Characterization of a Computationally Designed Water-soluble Human μ-Opioid Receptor Variant Using Available Structural Information

THE G-protein–coupled receptor (GPCR) superfamily contains many major pharmacological targets.¹  The study of the molecular mechanisms of GPCRs is hampered by limited information on structure–function relations, the difficulties of working with membrane proteins, various conformational states of the receptors, and the complexity in the mechanisms of selectivity and activation. Solving the crystal structures of GPCRs is a major scientific and technologic undertaking and often requires substantial molecular engineering.²  Revealing the crystal structures of the four major opioid receptors has been a major milestone in opioid-related research and drug development.^3–6 

From the opioid family, the μ-opioid receptor (MUR) is clinically important because it is the target of opioid analgesics. MUR is a class-A GPCR widely expressed in the central nervous system, and its signaling pathways are relevant in mediating analgesia, sedation, and euphoria. Despite recent advances, the detailed mechanisms of MUR activation and related protein dynamics are still missing because of technical limitations including the difficulty in obtaining large quantities of the functional receptor and the inherent low stability when these proteins are isolated from the membrane environment.

To overcome the difficulty of studying the MUR in a lipid environment, we successfully applied computational protein design to create the first version of a water-soluble variant of the human MUR (wsMUR-TM). The protein was expressed in Escherichia coli and retained structural and functional features consistent with those of the native MUR.⁷  wsMUR-TM was engineered using a comparative (homology) model based on the crystal structures of human β₂ adrenergic receptor and bovine rhodopsin.⁷  Although the homology model of the human MUR superimposed well with the crystal structure of murine MUR overall (root-mean-square deviation of the Cα atoms = 2.60 Å),⁸  there were several structural differences that would impact the selection of sites for designing to confer water solubility. Such differences were expected because there was only 27% sequence similarity between 288-residue transmembrane portions of the human β₂ adrenergic receptor and the human MUR.⁷  The murine MUR crystal structure provided critical information to locate the mutated positions in wsMUR-TM because the sequence similarity between the human and the mouse MURs was 94% for the entire sequence and 99% for the segment solved in the crystal structure. Such critical information was used in the current study to reengineer the wsMUR-TM by eliminating mutations that may be unnecessary to confer water solubility and avoid toxicity to the bacterial expression system. This restoration of native amino acid identities allowed us to explore possible deleterious expression, solubility, structure, and function and probe the sensitivity of structural and functional features of the designed variants to the choice of template structure. The resulting water-soluble variant (wsMUR-TM_v2) was characterized; structural and functional properties were compared with that of wsMUR-TM. Important findings regarding the stability of these water-soluble analogs were analyzed in detail as well as their relevance as suitable surrogates to study properties of the native MUR.

The SNAP-Lumi4-Tb-μ-opioid receptor (Cisbio Bioassays, Bedford, MA), which is composed of a SNAP-tag^® enzyme labeled with terbium cryptate (Lumi4^®-Tb, a long lifetime fluorescence resonance energy transfer donor) fused at the N-terminus of a native MUR, was expressed on human embryonic kidney 293 cells. Red-naltrexone, an analog of naltrexone and the MUR antagonist that contains the d2 dye, was also obtained from Cisbio Bioassays for receptor affinity determination. All other chemicals with high purity were obtained from Sigma-Aldrich (St. Louis, MO) except when specified otherwise.

The structure of the murine MUR (protein data bank accession code: 4DKL)⁶  was obtained from the protein data bank.^* The sequence of the previously designed water-soluble variant was mapped onto the crystal structure of the murine MUR, and structural analysis was performed using the Visual Molecular Dynamics program (Beckman Institute, Urbana, IL)⁹  and PyMOL (Version 1.3; Schrödinger, LLC, Portland, OR^†). Residue positions selected to be mutated to the native human residue identities were identified based on their relative location in the context of the murine μ-opioid crystal structure. Herein the Ballesteros and Weinstein indexing is adopted as a superscript in identifying particular residue positions in the receptors.¹⁰ 

The synthetic complementary deoxyribonucleotide encoding of the wsMUR-TM (previous version reported)⁷  and wsMUR-TM_v2 (the modified new version) was constructed by DNA2.0 Inc. (Menlo Park, CA) and GenScript Inc. (Piscataway, NJ), respectively. The sequences were subcloned between the NdeI and XhoI restriction sites of the expression plasmid pET-28b(+) (EMD/Novagen, Billerica, MA). This cloning strategy resulted in placement of a His-tag at the amino terminus of the protein and was not removed in subsequent characterization. Protein expression using E. coli BL21(DE3) cells (EMD/Novagen) and purification were performed as described previously.⁷ 

The purified proteins were run in a sodium dodecyl sulfate (SDS) polyacrylamide gel electrophoresis gel. The protein band of the expected molecular weight was excised and digested with trypsin for verification purposes. NanoLC proteomics experiments were run at 200 nl/min for 60 min with gradient elution by injecting digested peptides into a nano-LC mass spectrometry (10 cm C18 capillary column) to be separated by Eksigent (San Francisco, CA). Separated peptides were sprayed into LTQ by Nanospray (Thermo Fisher Scientific, Waltham, MA). Xcalibur (Xcalibur, Inc., Arlington, VA) was used for raw data acquisition. Sequest (Thermo Fisher Scientific^‡) was used to search the database Uniprot.^§ Scaffold 2.6 (Proteome Software, Inc., Portland, OR) was used to analyze the data quantitatively.

Tapping-mode atomic force microscopy in air (MFP-3D; Asylum Research, Goleta, CA) was used to visualize the protein particles under various solution conditions as we described previously.¹¹  Similar technology has been used to observe Alzheimer β-amyloid oligomers and monomers in solution conditions.¹²  To prepare a sample, protein solution was applied deposited onto the surface of an oxidized silicon wafer and allowed to incubate in a humid atmosphere for 1 h. The sample was then washed in two separate baths of deionized water for 3 min each to remove weakly bound protein molecules. From this series of experiments, we tested whether the protein remained predominantly in a monomeric state in the buffer condition.

Circular dichroism was used to determine the secondary structure content and the thermostability of the protein as described previously.¹³  Chirascan circular dichroism Spectrometer (AppliedPhotophysics Limited, Leatherhead, United Kingdom) was used with 1 mm path length and a scan speed of 1 nm/s. The experiments were performed at least three times to ensure repeatability of the experiments. Data sets were recorded three times and averaged to increase the signal-to-noise ratio. The CDNN circular dichroism spectra deconvolution software¹⁴  was used to determine the content of the secondary structure. Circular dichroism spectra for wsMUR-TM_v2 were also determined from 10° to 90°C with increments of 2°C per min (6 μM of the receptor in 5 mM sodium phosphate, pH = 7.0) in the absence and presence of 1 M sucrose. The effect of sucrose on the helical content of the wsMUR-TM_v2 was also tested in the presence and absence of 3 M urea. The cuvette was sealed with oil to avoid errors from evaporation. The helical content is plotted against temperature using GraphPad Prism (v 5.02; GraphPad Software, Inc., La Jolla, CA).

To determine the binding affinities of the modified new protein constructs, we used the same homogeneous time-resolved fluorescence methodology described in our previous report.⁷  In brief, Tag-lite^® MUR cells (Cisbio Bioassays) in culture medium were dispensed in white low-volume microplates (384 wells; Greiner Bio-One North America, Monroe, NC) at 3,700 cells per 10 μl per well Tag-lite^® μ-opioid–labeled cells with 60 nM of Tag-lite^® opioid receptors red-naltrexone, 5 μl of the engineered MUR with 1:2 serial dilutions from the micromolar to the nanomolar range was added into the sample with the final volume of 20 μl. After incubation at room temperature for 2 h, homogeneous time-resolved fluorescence signals were determined by FLUOstar plate reader (BMG, Cary, NC) with excitation at 337 nm at 620 and 665 nm emission. The experiments were repeated at least three times. Affinities for wsMURs were determined by fitting to concentration–responses curves using the GraphPad Prism program (v 5.02, GraphPad Software, Inc.).

To interpret the experimental ligand-binding data after the seven mutations were introduced, docking calculations for the ligand naltrexone using the structure of the murine MUR (4DKL)⁶  were carried out using the DockingServer^15 ‖ as previously described.^16,17  Affinity (grid) maps of 30 × 30 × 30 Å grid points with a spacing of 0.375 Å were used for the docking calculations.

The shift of the thermostability curve is determined by comparing the temperature difference where the normalized mean residue molar ellipticity is at 50%. Three repeats of data are obtained and analyzed using the t test. The affinities are presented as mean ± 1 SD from three repeated measurements. The results were analyzed using GraphPad Prism (version 5.02 Windows version). P values less than 0.05 are considered statistically significant.

From a comparative structural analysis of wsMUR-TM and the murine crystal structure, five positions (E120^2.54, K130^2.64, D232^5.36, G305^6.58, and K306^6.59) were identified whose variation could possibly disrupt either ligand binding or internal packing of amino acids (fig. 1A). On the basis of previous binding studies carried out in wsMUR-TM with a naltrexone analog, we found no significant disruption in the binding capabilities of wsMUR-TM compared with the native receptor. Nonetheless, native MUR also binds larger ligands, and these five mutations in the vicinity of the orthosteric binding site could affect the receptor’s binding to other ligands.⁷  In addition, two residues were mutated back to their native residue identities based on structural considerations (E297^6.50 and K303^6.56). Position E297^6.50 was originally a proline residue in wsMUR-TM, and its presence could be important in retaining the right geometry of transmembrane helix 6. At position K303^6.56, the native isoleucine appears to play an important role in the complementary packing of the transmembrane 5 and transmembrane 6 helices (fig. 1A). None of the seven positions were directly in contact with the β-funaltrexamine (β-FNA) ligand that was cocrystalized with murine MUR. These seven residues in wsMUR-TM, located in the transmembrane helical core and near the binding pocket, were mutated back to their native identities (E120T^2.54, K130Y^2.64, D232N^5.36, E297P^6.50, K303I^6.56, G305K^6.58, and K306A^6.59) to generate a modified new analog denominated wsMUR-TM_v2 (fig. 1, A and B).

Both variants of the wsMURs expressed well and were isolated with high purity using nickel affinity chromatography. There was no appreciable difference in the bacterial growth (as monitored by optical density) and the yields of both variants (approximately 20 mg/l of shake flask culture). Approximately 0.1% SDS was initially required to solubilize and purify the receptors. After removing nonbound SDS via dialysis, both variants were soluble at approximately 7.5 mg/ml in a buffer containing 130 mM NaCl and 20 mM sodium phosphate with a pH of 7.0 without appreciable precipitation. A thick dark band in the SDS-polyacrylamide gel electrophoresis presented in figure 2A belonged to wsMUR-TM_v2, as confirmed by mass spectroscopy. The coverage of the identified peptide segment was as high as 85%. One of the protein fragments of the wsMUR-TM_v2 variant, identified by mass spectrometry, is presented in figure 2B. It is important to note that there was a very faint band in the gel with a molecular weight corresponding to a dimer of the receptor (fig. 2A). This band was also studied by mass spectrometry and corroborated to be the sequence of wsMUR-TM_v2, suggesting the potential existence of covalent dimer of the protein.

From a series of atomic force microscopy experiments, we found that both wsMURs could be dispersed using buffer conditions and were predominantly monomeric (80 to 90%) when freshly prepared within a 1-week period. Consistent with the observation in SDS-polyacrylamide gel electrophoresis, the existence of features whose heights are consistent with those of the dimer was observed in the atomic force microscopy experiments. The receptor was observed to aggregate as a function of time, and 0.02% SDS and 5 mM 2-mercaptoethanol could reduce such aggregation significantly. As indicated in figure 3, A and B, 70% of the surface features are ascribed to be monomeric for a 2-month old sample of wsMUR-TM_v2 in a buffer containing 20 mM sodium phosphate, 130 mM sodium chloride, 0.02% SDS, and 5 mM 2-mercaptoethanol, pH = 7.0.

The circular dichroism spectra of wsMUR-TM_v2 indicated predominantly helical structure with a helical content of 57% estimated by the molar ellipticity over the range 205 to 260 nm. This value is consistent with similar measurements on wsMUR-TM (fig. 4). The helical content of wsMUR-TM_v2 and wsMUR-TM is presented in table 1. Upon addition of 1 M sucrose, no significant change in the helical content was observed. Also, the inclusion of 3 M urea reduced the helical content significantly, but such reduction was restored by the presence of 1 M sucrose (data not shown).

As monitored by circular dichroism, wsMURs started to lose ellipticity with increasing temperature and were nearly fully unfolded at 90°C (fig. 5A). The thermal stability curve of wsMUR-TM_v2 shifted to the left (lower temperature) by 5.8° ± 1.5°C relative to the wsMUR-TM (P = 0.007). It is important to note that the thermal stability curve shifted to higher temperature significantly in the presence of sucrose (fig. 5B).

Naltrexone binding was monitored using a competitive time-resolved fluorescence resonance energy transfer–based assay with fluorescently labeled wild-type MUR and a naltrexone-derived antagonist. The ratio of fluorescence emission at 620 and 665 nm decreased in a concentration-dependent manner with increasing concentrations of both variants of wsMUR. There was no significant difference for the affinity with naltrexone for both wsMURs (fig. 6). The K_d value for wsMUR-TM was similar to the previously reported value (67 ± 4 nM), whereas the K_d value for wsMUR-TM_v2 was 70 ± 7 nM. Both values are in closed agreement with values for native MUR reported using other techniques.^7,18 

To interpret the findings from the binding assay at a molecular level, docking calculations were carried out using the mouse crystal structure and naltrexone. Docking calculation results indicated that the naltrexone binding site overlapped well with that of the β-FNA as demonstrated in figure 7A. The consistency in the affinity values for both wsMURs could be ascribed to the fact that there is no direct contact of any of the seven mutated residues with the ligand naltrexone (fig. 7B).

The crystal structure of the murine MUR complexed with the irreversible antagonist β-FNA⁶  provided valuable information to reengineer our previously designed water-soluble variant wsMUR-TM.⁷  On the basis of structural analysis, seven positions in wsMUR-TM were returned to their native residue identities. Residues were selected based on their relative position to the surface of the protein and the orthosteric binding pocket in the structure of the murine MUR. The resulting protein, wsMUR-TM_v2, expressed well in E. coli and was purified in large amounts sufficient to perform biophysical experiments without significant change in the protein’s water solubility. Interestingly, although the affinity of the protein for naltrexone did not change significantly, the thermostability of the protein decreased slightly. Addition of the osmolyte sucrose increased the thermal stability of the protein significantly.

In the quest to facilitate the characterization of different membrane proteins, computational design has been used to engineer water-soluble variants of membrane proteins. Previously, variants have been engineered of the potassium channel from Streptomyces lividans (KcsA)^19–21  and the transmembrane domain of the α1 subunit of the nicotinic acetylcholine receptor,²²  with mutations ranging from 18 to 30% of the total number of residues. In our first version of the water-soluble analog of the human MUR, wsMUR-TM, 18% of the residues (53 of 288) was mutated. An important question that remained was to what extent to which the number of mutations could be reduced without compromising desired properties such as water solubility and protein production in E. coli. As we discussed previously, the toxic effects of the human MUR to cell lines have limited the capability to express the native receptor in this heterologous bacterial expression system.⁷  The cytotoxicity of the native MUR is possibly a result of hydrophobic residues located within the receptor’s transmembrane region, as mutations of mainly hydrophobic residues in the designed water-soluble variants significantly reduced the cytotoxicity and facilitated overexpression in E. coli. Although we cannot conclude that all the mutated residues in the current study were required for high protein expression and formation of a variant with structural and functional properties comparable with those observed in the native receptor, we achieved the lowest number of mutations for any water-soluble variants we have designed to date (16%, 46 of 288 mutated residues), an important technical achievement.

The production of a soluble and functional water-soluble MUR for its biophysical characterization involves computationally redesigning the protein, and residues in the transmembrane region are the focus. Mutating residues, including mutating redesigned sites back to their native identities, could disrupt any of the steps associated with production of native-like protein, including cytotoxicity of the heterologous system (failure of expression), protein purification, protein stability, water solubility, and receptor’s binding properties. However, even after mutating seven residues located in transmembrane helical core of the receptor, the protein can still be overexpressed in E. coli and retains intrinsic structural and binding features as a MUR. The previous protein variant created using a previously created homology model alone (wsMUR-TM) and the new variant created by an analysis of the crystallographic murine receptor to restore wild-type identities at seven residues (wsMUR-TM_v2) had essentially identical helical content and naltrexone affinities. The information of the mutations introduced in wsMUR-TM_v2 offers additional structure-sequence details to identify the features involved in cytotoxicity, solubility, and sequence–structure–function relation studies for MUR and other GPCRs in the same family. These observations further confirm our previous finding that at least for the GPCR family of proteins, homology modeling can yield high-quality structures suitable for the computational design of water-soluble variants. Although the properties of the two designed proteins are similar, it is unclear whether there are significant tertiary structure differences. Because both are water soluble that can be overexpressed in E. coli, they can be used as parallel alternative tools to access sequence–structure–function relations using nuclear magnetic resonance in solution conditions, which is an ongoing effort within our group.

Atomic force microscopy provides a very useful tool to evaluate our water-soluble proteins. The data are consistent with previous work indicating that the engineered wsMURs are soluble in buffer conditions and stay predominantly monomeric.⁷  Although the protein exists largely in the monomeric form, the formation of oligomers, particularly dimers, is also confirmed. The presence of different oligomeric states is consistent with the SDS-polyacrylamide gel electrophoresis data. It is known that the MUR can form SDS-resistant dimers in living organisms, and dimerization may play critical roles in physiological and pharmacological functions of the receptor.^23,24  This self-association appears not to be eliminated by the large numbers of mutations involving exposed hydrophobic residues described in our study. Such dimerization persists but is diminished in the presence of 5 mM of the reducing agent 2-mercaptoethanol. The presence of SDS and 2-mercaptoethanol appears to reduce aggregation to improve the overall shelf life of the protein sample. Using crystallography, we have previously demonstrated that low concentrations of SDS do not change the protein structure in the natural mammalian protein, apoferritin.²⁵  Nonetheless, further studies are needed to determine the extent to which the tertiary structure of the wsMUR is altered in such low SDS concentrations. Further studies are also needed to eliminate SDS in the purification process.

The secondary structure of the modified new version wsMUR-TM_v2 is consistent with that of the previously reported wsMUR-TM variant. As intended, the introduction of the mutations K130Y^2.64, K303I^6.56, and K306A^6.59 seem to induce better protein packing as the helical content improved slightly in wsMUR-TM_v2 when compared with wsMUR-TM. The remaining mutations could also play similar roles in improving the overall protein packing of the new water-soluble variant. However, it is interesting to note that the thermal stability curve of the modified new design shifted to lower temperature by 5°C as compared with that of the first version, suggesting a slight decrease in the thermal stability. One possible explanation is the reintroduction of a proline residue at position 297^6.50. Although this mutation was intended to restore and facilitate interactions that stabilize the irregularities in transmembrane helix 6, the helix propensity of proline is the lowest of all the amino acid residues, and in the absence of the lipid bilayer, this could decrease the thermal stability of the protein.

With the exception of rhodopsin, GPCRs are generally unstable, one of the major obstacles for structure–function relation studies.²⁶  We recently demonstrated that the hydrophobic molecule cholesterol improves the thermostability of the wsMUR-TM.⁷  It is critical to identify a water-soluble compound that could be used to stabilize the protein thermodynamically. Sucrose stabilizes some proteins and forces some unfolded protein to folded states.²⁷  Consistent with these reports, sucrose was able to restore the helical content reduction due to presence of urea. Moreover, addition of 1 M sucrose increased the thermal stability of wsMUR-TM_v2 significantly. This finding is critical for further structural and functional studies of these opioid receptor analogs, because the addition of this osmolyte is able to stabilize the protein without precipitation. Sucrose also does not generally interference with protein nuclear magnetic resonance signals, which is desired for the study of the receptor’s structure and dynamics in solution conditions.

In this study, we adopted the same methodology reported recently to quantify the affinity of the wsMURs with naltrexone, a potent antagonist of MUR.⁷  The results indicated no significant change in the affinity of wsMUR-TM_v2 relative to wsMUR-TM for naltrexone despite the introduction of seven mutations in proximity to the binding pocket. These findings are consistent with the observations that none of these residues have direct contact with β-FNA in the murine crystal structure of the MUR. In addition, our docking calculations suggest that naltrexone binds in essentially the same binding pocket as the morphian group of β-FNA. These findings corroborate our conclusions from our previous report that demonstrated a close agreement of the naltrexone affinity of the wsMUR-TM with that of the native MUR.

The study presented here demonstrated that available crystal structure allowed us to reliably reduce the number of mutated residues in the water-soluble variant wsMUR-TM_v2 without affecting the protein production in E. coli, the affinity for naltrexone, and the solubility of the protein in the buffer solution relative to its predecessor wsMUR-TM. However, the thermostability study revealed that some of the mutated residues play important roles in the protein stability, which is an important consideration for future mutagenesis studies and for structure–function investigations. Sucrose could be an important alternative molecule to be used to stabilize these water-soluble variants for structural studies. This study not only provides a novel variant of water-soluble MUR as a useful tool for MUR studies but also offers valuable information to engineer water-soluble versions of GPCR, a clinically relevant membrane protein superfamily.

The authors thank the outstanding technical support from Qingchen Meng, Ph.D., and Jingyuan Ma, B.A. (Department of Anesthesiology and Critical Care, University of Pennsylvania, Philadelphia, Pennsylvania); Dr. Liu thanks the mentorship and valuable discussions of Roderic G. Eckenhoff, M.D. (Department of Anesthesiology and Critical Care, University of Pennsylvania), and the support from the Department of Anesthesiology and Critical Care, University of Pennsylvania.

This work was supported by the Foundation for Anesthesia Education and Research, Rochester, Minnesota; National Institutes of Health, Bethesda, Maryland (grant no. K08-GM-093115); the Mary Elizabeth Groff Foundation, Radnor, Pennsylvania; and the Department of Anesthesiology and Critical Care, University of Pennsylvania, Philadelphia, Pennsylvania (to Dr. Liu). This work was also supported by the National Science Foundation, Arlington, Virginia (grant nos. PFI-AIR 13-12202 to Drs. Johnson and Saven, MRSEC DMR-1120901 to Dr. Saven, and NSEC DMR-0425780 to Dr. Johnson).

The authors declare no competing interests.