Multiple modes of cholesterol translocation in the human Smoothened receptor

Smoothened (SMO), a member of the G Protein-Coupled Receptor superfamily, mediates Hedgehog signaling and is linked to cancer and birth defects. SMO responds to accessible cholesterol in the ciliary membrane, translocating it via a longitudinal tunnel to its extracellular domain. Reaching a complete mechanistic understanding of the cholesterol translocation process would help in the development of cancer therapies. Competing hypotheses based on available structures support entry of cholesterol from outer and inner membrane leaflets, but the exact mechanism of translocation remains unclear. Using atomistic molecular dynamics simulations (∼2 millisecond simulations) and biochemical assays of SMO mutants, we assess the energetic feasibilities of proposed hypotheses. We show that the energetic barriers for cholesterol translocation from either leaflets are comparable. Mutagenesis experiments and complementary simulations of SMO mutants validate the role of critical amino acid residues along the translocation pathways. Our data suggests that cholesterol can take either pathway to enter SMO, thus explaining contradictory experimental observations in literature. Thus, our results illuminate the energetics and provide a first molecular description of cholesterol translocation in SMO.

Methods
Dataset was collected using Long-Timescale Molecular Dynamics Simulations (MD). Simulations were started from four starting points that consisted of Smoothened (SMO) protein embedded in a membrane. Each starting point contained the cholesterol at a different site in SMO. Simulations were performed in an adaptive manner, consisting of multiple rounds of sampling with hundreds of trajectories run in parallel for each round. After every round of sampling, we gathered the entire dataset until that point, clustered the trajectory frames according to metrics that would differentiate between the different stages of cholesterol transport. Then simulations were seeded from clusters with the least population, and the next round of simulations was run. In total, the entire dataset consists of 2 ms of unbiased simulations. The files deposited in this repository contain the .npy files computed from each MD trajectory, which were used to plot the free energy landscapes in the manuscript. Simulations were run using AMBER and OpenMM MD simulation engines.

Another set of biased simulations were run with different SMO mutants to describe the overall transport process for cholesterol, and compared to WT SMO. The .pmf files generated by NAMD are submitted to this repository for easy replication. 

Python was extensively used to analyse the dataset generated. This work has been performed by Prateek Bansal and Diwakar Shukla at the University of Illinois Urbana-Champaign.

This Dryad dataset has been made available in compliance with the "minimum dataset"/source data requirement per the Data Availability guidelines by Nature Communications.