Title:	Modelling isoforms and conformations of Kv7 channels through AlphaFold
DNr:	Berzelius-2023-71
Project Type:	LiU Berzelius
Principal Investigator:	Sara Liin <sara.liin@liu.se>
Affiliation:	Linköpings universitet
Duration:	2023-03-31 – 2023-10-01
Classification:	10603
Keywords:

Abstract

The Kv7 family of proteins are ion channels with a key role in cellular signal transduction. Specifically, they facilitate the efflux of potassium ions in-response to a depolarization of the transmembrane potential. Five isoforms termed Kv7.1 to Kv7.5 makeup this family of ion channels and are distributed across a range of tissues where they perform a multitude of physiological roles. This includes Kv7.1 channels in the heart, Kv7.2 and Kv7.3 in the brain, Kv7.4 in the ear and Kv7.5 in smooth muscles. As such, loss of function in Kv7 channels contributes to serious disorders including long-QT syndrome, epilepsy, deafness, and loss of bladder control. This positions them as attractive targets for drug development, especially considering there are no currently approved drugs targeting them. Drug development initiatives have been hampered by off-target effects with the two drugs, Retigabine and Flupirtine being withdrawn from the market as a result. Rational drug development require insight into protein ligand-interactions, therefore, the recent Cryo-EM structures of Kv7.1, Kv7.2 and Kv7.4, represent a significant step in drug development. Crucially however, structures of the Kv7.3 and Kv7.5 are not yet resolved, nor are physiologically relevant heteromeric channels i.e. Kv7.2/Kv7.3 and Kv7.4/Kv7.5. Additionally, the resolved structures lack crucial structural information. Firstly, the voltage sensors of Kv7 channels have only been resolved in the “up” activated conformation and not the “down” deactivated conformation. Secondly, there are regions of the channels that are unresolved due to their inherent flexibility. Together, these characteristics preclude the use of traditional homology modelling tools like Modeller or SwissModel. In this project, we thus propose use of the artificial intelligence program, AlphaFold2 (AF2) which allows for de-novo prediction of protein structure. Models generated by AF2 tend to have similar conformations, recently however, is has been shown that AF2 can predict protein structures in multiple conformations [Del Alamo et al. eLife 2022]. This was achieved by varying the depth of the input multiple sequence alignment. Therefore, AF2 can provide models of the Kv7 channels in multiple ‘up’ or ‘down’ conformations. Additionally, AlphaFold-Multimer [Evans et al. bioRxiv 2022] – an adaptation of AlphaFold2 permits prediction of multimeric proteins with multiple chains and thus ideal for use in predicting the unresolved homomers and heteromers. The proposed work requires the Berzelius supercomputer to run AF2 and generate the multiple models of the Kv7 channels. Subsequently, we will use our existing NAISS allocations within Dardel and Sigma HPC resources to run molecular dynamics simulations of the predicted models. These simulations will be used to validate the models generated by AF2 and identify potentially unique drug binding sites in the Kv7 isoforms. To support experimental validation of the computational predictions our research group has an extensive background in electrophysiology experiments with access to facilities including membrane protein overexpression and two-electrode voltage clamp and automated patch-clamp electrophysiology.