terms of potential ligand binding, catalysis, regulation or other protein-protein interactions?

4. Fig. S1&3 C: The particles are not visible. Please add a more strongly low-pass filtered image as a new panel to enhance visibility. Denoising (like cryoSPARC’s denoising model) might be another option. The scale bar should be included as well.

5. Given the uncertainty in reported resolution, the authors should refrain from using more than one significant digit in the main text, except where reporting FSC values in the methods.

6. Line 64: Reference to SEC chromatogram in Fig. S2A is missing. Also missing for Fig. S4A (line 96).

7. Line 68. Based on Fig. S3 B, micrographs were curated based on CTF metrics or by hand. Please provide applied cutoffs and state the number of accepted micrographs.

8. Line 71 and Tab S1.: There is a high clash score for both models (>10). The authors are encouraged to rebuild their models.

9. Line 73: It is not immediately clear which comparison was used; please provide PDB IDs. Also provide map-model fit metrics for ligands (Q-score) and revisit placement of ligands and water using appropriate cutoffs with tools like Phenix phenix.douse. Over-sharpening tends to favor false-positive placement of waters (https://phenix-online.org/phenixwebsite_static/mainsite/files/presentations/water.pdf). Additionally, the authors should provide a rationale for why they built the water molecules, and if they have physiological relevance for the MBP function. Are there differences to the X-ray reference structure in terms of structured water molecules?

10. Line 146: Add appropriate citation for the used Addgene construct.

11. Line 95: Please provide a rationale for using the Baculovirus expression system.

12. Line 163: What is the relevance of the selected mutation T120V? It was not mentioned in the main text. Please specify.

13. Provide Uniprot IDs for proteins of interest.

14. Fig. S2/4: Please include the cFAR FSC report from cryoSPARC and the confidence plot of angular assignment.

15. Fig. S4: The selected view of the PLK1 reconstruction is particularly misleading and does not fully represent the extent of anisotropy present (an orthogonal view should also be shown in this case).

16. Table. S1: Please provide number of accepted movies, not only collected, as well as the resolution range. The number of waters for MBP does not match the reported number in line 73. Also include the Q-score and EM-Ringer scores.

Competing interests

The authors declare that they have no competing interests.