Harnessing AlphaFold to reveal hERG channel conformational state secrets

Reviewer #1 (Public review):

Summary:

Ngo et. al use several computational methods to determine and characterize structures defining the three major states sampled by the human voltage-gated potassium channel hERG: the open, closed, and inactivated state. Specifically, they use AlphaFold and Rosetta to generate conformations that likely represent key features of the open, closed, and inactivated states of this channel. Molecular dynamics simulations confirm that ion conduction for structure models of the open but not the inactivated state. Moreover, drug docking in silico experiments show differential binding of drugs to the conformation of the three states; the inactivated one being preferentially bound by many of them. Docking results are then combined with a Markov model to get state-weighted binding free energies that are compared with experimentally measured ones.

Strengths:

The study uses state-of-the art modeling methods to provide detailed insights into the structure-function relationship of an important human potassium channel. AlphaFold modeling, MD simulations, and Markov modeling are nicely combined to investigate the impact of structural changes in the hERG channel on potassium conduction and drug binding.

Weaknesses:

(1) The selection of inactivated conformations based on AlphaFold modeling seems a bit biased. The authors base their selection of the "most likely" inactivated conformation on the expected flipping of V625 and the constriction at G626 carbonyls. This follows a bit of the "Streetlight effect". It would be better to have selection criteria that are independent of what they expect to find for the inactivated state conformations. Using cues that favour sampling/modeling of the inactivated conformation, such as the deactivated conformation of the VSD used in the modeling of the closed state, would be more convincing. There may be other conformations that are more accurately representing the inactivated state. I see no objective criteria that justify the non-consideration of conformations from cluster 3 of the inactivated state modeling. I am not sure whether pLDDT is a good selection criterion. It reports on structural confidence, but that may not relate to functional relevance.

(2) The comparison of predicted and experimentally measured binding affinities lacks an appropriate control. Using binding data from open-state conformations only is not the best control. A much better control is the use of alternative structures predicted by AlphaFold for each state (e.g. from the outlier clusters or not considered clusters) in the docking and energy calculations. Using these docking results in the calculations would reveal whether the initially selected conformations (e.g. from cluster 2 for the inactivated state) are truly doing a better job in predicting binding affinities. Such a control would strengthen the overall findings significantly.

(3) Figures where multiple datapoints are compared across states generally lack assessment of the statistical significance of observed trends (e,g. Figure 3d).

(4) Figure 3 and Figures S1-S4 compare structural differences between states. However, these differences are inferred from the initial models. The collection of conformations generated via the MD runs allow for much more robust comparisons of structural differences.

Reviewer #2 (Public review):

Summary:

Ngo et al. use AlphaFold2 and Rosetta to model closed, open, and inactive states of the human ion channel hERG. Subsequent MD simulations and comparisons with experiments support the plausibility of their models.

Strengths:

This is thorough work studied from many different angles. It provides a self-consistent picture of how conformational changes in hERG may affect its function and binding to different targets.

Weaknesses:

Though this work claims the methodologies can be generalized to other systems, it is not obvious how. Many modeling choices seem arbitrary and also seem to have required extensive expert knowledge of the system. This limits the applicability of the modeling strategy.

Reviewer #3 (Public review):

Summary:

The authors use Alphafold2, Rosetta, and Molecular Dynamics to model structures of the hERG K channel in open, inactive, and closed states. Experimental CryoEM data for open hERG (Wang and Mackinnon 2017), and closed EAG (Mandala and Mackinnon, 2002) were used as the main templates for channel models presented here. Given the importance of hERG as a safety pharmacology target, the identification of a robust simulation method to assess drug block is an important addition to the field.

Strengths

The key findings here are new inactivated and closed hERG channel conformations and hERG channel conformations with drugs docked in the inner vestibule below the selectivity filter. Amino acid pathways and interaction networks for different states are also presented.

The inactive state and drug block models are carefully correlated with experimental data for the inactivated state of hERG (Lau et al, 2024) and with experimental free energy data for drug binding and have overall good agreement.

It is remarkable that using cytoplasmic domain structures of hERG as a starting point revealed inactivation state structures in the hERG selectivity filter in Figures 2,3.

Weaknesses

Figure 6, if each data point is for a different drug, then perhaps identify each point.

The PAS domain was not included in the models as stated in Methods page 14 but the PAS does appear in some of the templates used as starting points for models in Figure 1 a,b,c. Perhaps mentioning that the PAS was not included in some (all?) of the final models should be moved into the main text and discussed.

The drug block of 1b channels (which do not contain PAS) has been reported to be slightly different than that for 1a channels (which contain PAS) and for 1a/1b channels (see London et al., 1997; https://doi.org/10.1161/01.RES.81.5.870 and Abi-Gerges et. al., 2011; DOI: 10.1111/j.1476-5381.2011.01378.x) and this should be discussed since the models presented here appear to be performed in the absence of the PAS.

It also appears that the N-linker region (between PAS and the S1) and distal C region of hERG (post CNBHD-COOH) are not included in models, please state this if correct, and discuss.

Author response:

Reviewer #1:

Summary:

Ngo et. al use several computational methods to determine and characterize structures defining the three major states sampled by the human voltage-gated potassium channel hERG: the open, closed, and inactivated state. Specifically, they use AlphaFold and Rosetta to generate conformations that likely represent key features of the open, closed, and inactivated states of this channel. Molecular dynamics simulations confirm that ion conduction for structure models of the open but not the inactivated state. Moreover, drug docking in silico experiments show differential binding of drugs to the conformation of the three states; the inactivated one being preferentially bound by many of them. Docking results are then combined with a Markov model to get state-weighted binding free energies that are compared with experimentally measured ones.

Strengths:

The study uses state-of-the art modeling methods to provide detailed insights into the structure-function relationship of an important human potassium channel. AlphaFold modeling, MD simulations, and Markov modeling are nicely combined to investigate the impact of structural changes in the hERG channel on potassium conduction and drug binding.

We appreciate the reviewer’s recognition of our integration of state-of-the-art computational methods, including AlphaFold2, Rosetta, MD simulations, and Markov modeling. We are pleased that the reviewer found our approach to investigating the structure-function relationship of the hERG channel insightful.

Weaknesses:

(1) The selection of inactivated conformations based on AlphaFold modeling seems a bit biased. The authors base their selection of the "most likely" inactivated conformation on the expected flipping of V625 and the constriction at G626 carbonyls. This follows a bit of the "Streetlight effect". It would be better to have selection criteria that are independent of what they expect to find for the inactivated state conformations. Using cues that favour sampling/modeling of the inactivated conformation, such as the deactivated conformation of the VSD used in the modeling of the closed state, would be more convincing. There may be other conformations that are more accurately representing the inactivated state. I see no objective criteria that justify the non-consideration of conformations from cluster 3 of the inactivated state modeling. I am not sure whether pLDDT is a good selection criterion. It reports on structural confidence, but that may not relate to functional relevance.

We acknowledge the concern regarding the selection criteria for the inactivated state models. In the revised manuscript version, we plan to broaden our selection approach and explicitly include conformations from different clusters beyond those highlighted in the initial submission (e.g., from cluster 3). We will also incorporate structural metrics that do not solely depend on the known channel inactivation hallmarks or reply on the pLDDT scores to further justify our chosen representative inactivated state models.

(2) The comparison of predicted and experimentally measured binding affinities lacks an appropriate control. Using binding data from open-state conformations only is not the best control. A much better control is the use of alternative structures predicted by AlphaFold for each state (e.g. from the outlier clusters or not considered clusters) in the docking and energy calculations. Using these docking results in the calculations would reveal whether the initially selected conformations (e.g. from cluster 2 for the inactivated state) are truly doing a better job in predicting binding affinities. Such a control would strengthen the overall findings significantly.

We agree that a more rigorous control for our drug-binding predictions is desirable. To address this, we will include molecular docking simulations and associated drug binding affinity estimations for more hERG channel models, including alternate conformations from the initial clustering that were not chosen as the final models. This will allow us to test whether our inactivated state structure from cluster 2 indeed outperforms or differs significantly from other possible inactivated hERG channel conformations in reproducing experimental drug potencies.

(3) Figures where multiple datapoints are compared across states generally lack assessment of the statistical significance of observed trends (e,g. Figure 3d).

(4) Figure 3 and Figures S1-S4 compare structural differences between states. However, these differences are inferred from the initial models. The collection of conformations generated via the MD runs allow for much more robust comparisons of structural differences.

We will incorporate statistical analyses and measures of uncertainty for key comparisons. In Figures 3 and S1-S4 the consensus structural hERG channel models for open, inactivated and closed states are being compared, i.e. one representative model for each state. We believe this is a valid comparison, and the statistical analysis of the observed trends based on those models (e.g., in the bar plot of Figure 3d) alone might not be possible. However, we agree with the reviewer that instead of relying solely on those initial static models, we will also draw on the ensemble of states sampled during the MD simulations to quantify structural differences between different putative hERG channel states. Specifically, we will present ensemble-averaged measurements and highlight how these distributions differ significantly between states.

Reviewer #2:

Summary:

Ngo et al. use AlphaFold2 and Rosetta to model closed, open, and inactive states of the human ion channel hERG. Subsequent MD simulations and comparisons with experiments support the plausibility of their models.

Strengths:

This is thorough work studied from many different angles. It provides a self-consistent picture of how conformational changes in hERG may affect its function and binding to different targets.

We are grateful for the reviewer’s recognition of the thoroughness and multi-faceted nature of our study.

Weaknesses:

Though this work claims the methodologies can be generalized to other systems, it is not obvious how. Many modeling choices seem arbitrary and also seem to have required extensive expert knowledge of the system. This limits the applicability of the modeling strategy.

We appreciate the reviewer’s comment on the generalizability of our approach. In the revision, we will more explicitly discuss the rationale behind the modeling choices and the extent to which they reflect system-specific knowledge. We will clarify how the strategies we developed (e.g., iterative refinement with AlphaFold2 and Rosetta, followed by MD simulation validation) can be adapted to other ion channels or related proteins. We will also outline a more generalizable workflow, specifying which steps require system-specific information and which steps are broadly applicable.

Reviewer #3:

Summary:

The authors use Alphafold2, Rosetta, and Molecular Dynamics to model structures of the hERG K channel in open, inactive, and closed states. Experimental CryoEM data for open hERG (Wang and Mackinnon 2017), and closed EAG (Mandala and Mackinnon, 2002) were used as the main templates for channel models presented here. Given the importance of hERG as a safety pharmacology target, the identification of a robust simulation method to assess drug block is an important addition to the field.

Strengths

The key findings here are new inactivated and closed hERG channel conformations and hERG channel conformations with drugs docked in the inner vestibule below the selectivity filter. Amino acid pathways and interaction networks for different states are also presented.

The inactive state and drug block models are carefully correlated with experimental data for the inactivated state of hERG (Lau et al, 2024) and with experimental free energy data for drug binding and have overall good agreement.

It is remarkable that using cytoplasmic domain structures of hERG as a starting point revealed inactivation state structures in the hERG selectivity filter in Figures 2,3.

We thank the reviewer for highlighting the novelty and importance of our work, particularly regarding the identification of new inactivated and closed hERG channel conformations and the modeling of drug block. We are also pleased that the reviewer found the correlation with experimental data to be strong and the structural insights to be valuable.

Weaknesses

Figure 6, if each data point is for a different drug, then perhaps identify each point.

Thank you so much for this suggestion. Please note that Table 3 contains drug-specific data plotted in Figure 6 including drug names. We will provide a reference to Table 3 in the revised Figure 6 caption. We will also revise Figure 6 (and any similar figures) to clearly identify each data point with the corresponding drug and/or include a corresponding key in the Figure legend. This will make it easier to correlate each data point’s binding prediction with the experimental datasets.

The PAS domain was not included in the models as stated in Methods page 14 but the PAS does appear in some of the templates used as starting points for models in Figure 1 a,b,c. Perhaps mentioning that the PAS was not included in some (all?) of the final models should be moved into the main text and discussed.

The drug block of 1b channels (which do not contain PAS) has been reported to be slightly different than that for 1a channels (which contain PAS) and for 1a/1b channels (see London et al., 1997; https://doi.org/10.1161/01.RES.81.5.870 and Abi-Gerges et. al., 2011; DOI: 10.1111/j.1476-5381.2011.01378.x) and this should be discussed since the models presented here appear to be performed in the absence of the PAS.

It also appears that the N-linker region (between PAS and the S1) and distal C region of hERG (post CNBHD-COOH) are not included in models, please state this if correct, and discuss.

We appreciate the reviewer’s insightful comment regarding the PAS domain and the potential influence of other regions, such as the N-linker and distal C-region, on hERG channel drug binding and state transitions.

The PAS domain did appear in the starting templates used for initial structural modeling (as shown in Figure 1a, b, c), but it was not included in the final models used for subsequent analyses. Similarly, the N-linker and the distal C-region were also omitted from the final models. These omissions were primarily due to hardware constraints used for AlphaFold structural modeling, as including these additional protein regions would exceed the memory capacity of graphical processing unit (GPU) cards on our available intramural, external and cloud high-performance computing resources, leading to failures during the protein structure prediction step.

The PAS domain of hERG 1a isoform, even if not serving as a direct drug-binding site, can influence the gating kinetics of hERG channels as the reviewer pointed out. By altering the probability and duration with which those ion channels occupy specific conformational states, it can indirectly affect how well drugs bind. For example, if the presence of the PAS domain shifts channel gating so that more channels enter (and remain in) the inactivated state, drugs with a higher affinity for that state would appear to bind more potently, as observed in electrophysiological experiments. It is also plausible that the PAS domain could exert allosteric effects that alter the conformational landscape of the ion channel during gating transitions, potentially impacting drug accessibility or binding stability. This is an intriguing hypothesis and an important avenue for future research.

With access to more powerful computational resources, it would be valuable to explore the full-length hERG 1a channel, including the PAS domain and associated regions, to assess their potential contributions to drug binding and gating dynamics. We will incorporate a discussion of these points into the main text, acknowledging the limitations of our current models, citing the references provided by the reviewer, and highlighting the need for future studies to explore these protein regions in greater detail.