|Home | About | Journals | Submit | Contact Us | Français|
The role of cholesterol in eukaryotic membrane protein function has been primarily attributed to an influence on membrane fluidity and curvature. Herein, we present the 2.8 Å resolution crystal structure of a thermally-stabilized human β2-adrenergic receptor bound to cholesterol and the partial inverse agonist timolol. The receptors pack as monomers in an antiparallel association with two distinct cholesterol molecules bound per receptor, but not in the packing interface, thereby indicating a structurally relevant cholesterol binding site between helices I, II, III, and IV. Thermal stability analysis using isothermal denaturation confirms that cholesterol significantly enhances the stability of the receptor. A consensus motif is defined that predicts cholesterol binding for 26% of all class A receptors suggesting that specific sterol binding is important to the structure and stability of other G protein-coupled receptors, and that this site may provide a novel target for allosteric therapeutic discovery.
The β2-adrenergic receptor (β2AR) is a well-characterized member of the class A G protein-coupled receptors (GPCRs) and is expressed in pulmonary and cardiac myocyte tissue (Milligan et al., 1994; Takeda et al., 2002). The β2AR, and its close relative β1AR sense epinephrine and norepinephrine in bronchial vasculature and cardiac muscle, respectively, and a positive inotropic response is elicited with β1AR and bronchial dilation associated with β2AR. One class of small molecules, known as beta-blockers, have been used clinically in the management of cardiac arrhythmias where they are thought to evoke their antagonistic effect mainly through β1AR, but also bind effectively to β2AR. The high-resolution crystal structure of the β2AR fused to T4-lysozyme (β2AR-T4L) and a low-resolution partial structure of β2AR in complex with a specific Fab fragment have recently been determined in the presence of the beta-blocker and partial inverse agonist carazolol (Cherezov et al., 2007; Rasmussen et al., 2007; Rosenbaum et al., 2007). Timolol, a member of the first generation of beta-blockers, has found extensive use in the treatment of glaucoma by reducing intraocular pressure, as well as in the treatment of high-blood pressure and heart disease. Timolol has been characterized as a partial inverse agonist of the β2AR where its efficacy varies from 28 to 90% inhibition of basal activity depending on the assay conditions. It also has high water-solubility, thus making it a useful ligand for structural studies (Chidiac et al., 1994; Zimmerman and Kaufman, 1977).
Cholesterol is an essential component regulating the structure and function of eukaryotic membranes and is thought to act primarily through the modulation of membrane fluidity and maintenance of sphingolipid rafts and membrane microdomains (Simons and Ikonen, 2000). In addition, cholesterol plays a well-established regulatory role in a number of membrane proteins, either indirectly through its ability to modulate the physical properties of lipid membranes, or directly through specific interactions with select protein systems (Burger et al., 2000; Lee, 2004). It has been postulated that cholesterol may also play an important role in GPCR function and pharmacology where certain receptors are thought to partition into or out of cholesterol-rich caveolae (Pucadyil and Chattopadhyay, 2006). In addition, direct protein-cholesterol interactions have been demonstrated for the oxytocin, galanin and serotonin1A receptors (Gimpl et al., 1997; Pang et al., 1999; Pucadyil and Chattopadhyay, 2004).
Herein, we report the 2.8 Å resolution crystal structure of an engineered β2AR in complex with timolol and two molecules of cholesterol. The β2AR was modified both for enhanced crystallizability by incorporation of T4-lysozyme between TM-V and TM-VI (β2AR(E122W)-T4L), and for stability through the introduction of a point mutation in the transmembrane domain (E1223.41W). Ballesteros-Weinstein numbering is used throughout the text as superscripts to the protein numbering (Ballesteros and Weinstein, 1995). Despite these modifications, the engineered receptor retains nearly native ligand binding and signaling properties (Rosenbaum et al., 2007; Roth et al., 2008).
In comparison to our previous β2AR-T4L structure, we reveal unambiguous structural evidence of a specific cholesterol-binding site for a membrane protein. In contrast to the high-resolution structure with carazolol (carβ2AR-T4L), the cholesterol molecules here are not involved in the crystal-packing interface and do not appear to be associated with the covalently attached palmitic acid. We further demonstrate that the water-soluble cholesterol analogue cholesteryl hemisuccinate (CHS) has an effect on thermal stability of the detergent-solubilized receptor in the presence and absence of timolol, and that cholesterol incorporation into native insect cell membranes modifies the ligand binding properties of the heterologously expressed receptor. We also report the binding mode for the partial inverse agonist timolol, which is bound at the orthosteric binding site of the receptor (timβ2AR(E122W)-T4L) and engages in many of the same key interactions with the protein as carazolol, but with some important differences.
High-level expression of β2AR(E122W)-T4L was carried out in Sf9 insect cells using standard protocols. A simplified purification scheme was enabled by the presence of the E122W mutation, which resulted in a higher yield of functionally active receptor in the folded state than for the wild-type (85% for E122W vs. 25% for wild-type), as judged by size-exclusion chromatography using a fluorescent-labeled alprenolol probe and traditional binding assays (Roth et al., 2008). This shift to a functionally folded receptor eliminates the need for a ligand affinity chromatography step in the purification. In addition, the high expression levels of 2 mg of receptor per liter of cell culture enabled a single metal-affinity chromatography step to achieve greater than 90% homogeneity, thus mitigating the effects of delipidation on the final purified protein. We purified and crystallized β2AR(E122W)-T4L in the presence of a saturating concentration of S-(−)-1-(t-Butylamino)-3-[(4-morpholino-1,2,5-thiadiazol-3-yl)oxyl]-2-propanol maleate (timolol) Table 1. Comparison between timolol bound β2AR(E122W)-T4L (timβ2R(E122W)-T4L) and carazolol bound carβ2AR-T4L (PDB ID: 2RH1) (Overall RMSD: 1.8 Å, receptor RMSD: 0.34 Å) reveals a change in the relative tilt angle between T4L and the receptor, as well as a few minor shifts in ICL2 and ECL3 that may be explained by the monomers packing in a higher symmetry space group with different crystal packing interactions (Figure 1A). In addition, the relative orientation between the two domains of T4L has shifted significantly so that when domain 1 (residues 2–11, 59–161) is aligned between the two structures (RMSD = 0.6 Å), domain 2 (residues 12–28) has an RMSD of 50.0 Å. However, as indicated, the receptor portion of the fusion protein is very similar between the two crystal structures. As such it is difficult to explain the occurrence of the altered space group as the conditions used for crystallization of each species are very similar. Thus, it is necessary to examine the differences in the protein and in its preparation. There are three possibilities for the observed shift in intermolecular interactions. The stabilizing mutant is altering the packing environment either through loss of a buried charge or through increased interactions with the monoolein lipid membrane. This possibility is supported by the presence of electron density for a lipid species adjacent to the W1223.41 modeled as monoolein in the current structure. No corresponding density was seen in this region for the carβ2AR-T4L structure despite higher resolution data. A second related possibility is the presence of endogenous insect plasma membrane lipid being retained through the shortened purification protocol. The final possibility is the shortened length of the C-terminus in the current structure enables an altered crystal packing environment despite the fact that the same number of residues are ordered at the C-terminus in both structures. Given the current information it is impossible to distinguish between these three possibilities although the first two seem more plausible as the majority of the interreceptor packing interactions occur within the lipid membrane.
Given the established equivalence between β2AR and β2AR(E122W) (Roth et al., 2008), as well as between β2AR and β2AR-T4L (Rosenbaum et al., 2007), we feel confident in drawing relevant conclusions from studies focusing on the molecular interactions associated with β2AR and its small molecule effectors using the β2AR(E122W)-T4L system, which we show here is structurally equivalent to carβ2AR-T4L. The binding orientation of timolol is similar to that of carazolol, where the oxypropanolamine tail forms strong interactions with the polar triad (Asp1133.32, Asn3127.39 and Trp3167.43), and the morpholino-thiadiazole head group binds in a similar orientation to the carbazole head group of carazolol (Figure 1B). However, two subtle yet relevant differences occur between the carazolol and timolol binding modes. The thiadiazole ring of timolol binds deeper into the receptor pocket allowing an additional hydrogen bonding interaction with Thr1183.37. Secondly, an additional polar interaction is formed between the ether moiety of the morpholino ring of timolol and Asn2936.55, which has been implicated in agonist binding and the transfer of conformational information to the G protein coupling interface (Hannawacker et al., 2002; Wieland et al., 1996). Due to hydrogen bonding between Asn2936.55 and Tyr3087.35, both residues have a slightly altered side-chain rotamers relative to carβ2AR-T4L to accommodate the receptor-timolol interaction (Figure 1B). This interaction may result in timolol being a more effective inverse agonist due to interactions formed with helix VI that restrict the receptor from forming an active conformation. Although a direct comparison of inverse agonism between timolol and carazolol has not been reported, a study describing the suppression of basal activity of β2AR by various ligands resulted in 28% suppression by timolol and partial enhancement of signaling by carvedilol, a chemically homologous ligand to carazolol (Baker et al., 2003). Another study of the suppression of basal activity of β2AR indicated a much stronger inverse agonism effect of timolol (Chidiac et al., 1994) and carazolol itself suppressed 50% basal activity in a reconstituted liposome system (Rasmussen et al., 2007). These data indicate that the magnitude of inverse agonism is context dependent, although one would expect the rank order of inverse agonist efficacy to remain constant.
Over the past few years, studies highlighting the effect of cholesterol depletion on ligand binding characteristics of a few receptors in membranes have been reported and recently reviewed (Pucadyil and Chattopadhyay, 2006). In addition, the thermal stability of both the oxytocin receptor and the β2AR is improved in the presence of cholesterol and cholesteryl hemisuccinate (CHS), respectively (Gimpl and Fahrenholz, 2002; Yao and Kobilka, 2005). For the oxytocin receptor, cholesterol or cholesterol analogues that enhance thermal stability also shift the receptor to the high-affinity agonist binding state implying allosteric modulation by cholesterol.
The structure of carβ2AR-T4L had interpretable density for three molecules of cholesterol per monomer of protein with visible density for the palmitate moiety that is post-translationally attached to Cys341 and located between cholesterol molecules 2 and 3 of a symmetry- related monomer (Figure 2A). Due to the position of the cholesterol molecules relative to the crystal-packing interface, the biological relevance of the binding sites could not be decoupled from potential crystal packing artifacts (Figure 2A) and the orientation of the sterol ring relative to the receptor could not be established. With the generation of crystals for timβ2AR(E122W)-T4L in a higher symmetry space group and the antiparallel orientation of the receptor molecules within the asymmetric unit, the role of cholesterol in crystal packing is reduced (Figure 2B). The retention of binding sites for cholesterol 1 and 2 in the current structure indicate that two sites are not experimental artifacts and that cholesterol binding in this region may be physiologically relevant. The electron density for cholesterol 1 in the timβ2AR(E122W)-T4L does not support the position seen in the carβ2AR-T4L structure, rather a 90° rotation and slight translation along the long axis of the molecule are necessary for optimal fitting of the experimental electron density (Figure 2C and 2D). This alternate orientation results in the cholesterol molecules packing with the sterol ring system of cholesterol 2 related to cholesterol 1 by an approximate two-fold rotation and a slight translational shift along an axis parallel to the ring system. Both cholesterol molecules bind in a shallow surface groove formed by segments of helices I, II, III and IV and, thus, provide an increase in the intramolecular occluded surface area, a parameter linked to the enhanced stability of proteins from thermophilic organisms and used to compare the internal packing of helices in membrane proteins relative to soluble protein (DeDecker et al., 1996; Eilers et al., 2002; Eilers et al., 2000). We use the occluded surface area method to analyze packing value (PV) differences between the transmembrane helices compared to each other and to their equivalent helices in rhodopsin (PDBID: 1U19) (Pattabiraman et al., 1995). It is interesting to note that overall the rhodopsin helical bundle has slightly tighter packing interactions than β2AR (PV: 0.43 vs. 0.42), perhaps reflective of its greater stability. In the absence of cholesterol, helix IV has the loosest packing of all the transmembrane helices in β2AR (PV without cholesterol: 0.37), whereas it is tightly packed in rhodopsin (PV: 0.45). Binding interactions with cholesterol 1 increase the packing for this helix (PV with cholesterol: 0.39), which is indicative of a decrease in mobility (Figure 3A, B). Both cholesterol molecules contribute equally to the increase in packing of helix II upon cholesterol binding (Figure 3A, B). However in the absence of cholesterol, helix II has the second highest occluded surface area of the transmembrane helices so that the marginal gain from cholesterol interactions on this helix would not significantly affect the mobility of the helix or the overall thermal stability of the protein. We show that cholesterol increases the packing interactions for both helix II and IV explaining structurally the observed increase in thermal stability, however helix IV is still unable to achieve the high degree of intramolecular packing interactions associated with rhodopsin, indicating that this helix is perhaps one of the weakest points in the β2AR fold. While the overall effect on normalized occluded surface area upon cholesterol binding appears small, the differences between rhodopsin and β2AR over the 7TM bundle are also very small even though the resulting change in thermostability appears quite significant. Similarly, comparison of hyperthermophilic proteins to their mesophilic counterparts indicates an increase in normalized occluded surface area of 4% at the most, putting the change in packing and its relationship to stability on a more absolute scale (Eilers et al., 2002). It is likely that the main effect of cholesterol on stability is mediated through its effect on helix IV where cholesterol 1 serves as a bridge to helix II, which forms many intramolecular interactions contributing to the core of the receptor fold (Figure 3B and 3C).
We have further characterized the thermal stability of β2AR(E122W)-T4L solubilized in the detergent dodecylmaltoside (DDM) in the presence and absence of CHS and timolol by monitoring the extent of covalent coupling of receptor cysteines to the thiol-reactive dye, 7-diethylamino-3-(4'-maleimidylphenyl)-4-methylcoumarin (CPM), as a function of time and guanidinium hydrochloride (GnHCl) concentration (Figure 3D) (Alexandrov et al., 2008). The half-life of thermal denaturation of β2AR(E122W)-T4L at 35 °C changes significantly upon incorporation of both timolol and CHS (Figure 3D). The overall stabilizing effect of CHS in 1 M GnHCl results in a 5-fold increase in the half-life of protein denaturation in the absence of timolol and a 2-fold increase over protein bound to timolol alone. Relative to the apo protein, with a half-life of 8 minutes, incorporation of both ligands simultaneously increases the half-life to 127 minutes, which is almost a 16-fold improvement in stability (Figure 3D).
The specific interaction of other membrane proteins with cholesterol has previously been postulated and a consensus sequence for cholesterol binding has been proposed within the context of the peripheral-type benzodiazepine receptor (Li and Papadopoulos, 1998). Cholesterol binding appears to be dependent on the presence of an incipient cleft located at a membrane interfacial region, which contains at least one aromatic residue and a positively charged residue capable of participating in electrostatic interactions with apical hydroxyl group (Epand et al., 2006; Jamin et al., 2005). The cleft formed by β2AR helices I, II, III and IV is capable of accommodating two cholesterol molecules although there are a reduced number of interactions between β2AR and cholesterol 2. Interactions with helix IV are perhaps most analogous to the previously defined cholesterol binding motif and together with an additional site on helix II form a receptor cholesterol consensus motif (CCM) defined by four spatially distributed interactions with cholesterol 1. The aromatic Trp1584.50 is almost universally conserved (94%) among class A GPCRs and appears to contribute the most significant interaction with the sterol ring of cholesterol 1 through a CH—π interaction and the edge of ring D. The hydrophobic Ile1544.46 interacts with rings A and B and is 60% homology conserved (35% by identity). An aromatic residue from helix II, Tyr702.41 in β2AR, forms Van der Waals interactions with ring A of cholesterol 1 and hydrogen bonds to Arg1514.43. Importantly, a positive charge at an analogous position to Arg1514.43 is only 22% conserved with either arginine or lysine occupying the position. However, due to the non-specific nature of electrostatic interactions in the interfacial region of the membrane, nearby positions with positive charge may also serve the role of interacting with the cholesterol hydroxyl group, the limits of which will need to be established structurally. Consideration of the spatial distribution of conserved residues that are important for cholesterol binding in β2AR allow the definition of the CCM for cholesterol interaction within the class A receptors based on the Ballesteros-Weinstein numbering scheme as follows: [4.39–4.43(R,K)]---[4.50(W,Y)]---[4.46(I,V,L)]---[2.41(F,Y)] where 26% of class A receptors (21% for human class A receptors) (Table 2) are predicted to bind cholesterol at the same site as β2AR (Figure 4A). The four positions within the CCM are listed according to their perceived rank order of binding interactions with the positively charged position contributing the most binding energy and the aromatic residue on helix II contributing the least.
The effect of cholesterol binding on the properties of any receptor will likely vary across the class and will need to be determined for each member. For instance, both the oxytocin and serotonin1A receptors contain the CCM sequence and are affected by cholesterol. In both receptors cholesterol incorporation led to a dramatic increase in agonist affinity and for oxytocin an increase in thermal stability was detected (Gimpl et al., 1997; Gimpl and Fahrenholz, 2002; Pucadyil and Chattopadhyay, 2004). On the other hand, cholesterol stabilizes rhodopsin in the inactive state through indirect effects on the plasma membrane curvature and molecular simulation argues for a nonspecific binding of cholesterol (Grossfield et al., 2006; Niu et al., 2002). In β2AR, addition of cholesterol induces an increase in the affinity for the partial inverse agonist timolol, but no change is observed for the full agonist isoproterenol (Figure 4B). Furthermore, the stability enhancement observed upon cholesteryl hemisuccinate incorporation in no way precludes the production of a sufficiently stable and active receptor in Spodoptera frugiperda which have significantly less cholesterol than mammalian cells (Marheineke et al., 1998). While the physiological effect of cholesterol binding to β2AR is unknown it has been established that β2AR preferentially sequesters to cholesterol rich caveolae in neonatal rat cardiomyocyte cultures (Ostrom et al., 2001; Rybin et al., 2000; Steinberg, 2004), and partitions out of the caveolae upon stimulation (Rybin et al., 2000). While this compartmentalization has been shown to be due in part to the presence of a C-terminal PDZ binding domain that directs the trafficking of β2AR to caveolae (Xiang and Kobilka, 2003), receptor cholesterol interactions may also play a role in defining the trafficking properties of β2AR and other GPCRs. Furthermore sequence disparities in the cholesterol binding site among receptors may potentiate the trafficking effect of cholesterol binding, analogous to the trafficking and cholesterol sequestration effects observed for caveolin and other cholesterol binding proteins (Epand et al., 2005).
The presence of an aromatic residue at postion 2.41 is the most restrictive of the four rules and also appears to be the least important for binding of cholesterol based on the structure. In the absence of this requirement 44% of human class A receptors would contain the revised CCM (rCCM) and should bind cholesterol in a similar manner as β2AR (Figure 5, Table 2). While, future mutagenesis studies will illuminate the actual rank order importance of these positions in cholesterol binding, it is increasingly apparent that sterols play an important direct role GPCR stability and pharmacology.
See the Supplemental Data for additional experimental procedures.
β2AR(E122W)-T4L was identical to β2AR-T4L in all but two features. (1) Residue 122 was mutated from a glutamate to a tryptophan using standard site-directed mutagenesis protocols, and (2) the C-terminus was further truncated to residue 348 eliminating a total of 65 residues. High-titer recombinant baculovirus was obtained using standard protocols in the Bac-to-Bac system (Invitrogen) and used to infect spodoptera frugiperda (Sf9) insect cells at a multiplicity of infection of five and the expression was allowed to proceed for 48 hours. Insect cell membranes were initially disrupted by nitrogen cavitation in a hypotonic buffer containing 10 mM Hepes pH 7.5, 20 mM KCl and 10 mM MgCl2. Extensive washing of the membranes was carried out by repeated centrifugation and dounce homogenization to strip the membranes of soluble and membrane associated proteins. Membranes were flash frozen and stored at −80 °C until further use. Prior to solubilization, prepared membranes were thawed on ice in the presence of 1 mM timolol, 2 mg/mL iodoacetamide, and protease inhibitors. Membranes were then solubilized by incubation in the presence of 0.5%w/v dodecylmaltoside (DDM) for 2–4 hours at 4 °C. After solubilization, the solution was clarified at 100,000 g and the resulting supernatant incubated with Co2+ charged TALON IMAC resin overnight at 4 °C. The resin was washed with 20 mM imidazole to remove impurities followed by an elution of the receptor with 200 mM imidazole. An additional IMAC step after desalting was used to concentrate and deglycosylate (PNGase) the receptor (increasing purity up to 98%). The protein was maintained in 20 mM Hepes pH 7.5, 150 mM NaCl, 0.05% DDM, 0.01% CHS and 1 mM timolol throughout the purification unless otherwise indicated. Timolol concentration was increased to 5 mM on the second IMAC column to avoid ligand depletion in the subsequent concentration step which utilized a 600 µL Vivaspin cartridge with a 100 kDa molecular weight cutoff.
Crystals were generated from a 10% w/w cholesterol/monoolein lipidic cubic phase after reconstituting the protein from a 30 mg/mL solution and overlaying the lipid phase with a solution of 28% w/v PEG 400, 300 mM K Formate, 100 mM Bis-tris propane pH 7.0 and 2 mM timolol. Crystals were obtained from 25 nL of cubic phase and harvested directly from glass sandwich plates in which they were grown (Cherezov et al., 2004).
Thermal stability data was collected by a modified procedure utilizing a thiol-reactive coumarin maleimide, 7-diethylamino-3-(4'-maleimidylphenyl)-4-methylcoumarin (CPM). The data were collected at 35 °C in the presence of guanidinium hydrochloride (GnHCL) at three different concentrations. The chemical denaturant was included to speed the process of unfolding thereby minimizing the effects of dehydration on data processing error. Protein was purified as according to standard procedure until the final Ni-NTA purification and concentration step at which point timolol and cholesteryl hemisuccinate (CHS) were washed from the protein with five column volumes of buffer. The protein was then eluted at approximately 2 mg/mL as a 10x stock. The protein was then diluted 3-fold into a 50 mM Hepes pH 7.5, 150 mM NaCl buffer containing 2 µg of CPM dye (diluted from a 4 mg/mL stock in dimethyl formamide) and incubated at 4 °C for 10 minutes. Various combinations of timolol and CHS in dodecylmaltoside (DDM) were then added to the mixture at 2x final concentration and allowed to equilibrate for 30 minutes at 4 °C before the final dilution to working volume with the appropriate concentration of GnHCl in the assay buffer. Data was collected as soon as possible after the addition of the chemical denaturant to reduce the degree of unfolding prior to measurement. Fluorescence of the CPM dye was measured on a Tecan Genios Pro fluorescent plate reader using 340 nM excitation filter with a 35 nM bandpass and a 475 nM emission filter with a 20 nM bandpass. Readings were measured every minute for three hours. The data were fit to a single exponential decay curve using GraphPad Prism Software (San Diego, CA).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.