Multiple modes of cholesterol translocation in the human Smoothened receptor

Reviewer #1 (Public review):

Summary:

This manuscript uses primarily simulation tools to probe the pathway of cholesterol transport with the smoothened (SMO) protein. The pathway to the protein and within SMO is clearly discovered, and interactions deemed important are tested experimentally to validate the model predictions.

Strengths:

The authors have clearly demonstrated how cholesterol might go from the membrane through SMO for the inner and outer leaflets of a symmetrical membrane model. The free energy profiles, structural conformations, and cholesterol-residue interactions are clearly described.

Weaknesses:

(1) Membrane Model:

The authors decided to use a rather simple symmetric membrane with just cholesterol, POPC, and PSM at the same concentration for the inner and outer leaflets. This is not representative of asymmetry known to exist in plasma membranes (SM only in the outer leaflet and more cholesterol in this leaflet). This may also be important to the free energy pathway into SMO. Moreover, PE and anionic lipids are present in the inner leaflet and are ignored. While I am not requesting new simulations, I would suggest that the authors should clearly state that their model does not consider lipid concentration leaflet asymmetry, which might play an important role.

(2) Statistical comparison of barriers:

The barriers for pathways 1 and 2 are compared in the text, suggesting that pathway 2 has a slightly higher barrier than pathway 1. However, are these statistically different? If so, the authors should state the p-value. If not, then the text in the manuscript should not state that one pathway is preferred over the other.

(3) Barrier of cholesterol (reasoning):

The authors on page 7 argue that there is an enthalpy barrier between the membrane and SMO due to the change in environment. However, cholesterol lies in the membrane with its hydroxyl interacting with the hydrophilic part of the membrane and the other parts in the hydrophobic part. How is the SMO surface any different? It has both characteristics and is likely balanced similarly to uptake cholesterol. Unless this can be better quantified, I would suggest that this logic be removed.

Reviewer #2 (Public review):

Summary:

In this work, the authors applied a range of computational methods to probe the translocation of cholesterol through the Smoothened receptor. They test whether cholesterol is more likely to enter the receptor straight from the outer leaflet of the membrane or via a binding pathway in the inner leaflet first. Their data reveal that both pathways are plausible but that the free energy barriers of pathway 1 are lower, suggesting this route is preferable. They also probe the pathway of cholesterol transport from the transmembrane region to the cysteine-rich domain (CRD).

Strengths:

(1) A wide range of computational techniques is used, including potential of mean force calculations, adaptive sampling, dimensionality reduction using tICA, and MSM modelling. These are all applied in a rigorous manner, and the data are very convincing. The computational work is an exemplar of a well-carried out study.

(2) The computational predictions are experimentally supported using mutagenesis, with an excellent agreement between their PMF and mRNA fold change data.

(3) The data are described clearly and coherently, with excellent use of figures. They combine their findings into a mechanism for cholesterol transport, which on the whole seems sound.

(4) The methods are described well, and many of their analysis methods have been made available via GitHub, which is an additional strength.

Weaknesses:

(1) Some of the data could be presented a little more clearly. In particular, Figure 7 needs additional annotation to be interpretable. Can the position of the cholesterol be shown on the graph so that we can see the diameter change more clearly?

(2) In Figure 3C, it doesn't look like the Met is constricting the tunnel at all. What residue is constricting the tunnel here? Can we see the Ala and Met panels from the same angle to compare the landscapes? Or does the mutation significantly change the tunnel? Why not A283 to a bulkier residue? Finally, the legend says that the figure shows that cholesterol can still pass this residue, but it doesn't really show this. Perhaps if the HOLE graph was plotted, we could see the narrowest point of the tunnel and compare it to the size of cholesterol.

(3) The PMF axis in 3b and d confused me for a bit. Looking at the Supplementary data, it's clear that, e.g., the F455I change increases the energy barrier for chol entering the receptor. But in 3d this is shown as a -ve change, i.e., favourable. This seems the wrong way around for me. Either switch the sign or make this clearer in the legend, please.

(4) The impact of G280V is put down to a decrease in flexibility, but it could also be a steric hindrance. This should be discussed.

(5) Are the reported energy barriers of the two pathways (5.8{plus minus}0.7 and 6.5{plus minus}0.8 kcal/mol) significantly and/or substantially different enough to favour one over the other? This could be discussed in the manuscript.

(6) Are the energy barriers consistent with a passive diffusion-driven process? It feels like, without a source of free energy input (e.g., ion or ATP), these barriers would be difficult to overcome. This could be discussed.

(7) Regarding the kinetics from MSM, it is stated that the values seen here are similar to MFS transporters, but this then references another MSM study. A comparison to experimental values would support this section a lot.

Reviewer #3 (Public review):

This manuscript presents a study combining molecular dynamics simulations and Hedgehog (Hh) pathway assays to investigate cholesterol translocation pathways to Smoothened (SMO), a G protein-coupled receptor central to Hedgehog signal transduction. The authors identify and characterize two putative cholesterol access routes to the transmembrane domain (TMD) of SMO and propose a model whereby cholesterol traverses through the TMD to the cysteine-rich domain (CRD), which is presented as the primary site of SMO activation.

The MD simulations and biochemical experiments are carefully executed and provide useful data. However, the manuscript is significantly weakened by a narrow and selective interpretation of the literature, overstatement of certain conclusions, and a lack of appropriate engagement with alternative models that are well-supported by published data-including data from prior work by several of the coauthors of this manuscript. In its current form, the manuscript gives a biased impression of the field and overemphasizes the role of the CRD in cholesterol-mediated SMO activation. Below, I provide specific points where revisions are needed to ensure a more accurate and comprehensive treatment of the biology.

Major Comments:

(1) Overstatement of the CRD as the Orthosteric Site of SMO Activation

The manuscript repeatedly implies or states that the CRD is the orthosteric site of SMO activation, without adequate acknowledgment of alternative models. To give just a few examples (of many in this manuscript):

a) "PTCH is proposed to modulate the Hh signal by decreasing the ability of membrane cholesterol to access SMO's extracellular cysteine-rich domain (CRD)" (p. 3).

b) "In recent years there has been a vigorous debate on the orthosteric site of SMO" (p. 3).

c) "cholesterol must travel through the SMO TMD to reach the orthosteric site in the CRD" (p. 4).

d) "we observe cholesterol moving along TM6 to the TMD-CRD interface (common pathway, Fig. 1d) to access the orthosteric binding site in the CRD" (p. 6).

While the second quote in this list at least acknowledges a debate, the surrounding text suggests that this debate has been entirely resolved in favor of the CRD model. This is misleading and not reflective of the views of other investigators in the field (see, for example, a recent comprehensive review from Zhang and Beachy, Nature Reviews Molecular and Cell Biology 2023, which makes the point that both the CRD and 7TM sites are critical for cholesterol activation of SMO as well as PTCH-mediated regulation of SMO-cholesterol interactions).

In contrast, a large body of literature supports a dual-site model in which both the CRD and the TMD are bona fide cholesterol-binding sites essential for SMO activation. Examples include:

a) Byrne et al., Nature 2016: point mutation of the CRD cholesterol binding site impairs-but does not abolish-SMO activation by cholesterol (SMO D99A, Y134F, and combination mutants - Fig 3 of the 2016 study).

b) Myers et al., Dev Cell 2013 and PNAS 2017: CRD deletion mutants retain responsiveness to PTCH regulation and cholesterol mimetics (similar Hh responsiveness of a CRD deletion mutant is also observed in Fig 4 Byrne et al, Nature 2016).

c) Deshpande et al., Nature 2019: mutation of residues in the TMD cholesterol binding site blocks SMO activation entirely, strongly implicating the TMD as a required site, in contrast to the partial effects of mutating or deleting the CRD site.

Qi et al., Nature 2019, and Deshpande et al., Nature 2019, both reported cholesterol binding at the TMD site based on high-resolution structural data. Oddly, Deshpande et al., Nature 2019, is not cited in the discussion of TMD binding on p. 3, despite being one of the first papers to describe cholesterol in the TMD site and its necessity for activation (the authors only cite it regarding activation of SMO by synthetic small molecules).

Kinnebrew et al., Sci Adv 2022 report that CRD deletion abolished PTCH regulation, which is seemingly at odds with several studies above (e.g., Byrne et al, Nature 2016; Myers et al, Dev Cell 2013); but this difference may reflect the use of an N-terminal GFP fusion to SMO in the Kinnebrew et al 2022, which could alter SMO activation properties by sterically hindering activation at the TMD site by cholesterol (but not synthetic SMO agonists like SAG); in contrast, the earlier work by Byrne et al is not subject to this caveat because it used an untagged, unmodified form of SMO.

Although overexpression of PTCH1 and SMO (wild-type or mutant) has been noted as a caveat in studies of CRD-independent SMO activation by cholesterol, this reviewer points out that several of the studies listed above include experiments with endogenous PTCH1 and low-level SMO expression, demonstrating that SMO can clearly undergo activation by cholesterol (as well as regulation by PTCH1) in a manner that does not require the CRD.

Recommendation:

The authors should revise the manuscript to provide a more balanced overview of the field and explicitly acknowledge that the CRD is not the sole activation site. Instead, a dual-site model is more consistent with available structural, mutational, and functional data. In addition, the authors should reframe their interpretation of their MD studies to reflect this broader and more accurate view of how cholesterol binds and activates SMO.

(2) Bias in Presentation of Translocation Pathways

The manuscript presents the model of cholesterol translocation through SMO to the CRD as the predominant (if not sole) mechanism of activation. Statements such as: "Cholesterol traverses SMO to ultimately reach the CRD binding site" (p. 6) suggest an exclusivity that is not supported by prior literature in the field. Indeed, the authors' own MD data presented here demonstrate more stable cholesterol binding at the TMD than at the CRD (p 17), and binding of cholesterol to the TMD site is essential for SMO activation. As such, it is appropriate to acknowledge that cholesterol may activate SMO by translocating through the TM5/6 tunnel, then binding to the TMD site, as this is a likely route of SMO activation in addition to the CRD translocation route they highlight in their discussion.

The authors describe two possible translocation pathways (Pathway 1: TM2/3 entry to TMD; Pathway 2: TM5/6 entry and direct CRD transfer), but do not sufficiently acknowledge that their own empirical data support Pathway 2 as more relevant. Indeed, because their experimental data suggest Pathway 2 is more strongly linked to SMO activation, this pathway should be weighted more heavily in the authors' discussion. In addition, Pathway 2 is linked to cholesterol binding to both the TMD and CRD sites (the former because the TMD binding site is at the terminus of the hydrophobic tunnel, the latter via the translocation pathway described in the present manuscript), so it is appropriate that Pathway 2 figure more prominently than Pathway 1 into the authors' discussion.

The authors also claim that "there is no experimental structure with cholesterol in the inner leaflet region of SMO TMD" (p 16). However, a structural study of apo-SMO from the Manglik and Cheng labs (Zhang et al., Nat Comm, 2022) identified a cholesterol molecule docked at the TM5/6 interface and also proposed a "squeezing" mechanism by which cholesterol could enter the TM5/6 pocket from the membrane. The authors do not take this SMO conformation into account in their models, nor do they discuss the possibility that conformational dynamics at the TM5/6 interface could facilitate cholesterol flipping and translocation into the hydrophobic conduit, even though both possibilities have precedent in the 2022 empirical cryoEM structural analysis.

Recommendation:

The authors should avoid oversimplification of the SMO cholesterol activation process, either by tempering these claims or broadening their discussion to better reflect the complexity and multiplicity of cholesterol access and activation routes for SMO, and consider the 2022 apo-SMO cryoEM structure in their analysis of the TM5/6 translocation pathway.

(3) Alternative Possibility: Direct Membrane Access to CRD

The possibility that the CRD extracts cholesterol directly from the membrane outer leaflet is not considered. While the crystal structures place the CRD in a stable pose above the membrane, multiple cryo-EM studies suggest that the CRD is dynamic and adopts a variety of conformations, raising the possibility that the stability of the CRD in the crystal structures is a result of crystal packing and that the CRD may be far more dynamic under more physiological conditions.

Recommendation:

The authors should explicitly acknowledge and evaluate this potential mechanism and, if feasible, assess its plausibility through MD simulations.

(4) Inconsistent Framing of Study Scope and Limitations

The discussion contains some contradictory and misleading language. For example, the authors state that "In this study we only focused on the cholesterol movement from the membrane to CRD binding site." and then several sentences later state that "We outline the entire translocation mechanism from a kinetic and thermodynamic perspective.". These statements are at odds. The former appropriately (albeit briefly) notes the limited scope of the modeling, while the latter overstates the generality of the findings.

In addition, the authors' narrow focus on the CRD site constitutes a major caveat to the entire work. It should be acknowledged much earlier in the manuscript, preferably in the introduction, rather than mentioned as an aside in the penultimate paragraph of the conclusion.

Recommendation:
The authors should clarify the scope of the study and expand the discussion of its limitations. They should explicitly acknowledge that the study models one of several cholesterol access routes and that the findings do not rule out alternative pathways.

Summary:

This study has the potential to make a useful contribution to our understanding of cholesterol translocation and SMO activation. However, in its current form, the manuscript presents an overly narrow and, at times, misleading view of the literature and biological models; as such, it is not nearly as impactful as it could be. I strongly encourage the authors to revise the manuscript to include:

(1) A more balanced discussion of the CRD vs. TMD binding sites.

(2) Acknowledgment of alternative cholesterol access pathways.

(3) More comprehensive citation of prior structural and functional studies.

(4) Clarification of assumptions and scope.

Of note, the above suggestions require little to no additional MD simulations or experimental studies, but would significantly enhance the rigor and impact of the work.