The palmitoylation assay was carried out as reported by Drisdel et al. . Briefly, receptor was immunoprecipitated with protein G agarose beads. The beads were then sequentially incubated with 50 mM NEM to block free sulfhydryl groups, 1 M hydroxylamine to remove thioester-linked palmitic acid, and 40 μM btn-BMCC to conjugate biotin to the depalmitoylated cysteines. To assess the receptor palmitoylation level, the amount of conjugated biotin was determined by immunoprecipitation and immunoblotting [12, 26]. Protein concentrations and receptor expression levels were measured to ensure that equal amounts of receptor were loaded in each lane of the gel.
Membrane purification and cholesterol assay
Cells were homogenized in 0.32 M sucrose and 10 mM HEPES (pH 7.7). The crude lysate was then centrifuged at 1,000 × g for 10 min at 4°C, the supernatant was collected, and the pellet was re-homogenized. These processes were repeated until the pellet appeared translucent. The collected supernatant was centrifuged at 100,000 × g for 60 min at 4°C. The pellet was re-suspended and used to determine the cholesterol content in cell membranes. The results were normalized against cholesterol levels in cells under control condition. Cholesterol concentrations were determined by using the Amplex Red Cholesterol Assay Kit (Invitrogen, Carlsbad, CA) on the cell membrane preparation according to manufacturer's instructions.
To determine the amount of cholesterol associated with the OPRM1 complex, a new method was used. Cells were first treated with lysis buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.25% sodium deoxycholate, 0.1% Nonidet P-40, 0.5% Triton X-100, 0.1% digitonin, 50 mM NaF, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 50 mM sodium pyrophosphate, 10 mM sodium vanadate, and 1X protease inhibitor cocktail; Roche, Indianapolis, IN). The supernatants from the cell lysates were divided into three equal aliquots. These aliquots were used to perform co-immunoprecipitation with PBS (control), HA antibody (Convance, 1:1000) (to detect HA-tagged OPRM1 and C170A), or FLAG antibody (Sigma, 1:1000). After antibody incubation, protein G agarose (Invitrogen, Carlsbad, CA) was added for an additional overnight incubation. The resulting agarose was used to determine the amount of precipitated cholesterol with the Amplex Red Cholesterol Assay Kit (Invitrogen, Carlsbad, CA). The FLAG antibody was used as a control antibody to exclude any possible influence of antibody usage. The greater amount of cholesterol precipitated by HA antibody compared with PBS or FLAG antibody reflects the cholesterol associated specifically with OPRM1 signaling complex. Although this method does not directly determine cholesterol's interaction with the receptor, it does specifically detect cholesterol's interaction with the receptor signaling complex.
CFP and YFP were fused to the C terminus of wild-type OPRM1 or the C170A mutant of OPRM1. YFPGαi2 has YFP inserted between residues 91 and 92 of Gαi2 . Throughout the studies, all FRET values are expressed as the normalized net FRET using the following formula: IFRET = [(ICFP × CoA) - (IYFP × CoB)]/[the square root of (ICFP × IYFP)]. IFRET is the fluorescence intensity when a CFP-YFP (excitation-emission) filter set is used, ICFP is the fluorescence intensity when a CFP-CFP filter set is used, and IYFP is the fluorescence intensity when a YFP-YFP filter set is used. CoA was determined in the cells transfected with only CFP constructs by the following formula: CoA = IFRET /ICFP. CoB was determined similarly. Including "square root" in the formula eliminates the influence from the differential expression of CFP- and YFP-conjugated protein. Briefly, more than twenty individual regions on the cell membrane of a single cell were analyzed, and more than twelve individual cells were analyzed for each sample.
OPRM1 binding assay
The amounts of receptor on the cell membrane and the affinity of agonists for receptors were determined by binding assay . Briefly, purified cell membrane was incubated with [3H]-diprenorphine and agonists/antagonists. After incubation, PEG8000 and NaCl were added to trap the receptors on Whatman GF/B filters for final radioactive reading. Scatchard analyses were carried out to determine the level of wild-type or mutant OPRM1 expressed on cell membranes. To determine the affinities of various ligands, the cell membranes were incubated with 2 nM [3H]-diprenorphine and various concentrations of morphine, naloxone, and CTOP (0.01 nM - 10 μM). These competition binding studies were analyzed with one- or two-site curve-fitting models in GraphPad 5.0.
The pCMV-shuttle vector (Stratagene) was used in all studies. cDNA generation from the receptors, Gαi2, and their fluorescence-conjugated constructs was controlled by the CMV promoter. Transient transfections were performed with Lipofectamine 2000 (Invitrogen) following manufacturer's instructions. Cells were allowed to rest for 24 h before further treatment.
Assays based on antibodies
Immunoblotting and co-immunoprecipitation assays were performed as described previously . The same Confocal Imager used for FRET was used to analyze colocalization. Adenylyl cyclase inhibition was measured as previously reported . ERK phosphorylation was determined by immunoblotting .
Colocalization studies were performed as reported previously . Briefly, cells were cultured on poly-lysine-coated coverslip in six-well plates. After transient transfection and various treatments, cells were fixed with 2% formaldehyde for 30 min. HA, Flag, and Gαi2 antibodies were used as primary antibodies (1:1000). The confocal images were captured with a BD CARV II Confocal Imager and a Leica DMIRE2 fluorescence microscope. Colocalization of the fluorescence pixels was calculated with IPlab 4.0 software (BD Biosciences-Bioimage) and the following formula: 2 × Nyellow /(Nred + Ngreen), where N represents the number of pixels with fluorescence intensity over a pre-defined threshold.
Development of the OPRM1 homodimer model
A computational model of the OPRM1 inactive state was developed using the β2-AR crystal structure as a template  with two major modifications. First, the TMH 7/elbow/Hx8 region of the β2-AR was replaced with that of the adenosine A2A crystal structure  because the "elbow region" between TMH7 and the C terminus Helix 8 contains only two residues (P7.57 and D7.58). This results in an elbow region of β2-AR that is stretched. We would not expect OPRM1 to have a similar conformation since it has three elbow residues (D7.57 E7.58 N7.59). We therefore used the elbow conformation in the A2A crystal structure, which also has three residues , in the OPRM1 model. Second, the Monte Carlo/simulated annealing technique CM  was used to study the conformations of three OPRM1 TMHs with important sequence divergences from the β2-AR template: TMH2 (P2.58 OPRM1 vs. P2.59 β2-AR), TMH4 (P4.59 OPRM1 vs. P4.60 β2-AR), and TMH6 (CWTP OPRM1 vs. CWLP β2-AR). The CM technique explores the low free energy conformations possible for a helix of interest using Monte Carlo simulated annealing. The method of CM, developed by Guarnieri and Wilson  and extended by Guarnieri and Weinstein , efficiently and completely explores the dihedral conformational space of a molecule, independent of the dihedral conformation of the initial molecular structure. The CM method combines Monte Carlo exploration of the dihedral angle space with simulated annealing (MC/SA) to determine the range of values that each dihedral angle is capable of reaching in a broad temperature range. The CM method has been expanded to allow variation of bond angles in addition to dihedral angles .
In the CM calculations reported here, the backbone dihedrals of each helix were set to the standard φ (-63°) and ψ (-41.6°) for transmembrane helices. Our established protocol is to allow all torsion angles to vary ± 10°, and to allow a larger variation of ± 50° in regions containing flexible areas. These flexible areas are regions where there are known helix bending residues such as prolines, glycines, serines and/or threonines . The OPRM1 TMH regions considered flexible were the following: TMH2: region of i (P2.58) to i-4 (T2.54); TMH4 region of i (P4.59) to i-4 (A4.55) and TMH6 region of i (P6.50) to i-4 (V6.46). Individual bond angles were allowed to vary ± 8°.
CM calculations are performed in two phases: an exploratory phase and a biased annealing phase. In the exploratory phase, a random walk is used to first identify the region of conformational space most probable for each torsion angle and bond angle. Specifically, each step consists of varying two dihedral angles and one bond angle chosen at random from the entire set of variable angles. The torsion angles and bond angles are randomly picked at each temperature and each move is accepted or rejected using the Metropolis criterion . Accepted conformations in the exploratory phase are used to create "memories" of torsion angles and bond angles that were accepted. This information provides a map of the accessible conformational space of each TMH as a function of temperature. In the biased annealing phase, the only torsion angle and bond angle moves attempted are those that would keep the angle in the "populated conformational space" mapped at 310 K in the exploratory phase.
Here, the initial temperature for each run was 3000 K with 50,000 Monte Carlo steps applied to each torsion or bond angle variation, with cooling in 18 steps to a final temperature of 310 K. The biased annealing phase for the calculations began at 749.4 K, and the cooling to 310 K was performed in 7 steps. 105 structures were output at 310 K. The output from each TMH study was superimposed on the corresponding template helix in the β2-AR template that had been mutated to the sequence of OPRM1. A helix was selected for inclusion in the revised OPRM1 that fit in the bundle with no VDW overlaps with residues on other TMHs.
The CM helices chosen for substitution into the TMH bundle had the following helix bend angles, wobble angles, and face shifts: TMH2 (35.2°, -105.8°, 40.3°), TMH4 (14.8°, -126.1°, 25.9°), and TMH6 (30.6°, -129.9°, 45.6°). Extracellular and intracellular loops were added using MODELLER v8.2 . Energy minimizations were performed using Macromodel and the OPLS2005 all-atom force field (version 9.8, Schrödinger, LLC, New York, NY). A distance-dependent dielectric, 8.0 Å extended nonbonded cutoff, 20.0 Å electrostatic cutoff, and 4.0 Å hydrogen bond cutoff were used. A palmitoyl was added to C3.55(170) and a cholesterol was docked between palmitoylated C3.55(170) and TMH3. Interactive docking studies in Maestro (version 9.1, Schrödinger, LLC, New York, NY) were used to orient two OPRM1/cholesterol protomers at the symmetric TMH4 interface of mouse dark state rhodopsin . In this orientation, the OPRM1 protomers form an interface analogous to that shown by Guo and co-workers to characterize the inactive state homodimer interface of the dopamine D2 receptor . This interface in OPRM1 involves N4.41, I4.44, C4.48, I4.51, A4.55, and P4.59 on each protomer. In the resultant dimer, cholesterol is packed against the TMH4 interface and TMH3. The palmitoyl at C3.55(170) is packed against the cholesterol and TMH5, blocking cholesterol from leaving the interface.
The energy of the OPRM1 homodimer complex was minimized using the same force field, dielectric, and cutoffs as described above. In the first stage of the calculation, Polak-Ribier conjugate gradient minimization was employed until a gradient of 0.1 kcal/mol A2 was reached. A force constant of 250 kcal/mol was used on the loop backbone atoms. All charged residues in the loop regions and at the ends of the TMHs that face toward lipid headgroups were mutated to neutral forms. Non-moving fixed atom restraints were applied to the C-alpha atoms of TMH3 in both protomers, restraining the protomers from moving apart. The protocol was repeated with TMH3 non-moving fixed restraints removed.
Macromodel was used to output the pair-wise interaction energy (VDW and coulombic) for a given pair of atoms. The nonbonded interactions are represented in OPLS2005 (as implemented in Macromodel through Coulomb and Lennard-Jones terms) interacting between sites centered on nuclei. Thus, the intermolecular interaction energy between molecules a
is given by the sum of interactions between the sites on the two molecules [37
], as represented in the following equation:
where a and b are collected by Macromodel as atom sets representing all atoms of a single residue for a and all atoms of a nearby residue for b. The residue represented by a is evaluated separately against all residues within a 7.0Å radius of residue a. With cholesterol A defined as group 1, and with all of the atoms of any residue within 7.0 Å of protomers A or B defined as group 2, the pair-wise interaction energies were calculated. The interaction energy at the homodimer interface was calculated as the sum of the interaction energies between protomers A and B at the homodimer interface plus the interaction energy of cholesterol A with protomer B and the interaction energy of cholesterol B with protomer A. Cholesterols A and B were blocked from interacting with each other by the close interactions and steric bulk of protomer A's TMH4 and protomer B's TMH4; the palmitoyls could not interact with each other for the same reason.