A Sterol-Binding Cavity Underlies Sterol Recognition and Differential Activation of the ABCG5/G8

Reviewer #1 (Public review):

Summary:

In this manuscript by Rezaei F. et al., the authors determine a cryo-EM structure of the human ABCG5/G8 ABC transporter that was expressed in Pichia Pastoris. The structure was nearly identical to previous crystal and cryo-EM structures of ABCG5/G8, and displayed a non-protein density in the central transport pocket between both halves of the transporter. Since the expression host Pichia pastoris natively makes ergosterol rather than cholesterol, and cholesterol hemisuccinate (CHS) was added during protein solubilization and purification, the authors used molecular dynamics simulations to probe the stability of binding for cholesterol, CHS, and ergosterol in the central pocket where the density was identified. These simulations suggested that ergosterol, rather than CHS, occupies the central pocket in the cryo-EM structure.

The authors then went on to investigate the effect of ergosterol on ABCG5/G8 ATPase activity. Interestingly, ergosterol was more potent than cholesterol or CHS in stimulating ABCG5/G8 ATPase activity. It was further demonstrated that the hydroxyl moiety of Y342 in ABCG5 is essential to achieve full stimulation of ATPase activity by ergosterol, but not cholesterol. This assay highlights a key residue (Y342) in ABCG5 that contributes to differential sterol sensitivity of ABCG5/G8.

Overall, this is a compelling story with solid biochemical kinetic data that largely support the interpretation that binding of ergosterol gives rise to the sterol-like density in the transport pocket of ABCG5/G8.

Strengths:

(1) The finding that ABCG5/G8 responds to ergosterol more strongly than cholesterol is very interesting. This suggests a plausible role for ABCG5/G8 in regulating absorption of exogenous sterols from the human diet.

(2) The biochemical kinetics data appear to be quite strong, with appropriate replicates and statistical analyses in most locations.

(3) The cryo-EM density map appears to be of good quality despite an overall resolution of ~3.6 Å.

(4) The manuscript and figures are well prepared and easy to follow.

Weaknesses:

(1) While I agree with the conclusion that the central density is likely not CHS, how can the authors be sure that this is not cholate (which was included throughout solubilization and purification)?

(2) The authors do comment on the role of the hydroxyl on Y432 being important for differential regulation of the transporter by sterols. However, it was not clear to me how the hydroxyl of Y342 could contribute to enhanced activity with ergosterol and not cholesterol. These two molecules are very similar and bind in essentially the exact same location. Could it be that the additional double bond (creating a conjugated double-bond system) in the B-ring of ergosterol interacts with Y342 in a favorable manner that is not possible with cholesterol, which lacks this double bond? It seems like more discussion on the chemical similarities and differences between ergosterol and cholesterol is warranted in order to potentially explain why the Y342 hydroxyl moiety is so critical for stimulation with ergosterol but not cholesterol.

(3) Minor weakness: Enthusiasm for the cryo-EM structure is slightly tempered by the fact that the structure is essentially identical to a previous ABCG5/G8 cryo-EM structure bound to cholesterol.

(4) I am a little confused by Figure 1D. If the RMSD is calculating the ligand position relative to its initial position in the first frame, then shouldn't the RMSD value for all ligands start at zero when the time is zero? Why does RMSD not start at 0 for all ligands and then gradually increase as the simulation progresses? Similarly, why is the starting RMSD at time 0 so different for ergosterol, cholesterol, and CHS?

(5) It would be good to include a "no sterol" control for the experiments in Figure 3B. Is the basal ATPase activity without sterols the same for WT and Y342F constructs?

(6) In Figure 4D and Table 1 (bottom), ATPase activity is measured as a function of sterol concentration, and the data are fit using a Michaelis-Menten equation, with Vmax and Km reported. I do not think that interpreting these parameters as Vmax and Km is appropriate for this experiment. Although a saturating Michaelis-Menten-type equation can be used to describe the dependence of ATPase activity on sterol concentration, sterol is not the substrate being directly measured in the ATPase reaction. Rather, the experiment measures sterol-dependent stimulation of ATP hydrolysis. Thus, these data are more appropriately described as a sterol dose-response or activation curve, and an apparent EC50 would be a more appropriate parameter to report. A true Km for sterol would require measurement of the sterol transport reaction as a function of sterol concentration.

(7) It would be good to include statistical analyses of the interaction energy decompositions from molecular dynamics (Figure 2C) where appropriate. For instance, is the difference between interaction energies for (G5)-Y424 really different for cholesterol and ergosterol? The error bar for ergosterol makes it hard to tell if this difference is statistically significant.

(8) The changes in side-chain orientations noted in the paper (Figure 2B) are really very, very small changes. I am not quite sure that these represent true changes due to ligand binding, versus small changes in model refinement into intermediate resolution range maps (~3.6Å cryoEM and ~4Å crystal structure).

(9) The discussion surrounding an alternative sterol binding pocket, and particularly its involvement in sterol flip-flop (Figure 4B), seems very speculative.

(10) The authors included AMP-PNP and MgCl2 in their sample for cryo-EM, but did not observe a closed state in 3D classification. Any speculation as to why the open conformation was obtained, or, more importantly, why AMP-PNP was included at all?

Reviewer #2 (Public review):

Summary:

Rezaei et al. determined the cryo-electron microscopy structure of human ABCG5/8 in complex with ergosterol, revealing a conserved sterol-binding site at the transmembrane domain (TMD) interface, signifying the role of a specific key residue, Y432, in the interaction. ATPase assays showed the effective stimulation of ATPase activity of ABCG5/8 in the presence of ergosterol than cholesterol and CHS. The role of Y432 in the interaction with the transporter and stimulation of its ATPase activity was further supported by a mutagenesis study.

Strengths:

The study has been supported well with the structural and biochemical data combined with molecular dynamics simulations. The methods are very well detailed, which is helpful for the reader.

Weaknesses:

A few more experiments along similar lines (structures and mutagenesis and ATPase experiments) would have strengthened the molecular mechanism of sterol transport. Also, the physiological relevance of this sterol-ABCG5/8 structure has not been elaborated or connected via any experiments.

Reviewer #3 (Public review):

Summary:

This study provides structural and functional insights into the sterol recognition mechanism of ABCG5/G8.

Strengths:

The modeling of ergosterol into the density map is plausible, considering that the protein was heterologously expressed in yeast, only CHS was added during purification, and MD simulations rendered CHS binding to this site unlikely. The identified ergosterol-binding site in the central TMD cavity closely aligns with the location previously observed in a cholesterol-soaked crystal structure, strongly supporting the notion that this cavity serves as the primary substrate-transport pocket. Furthermore, mutagenesis of the key Tyr residue abolishes the ergosterol-specific enhancement of ATPase activity, providing robust functional support for the structural findings.

Weaknesses:

However, several caveats should be considered when interpreting these results:

First, while local resolution in the TMD region appears sufficient to trace side chains, the overall resolution of 3.5 Å poses challenges for precisely discerning subtle side-chain movements or rotamer changes. As the structural comparison with the apo and CRL-bound forms reveals no major helical shifts-remaining confined to side-chain rearrangements-caution is warranted when attributing mechanistic significance to these localized variations without exhaustive functional validation.

Second, a major limitation lies in the mechanistic model proposed by the authors. The manuscript suggests that the peripheral cholesterol-binding site (CRL) observed exclusively in previous crystal structures represents an intermediate state of sterol flip-flop. However, the authors provide no new experimental or computational evidence in this study to substantiate this hypothesis. Consequently, this model remains speculative and lacks sufficient empirical backing at this stage.

Author response:

We thank the editors and reviewers for their careful evaluation of our manuscript and their constructive comments. We are pleased that the reviewers found the identification of a sterol-binding cavity in ABCG5/G8, the stronger stimulation of ATPase activity by ergosterol, and the role of ABCG5_Y432 in differential sterol activation to be of interest to the greater transporter community.

Reviewer #1:

(1) We agree that cholate should be considered, as it was present throughout protein purification. We acknowledge that at the current resolution, the identity of the bound ligand cannot be determined unambiguously from the cryo-EM density alone, particularly among sterol-like molecules of similar size and shape. For this reason, we described the cavity as a potential sterol-binding site and ergosterol as the most likely ligand and did not rely on the structure alone to define sterol-protein interactions through this cavity. Here, we combined ligand fitting, molecular dynamics simulations, site-directed mutagenesis, and in vitro functional assays to evaluate the role of this cavity in sterol selectivity. In the revision, we will fit cholate into the density, compare its fit with those of ergosterol, cholesterol, and CHS, and include MD simulation data for cholate as described for these molecules. Furthermore, previous studies showed that cholate only contributed to the basal ATPase activity [1, 2], so we will include ATPase data for the ABCG5_Y432F mutant under sterol-free conditions and compare it with WT under the same conditions, since our assays included cholate.

(2) In the current manuscript, we discuss how the hydroxyl group of Y432 may provide an additional interaction capability within the binding cavity. In the revision, we will add a discussion of the chemical differences between ergosterol and cholesterol, including the double bond in the B ring and the additional double bond and methyl group in the side chain, and how these features may favor interactions with the Y432 hydroxyl and aromatic ring.

(3) We agree that the overall fold is similar to previous ABCG5/G8 structures. We want to clarify that this is the first structural evidence showing a non-cholesterol bound to the central cavity and its link to sterol-specific activation of the transporter’s catalysis.

(4) The ligand RMSD values were calculated relative to the docked starting poses. Because the systems underwent equilibration and the first 25 ns of each production simulation were excluded from analysis, the ligands had already relaxed from their initial poses at the start of the analyzed trajectories, which explains why the values do not start at zero and differ among sterols. We will clarify this in the Methods and figure legend in the revision.

(5) We thank the reviewer for this suggestion. The basal activity of WT ABCG5/G8 is currently shown in Figure 3A. In the revision, we will measure the basal activity of ABCG5_Y432F under the same sterol-free conditions and add both WT and Y432F sterol-free controls to Figure 3B for direct comparison.

(6) We agree that sterol acts as an activator rather than the directly measured substrate in this assay. In the revision, we will describe these data (Figure 3D and Table 1) as sterol-dependent activation curves and report apparent EC50 and maximal activity values in place of Km and Vmax.

(7) We thank the reviewer for bringing up this important point. In the original submission, Fig. 2C shows per-residue MMGBSA decomposition as the mean +/- SD of three independent replicates. We performed a Welch’s t-test for all three pairs with a Benjamini-Hochberg FDR for all 48 comparisons, the residues which had mentioned were in direction but were not statistically significant. This remains relevant since our simulations showed that in all replicates, CHS would move away from the docked site which explains why the standard deviations for the MMGBSA calculations were so large. We equally re-analyzed each trajectory from the point where CHS’s RMSD stabilized which equally yielded large SD binding free energy decomposition, however, no residues passed the BH FDR. For this reason, we will modify our language within the text to reflect the comparisons between sterol groups as trends rather than significant differences.

(8) We agree that these side chain differences are small relative to the resolution of the structures. We will revise the text to describe them more cautiously.

(9) We agree that the proposed role of the surface-exposed site in sterol flip-flop is speculative. The model is based on the cholesterol molecule observed at the ABCG8 to ABCG5 transmembrane interface in our previous crystal structure (PDB ID: 8CUB) [3]. We discuss this site as having a possible role in sterol sensing and regulation or in sterol movement between membrane leaflets, as suggested for other ABC transporters [4]. We acknowledge that neither role is directly tested in this study. In the revision, we will rephrase this section, clearly present both possibilities as hypotheses for future investigation, and revise Figure 4B and its legend to indicate that it represents a proposed model.

(10) AMPPNP and MgCl₂ were included to match the conditions under which we previously obtained a cholesterol-bound structure of ABCG5/G8 [3]. Our aim was to capture the sterol-bound state of the transporter rather than an ATP hydrolysis or nucleotide-bound closed state. Consistent with our previous structure, no nucleotide density was observed in the nucleotide binding sites, and the transporter adopted an inward-facing conformation. We will clarify this rationale in the Methods of the revised manuscript and discuss possible role of nucleotide in this report.

Reviewer #2:

We agree that additional structures and mutants would further strengthen the mechanistic model, and we will discuss these as directions for future work. We will also expand the Discussion on the physiological relevance of differential sterol recognition by ABCG5/G8, including dietary non-cholesterol sterols.

Reviewer #3:

(1) We agree that the observed side chain differences should be interpreted with caution given the resolution. We will revise the text to avoid attributing mechanistic significance to these localized variations beyond what is supported by our mutagenesis and functional data.

(2) We agree that the present study does not provide new experimental or computational evidence that the CRL site represents an intermediate in sterol flip-flop. This model is based on the cholesterol-bound site observed in our previous crystal structure (PDB ID: 8CUB) [3], and in the current manuscript we present it as one of two possible roles for this site, alongside sterol sensing or regulation. In the revision, we will state the basis of the model more clearly, present both possibilities as hypotheses for future investigation, and revise the related text and Figure 4B accordingly (see also our response to Reviewer #1, point 9).

We look forward to submitting a revised manuscript that addresses these points.

References:

(1) B. M. Xavier, A. A. Zein, A. Venes, J. Wang, J. Y. Lee, Transmembrane Polar Relay Drives the Allosteric Regulation for ABCG5/G8 Sterol Transporter. International Journal of Molecular Sciences 2020, Vol. 21, Page 8747 21, 8747 (2020).

(2) B. J. H. Johnson, J. Y. Lee, A. Pickert, I. L. Urbatsch, Bile acids stimulate ATP hydrolysis in the purified cholesterol transporter ABCG5/G8. Biochemistry 49, 3403–3411 (2010).

(3) D. Farhat, F. Rezaei, M. Ristovski, Y. Yang, A. Stancescu, L. Dzimkova, S. Samnani, J. F. Couture, J. Y. Lee, Structural Analysis of Cholesterol Binding and Sterol Selectivity by ABCG5/G8. J. Mol. Biol. 434, 167795 (2022).

(3) S. Thangapandian, K. Kapoor, E. Tajkhorshid, Probing cholesterol binding and translocation in P-glycoprotein. Biochimica et Biophysica Acta (BBA) - Biomembranes 1862, 183090 (2020).