Two Classes of Cholesterol Binding Sites for the β₂AR Revealed by Thermostability and NMR

Abstract

Cholesterol binding to G protein-coupled receptors (GPCRs) and modulation of their activities in membranes is a fundamental issue for understanding their function. Despite the identification of cholesterol binding sites in high-resolution x-ray structures of the β₂ adrenergic receptor (β₂AR) and other GPCRs, the binding affinity of cholesterol for this receptor and exchange rates between the free and bound cholesterol remain unknown. In this study we report the existence of two classes of cholesterol binding sites in β₂AR. By analyzing the β₂AR unfolding temperature in lipidic cubic phase (LCP) as a function of cholesterol concentration we observed high-affinity cooperative binding of cholesterol with sub-nM affinity constant. In contrast, saturation transfer difference (STD) NMR experiments revealed the existence of a second class of cholesterol binding sites, in fast exchange on the STD NMR timescale. Titration of the STD signal as a function of cholesterol concentration provided a lower limit of 100 mM for their dissociation constant. However, these binding sites are specific for both cholesterol and β₂AR, as shown with control experiments using ergosterol and a control membrane protein (KpOmpA). We postulate that this specificity is mediated by the high-affinity bound cholesterol molecules and propose the formation of transient cholesterol clusters around the high-affinity binding sites.

Introduction

Effects of cholesterol on the stability and function of G protein-coupled receptors (GPCRs) have been known for at least two decades (1–3). However, there is often no clear consensus as to whether the effects of cholesterol on individual protein species arise from a specific GPCR-cholesterol interaction, from the effect of cholesterol on the bulk properties of the lipid bilayer, or from a combination of both (4). The publication of crystal structures of the β₂ adrenergic receptor (β₂AR) in 2007–2008 allowed the first visualization of direct interactions between this receptor and cholesterol (5,6). Since that time, cholesterol binding sites have been identified in the crystal structures of six different GPCRs: the β₁ adrenergic receptor (7), the adenosine A_2a receptor (8), the μ-opioid receptor (9), the serotonin 2B receptor (10), the metabotropic glutamate receptor 1 (11), and the purinergic receptor P2Y₁₂ (12). Crystallography, however, provides only static pictures, whereas it is important to determine affinities and exchange rates to fully understand how cholesterol interacts with GPCRs in a dynamic environment.

Lipidic cubic phase (LCP) represents a convenient membrane mimetic matrix for studying effects of lipids on membrane protein structure, function, and stability (13). In particular, LCP can be used to reconstitute active membrane proteins within a local highly curved bilayer-like structure containing certain lipids or sterols that are necessary for protein function. LCP has the consistency of a transparent gel, which is amenable for variety of spectroscopic approaches (6,14). Moreover, LCP supports the crystallization of membrane proteins directly from the lipidic environment (15); this method has been instrumental for structure determination of ∼ 60 different membrane proteins, including 25 different GPCRs. LCP also presents several interesting properties from the perspective of NMR spectroscopy. First, LCP is composed of membrane-like lipid bilayers at high density (lipid content is ∼ 50%), in which different classes of integral membrane proteins can be reconstituted in a functional state (see reviews (16,17) and references therein). Second, the bulk LCP is isotropic, with rapid diffusion of lipid molecules (18,19). This reorientation of lipid molecules on the NMR timescale leads to averaging of the anisotropic components of the NMR signal that are responsible for broadening of peaks in anisotropic lipid phases (e.g., lamellar or hexagonal phases). Therefore, it is possible to acquire almost liquid-like spectra of the lipid components of LCP in terms of line-width and intensity, and the NMR spectra of LCP samples have even been recorded in a solution state probe (20), although we observed that high resolution–magic angle spinning (HR-MAS) improves resolution. NMR methods that rely on parameters accessible in isotropic systems, such as polarization transfer rates (21), relaxation times (22), and line-shapes (23) have also been applied successfully in LCP.

Saturation transfer difference (STD) is a well-established technique in NMR that has found use predominantly as a method to screen soluble small-molecule ligands for proteins of pharmaceutical interest (24,25). The basis for the method is to measure the transfer of magnetization to a ligand following selective irradiation of the protein. Such an incoherent Nuclear Overhauser Effect (NOE) process has an r⁻⁶ distance dependence and is therefore strongly correlated with interaction between the protein and the ligand (structure of the complex and contact probability), as well as with the exchange rates between bound and free states. Several variations of the original solution-state ¹H STD method have been published, including heteronuclear experiments with irradiation of the protein ¹H atoms and measurement of the ¹³C ligand resonances (26,27), and group-selective techniques to measure weak interactions (28). Other advances in the application of STD, including the use of ¹H-¹⁹F heteronuclear STD, are described in a recent review (29). STD NMR has been used extensively in the context of membrane protein-lipid interactions by Soubias, Gawrisch, and collaborators to demonstrate that rhodopsin preferentially interacts with docosahexaenoic acid (DHA) containing lipids as well as with phosphatidyl ethanolamine (PE) containing lipids, as reviewed in (30).

In this study, we present to our knowledge, new insights into cholesterol-β₂AR interactions obtained by LCP-Tm thermostability measurements (14) and STD NMR. LCP-Tm suggests the existence of high-affinity cholesterol-specific binding sites, whereas STD experiments reveal the presence of a second class of low-affinity cholesterol binding sites. Using ergosterol, a structurally related yeast sterol (see Fig. 1), and Klebsiella pneumoniae outer membrane protein A (KpOmpA) (31), we confirm that both the low- and high-affinity binding sites of cholesterol on β₂AR are specific to the fine molecular details of cholesterol.

Figure 1.

Sterol structures with numbered carbon positions.

Materials and Methods

Protein expression and purification

β₂AR

The same engineered human β₂-adrenergic receptor-T4 lysozyme chimeric construct, β₂AR-T4L, as the one previously crystallized, was used in this study, and was expressed and purified as previously described (see (6) and Supplementary Material). Final protein solutions contained 50 mg/ml β₂AR with 20 mM HEPES pH 7.5, 150 mM NaCl, 200 mM imidazole, 0.05% w/v n-dodecyl-β-d-maltoside (DDM), 0.01% w/v cholesteryl hemisuccinate (CHS), and 50 μM carazolol (for NMR and LCP-Tm) or 500 μM timolol (for LCP-Tm). Protein concentration was determined by integration of the 280 nm absorption peak on the size exclusion chromatography (SEC) elution profile. For NMR samples, the last wash (20× column volume) and elution from Ni column was done using buffers prepared with D₂O (Sigma, 99.99% purity).

KpOmpA

The cloning, expression and purification of the Klebsiella pneumoniae outer membrane protein A (KpOmpA) transmembrane domain were performed as previously described (see (31) and Supplementary Material). The final solution thus contained 25 mg/ml KpOmpA in 50 mM HEPES (pH 7.5—corrected for deuterated solvent (32)), 150 mM NaCl, 0.05% w/v DDM, 0.01% w/v CHS, and 200 mM imidazole in D₂O. The final protein concentration was also verified spectrophotometrically.

Sample preparation for NMR

Monoolein (MO) was purchased from NU-CHECK Prep, Inc., Elysian, MN. Cholesterol (including 25,26,27-¹³C₃-cholesterol), 7-dehydrocholesterol, ergosterol, stigmasterol, HEPES, imidazole, and CHS were purchased from Sigma-Aldrich Chemie Sarl, Saint-Quentin Fallavier, France; sodium chloride from EUROMEDEX, Souffelweyersheim, France; and DDM and Zwittergent 3-14 (N-tetradecyl-N,N-dimethyl-3-ammonio-1-propanesulfonate) from MERCK KGaA, Darmstadt, Germany. D₂O (99.85 atom%) was obtained from EURISO-TOP SA, Saint-Aubin, France; and chloroform (used in sample preparation) from VWR, Fontenay sous Bois, France. All materials were used without further purification. The standard buffer preparation included 50 mM HEPES at pH 7.5, 150 mM NaCl, 200 mM imidazole, 0.05% w/v DDM, and 0.01% w/v CHS in D₂O.

MO, cholesterol at the specified concentration, and ergosterol (if required) were dissolved in chloroform and mixed in the desired molar ratio. Chloroform was evaporated to dryness under a constant flow of N₂ gas, and then under vacuum overnight. LCP samples were prepared at 40% w/w hydration by mechanical mixing of the appropriate mass of lipid fraction and buffer solution with or without protein, according to the established protocol (16), transferred directly into the NMR rotor and then centrifuged in a desktop centrifuge (∼ 5 min at 3354 g). Approximately 30 to 40 mg of material was used in each NMR experiment, and, where included, protein concentrations in these samples were 0.60 mM (β₂AR) or 1 mM (KpOmpA). ¹H and ¹³C spectra of samples prepared with and without this centrifugation step were compared, and there was no evidence that the LCP was disrupted by centrifugation. Once prepared, samples were maintained at 20°C. All experiments were conducted within 5 days of sample preparation, and the stability of the samples over this period was monitored by ¹H NMR.

LCP-Tm thermostability assay

β₂AR-T4L thermostability in LCP was measured following the published LCP-Tm protocol (14) using intrinsic protein fluorescence as a reporter for the protein folding state for samples containing timolol and CPM (7-diethylamino-3-(40-maleimidylphenyl)-4-methylcoumarin, Invitrogen, Carlsbad, CA) for samples containing carazolol. A full description of the LCP procedure and of the fitting of the Tm data to obtain an estimate for n and K_D is given in the Supplementary Material.

NMR spectroscopy

All NMR experiments were performed on a BRUKER Avance narrow bore spectrometer operating at a ¹H Larmor frequency of 500.13 MHz and using Topspin 1.3 software (BRUKER, Billerica, MA). Spectra were acquired using a 4 mm BRUKER HR-MAS gradient probe with a deuterium lock, and unless otherwise specified, data was acquired at a MAS frequency of 5 kHz and a temperature of 20°C (calibrated using ethylene glycol). Direct and INEPT transfer heteronuclear STD pulse sequences and phase cycling (for I₀–I_sat and I₀ experiments) were constructed according to the published schemes (26), with 2.5 kHz waltz16 power-gated ¹H decoupling during acquisition time and a 0.5 s interscan (d1) delay. The presaturation duration (3.0 s unless otherwise specified) consisted of a train of Gaussian pulses of length 60 ms with a radio-frequency field of 300 Hz, separated by 1 ms. The irradiation frequency of the presaturation pulse in the ¹H spectrum was either 8.6 ppm (on protein amide protons, i.e., on-resonance) or 100.0 ppm (i.e., off-resonance). The typical duration of a pair of I₀–I_sat (namely STD experiment) and I₀ experiments was 3 days for a single sample containing natural abundance cholesterol (32k and 16k scans for I_sat and I₀ experiments, respectively), or 24 h for a sample containing ¹³C-labeled cholesterol (8k and 4k scans for I_sat and I₀ experiments, respectively). All NMR data analysis was performed using Topspin (version 2.1 or 3.2) software (BRUKER, Billerica, MA).

The STD amplification factor (A Factor)—which describes the amount of magnetization transferred from the saturated protein to the lipid—was calculated according to the following Eq. 1, as reported by Mayer et al. (24), where I₀ and I_sat are the integral of a given resonance for saturation frequency applied off and on-resonance, respectively, and LE is the molar ligand excess:

$$AFactor=\frac{I_0-I_{sat}}{I_0}\times LE$$ (1)

The A Factor is proportional to the sterol bound fraction (see data analyses in Supplementary Material). Note that it is not necessarily valid to compare the A Factor values between different peaks because of differences in relaxation times.

NMR control experiments and sample characterization

The stability and homogeneity of the LCP at this frequency was determined by comparing the shape and intensity of the H₂O peak in the ¹H spectrum before and after each experiment. At a MAS frequency of 5 kHz, no significant broadening or diminution of the water peak was observed, indicating that there was no introduction of inhomogeneity or dehydration of the sample during the experiments. The 5 kHz frequency was selected as the minimum frequency that avoided any overlap of the STD presaturation pulses with spinning side bands from the MO and cholesterol peaks in the ¹H spectra. As a control, the STD experiments were also performed without protein to check the selectivity and efficiency of signal subtraction of presaturation off or on-resonance, and corrected when necessary (i.e., for intense MO peaks) before the A Factor calculation.

1D ¹H and ¹³C spectra were acquired for all samples, to confirm that there were no significant changes (i.e., < 0.001 ppm) in relative chemical shifts in samples with different cholesterol concentrations or protein compositions. Similarly, ¹³C and ¹H line-widths, and ¹H T₁ relaxation times do not change significantly for sterol or MO peaks in samples containing 0.5, 1.5, 2.5, 3.5, or 5 mol% cholesterol ± β₂AR (data not shown). Therefore, variations in A Factor could not be interpreted in term of alterations in internal dynamics of the LCP because of cholesterol content but rather as direct probe of the interaction between membrane proteins and either MO or sterols and thus on the contact probability between these molecules.

Partial assignments of ¹³C chemical shifts for the sterols were carried out with reference to published data (33–37). The most intense and well-resolved sterol peaks in the ¹³C NMR spectra of the various LCP samples were those of the methyl peaks in the sterol side chains (see Fig. 1 for sterol structures with carbon numbering). The chemical shifts and assignments of the sterol methyl peaks in LCPs, and in chloroform solution for comparison, are given in Table 1. Detailed comparisons of the chemical shifts of cholesterol in CDCl₃ solution, lipid bilayers and in LCP were previously reported (33,34,38). Although the addition of proteins or peptides to the LCP can potentially affect diffusion rates of lipids (18) and phase transitions, β₂AR did not significantly alter any of the cholesterol ¹³C chemical shifts in our samples, indicating that the protein did not have a particular effect on the cholesterol-MO-water hydrogen-bonding network at the lipid-water interface.

Table 1.

Average chemical shifts and assignments of sterol methyl peaks in LCP and chloroform

Sterol	Carbon number	δ (ppm) in LCP	δ (ppm) in CDCl3
Cholesterol	18	11.84	11.84
	19	19.31	19.38
	21	18.69	18.70
	26 and 27	22.41 and 22.61	22.55 and 22.81
Ergosterol	18	11.96	12.01
	19	15.88	16.25
	21	–a	21.07
	26 and 27	19.65 and 19.81	19.61 and 19.91
	28	17.69	17.57

All chemical shifts for LCP spectra were recorded in samples containing 2.5 mol% cholesterol and 2.5 mol% ergosterol in monoolein-based LCP at 40% w/w hydration. Chemical shifts were calibrated to the chemical shift of cholesterol C18 at 11.84 ppm. The uncertainty on the chemical shifts is ± 0.05 ppm.

^aOverlap with cholesterol 11 and ergosterol 21 at ∼ 21.15 ppm. Additional information regarding the chemical shifts of cholesterol in various phases can be found in (33,34,38).

Results and Discussion

Specificity of the cholesterol-β₂AR interaction from STD and thermal denaturation data

For the purposes of identifying selectivity in the cholesterol-β₂AR interaction, STD experiments were conducted on two sets of samples. The first set of samples used non-¹³C-labeled sterol, and each sample contained a total of 5 mol% sterol (2.5 mol% cholesterol and 2.5 mol% ergosterol). Within this set, three samples were prepared: a sample without protein (to estimate nonspecific magnetization transfer or background error); a sample containing KpOmpA (control protein); and a sample containing β₂AR. The following comparisons could then be made: first, the difference (if any) between the A Factor for equivalent carbon positions in the two sterols in any particular sample, and second the difference (if any) between the A Factor for the equivalent carbon position of a particular sterol in the sample containing β₂AR and in the sample containing KpOmpA. The STD spectra (I₀ and I₀–I_sat) for the sample containing β₂AR are shown in Fig. S1 in the Supporting Material, and the calculated A Factor values are given in Fig. 2. Note that the A Factor was only calculated for positions C26/27 where there was no overlap between two sterol peaks or between sterol and MO peaks for both the peak on cholesterol and the equivalent peak on ergosterol, and where there was sufficient signal-to-noise.

Figure 2.

Saturation transfer difference A Factors for samples containing β₂AR with natural-abundance ¹³C cholesterol and ergosterol. The A Factor values for sterol methyl peaks in samples containing 2.5 mol% cholesterol and 2.5 mol% ergosterol are plotted. The A Factor reports the amount of magnetization transferred from the protein to the sterol, and thus the contact probability between the two species. Samples contained 95 mol% monoolein in the lipid mix and were hydrated with 40% w/w buffer containing either KpOmpA (dark gray bars) or β₂AR (light gray bars). Experiments were conducted at 25°C. Error bars were calculated with reference to differences between experiments repeated on the same sample.

In the case of the samples containing cholesterol and ergosterol, an increase in the A Factor value for cholesterol was observed in samples containing β₂AR, compared with the value in samples containing KpOmpA (Fig. 2). In contrast, no such increases in A Factor values were observed for the equivalent carbon positions in ergosterol in samples containing β₂AR compared with those containing KpOmpA. Similarly, the A Factor values for the equivalent carbon positions in cholesterol were larger in samples containing β₂AR than the A Factor values for the same carbon positions in ergosterol. These data imply that there is a selective interaction between cholesterol and the β₂AR. The interaction is protein-specific, because the increase in A Factor occurs only in the sample containing β₂AR and not in the sample containing KpOmpA. It is also cholesterol-specific, because the increase in A Factor in the β₂AR sample occurs only for cholesterol and not for ergosterol.

In a second set of experiments we analyzed competition for binding between cholesterol and ergosterol by varying the molar concentration of ergosterol in the sample. We used 25,26,27-¹³C₃-cholesterol (2.5 mol%) in MO for sensitivity reasons, with either 0, 1.5, or 2.5 mol% unlabeled ergosterol. In agreement with the data obtained using unlabeled sterols, the presence of 0 mol%, 1.5 mol%, or 2.5 mol% ergosterol in a sample containing 2.5 mol% cholesterol with β₂AR did not significantly alter the A Factor of the cholesterol in the presence of β₂AR. The fact that ergosterol does not compete with the cholesterol in binding reveals a notable structural specificity considering the similarity of the cholesterol and ergosterol structures (see Fig. 1).

As an additional control, the STD values for the MO methyl peak were also measured for samples containing 5 mol% cholesterol with either β₂AR or KpOmpA. There was no significant difference between the A Factor for MO with KpOmpA compared with β₂AR, indicating that, similar to ergosterol, there is no selective interaction between MO and β₂AR (Fig. S2 in the Supporting Material). It was further observed that the initial build-up rate of STD was faster for cholesterol than for MO resonances (Fig S3), again indicating a stronger affinity of the β₂AR for cholesterol.

The selectivity of interaction between β₂AR and cholesterol with respect to other sterols was also observed by LCP-Tm thermal denaturation data (Fig. 3 A and B): although cholesterol has a strong, concentration-dependent effect on the thermal stability of β₂AR with saturation reached at 1 mol%, ergosterol (a yeast sterol) and the structurally related sterols stigmasterol (a plant sterol) and 7-dehydrocholesterol have no substantial effect on the melting transition (T_m) of the protein up to 10 mol%.

Figure 3.

Effect of sterol concentration on thermal denaturation of β₂AR. The thermal denaturation temperature (T_m) of β₂AR in LCP samples. (A) Samples containing cholesterol (diamonds), ergosterol (squares), stigmasterol (triangles), or 7-dehydrocholesterol (circles), over a 0 to 10 mol% range, for β₂AR bound to carazolol. (B) Expansion of thermal denaturation curves for β₂AR bound to carazolol in the 0 to 1 mol% cholesterol range, with equilibration time of 2 min (squares), 5 min (diamonds), or 30 min (triangles). A similar result was obtained for β₂AR bound to timolol.

Estimating K_D for the cholesterol-β₂AR interaction from STD data and LCP-Tm

A Factors were measured for the cholesterol C26/27 peak at 22.3 ppm based on data from five samples containing β₂AR with cholesterol compositions of 0.5, 1.5, 2.5, 3.5, or 5 mol% in MO (Fig. 4). Note that a cholesterol concentration of 0.5 mol% corresponds to a ligand excess of ∼ 23.5 (235 for 5 mol%), so that over the entire titration range cholesterol is in large excess with respect to β₂AR. The (I₀–I_sat) term in the A Factor equation is directly proportional to the cholesterol bound fraction in the sample. The A Factor is expected to increase with cholesterol concentration, but what is significant in Fig. 4 is that it increases linearly (R² = 0.966), which implies weak binding (see data Analysis 1 in the Supporting Material and (24)). By varying the dissociation constant K_D, and plotting the bound fraction over the same range of cholesterol concentrations, the minimum value of K_D for which a predominantly linear trend is seen, considering errors, can be estimated as 0.1 M. Fig. S4 shows that for a K_D ≤ 0.05 M, a highly nonlinear dependence of A Factor on cholesterol concentration should be observed.

Figure 4.

A Factors versus cholesterol concentration for samples containing β₂AR. Plot of A Factors calculated from STD data versus cholesterol concentration for samples containing β₂AR. Samples were hydrated with 40% w/w buffer containing β₂AR. Experiments were conducted at 20°C. Error bars were calculated with reference to differences between experiments repeated on the same sample.

The LCP-Tm denaturation data provide a very distinct picture of the cholesterol effect. A closer look at the 0 to 1 mol% range of concentrations (Figs. 3B and S5) reveals a sigmoidal curve, which is characteristic of a certain degree of cooperative binding. The relevant timescale for the denaturation experiment is 5 min per temperature point, and we have shown that the Tm values do not vary with equilibration times between 2 and 30 min (Fig. 3B). In the case of fast exchange of cholesterol on this timescale, the denaturation of cholesterol-free receptor would drive the whole population toward denaturation by shifting the binding equilibrium. In the case of slow exchange, the observed Tm is proportional to the cholesterol-bound receptor fraction (see Analysis 2 in the Supporting Material and Fig. S5 for additional material). The Tm curve can thus be fitted by a standard Hill binding law (cholesterol in excess, n equivalent binding sites), with n = 3 to 5 (n < 3 and n > 5 clearly give poor fit to the data points; whereas n = 3 to 5 gives reasonable fit considering experimental errors), and a K_D in the subnanomolar range.

Discussion

Specificity of a GPCR for a particular sterol structure has been observed previously with the serotonin_1A receptor, whose ligand binding function can be restored by cholesterol or desmosterol (39), but not by 7-dehydrocholesterol (40). Similar effects have been observed with the oxytocin receptor (41,42). In this study, we have revealed two distinct classes of cholesterol binding sites on β₂AR. The high-affinity binding sites observed by LCP-Tm correspond to cholesterol molecules in slow exchange on the minute timescale. These could not be observed by STD-NMR, which requires fast exchange on T₁ timescales, (i.e., exchange rates faster than 1 Hz). What is striking in this situation is that both of the observed binding affinities are specific for cholesterol, even that corresponding to a low-affinity–very fast exchange situation. The specificity was confirmed for the low affinity both on the part of the protein (KpOmpA versus β₂AR) and on the part of the sterol (cholesterol versus other ergosterol).

One possible explanation for this finding with cholesterol and β₂AR would be that the tightly bound (nonannular) cholesterol molecules create an environment suitable for lower-affinity (annular) binding of sterols (43). This leads to the idea of a cluster of cholesterol molecules around the β₂AR, certain molecules being in slow exchange and others in fast exchange, with the former contributing to the specific binding of the latter. This may have implications on the effects of cholesterol on the oligomerisation state of β₂AR (44) and on the recruitment of β₂AR in cholesterol-rich domains (45). Although this concept is still speculative at this stage, we dare to propose it as it provides a framework for devising future experiments.

The potential existence of three to five high-affinity cholesterol binding sites is supported by x-ray structures of the β₂AR (5,6). Overall at least seven distinct cholesterol binding sites have been observed in crystal structures of seven different GPCRs (Fig. 5 and Table 2) and recent molecular dynamics simulations revealed seven potential cholesterol binding sites on the surface of β₂AR (46). All these results agree well with our estimation of the number of up to three to five high-affinity cholesterol binding sites based on LCP-Tm data. Physiologically, the disruption of native cell membranes by cholesterol-depleting agents such as methyl-β-cyclodextrin (MβCD) results in an increase in β₂AR signaling (47) but has no apparent effect on protein clustering (48), which is more sensitive to the actin cytoskeleton. However, the effect of cholesterol depletion on signaling of the nonraft-associated β₂AR in vitro has been linked to the release of the G_αs and AC proteins from sequestration in lipid rafts, rather than on a direct interaction between β₂AR and cholesterol (49), and thus additional work is required to link the data presented here with the in vitro studies to more fully elucidate the role of cholesterol in β₂AR function.

Figure 5.

Binding sites of cholesterol in GPCR x-ray structures. A representative GPCR structure with a composite map of cholesterol binding sites, derived from the structures shown in Table 2. Cholesterol (blue carbons) and palmitate (yellow carbons) are shown in spheres representation. Two horizontal lines correspond to membrane boundaries. The extracellular side (EC) is on top, and the intracellular side (IC) is bottom. Cholesterol molecules participating in crystal contacts or dimerization interface are indicated in Table 2. It should be stressed that the LCP-Tm data presented confirm the presence of three to five high-affinity binding sites but do not address the issue of their localization on the receptor surface. To see this figure in color, go online.

Table 2.

Summary of reported cholesterol-GPCR binding sites shown in Fig. 5

Site no.	GPCR	PDB ID	No. of cholesterol	Binding interface	Crystal contacts (C) or dimer interface (D)
1	β2AR	2RH1, 3D4S	2	IC; helices I–IV	C
	β1AR	2Y00	1 (CHS)	IC; helices II–IV, ICL1
	P2Y12	4PXZ	1	IC; helices II–IV, ICL1
2	β2AR	2RH1	1	IC; helices I, VIII, palmitate	D
	5HT2B	4IB4	1	IC; helices I, VIII, palmitate
3	β1AR	2Y00	1 (CHS)	IC; helices III–V	C
	P2Y12	4NTJ	1	IC; helices III,V	C
4	A2AAR	4EIY	1	EC; helices V–VI	C
	β1AR	2Y00	1 (CHS)	EC; helix V	C
5	A2AARμ-OR	4EIY4DKL	11	EC; helices VI–VII, ECL3EC; helices VI–VII, ECL3
6	P2Y12	4NTJ	1	EC; helices VII, I
7	A2AAR	4EIY	1	EC; helices II–III, ECL1	C
	mGlu1	4OR2	3	EC; N-term, helices I–III, ECL1	D

Although the physiological consequences of β₂AR-cholesterol interactions remain to be further clarified, there is considerable evidence for both specific and nonspecific effects of cholesterol on the function of various other GPCRs. For example, in the case of rhodopsin (recently reviewed in (50)), the thermal stability of rhodopsin is increased by the addition of cholesterol in both egg-phosphatidylcholine bilayers (51) and in intact disk membranes (1). High cholesterol contents also reduce rhodopsin signaling via cGMP by altering the meta I/meta II (MI/MII) conformational equilibrium, with increasing cholesterol concentrations favoring the signaling-inactive MI state (1,52,53). These effects have been suggested (50) to result primarily from the effect of cholesterol on the fluid properties of the bilayer, including curvature forces (54,55), rather than from a specific interaction between cholesterol and the protein, although indications of direct interaction have also been described in the literature (1,56). Conversely, in the case of the oxytocin receptor, cholesterol affects the ligand binding affinity (57,58) and downstream signaling (59), and increases protein thermal and pH stability (41). Cholesterol also induces a more compact conformational state of the oxytocin receptor (60). Hill analysis of oxytocin binding versus cholesterol content suggest that at least six cholesterol molecules bind to oxytocin receptor in a cooperative manner (61). The effect of cholesterol on the oxytocin receptor has been proposed to result from a specific interaction, rather than from a general effect of cholesterol on the bulk physical properties of the membrane (57), and mutagenesis studies have identified regions of the oxytocin receptor sequence that are important for cholesterol sensitivity (62). Many other GPCRs also exhibit sensitivity to cholesterol, although the mechanisms for these responses are generally not yet well understood.