Publications

Cholesterol is a key modulator of membrane protein structure and function, yet predicting cholesterol-binding sites remains challenging due to its non-druglike physicochemical properties. Here, we curated more than 800 high-resolution transmembrane protein structures containing cholesterol, and developed an interpretable, atom-based graph neural network, called CholBindNet. A positive-unlabeled (PU) training strategy was employed to address the scarcity of negative samples due to the promiscuous nature of cholesterol binding. We show that CholBindNet substantially outperforms existing machine learning models trained on general ligand-binding datasets, including AlphaFold3, P2Rank, and DiffDock. The performance and generalizability of the model on unseen membrane proteins were further demonstrated by rapidly assessing cholesterol-binding sites in the PIEZO2 ion channel against all-atom molecular dynamics (MD) simulations conducted on Anton3 supercomputer. Additionally, strong model interpretability was achieved for CholBindNet through atom-level feature encoding, Grad-CAM visualization, and attention-based scoring analysis. Overall, CholBindNet provides an efficient and scalable approach for classifying and ranking cholesterol-binding sites on membrane proteins, achieving performance comparable to computationally expensive MD simulations while offering rich biophysical insights into the atomic-level spatial patterns beyond amino-acid sequence. This work lays the foundation for future deep-learning models targeting membrane protein drug-binding sites and cholesterol-modulated therapeutics.

PIEZO proteins (PIEZO1 and PIEZO2) are essential mechanosensitive channels. PIEZO1 is thought to be selectively activated by Yoda molecules (Yoda1 and Yoda2). Although a structural framework for PIEZO1 activation by Yoda1 exists, a molecular mechanism underlying this selective activation is lacking. Here, using electrophysiology and calcium imaging, we show that Yoda1 increases PIEZO2 open probability and stretch sensitivity as efficaciously as PIEZO1 but elicits weaker PIEZO2-dependent calcium entry, rationalizing why its effect on PIEZO2 has been overlooked. Both Yoda1 and its more potent Yoda2 analog slow down inactivation of PIEZO2 currents with potency similar to PIEZO1 but with lower efficacy. Using mutagenesis and molecular dynamics simulations, we further show that Yoda2's benzoic acid group forms a transient salt bridge with a conserved arginine in the Yoda binding site, providing a molecular basis for Yoda2's increased potency. Our study cautions a reevaluation of studies using these molecules to untangle biological functions mediated by PIEZO channels.

Mechanosensitive PIEZO channels are thought to open via tension-induced flattening of peripheral transmembrane arm domains, yet the structural basis of this activation remains unclear. Here, by leveraging hybrid-resolution molecular dynamics simulations, we uncover how large-scale PIEZO2 arm movements funnel into subtle gating motions in the central pore under physiological tension. Arm flattening correlates with anticlockwise rotation of the pore relative to the arms and with clockwise twisting of inner pore helices. These clockwork motions open the pore in a two-step fashion, yielding a fully conducting state and a stable subconducting state populated at a low tension, which was detected electrophysiologically. The fully open PIEZO2 pore is walled by both lipids and amino acids and recapitulates minimal pore size, conductance, ion selectivity and outward rectification of chloride currents measured electrophysiologically. These findings provide structural insights into PIEZO2 gating and demonstrate hybrid-resolution molecular dynamics as a powerful approach to study large-scale membrane protein dynamics and guide drug discovery.

Membrane curvature is a fundamental property of biological membranes, driving essential processes such as endocytosis, vesicle formation, and mechanotransduction. Molecular dynamics (MD) simulations have become a powerful approach for studying curved membrane systems, providing atomistic insights into curvature-driven phenomena and protein-membrane interactions. However, online platforms like CHARMM-GUI and CGMD focus on constructing flat bilayers or vesicles and lack support for generating curved membranes with defined geometries. Local tools, while more flexible, often do not incorporate protein-specific curvature features, such as those from the Orientations of Proteins in Membranes (OPM) database, which are critical for accurately modeling protein-lipid interactions in curved environments. To address these limitations, we developed Cmem Builder, a novel and user-friendly web server for automating the generation of curved lipid membranes and membrane-protein complexes for coarse-grained (CG) MD simulations using the MARTINI force field. Cmem Builder specializes in generating Z-axis symmetric curved membrane shapes, supports curvature profiles derived from OPM database or custom geometries, allows extensive control over lipid composition, and ensures lipid placement through geometric sampling. The tool has been successfully applied to classical curved membrane systems, including Piezo1 and BAR proteins, as well as plasma membranes with asymmetric lipid compositions, demonstrating its accuracy and efficiency. In total, Cmem Builder provides a robust and accessible platform for exploring the complex dynamics of curved membrane systems. The tool is freely available at https://cmembuilder.com.

GJC2 encodes connexin 47 (Cx47), a gap junction protein expressed by oligodendrocytes that forms gap junction channels (GJCs) between adjacent oligodendrocytes (or astrocytes, via heterotypic Cx47-Cx43 GJCs). Autosomal recessive mutations of GJC2 lead to at least three central nervous system phenotypes: Pelizaeus-Merzbacher-like disease 1 (PMLD1), spastic paraparesis 44 (SPG44), and a minimal leukodystrophy. Here, we describe the clinical, functional, and molecular effects of two mutations in GJC2, p.G40S, and p.R244P, identified in two different families with GJC2-related disorders. Expressed exogenously, p.G40S forms GJC plaques like WT but does not functionally couple with WT nor with Cx43. p.R244P also fails to demonstrate functional coupling. Moreover, plaque formation is absent, concomitant with intracellular connexin accumulation. When the two mutants are co-expressed in a compound heterozygous state, plaques form, but no GJC coupling is detected in any configuration. MD simulations demonstrate that p.G40S modifies secondary structure of the pore-lining α-helix, disrupting supersecondary interactions with the N-terminal helix and predicting channel closure. p.R244P simulations are characterized by partial loss of the extracellular β-sheet domains and a marked reduction of electrostatic interactions between the connexin and lipid headgroups of the plasma membrane, suggesting pathways by which p.R244P mutation impairs GJC formation. This combination of in vitro assays and molecular simulations provides mechanistic insight into the pathogenesis of GJC2-related disease.

Computational methods such as molecular dynamics (MD) have illuminated how single-atom ions permeate membrane channels and how selectivity among them is achieved. Much less is understood about molecular permeation through eukaryotic channels that mediate the flux of small molecules (e.g. connexins, pannexins, LRRC8s, CALHMs). Here we describe computational methods that have been profitably employed to explore the movements of molecules through wide pores, revealing mechanistic insights, guiding experiments, and suggesting testable hypotheses. This review illustrates MD techniques such as voltage-driven flux, potential of mean force, and mean first-passage-time calculations, as applied to molecular permeation through wide pores. These techniques have enabled detailed and quantitative modeling of molecular interactions and movement of permeants at the atomic level. We highlight novel contributors to the transit of molecules through these wide pathways. In particular, the flexibility and anisotropic nature of permeant molecules, coupled with the dynamics of pore-lining residues, lead to bespoke permeation dynamics. As more eukaryotic large-pore channel structures and functional data become available, these insights and approaches will be important for understanding the physical principles underlying molecular permeation and as guides for experimental design.

Since its inception nearly a half century ago, CHARMM has been playing a central role in computational biochemistry and biophysics. Commensurate with the developments in experimental research and advances in computer hardware, the range of methods and applicability of CHARMM have also grown. This review summarizes major developments that occurred after 2009 when the last review of CHARMM was published. They include the following: new faster simulation engines, accessible user interfaces for convenient workflows, and a vast array of simulation and analysis methods that encompass quantum mechanical, atomistic, and coarse-grained levels, as well as extensive coverage of force fields. In addition to providing the current snapshot of the CHARMM development, this review may serve as a starting point for exploring relevant theories and computational methods for tackling contemporary and emerging problems in biomolecular systems. CHARMM is freely available for academic and nonprofit research at https://academiccharmm.org/program.

A major driver of neuronal hyperexcitability is dysfunction of K+ channels, including voltage-gated KCNQ2/3 channels. Their hyperpolarized midpoint of activation and slow activation and deactivation kinetics produce a current that regulates membrane potential and impedes repetitive firing. Inherited mutations in KCNQ2 and KCNQ3 are linked to a wide spectrum of neurodevelopmental disorders (NDDs), ranging from benign familial neonatal seizures to severe epileptic encephalopathies and autism spectrum disorders. However, the impact of these variants on the molecular mechanisms underlying KCNQ3 channel function remains poorly understood and existing treatments have significant side effects. Here, we use voltage clamp fluorometry, molecular dynamic simulations, and electrophysiology to investigate NDD-associated variants in KCNQ3 channels. We identified two distinctive mechanisms by which loss- and gain-of function NDD-associated mutations in KCNQ3 affect channel gating: one directly affects S4 movement while the other changes S4-to-pore coupling. MD simulations and electrophysiology revealed that polyunsaturated fatty acids (PUFAs) primarily target the voltage-sensing domain in its activated conformation and form a weaker interaction with the channel's pore. Consistently, two such compounds yielded partial and complete functional restoration in R227Q- and R236C-containing channels, respectively. Our results reveal the potential of PUFAs to be developed into therapies for diverse KCNQ3-based channelopathies.