(a) The binding sites of the sterols along the hypothesized tunnel in SMO. Sterol binding sites have been identified deep in the TMD (6XBL) (Qi et al., 2020), at the interface of CRD and TMD (6XBM) (Qi et al., 2020), at the CRD sterol-binding site (5L7D) (Byrne et al., 2016), and in a dualbound mode where cholesterol is bound to both the TMD and CRD (6O3C) (Deshpande et al., 2019). (b) Example simulation system showing SMO (5L7D, cyan) embedded in a membrane (white/magenta). Water is shown as a white surface, while sodium (purple) and chloride (green) ions are shown as spheres. (c) Pathway 1 and Pathway 2 investigate the translocation of cholesterol from the membrane to SMO’s TMD. (d) The Common Pathway follows the translocation of cholesterol from the TMD to the CRD. Snapshots in (a) are made from structures in the PDB, while (b-d) are frames taken from MD simulations.

The molecular events as cholesterol enters the core of SMO’s TMD from the outer leaflet of the membrane.

(a) Free energy plot showing the angle of cholesterol with the x-y plane, the plane of the membrane, versus the z-coordinate of cholesterol for Pathway 1. The pathway followed by cholesterol is α → β → α∗. The experimental structures of SMO are shown as black polygons. (b) Free energy landscape of cholesterol’s y-coordinate plotted versus cholesterol’s x-coordinate. Cholesterol interacts with residues in TM2-TM3 while entering the core TMD of SMO. (c-f) Insets show cholesterol’s interactions with residues at the membrane-protein interface for Pathway 1. (c,e) show cholesterol outside the protein (α), while (d,f) show cholesterol entering the protein (β). All snapshots presented are frames taken from MD simulations.

Effects of mutations along Pathways 1 and 2 on the activation of SMO.

(a) Gli1 mRNA fold changes show the responsiveness of SMO mutants to SHH. Untreated Gli1 levels indicate low SMO activity, while SHH-treated values correspond to the level of SMO activation induced by SHH ligand. A t-test with Welch’s correction was used to compute statistical significance. (P values: untreated vs treated: WT: 1.327 × 10−3, G2.57f V: 9.212 × 10−3, IECL2A: 4.2 × 10−5, A2.60f M: 7.1 × 10−5, R5.64f A: 2.062 × 10−3, R5.64f Q: 1.192 × 10−3, F6.36f I: 2.163 × 10−3, L5.62f A: 1.948 × 10−3, treated WT vs treated mutant: G2.57f V: 9.1 × 10−3, IECL2A: 0.02734, A2.60f M: 0.7477, R5.64f A: 0.08858, R5.64f Q: 0.02766, F6.36f I: 1.923 × 10−3, L5.62f A: 2.306 × 10−3 key: Not significant (ns) P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001, N=4 for all experiments.) (b) ΔGli1 mRNA fold change (SHH vs untreated) and Δ PMF (difference of peak PMF, calculated as P MFW T - P MFmutant) plotted for the mutants in Pathway 1. (c) Example mutant A2.60f M shows that cholesterol is able to enter SMO through Pathway 1 even on a bulky mutation. (d) Same as (b) but for Pathway 2 (e) Example mutant L5.62f A shows that cholesterol can enter SMO through Pathway 2 due to lesser steric hindrance. All snapshots presented are frames taken from MD simulations.

The molecular events as cholesterol enters the TMD from the inner leaflet in Pathway 2.

(a) Free energy plot showing the angle of cholesterol with the x-y plane, the plane of the membrane, versus the z-coordinate of cholesterol for Pathway 2. The pathway followed by cholesterol is η → θ → α∗. The experimental structures of SMO are shown as black polygons. (b) Free energy landscape of cholesterol’s y-coordinate plotted versus cholesterol’s x-coordinate for Pathway 2. Cholesterol interacts with residues in TM5-TM6 for Pathway 2 while entering the SMO core TMD. (e-h) Insets show cholesterol’s interactions with residues at the membrane-protein interface for Pathway 2. (c,e) show cholesterol outside the protein (η), while (d,f) show cholesterol entering the protein (θ) for Pathway 2. All snapshots presented are frames taken from MD simulations.

Multiple positions of cholesterol as it translocates through the Common Pathway, including the off-pathway intermediate.

(a) upright (δ), (b) tilted (𝜸), (c) the overtilted off-pathway intermediate(𝝐), and (d) cholesterol at the CRD binding site (𝜻). All snapshots presented are frames taken from MD simulations.

The effects of mutations along the translocation pathway connecting the TMD and CRD binding sites on the activation of SMO.

(a) Gli1 mRNA fold changes show the responsiveness of SMO mutants to SHH. Untreated Gli1 levels indicate low SMO activity, while SHH-treated values correspond to the level of SMO activation induced by SHH ligand. A t-test with Welch’s correction was used to compute statistical significance. (P values: untreated vs treated: WT: 3 × 10−6, YLDA: 2.46 × 10−4, F6.65f A: 1.08 × 10−3, IECL3A: 1.12 × 10−4, treated WT vs treated mutant: F6.65f A: 1.6 × 10−5, IECL3A: 1.6 × 10−5, YLDA: 1.4 × 10−5, key: Not significant (ns) P > 0.05, *P ≤ 0.05, **P ≤ 0.01, ***P ≤ 0.001, and ****P ≤ 0.0001, N=4 for all experiments.) (b) ΔGli1 mRNA fold change (SHH vs untreated) and Δ PMF (difference of peak PMF, calculated as P MFW T - P MFmutant) are plotted for mutants along the TMD-CRD pathway. (c, d) Example mutants YLDA and F6.65f A show that cholesterol is unable to translocate through this pathway because of the loss of crucial hydrophobic contacts provided by Y207 and F484 and along the solvent-exposed pathway.

The tunnel profile during cholesterol translocation in SMO.

(a) Free energy plot of the z-coordinate versus the tunnel diameter when cholesterol is present in the core TMD. The tunnel shows a spike in the radius in the TMD domain, indicating the presence of a cholesterol-accommodating cavity. (b) Representative figure for the tunnel when a cholesterol molecule is in the TMD. (c) Same as (a), when cholesterol is at the TMD-CRD interface. (d) same as (b), when cholesterol is at the TMD-CRD interface. (e) same as (a), when cholesterol is at the CRD binding site. (f) same as (b), when cholesterol is at the CRD binding site. Tunnel diameters are shown as spheres. Cholesterol positions are marked on plots using dotted lines. All snapshots presented are frames taken from MD simulations.

The timescales associated with the translocation of cholesterol through SMO.

Each major intermediate state has been marked (a-f). Timescales were obtained by calculating the mean first passage time (MFPT) using the Markov state model. Errors in timescales are shown as subscripts. The arrows represent the relative flux for the translocation between subsequent steps. The overall process occurs at a timescale of ∼ 1ms.

Comparison of TM6 length for SMO (A) and β2AR (B).

SMO shows an elongated TM6, which emerges into the extracellular space.

Error in Free Energies for Fig. 2 and Fig. 4.

Errors were computed using a bootstrapping strategy, where the MSM probabilities were computed from 80% of the data 200 times. (a) Error in Fig. 2(a). (b) Error in Fig. 2(b). (c) Error in Fig. 4(a). (d) Error in Fig. 4(b)

TICA (Time-lagged Independent Component Analysis) plot for SMO Cholesterol transport - Pathway 1.

(A) The entire data projected along the first two time-lagged independent components (tICs). Free energies were calculated for each datapoint and reweighed using the Markov State Model probabilities. (B) The same data as (A), except the z-coordinate of the cholesterol plotted against the first two tICs. The translocation of cholesterol is the slowest process observed, as shown by the gradient in the plot. (C) Same as (A), but for Pathway 2. (D) Same as (B), but for Pathway 2.

Minimum energy path taken by cholesterol for both pathways.

(a) Minimum energy pathway for Pathway 1 (black), marked on Figure 2(a). Note that the order of transitions is α → β → α∗ → 𝜸 → 𝜻. (b) Free energy profile as cholesterol moves along Pathway 1. The progress is mapped by the cholesterol translocation progress variable. (c) Minimum energy pathway for Pathway 2 (black), marked on Figure 4(a). Note that the order of transitions is η → θ → α∗ → 𝜸 → 𝜻. (d) Free energy profile as cholesterol moves along Pathway 2. The progress is mapped by the cholesterol translocation progress variable.

Experimental data for mutants along Pathway 1.

(a) Immunoblotting was used to measure abundance of mSMO and GLI1 proteins in SMO−∕− cells stably expressing either mSMO-WT or Pathway 1 mutants after treatment with SHH. (b) Gli1 mRNA fold change plotted for SMO mutants, showing fold change when the mutants are untreated, treated with a low concentration of SHH, and treated with a saturating concentration of SHH.

PMF data for mutants along Pathway 1.

(a) PMF for cholesterol translocation was computed for hSMO G280V and compared with WT hSMO using adaptive biasing force-based simulations. Representative figures showing the mutation for WT G280 (b) versus V280 (c). (d, g) - Same as (a) but for I389A and A283M, respectively. (e, h) same as (b) but for I389 and A283, respectively. (f, i) - same as (c) but for A389 and M283, respectively.

Tunnel diameter profile for WT versus mutant A283M.

Experimental data for mutants along Pathway 2.

(a) Immunoblotting was used to measure abundance of mSMO and GLI1 proteins in SMO−∕− cells stably expressing either mSMO-WT or Pathway 2 mutants after treatment with SHH. (b) Gli1 mRNA fold change plotted for SMO mutants, showing fold change when the mutants are untreated, treated with a low concentration of SHH, and treated with a saturating concentration of SHH.

PMF data for mutants along Pathway 2.

(a) PMF for cholesterol translocation was computed for hSMO R421A/R421Q and compared with WT hSMO using adaptive biasing force-based simulations. Representative figures showing the mutation for WT R421 (b) versus A421 (c) and Q421 (d). (e, h) - Same as (a) but for F455I and L419A, respectively. (f, i) same as (b) but for F455 and L419, respectively. (g, j) - same as (c, d) but for I455 and A419, respectively.

Comparison of cholesterol entry with existing resolved structure.

Frame from MD simulation showing cholesterol entry (cyan-blue), and resolved structure (PDB 8CXO, Zhang et al. (2022), orange). Cholesterol from the MD simulation frame is magenta, cholesterol from the resolved structure is white.

Hydrophobic Tunnel inside SMO core TMD.

Cholesterol interacts with the hydrophobic core of the TMD along the transport pathway. Figure corresponds to pose δ in Figure 4.

Cholesterol (purple) forms a tilted pose to enter the CRD binding site.

This is due to the presence of the Linker Domain (yellow).

Experimental data for mutants along the Common Pathway between the TMD and the CRD.

(a) Immunoblotting was used to measure the abundance of mSMO and GLI1 proteins in SMO−∕− cells stably expressing either mSMO-WT or Common Pathway mutants after treatment with SHH. (b) Gli1 mRNA fold change plotted for SMO mutants, showing fold change when the mutants are untreated, treated with a low concentration of SHH, and treated with a saturating concentration of SHH.

PMF data for mutants along the Common Pathway between the TMD and the CRD.

(a) PMF for cholesterol translocation was computed for hSMO Y207A and compared with WT hSMO using adaptive biasing force-based simulations. Representative figures showing the mutation for WT Y207 (b) versus A207 (c). (d, g) - Same as (a) but for F484A and I509A, respectively. (e, h) same as (b) but for F484 and I509, respectively. (f, i) - same as (c) but for A484 and A509, respectively.

Cholesterol positions along the entire transport pathway from membrane to CRD.

Pathway 1 is shown as (α → β → α∗), while Pathway 2 is shown as (η → θ → α∗) from the membrane to the TMD are shown separately in the left. The Common Pathway from the TMD to the CRD is α∗ → 𝜸 → 𝜻, with off-pathway intermediates 𝝐 and δ.

Error in Free Energies for Figure 7.

Errors were computed using a bootstrapping strategy, where the MSM probabilities were computed from 80% of the data 200 times. (a) Plot showing error for Figure 7(a). (b) Plot showing error for Figure 7(b). (c) Plot showing error for Figure 7(c).

Average tunnel radius for Figure 7.

The different positions of the cholesterol are marked in bold, and the overall tunnel profile has been marked as translucent.

Modelled residues in 5L7D-inactive-Apo-SMO starting structure.

The helical content of K440–I445 was modelled based on the structure of SANT1-bound SMO (PDB: 4N4W) (Wang et al., 2014).

Composition of the membrane used for embedding the protein in simulations.

Adaptive Sampling metrics used for clustering during the iterative landscape exploration process.

Round wise data collection for Cholesterol transport of SMO.

Roundwise data collection for SMO Cholesterol transport as projected along the first two time-lagged independent components (tICs).

Round numbers are specified in the respective plots.

MSM Construction for SMO Cholesterol transport - Pathway 1.

(A) Implied Timescales versus MSM Lagtime plot shows the convergence of timescales. A lagtime of 30ns was chosen for the MSM construction. (B) VAMP2 score versus Number of Clusters used for clustering the TICA-reduced data at three different variational cutoffs. The final MSM was made using 200 clusters and a 95 % cutoff (corresponding to 42 tIC dimensions).

MSM Construction for SMO Cholesterol transport - Pathway 2.

(A) The Implied Timescales versus MSM Lagtime plot shows the convergence of timescales. A lagtime of 30ns was chosen for the MSM construction. (B) VAMP2 score versus Number of Clusters used for clustering the TICA-reduced data at three different variational cutoffs. The final MSM was made using 400 clusters and a 65 % cutoff (corresponding to 28 tIC dimensions).

Chapman Kolmogorov Test for MSM validation - Pathway 1.

Chapman Kolmogorov test was performed using the deeptime library (Hoffmann et al., 2021).

Chapman Kolmogorov Test for MSM validation - Pathway 2.

Chapman Kolmogorov test was performed using the deeptime library (Hoffmann et al., 2021).

Distance of the binding site of cholesterol (marked by D95) from the nearest lipid headgroups in the membrane.

Kernel density plotted for 2.1ms of simulation data. The binding site is at least 20 Å away from the nearest lipid headgroup in simulations.