Allosteric modulation of the adenosine A2A receptor by cholesterol
Curation statements for this article:-
Curated by eLife
Evaluation Summary:
Cholesterol has long been known to have significant effects on G protein-coupled receptor (GPCR) ligand binding properties and stability, and cholesterol/GPCR interactions have frequently been observed in high-resolution structures. However, relatively limited biophysical work has been done to investigate the mechanistic basis for cholesterol's effects. This manuscript describes the use of a sensitive 19F NMR probe to monitor conformational equilibria in a prototypical GPCR, the A2a adenosine receptor. These experiments, together with data from other NMR experiments, computational analysis, and G protein assays, show that the subtle effects of cholesterol derive in large part from modulation of membrane biophysical properties, in contrast to conventional allosteric modulators that exert their effects through direct long-lived receptor binding.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 agreed to share their name with the authors.)
This article has been Reviewed by the following groups
Listed in
- Evaluated articles (eLife)
Abstract
Cholesterol is a major component of the cell membrane and commonly regulates membrane protein function. Here, we investigate how cholesterol modulates the conformational equilibria and signaling of the adenosine A 2A receptor (A 2A R) in reconstituted phospholipid nanodiscs. This model system conveniently excludes possible effects arising from cholesterol-induced phase separation or receptor oligomerization and focuses on the question of allostery. GTP hydrolysis assays show that cholesterol weakly enhances the basal signaling of A 2A R while decreasing the agonist EC 50 . Fluorine nuclear magnetic resonance ( 19 F NMR) spectroscopy shows that this enhancement arises from an increase in the receptor’s active state population and a G-protein-bound precoupled state. 19 F NMR of fluorinated cholesterol analogs reveals transient interactions with A 2A R, indicating a lack of high-affinity binding or direct allosteric modulation. The combined results suggest that the observed allosteric effects are largely indirect and originate from cholesterol-mediated changes in membrane properties, as shown by membrane fluidity measurements and high-pressure NMR.
Article activity feed
-
-
Author Response:
Reviewer #2:
What the authors attempt to achieve, and their approaches:
The author attempt to establish by which mechanisms cholesterol influences the function of the GPCR A_{2A}R, an adenosine receptor. The role of cholesterol on GPCRs has been reported in a number of studies, primarily in cellular experiments, and the authors set out here to clarify the molecular mechanisms.
To this end, they build upon their recent achievements to produce this protein and reconstitute it in nanodiscs, i.e. discoidal objects comprised of the membrane protein (here: A_{2A}R), lipids (here: POPC, POPG and cholesterol) and a membrane-scaffold protein (MSP) which wraps around this disc of protein+lipid. Nanodiscs allow studying proteins in solution, and are thought to be much more native-like than e.g. detergent micelles.
The …
Author Response:
Reviewer #2:
What the authors attempt to achieve, and their approaches:
The author attempt to establish by which mechanisms cholesterol influences the function of the GPCR A_{2A}R, an adenosine receptor. The role of cholesterol on GPCRs has been reported in a number of studies, primarily in cellular experiments, and the authors set out here to clarify the molecular mechanisms.
To this end, they build upon their recent achievements to produce this protein and reconstitute it in nanodiscs, i.e. discoidal objects comprised of the membrane protein (here: A_{2A}R), lipids (here: POPC, POPG and cholesterol) and a membrane-scaffold protein (MSP) which wraps around this disc of protein+lipid. Nanodiscs allow studying proteins in solution, and are thought to be much more native-like than e.g. detergent micelles.
The authors first use GTP hydrolysis experiments to quantify the basal activity and agonist potency at cholesterol concentrations from 0 to 13%. The cholesterol effects are weak but detectable. Then they use a single 19F label that reports on the protein's conformation (active, inactive) to show that the protein populates slightly more active states with cholesterol. (again, weak effects). Then they investigate G-protein binding to A_{2A}R in the nanodisc, and find (very!) weak enhancement at 13% cholesterol. These data point to weak positive allosteric modulation by cholesterol. They then use molecular dynamics simulations to probe the allosteric communication, using a recently proposed framework (Rigidity-transmission allostery). Doing these simulations in the presence of cholesterol (postions of cholesterol from X-ray structure) and absence. This analysis shows again only very weak effects of cholesterol, and this time the effect is opposite, i.e. negative allosteric modulation by cholesterol. Then they use 19F-labeled cholesterol analogues to probe by NMR the state of cholesterol (bound to protein?). Lastly, they use Laurdan fluorescence experiments and pressure NMR to establish that (i) the lipids become more ordered when cholesterol is present, and (ii) if one achieves such ordering even without cholesterol - namely by pressure - one may achieve similar effects as those that cholesterol has.
Collectively, these data lead them to conclude that cholesterol has a (weak) positive allosteric effect on this receptor, and this effect is not a direct one, but goes via modulation of the membrane properties.
We thank the reviewer for his comments and critique. A lot of his comments have to do with the nanodisc as a model system. We have therefore included an additional paragraph as discussed above, highlighting the advantages and disadvantages of the nanodisc. We’ve also included references to papers that have characterized nanodiscs or membrane proteins in nanodiscs. In our hands, 31P NMR spectra of POPC/POPG nanodiscs and their melt behavior is very similar to liposomes. We’ve tried to add to the discussion on nanodiscs without distracting too much from the focus in the paper.
Major strengths and weaknesses of methods and results:
The study addresses an important question, which inherently is difficult to answer: the effect of cholesterol is poorly understood and such studies require to work in an actual membrane. The authors do a careful combination of different methods to achieve their goal of identifying the mechanisms.
Despite combining several methods, several of them have their inherent problems:
(i) the nanodisc is too small to properly mimic the membrane environment, and it does not allow reaching relevant cholesterol concentrations. Moreover, it is not clear (to me) if one can exclude e.g. interactions of the protein with the surrounding MSP, or of cholesterol with MSP (see (iii) below).
We agree. In principle, we should worry about MSP. On the other hand, this is a constant in all of the samples and we focus instead on the cholesterol-dependent effects. These nanodiscs are unarguably small. We’ve commented on this in the paper now. However, we’d expect that the confinement would if anything emphasize the cholesterol bound state. Yet, the NMR studies of F-cholesterol interactions at best identified transient bound states.
(ii) the state of the protein (inactive, active) is probed with a single NMR-active site. The effects are small and I am not convinced that one shall interpret changes as small as the ones in Figures 3 and 4. In particular, how does this single probe behave at high pressure? Does it reflect an active state at 2000 bar pressure - where possibly other effects (unfolding?) may occur?
Here we can be quite confident. The spectra are predicated on a recent paper (Huang, et al, 2021) published in Cell in the spring of this year. Each state was carefully correlated with specific functional assays and conditions in a self-consistent way. The labeling site used on TM6 was strategically chosen based on earlier crystallographic studies of inactive and active A2AR. We have other labeling sites (TM7 and TM5) but the point was to use the chemical shift signatures to talk about cholesterol-induced changes to the conformational ensemble assigned in the Cell paper. The differences are small, but the fact that PAM effects are observed across conditions (apo, inverse agonist-bound, agonist-bound, and G protein-bound) reassures us that the spectral differences between low and high cholesterol samples are real. Unfolding by 19F NMR is in this case easy to see – the effects become irreversible and independent of ligand and the chemical shift ends up as one upfield peak. We also see a stabilization of the A1 (active) state, and a slight downfield shift of the active ensemble with increased pressure, consistent with reduced exchange dynamics (and coalescence) associated with the active state. We’ve commented on this in the revised version while trying not to distract from the flow of the paper.
(iii) the data in Figure 6 (19F of cholesterol analogs) are hard to interpret. Is cholesterol bound to the protein? Does the 19F shift reflect binding to the protein? or interactions within the confined space of the disc? or with MSP? The two analogs do not tell a coherent story.
It is confusing. We agree. We were fully expecting to see a clear A2AR bound state of cholesterol either through a concentration-dependent shift or a new peak. We also looked for “hidden” bound states through 19F NMR CEST experiments. We never identified a bound state in the presence of a range of cholesterol concentrations, as a function of receptor drug. We did observe small shifts although often these effects were as prominent with inverse agonist as agonist, possibly pointing to the existence of multiple weak binding sites. We’ve added some of this to the conversation. It’s also certainly possible that cholesterol exhibits some interaction with MSP, although again MSP is a constant presence in all the samples while we are focusing on cholesterol-dependent effects. In any case, we never detected a bound signature characteristic of slow exchange. That’s significant to the study despite the ambiguity of the measurements.
(iv) the pressure NMR study (Fig 7D) has weaknesses. The authors implicitly assume that pressure acts on the membrane, leading to more ordering. (They do recognize the possibility that pressure may have an effect on the protein directly, but consider that this direct effect on the protein is minor.) I think that their arguments are possibly incorrect: they apply here pressure onto a sample of nanodiscs, but all studies they cite to justify the use of pressure on membranes dealt with extended lipid bilayers (liposomes). To me it is not clear what is the lateral effect of pressure onto a nanodisc. Can water laterally enter into the bilayer and thus modify the lipid structure? I also note that previous pressure-NMR studies on a GPCR in micelles (rather than nanodiscs) showed a shift toward the active state. As a micelle is a very different thing than a nanodisc, this suggests that the pressure effect is, at least in part or predominantly, on the protein itself.
On top of the weakness of the pressure NMR experiment to identify what actually happens to the disc, it is not clear either how to interpret the 19F shift at very high pressure (Fig 7D). Given that there is only a single NMR probe, far out in an artificial side chain, it is difficult to assess the state of the protein.
These are good questions. Firstly, lipid bilayers (be it in liposomes, bicelles, or nanodiscs) are super soft and compressible systems – all known to change in hydrophobic thickness to pressure much more readily than proteins – be they membrane embedded or soluble. Secondly, the 19F NMR spectra are well-known to be representative of fully functional receptor as discussed above. Thirdly, even detergent micelles are susceptible to pressure (much more so than the receptor itself) See J. Phys. Chem. B 2014, 118, 5698−5706 (now referenced in the paper). Pressure will enhance hydrophobic thickness, even in a detergent host, by ordering the acyl chains. The lower specific volume states, selected by higher pressure, have a larger hydrophobic dimension. Thus, the effects seen earlier are equally an effect of environment. In the revised version, we simply make the point that the protein isn’t unfolded and that both cholesterol or pressure give rise to enhanced hydrophobic thickness and corresponding shifts in equilibria to the active states.
-
Evaluation Summary:
Cholesterol has long been known to have significant effects on G protein-coupled receptor (GPCR) ligand binding properties and stability, and cholesterol/GPCR interactions have frequently been observed in high-resolution structures. However, relatively limited biophysical work has been done to investigate the mechanistic basis for cholesterol's effects. This manuscript describes the use of a sensitive 19F NMR probe to monitor conformational equilibria in a prototypical GPCR, the A2a adenosine receptor. These experiments, together with data from other NMR experiments, computational analysis, and G protein assays, show that the subtle effects of cholesterol derive in large part from modulation of membrane biophysical properties, in contrast to conventional allosteric modulators that exert their effects through direct …
Evaluation Summary:
Cholesterol has long been known to have significant effects on G protein-coupled receptor (GPCR) ligand binding properties and stability, and cholesterol/GPCR interactions have frequently been observed in high-resolution structures. However, relatively limited biophysical work has been done to investigate the mechanistic basis for cholesterol's effects. This manuscript describes the use of a sensitive 19F NMR probe to monitor conformational equilibria in a prototypical GPCR, the A2a adenosine receptor. These experiments, together with data from other NMR experiments, computational analysis, and G protein assays, show that the subtle effects of cholesterol derive in large part from modulation of membrane biophysical properties, in contrast to conventional allosteric modulators that exert their effects through direct long-lived receptor binding.
(This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 agreed to share their name with the authors.)
-
Reviewer #1 (Public Review):
Although the authors make persuasive arguments regarding the effect of cholesterol being indirect, much is based on comparison of what to expect for a classic allosteric modulator, where the binding affinity is typically far higher. By comparison, the local concentration of cholesterol in the bilayer is extremely high, and may even involve an interplay between different low-affinity sites that have been identified by prior structural studies.
One argument used by the authors that if cholesterol is an allosteric modulator, its binding to A2AR should depend on whether A2AR is in the active or inactive state (Fig. S4). The observation that the effect is more pronounced at low F7-chol may point to a not-insignificant effect, in particular when noting that the effect decreases at higher F7-chol. The observation …
Reviewer #1 (Public Review):
Although the authors make persuasive arguments regarding the effect of cholesterol being indirect, much is based on comparison of what to expect for a classic allosteric modulator, where the binding affinity is typically far higher. By comparison, the local concentration of cholesterol in the bilayer is extremely high, and may even involve an interplay between different low-affinity sites that have been identified by prior structural studies.
One argument used by the authors that if cholesterol is an allosteric modulator, its binding to A2AR should depend on whether A2AR is in the active or inactive state (Fig. S4). The observation that the effect is more pronounced at low F7-chol may point to a not-insignificant effect, in particular when noting that the effect decreases at higher F7-chol. The observation that the chemical shift change is in the same direction for agonist and inverse agonist would be incompatible with a single binding site, but perhaps not with the existence of two binding sites. The fact that there is any effect at all on the F7-chol NMR spectrum suggests that F7-chol senses the state of A2AR and therefore must involve transient binding. It appears likely that, as the authors point out, "subtle direct interaction with cholesterol" are in play at the same time as "indirect effects through the membrane".
-
Reviewer #2 (Public Review):
What the authors attempt to achieve, and their approaches:
The author attempt to establish by which mechanisms cholesterol influences the function of the GPCR A_{2A}R, an adenosine receptor. The role of cholesterol on GPCRs has been reported in a number of studies, primarily in cellular experiments, and the authors set out here to clarify the molecular mechanisms.
To this end, they build upon their recent achievements to produce this protein and reconstitute it in nanodiscs, i.e. discoidal objects comprised of the membrane protein (here: A_{2A}R), lipids (here: POPC, POPG and cholesterol) and a membrane-scaffold protein (MSP) which wraps around this disc of protein+lipid. Nanodiscs allow studying proteins in solution, and are thought to be much more native-like than e.g. detergent micelles.
The authors first …
Reviewer #2 (Public Review):
What the authors attempt to achieve, and their approaches:
The author attempt to establish by which mechanisms cholesterol influences the function of the GPCR A_{2A}R, an adenosine receptor. The role of cholesterol on GPCRs has been reported in a number of studies, primarily in cellular experiments, and the authors set out here to clarify the molecular mechanisms.
To this end, they build upon their recent achievements to produce this protein and reconstitute it in nanodiscs, i.e. discoidal objects comprised of the membrane protein (here: A_{2A}R), lipids (here: POPC, POPG and cholesterol) and a membrane-scaffold protein (MSP) which wraps around this disc of protein+lipid. Nanodiscs allow studying proteins in solution, and are thought to be much more native-like than e.g. detergent micelles.
The authors first use GTP hydrolysis experiments to quantify the basal activity and agonist potency at cholesterol concentrations from 0 to 13%. The cholesterol effects are weak but detectable. Then they use a single 19F label that reports on the protein's conformation (active, inactive) to show that the protein populates slightly more active states with cholesterol. (again, weak effects). Then they investigate G-protein binding to A_{2A}R in the nanodisc, and find (very!) weak enhancement at 13% cholesterol. These data point to weak positive allosteric modulation by cholesterol.
They then use molecular dynamics simulations to probe the allosteric communication, using a recently proposed framework (Rigidity-transmission allostery). Doing these simulations in the presence of cholesterol (postions of cholesterol from X-ray structure) and absence. This analysis shows again only very weak effects of cholesterol, and this time the effect is opposite, i.e. negative allosteric modulation by cholesterol. Then they use 19F-labeled cholesterol analogues to probe by NMR the state of cholesterol (bound to protein?). Lastly, they use Laurdan fluorescence experiments and pressure NMR to establish that (i) the lipids become more ordered when cholesterol is present, and (ii) if one achieves such ordering even without cholesterol - namely by pressure - one may achieve similar effects as those that cholesterol has.Collectively, these data lead them to conclude that cholesterol has a (weak) positive allosteric effect on this receptor, and this effect is not a direct one, but goes via modulation of the membrane properties.
Major strengths and weaknesses of methods and results:
The study addresses an important question, which inherently is difficult to answer: the effect of cholesterol is poorly understood and such studies require to work in an actual membrane. The authors do a careful combination of different methods to achieve their goal of identifying the mechanisms.
Despite combining several methods, several of them have their inherent problems:
(i) the nanodisc is too small to properly mimic the membrane environment, and it does not allow reaching relevant cholesterol concentrations. Moreover, it is not clear (to me) if one can exclude e.g. interactions of the protein with the surrounding MSP, or of cholesterol with MSP (see (iii) below).
(ii) the state of the protein (inactive, active) is probed with a single NMR-active site. The effects are small and I am not convinced that one shall interpret changes as small as the ones in Figures 3 and 4. In particular, how does this single probe behave at high pressure? Does it reflect an active state at 2000 bar pressure - where possibly other effects (unfolding?) may occur?
(iii) the data in Figure 6 (19F of cholesterol analogs) are hard to interpret. Is cholesterol bound to the protein? Does the 19F shift reflect binding to the protein? or interactions within the confined space of the disc? or with MSP? The two analogs do not tell a coherent story.
(iv) the pressure NMR study (Fig 7D) has weaknesses. The authors implicitly assume that pressure acts on the membrane, leading to more ordering. (They do recognize the possibility that pressure may have an effect on the protein directly, but consider that this direct effect on the protein is minor.) I think that their arguments are possibly incorrect: they apply here pressure onto a sample of nanodiscs, but all studies they cite to justify the use of pressure on membranes dealt with extended lipid bilayers (liposomes). To me it is not clear what is the lateral effect of pressure onto a nanodisc. Can water laterally enter into the bilayer and thus modify the lipid structure? I also note that previous pressure-NMR studies on a GPCR in micelles (rather than nanodiscs) showed a shift toward the active state. As a micelle is a very different thing than a nanodisc, this suggests that the pressure effect is, at least in part or predominantly, on the protein itself.
On top of the weakness of the pressure NMR experiment to identify what actually happens to the disc, it is not clear either how to interpret the 19F shift at very high pressure (Fig 7D). Given that there is only a single NMR probe, far out in an artificial side chain, it is difficult to assess the state of the protein.
Appraisal of whether the authors achieved their aims, and whether the results support their conclusions:
The manuscript presents a number of observations which can be interpreted in the way that is proposed here, but as stated above, several experiments have their own problems: from the small disc with little cholesterol to questions in interpreting 19F of cholesterol analogs and high-pressure NMR data. Collectively, the interpretations are somewhat tentative in my view.
Likely impact of the work on the field, and the utility of the methods and data to the community:
In my view, the conclusions of this manuscript are fairly tentative (see above). Nonetheless, given the difficulty of studying these effects by any experimental method, this work may have an impact in the GPCR field, and more broadly in the membrane-protein field. I hope that reviewers from those fields can clarify this question. From my NMR perspective I would say that the paper is a nice combination of methods, I feel that the authors carefully assembled the methods and thought about pitfalls -- but a number of issues remain, as listed above. They make it hard to reach final conclusions. Some of the data appear contradictory (e.g. negative allosteric modulation seen in MD, contradicts the positive allostery seen by other methods; the two cholesterol analogs are not consistent).
-
Reviewer #3 (Public Review):
There has been much interest in whether cholesterol modulates the function of G protein-coupled receptors and, if so, how. Here Huang, Prosser and colleagues examine this question for the Adenosine A2 receptor using purified receptor, biophysical measurements in nanodiscs of well-defined composition, and a variety of functional measurements. They convincingly show that cholesterol does modulate A2AR function, but not via direct binding. Instead cholesterol exerts its impact by altering the thickness and dynamics of the membrane milieu.
This paper is extremely well written and seems to be typo free. State-of-the-art methods are used and this paper is thorough and rigourous. While there have been many papers on whether and how cholesterol interacts with and modulates GPCR function, this paper distinguishes …
Reviewer #3 (Public Review):
There has been much interest in whether cholesterol modulates the function of G protein-coupled receptors and, if so, how. Here Huang, Prosser and colleagues examine this question for the Adenosine A2 receptor using purified receptor, biophysical measurements in nanodiscs of well-defined composition, and a variety of functional measurements. They convincingly show that cholesterol does modulate A2AR function, but not via direct binding. Instead cholesterol exerts its impact by altering the thickness and dynamics of the membrane milieu.
This paper is extremely well written and seems to be typo free. State-of-the-art methods are used and this paper is thorough and rigourous. While there have been many papers on whether and how cholesterol interacts with and modulates GPCR function, this paper distinguishes itself in it thoroughness and the use of NMR spectroscopy not only to quantitate the functional states of the receptor under various conditions, but also to look directly at fluorinated cholesterol analogs as they are titrated into the protein, an approach that revealed no evidence for long lifetime binding of these analogs to the receptor. I found this paper to be mostly compelling in its conclusions but recommend that the authors address concerns regarding the cholesterol analog data.
-
