Global Fold of Human Cannabinoid Type 2 Receptor Probed by Solid-State ¹³C-, ¹⁵N-MAS NMR and Molecular Dynamics Simulations

Abstract

The global fold of human cannabinoid type 2 (CB₂) receptor in the agonist-bound active state in lipid bilayers was investigated by solid-state ¹³C- and ¹⁵N magic-angle spinning (MAS) NMR, in combination with chemical-shift prediction from a structural model of the receptor obtained by microsecond-long molecular dynamics (MD) simulations. Uniformly ¹³C-, and ¹⁵N-labeled CB₂ receptor was expressed in milligram quantities by bacterial fermentation, purified, and functionally reconstituted into liposomes. ¹³C MAS NMR spectra were recorded without sensitivity enhancement for direct comparison of C_α, C_β, and C=O bands of superimposed resonances with predictions from protein structures generated by MD. The experimental NMR spectra matched the calculated spectra reasonably well indicating agreement of the global fold of the protein between experiment and simulations. In particular, the ¹³C chemical shift distribution of C_α resonances was shown to be very sensitive to both the primary amino acid sequence and the secondary structure of CB₂. Thus the shape of the C_α band can be used as an indicator of CB₂ global fold. The prediction from MD simulations indicated that upon receptor activation a rather limited number of amino acid residues, mainly located in the extracellular loop 2 and the second half of intracellular loop 3, change their chemical shifts significantly (≥1.5 ppm for carbons and ≥5.0 ppm for nitrogens). Simulated two-dimensional ¹³C_α(i)-¹³C=O(i) and ¹³C=O(i)-¹⁵NH(i+1) dipolar-interaction correlation spectra provide guidance for selective amino-acid labeling and signal assignment schemes to study the molecular mechanism of activation of CB₂ by solid-state MAS NMR.

Introduction

The cannabinoid CB₂ receptor is a member of the class A (rhodopsin family) of G protein-coupled receptors (GPCR). It is found primarily in tissues of immune and hematopoietic systems like spleen, tonsil, and thymus, where it controls function of immune cells.¹ Evidence has been presented that the CB₂ receptor is also present in brain, which makes the previous classification of CB₂ as peripheral cannabinoid receptor obsolete.² Experimental determination of the receptor structure is critical for elucidating molecular mechanisms of receptor function.

GPCR have a common structural motif of seven transmembrane α-helices which are alternately connected via intra- and extracellular loops. For several GPCR, the structure at atomic resolution was recently unveiled by x-ray crystallography including bovine rhodopsin,^3–5 adenosine A_2A receptor,⁶ β-adrenergic receptors,^7,8 CXCR4 chemokine receptor,⁹ dopamine D3 receptor,¹⁰ muscarinic acetylcholine receptors,^11,12 S1P₁ receptor,¹³ histamine H₁ receptor,¹⁴ nociceptin/orphanin FQ receptor,¹⁵ opioid receptors,^16–18 protease activated receptor,¹⁹ neurotensin receptor,²⁰ serotonin receptors,^21,22 smoothened receptor,²³ and glucagon receptor.²⁴

Crystallographic results provided valuable information, such as the configuration of amino-acid residues composing the ligand binding pocket, the arrangement of transmembrane helices, as well as the existence of amphipathic helix VIII near the intracellular end of helix VII. The recently published structure of a co-crystallized GPCR–G protein complex revealed the sites of interactions with G protein as well as the structural changes in GPCR upon activation.^25,26

The availability of crystallographic data has stimulated interest and highlighted the necessity of studying structure and function of GPCR in their natural environment, the lipid bilayer, without conformational constraints from crystal packing.²⁷ Such experimental conditions are accessible by solid-state NMR spectroscopy. Recently, the chemokine receptor CXCR1 in DMPC bilayers was studied by solid-state NMR.²⁸ The secondary structure was obtained from measured and assigned chemical shifts of the backbone ¹³C- and ¹⁵N resonances of the protein. The global fold of the receptor was obtained by energy minimization using a homology model to rhodopsin with extensive experimental restraints from orientation-dependent ¹H-¹³C_α and ¹H-¹⁵N dipolar couplings.²⁸

A prerequisite for structural studies by NMR is availability of purified, functional protein in the low milligram range. The expression levels of CB₂ receptor as well as of other GPCR in their natural host cells are quite low, except for visual rhodopsin that is highly concentrated in the disc membranes of retinal rod outer segment. Therefore, heterologous expression using mammalian-, insect-, yeast-, or bacterial cells as well as cell-free expression are used to produce the quantities of GPCR required for structural investigations.^29–31 Furthermore, for NMR studies the expression system must be adapted to efficiently incorporate stable ¹³C, ¹⁵N, isotopes into the target protein which requires modification of expression protocols. Isotope labeling of ligands, on the other hand, requires modified schemes of chemical synthesis.^27,32 Although a variety of expression hosts have been adapted to isotopic labeling, the use of a robust and efficient bacterial expression system is still the first choice for flexibility in planning of labeling schemes. E. coli cells grow well in minimal media of a defined composition which simplifies procedures for the incorporation of isotope-labeled precursors into the recombinant receptor.

Previously, we reported results on (i) E. coli expression of CB₂ receptor as a fusion with maltose binding protein (MBP) and removal of expression partners, (ii) chromatographic purification of the receptor at the level of milligrams,^33–35 (iii) stabilization of the receptor in micelles,³⁶ and (iv) detergent-free, homogeneous reconstitution into liposomes while maintaining function.²⁷ Large scale production of functional CB₂ receptor by fermentation of E. coli cells in a mineral salt medium (MSM) was also demonstrated.³⁷

Here we report on expression, purification, and functional reconstitution of agonist CP-55,940-activated, uniformly ¹³C-,¹⁵N-labeled cannabinoid receptor CB₂ and first structural studies by solid-state NMR. We optimized expression conditions to produce uniformly ¹³C-,¹⁵N-labeled CB₂ receptor while minimizing the use of ¹³C-labeled glucose. The latter is essential since yields of recombinant CB₂ per unit of biomass³⁷ are much lower than for soluble proteins or membrane proteins expressed as inclusion bodies. Typical expression protocols would have required quantities of ¹³C-labeled glucose that are cost-prohibitive. Uniform labeling of the receptor with ¹⁵N is easily achieved by supplementation of the MSM with the relatively inexpensive ¹⁵N-ammonium salts.

One-dimensional ¹³C-, and ¹⁵N-MAS NMR spectra of the labeled CB₂ receptor were acquired for a quantitative comparison with simulated spectra derived from the atomic coordinates of MD simulations of the receptor.³⁸ The isotropic chemical shifts of proteins for the backbone ¹³C_α-, ¹³C=O-, ¹⁵NH-, and the side-chain ¹³C_α-resonances reflect sensitively the aminoacid sequence, secondary structure, side-chain conformations, and through-space interactions between amino acid residues.^39–45 Therefore, the chemical shift distribution of the recombinant CB₂ receptor measured by MAS NMR, containing contributions from resonances of all the 400 amino acid residues, is tightly linked to the global fold of the protein. Spin-spin (T₂) relaxation times were measured as well to calculate the relaxation linewidth of resonances. We also acquired two-dimensional ¹³C-¹³C and ¹³C-¹⁵N correlation spectra with the intention to resolve and assign resonances. However, for reasons discussed later, we did not achieve the required sensitivity and resolution that would have enabled assignment. Nevertheless, the measured distribution of chemical shifts, in particular the distribution of the C_α resonances reveals valuable information on the global fold of the recombinant CB₂.

The model of the CB₂ receptor employed here is based on the crystal structure of the Class A GPCR, rhodopsin in the dark-adapted state.⁴ The rhodopsin structure was chosen as the template because no modifications were made to its structure in order to crystallize it. In addition, rhodopsin shares some unusual sequence motifs with the cannabinoid receptors, including the GWXC motif at the extracellular end of transmembrane helix IV that forms an aromatic stacking interaction with Y5.39. This interaction affects the extracellular positions of transmembrane helices III–V and is therefore important for CB₂. We calculated low free energy conformations for any transmembrane helices with an important sequence divergence from rhodopsin. Specifically, in transmembrane helix II there is a pair of glycines that distort the helix while CB₂ has no equivalent distortion; similarly, residue 5.50 is a proline in rhodopsin but not in CB₂. The resultant model has been tested by substituted cysteine accessibility studies to identify residues that face into the CB₂ binding pocket,^46,47 by mutation to identify key ligand interactions sites,^46–49 and by covalent labeling of CB₂⁵⁰ to identify the method of ligand entry into the CB₂ binding pocket.

The CB₂ receptor model was then inserted into a lipid bilayer composed of 123 molecules of 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphocholine (POPC) and 38 molecules of the endogenous ligand 2-arachidonyl-glycerol (2-AG).³⁸ A spontaneous binding of a 2-AG molecule from the lipid matrix to the ligand binding site observed in the simulations yielded the structure of agonist-activated CB₂ receptor that was compared with experimental results for agonist CP-55,940-bound receptor. The structure of CB₂ before 2-AG binding was defined as the inactive state of the receptor. Spectral prediction was performed using the programs SHIFTX ⁵¹ and SPARTA,⁵² which calculate the chemical shift values semi-empirically based on databases of previously reported protein structures with resonance assignments. The sensitivity of spectra to changes in the primary amino acid sequence and to secondary structure is demonstrated.

The experiments and spectral predictions are useful for developing strategies for specific labeling of amino acids that are likely to yield well-resolved and assigned resonances in multidimensional NMR experiments. The CB₂ model was also used to show that selective amino acid labeling schemes can be used for elucidation of mechanisms of receptor activation.

Materials and Methods

Information on chemicals, the protein expression protocol including composition of media, protein purification and reconstitution, and molecular dynamics calculations are provided in Supplemental Materials.

¹³C MAS NMR

The proteoliposomes containing uniformly ¹³C-, ¹⁵N-labeled CB₂ receptor were pelleted by ultracentrifugation at 417,200×g for 12 h at 4°C. The pellet was transferred to a 4-mm MAS rotor with a 50-μL sample volume sealed by a Kel-F insert (Bruker BioSpin, Inc., Billerica, MA). The ¹³C MAS NMR spectra of proteoliposomes were recorded at a ¹³C resonance frequency of 201.21 MHz on an AV800 spectrometer equipped with a 4-mm ¹H/¹³C/¹⁵N MAS variable-temperature probehead (Bruker BioSpin, Inc.). Measurements were carried out at MAS frequencies of 10 and 15 kHz. The sample temperature was varied from 9 to 26 °C. Spectra were acquired without sensitivity enhancement by cross-polarization or the nuclear Overhauser effect using a 5.0-μs 90° pulse on ¹³C and a ¹H decoupling power of (γ/2π)B₁= 56.7 kHz in a spinal-64 sequence applied during ¹³C detection. Typically 24,000 scans with 1024 data points were acquired at a dwell time of 8.3 μs and a 2-s delay time between scans. The free-induction decay signals were processed with an exponential window function corresponding to a line broadening of 50 Hz upon Fourier transformation. For subtraction of the natural abundance ¹³C resonances of lipids, spectra with a transverse-relaxation (T₂) filter (Hahn echo) were acquired. The echo delay time was varied from 0.5 to 5.0 ms. A delay time of 3.0 ms resulted in an almost complete decay of protein resonances with only modest attenuation of lipid resonances that decay much slower. After intensity adjustment to account for the relaxation decay, this lipid spectrum was subtracted from the spectrum recorded without the relaxation filter and yielded a ¹³C spectrum of CB₂ receptor almost free from contributions of lipid resonances. The ¹³C-chemical shift scale was calibrated in reference to DSS (sodium 2,2-dimethyl-2-silapentane-5-sulfonate) by setting the choline γ signal of PC to 57.6 ppm.^53,54

¹⁵N MAS NMR

A ¹H-¹⁵N cross-polarization (CP) -MAS NMR spectrum of the CB₂ proteoliposomes was recorded on an AV800 spectrometer at a ¹⁵N resonance frequency of 81.08 MHz, 10 kHz MAS, 18°C, 3.7-μs ¹H 90° pulse, 1-ms ¹H-¹⁵N cross-polarization at (γ/2π)B₁=50 kHz with a 100-80% linear ramp on the ¹H channel, (γ/2π)B₁= 60 kHz ¹H decoupling during ¹⁵N detection using a spinal-64 sequence, 30,000 scans, 1024 data points at a dwell time of 5 μs, and 2-s delay time between scans. The ¹⁵N-chemical shift scale was calibrated by setting the midpoint of the ¹⁵N-amide resonance to 120 ppm.

¹³C-, and ¹⁵N chemical-shift prediction

The atomic coordinates of the CB₂ receptor were obtained from microsecond-long molecular dynamics simulations of CB₂ receptor in a lipid matrix as explained in the Introduction section.³⁸ An abbreviated description of simulation procedures is provided in Supplemental Materials. The ¹³C-, and ¹⁵N-chemical-shifts of C_α-, C_β-, C=O-, and NH-resonances of the CB₂ receptor were predicted based on the atomic coordinates of the receptor obtained in MD simulations using the shift-prediction programs SHIFTX ⁵¹ and SPARTA.⁵² The programs yield the shift values referenced to 4,4-dimethyl-4-silapentane-1-sulfonic acid (DSS). The prediction was performed on the ligand-free inactive and agonist-bound activated states of CB₂ receptor from the simulations.³⁸ The chemical-shift value at each atomic site was calculated as the average from 5 or 6 time points taken at intervals of 0.9 or 1 ns of the simulations. The predicted chemical shift values are expected to be accurate within standard deviations of 1.4 ppm for ¹³C- and 3 ppm for ¹⁵N-resonances when applied to membrane proteins.⁵⁵ The ¹³C spectra were simulated using the predicted C_α, C_β, and C=O chemical shifts and assuming a linewidth of resonances of 100 Hz. In the chemical-shift prediction, the eleven histidines in wild-type CB₂ receptor were treated as deprotonated as required by programs for chemical shift prediction. For recombinant CB₂ receptor ³⁵ (see Fig. S4 for amino acid sequence), the 12 histidines in the N- and C termini were also treated as deprotonated, and all amino acids in the termini were assumed to be in random coil conformation. The number of histidines is modest compared to the total number of residues (11 out of 360 total in the wild type sequence, and 23 out of 400 in recombinant CB₂-130). A comparison of simulated spectra for wild-type CB₂ and recombinant CB₂-130 suggested that the influence of histidines and their state of protonation on the shape of ¹³C resonance bands is insignificant (see Fig. S3).

Results and Discussion

Expression, purification, and reconstitution of ¹³C-, ¹⁵N-CB₂ receptor

Fermentation of E. coli BL21-21(DE3) harboring plasmid pAY130 for expression of ¹³C- and ¹⁵N-labeled recombinant CB₂ receptor ³⁵ was performed in MSM containing ¹³C-labeled glucose as the sole carbon source and ¹⁵NH₄Cl as the sole nitrogen source, and produced high density biomass that yielded ~6 mg of uniformly ¹³C-, ¹⁵N-labeled, MBP-CB₂ fusion protein in the functional state. The fusion protein was solubilized from bacterial plasma membranes in a mixture of detergents CHAPS, n-Dodecyl-β-D-maltopyranoside (DDM), and cholesteryl hemisuccinate (CHS) with addition of cannabinoid agonist CP-55,940.³⁶ Upon purification of the fusion protein on a Ni-NTA column, the MBP expression partner was removed by treatment with TEV protease. CB₂ was further purified by affinity chromatography on a StrepTactin resin as described previously.³⁵

Reconstitution of the purified, uniformly ¹³C-,¹⁵N-labeled CB₂ receptor (2.8 mg) with phospholipids (28 mg; POPC/POPS=4/1, mol/mol) by the rapid-dilution method yields detergent free, homogeneous proteoliposomes.²⁷ The recovery of protein and lipids in the proteoliposome preparation was high: 2.6 mg (92% yield) of the ¹³C-, ¹⁵N-labeled CB₂ and 33.3 mg of lipids (nearly 100% yield). In addition to the POPC and POPS, the lipid matrix also contained the CHS that was added to micellar solutions of CB₂ to prevent denaturation of the receptor and was almost quantitatively incorporated into proteoliposomes upon detergent removal. The molar ratio of protein and lipids in proteoliposomes was CB₂/POPC/POPS/CHS, 1/505/120/150.²⁷ Ligand-binding experiments by solid-state ²H MAS NMR using selectively ²H-labeled CP-55,940-d₆ ³² revealed that >90% of reconstituted CB₂ receptor is ligand-binding competent.²⁷ The agonist-bound ¹³C-, ¹⁵N-CB₂ receptor activated G proteins accordingly in an in vitro coupled assay.^27,33,35

¹³C MAS NMR

The NMR experiments were conducted on proteoliposomes in 50 wt.% PBS buffer containing 2.6 mg of uniformly ¹³C-, ¹⁵N-labeled CB₂ receptor as described above, activated with agonist CP-55,940. Detection of ¹³C-resonances was performed without sensitivity enhancement such as cross-polarization, INEPT, or the nuclear Overhauser effect to ensure that intensities of all ¹³C-resonances of the protein are identical to enable a quantitative comparison of predicted spectra from molecular simulations and experimental results. The relaxation delay between acquisitions was sufficiently long to avoid intensity losses due to saturation.

The measured ¹³C-spectrum is presented in Fig. 1a. The spectrum shows bands of superimposed resonances; each band represents different carbon sites ^{39,42,43,54,56} of the CB₂ receptor: side-chain aliphatic, 10–45 ppm; C_α, 45–70 ppm; side-chain aromatic, 110–140 ppm; Arg C_ε, 160 ppm; C=O, 175–180 ppm. The bands are accompanied by MAS side bands of much lower intensity located at multiples (n=1,2,…) of the MAS frequency from the center bands. In Fig. S1, the locations of the side bands recorded at MAS frequencies of 10 and 15 kHz are shown. At 15 kHz, the C_α and C=O resonances which are highly sensitive to the global fold of the backbone, can be analyzed without interference from side bands of other carbon sites. The well-resolved signals on top of the resonance bands of CB₂ in Fig. 1a are natural-abundance ¹³C resonances of lipids in the fluid phase. The lipid resonances have much longer transverse (T₂) relaxation times than the protein resonances. Therefore, a spectrum showing only lipid resonances can be obtained by application of a T₂-filter that effectively suppresses protein resonances (Fig. 1b). The experiments also yielded the T₂ relaxation times of lipids which permitted correction of their intensities for relaxation decay. The corrected lipid spectrum was then subtracted from the combined spectrum of protein and lipids (Fig. 1a) to yield the ¹³C spectrum of pure CB₂ receptor (Fig. 1c). Experiments were conducted in the temperature range of 9–26°C, above the gel-fluid phase transition of the lipid matrix. No changes in spectral appearance as a function of temperature were observed (Fig. S2).

Figure 1.

(a) ¹³C MAS NMR spectrum of the agonist-bound active state of the ¹³C-,¹⁵N-CB₂ receptor in liposomes. (b) The lipid spectrum observed by application of a T₂ filter to remove the protein signals, was subtracted from the total spectrum (CB₂ receptor + lipids) to yield (c) the spectrum of CB₂ receptor. The measurements were conducted at 9°C and a MAS frequency of 15 kHz. The delay time in the Hahn echo sequence was set to 3.0 ms. The molar ratios of the protein and lipids in the proteoliposomes are CB2/POPC/POPS/CHS, 1/505/120/150.

The T₂ relaxation experiments indicated that the C_α and C=O bands have relaxation times of 2.8 and 4.2 ms, respectively (Fig. 2), equivalent to a resonance linewidth at half height of 0.55 and 0.40 ppm. The actual linewidth could be somewhat larger due to inhomogeneous line broadening. The intrinsic resolution on the order of 0.5 ppm is encouraging for designing schemes of selective-isotope labeling of CB₂ receptor that may yield resolved resonances with their assignments; see discussion below. The large number of resonances of the 44-kDa recombinant, uniformly ¹³C-, ¹⁵N-labeled protein in combination with the limited resolution resulted in heavy signal superposition. Nevertheless, the shape of the band of superimposed resonances reports important information on receptor fold as will be presented below. This applies in particular to the band of C_α resonances which has a rather wide distribution of chemical shifts.

Figure 2.

Spin-spin relaxation, T₂, of the ¹³C-resonance bands of the C_α and C=O of CB₂ receptor in liposomes.

¹⁵N MAS NMR

The ¹⁵N MAS NMR spectrum of ¹³C-, ¹⁵N-CB₂ receptor recorded with sensitivity enhancement by ¹H-¹⁵N cross-polarization is shown in Fig. 3 with signal assignments. Note that detection of the ¹⁵N resonances (sensitivity ∝ γ³, where γ is the gyromagnetic ratio) at the protein concentration of our experiments requires sensitivity enhancement by ¹H-¹⁵N cross-polarization. The spectrum is largely dominated by the band of backbone amide resonances centered at 120 ppm. Superimposed on the backbone amide resonances are side-chain ¹⁵N resonances of tryptophan indole, asparagine and glutamine amides, as well as backbone ¹⁵N resonances of the proline ring. The side-chain resonances of arginines and lysines upfield from the amide band were well-resolved, and the side-chain resonances of histidines appear downfield. The backbone amide band plays a key role for ¹³C resonance assignments when site-selectively isotope-labeled CB₂ receptor is investigated.

Figure 3.

¹⁵N MAS NMR spectrum of uniformly ¹³C-, and ¹⁵N-labeled CB₂ in liposomes recorded with ¹H-¹⁵N cross-polarization at a MAS frequency of 10 kHz.

Multidimensional NMR experiments

2D CP-DARR ¹³C-¹³C correlation experiments ⁵⁷ were acquired with mixing times from 50- to 300-ms. Furthermore, we conducted 2D ¹⁵N-¹³C correlation experiments⁵⁸ yielding a selective transfer of the ¹⁵N magnetization to directly bonded ¹³C nuclei (results not shown). Both, the ¹³C-¹³C- and the ¹⁵N-¹³C correlation spectra had insufficient resolution to resolve individual ¹³C- and ¹⁵N resonances. There was insufficient sensitivity to attempt 3D experiments as reported by Park et al. for CXCR1.²⁸ Several reasons may account for our lower sensitivity: (i) It appears that resolution of ¹³C and ¹⁵N resonances of CB₂ is lower which not only results in more signal superposition but also in lower sensitivity, and (ii) our lipid/protein molar ratio is higher by about a factor of three which resulted in a larger sample size for equal amounts of protein. The CB₂/lipid molar ratio in bilayers is limited by stringent requirements for the presence of a proper concentration of CHS in the micellar phase to protect the receptor from denaturation.³⁶ The CHS from the micelles ends up quantitatively in the lipid matrix which requires addition of a higher amount of phospholipids not to exceed solubility limits for CHS in proteoliposomes.²⁷ The use of a 3.2-mm, low-E MAS probe with a sample volume of 36 μL ²⁸ compared to a 4-mm MAS probe with twice the sample volume in our experiments has probably resulted in higher sensitivity for the CXCR1 experiment as well. Last but not least, functional studies on reconstituted CB₂ determined that the receptor at ambient temperature and higher denatures at a rate that prevents long-lasting three-dimensional NMR experiments.³⁶

Comparison to simulated spectra from a molecular model

The measured one-dimensional ¹³C MAS NMR spectra of the agonist-bound CB₂ receptor in liposomes were compared to predicted spectra from MD simulations of CB₂ in the inactive and agonist activated states. In the course of the simulations, a spontaneous binding event of 2-AG to CB₂ was observed. The ligand entered the binding pocket located on the extracellular side of the receptor through a gap between transmembrane helices VI and VII. Ligand binding resulted in breaking of the ionic lock between arginine R3.55 and aspartic acid D6.30 located near the ends of transmembrane helices III and VI on the intracellular side of the receptor. This was followed by a side-chain flip at W6.48 that may correspond to the so-called rotamer toggle switch. Hence, the MD simulations report on atomic coordinates of the protein structure related to the critical steps of receptor activation from ligand-free inactive to agonist-bound active states. Atomic coordinates of the receptor were extracted from before 2-AG binding, after binding and breakage of the ionic lock, as well as before and after activation of the toggle switch. Each period in the evolution of the CB₂ structure was probed using five to six coordinate sets taken at 1-ns or 0.9-ns intervals to account for rapid structural fluctuations. This allowed us to predict NMR spectral changes related to the full activation process, but also for ligand binding and breakage of the ionic lock, and for activation of the toggle switch selectively.

Chemical shifts of the backbone ¹³C_α-, ¹³C=O-, ¹⁵NH-, and the side-chain ¹³C_β resonances were determined from the pdb files using the programs SHIFTX and SPARTA and averaged for each time period. Differences in spectral predictions between SHIFTX and SPARTA were insignificant (Fig. S3). The comparison of measured and predicted spectra is shown in Fig. 4. The predicted distribution of ¹³C resonances of the fully activated state of CB₂ (black spectrum) is to be compared with the experimentally recorded ¹³C MAS NMR spectrum in blue. It was assumed that the simulated structure of CB₂ receptor activated by 2-AG is similar to the structure of CB₂ activated by CP-55,940 in the experiment. The predicted spectrum of ligand-free, inactive CB₂ is shown as well (grey spectrum). The green spectrum shows the predicted distribution of the C_β resonances (SHIFTX and SPARTA predict only C_α, C=O, and the side-chain C_β resonances). The predicted distribution of the C_α resonances from the simulation is in decent qualitative agreement with the experimental spectrum: the experimentally measured C_α band has its highest intensity at 58.1 ppm in agreement with the predicted value of 58.0 ppm. The downfield edge of the C_α band contains also the C_β resonances of serines at 63 ppm, and threonines at 69 ppm. Their resonances are shifted downfield due to the directly bonded hydroxyl group. Inclusion of those resonances is shown to be critical for predicting the shape of the C_α band. Some differences between prediction and measurement near the upfield edge of the C_α band near 50 ppm may stem from contributions of side-chain aliphatic carbons other than the predicted C_β resonances. Thus the distribution of resonances in the C_α band can be used as an indicator of CB₂ global fold. Several more examples that indicate sensitivity of chemical shifts of the C_α region to molecular structure are shown in Supplemental Fig. S5.

Figure 4.

Comparison of the measured ¹³C MAS NMR spectrum of the CB₂ receptor (blue) with a predicted spectrum from the MD simulations (black). The receptor is in the agonist-bound active state. The prediction was made for the backbone C_α, C=O, and side-chain C_β carbons by application of the chemical-shift prediction program SHIFTX. The predicted C_β signals were shown explicitly to specify signals from Ser (63 ppm) and Thr (69 ppm) that appear in the C_α region due to a directly bonded hydroxyl group (green). The predicted spectrum of the ligand-free inactive state of the receptor is also presented (grey) as well as the C_β signals (light green).

Both the predicted and measured band of C=O resonances are centered at 177 ppm and have a comparable width. However, the measured C=O band includes resonances from side-chain carbonyl/carboxyl groups of asparagine (Asn), aspartic acid (Asp), glutamine (Gln), and glutamic acid (Glu) that are absent in the predicted spectrum. The standard ¹³C chemical-shifts of side-chain carbonyl resonances (in DSS-referenced scale) measured on GGXA peptides, where X is N, D, Q, or E, are: Asn (178.3 ppm), Asp (181.2 ppm), Gln (181.7 ppm), and Glu (185.5 ppm).⁵⁶ Therefore, the visible shoulders in the measured C=O band at 181 and 185 ppm are likely to have significant contributions from those polar side chains.

The predicted chemical shift distribution indicated small but notable differences between inactive and active states of CB₂ receptor, particularly in the distribution of intensities in the band of C_α resonances. The band intensity at 58.0 ppm decreases somewhat upon activation and an increase of intensity is predicted near 61 ppm. However, those differences would be difficult to detect in experiments using uniformly labeled CB₂ receptor, considering that the deviations between predicted and measured spectra are of similar magnitude. The predicted shape of the C=O band of resonances is almost identical for the inactive and active states except for some minor increase in intensity near 178 ppm upon activation.

To scrutinize how the intensity distribution of ¹³C chemical shifts reflects secondary structure of the protein, we separately calculated contributions to the spectrum from (i) helices, (ii) loops, and (iii) N- and C-termini for C_α chemical shifts (Fig. 5). The helices include the transmembrane helices I–VII and the intracellular amphipathic helix VIII. The loops are rich in turns and somewhat ordered, while the N- and C-terminal domains are rich in random coil structure. It is obvious that the characteristic patterns of the experimental spectrum are determined by resonances of the helices of CB₂ that yield the characteristic maximum at 58 ppm. The upfield region of the C_α band, to the right from the maximum contains significant contributions from resonances of turns and disordered segments in the loops and the N-, and C-termini. Indeed, C_α resonances shift upfield upon a conformational transition from α-helix to random coil.^39,42,43

Figure 5.

The C_α region of ¹³C NMR spectra simulated for helices, N-, and C-termini, and loops of CB₂ compared to the full experimentally measured spectrum. The prediction was made by the program SHIFTX using the structure of activated CB₂ obtained in MD simulations.

It is concluded that the experimentally measured distribution of the ¹³C-resonance intensities in the C_α region (45–70 ppm) sensitively reports the global fold of CB₂ in liposomes. The remaining deviations between the measured and predicted ¹³C NMR spectra may have several causes: (i) The accuracy of prediction algorithms for chemical shifts of proteins has very much improved but is finite especially for comparison with chemical shifts measured by solid-state NMR.⁵⁵ (ii) The simulation of CB₂ receptor was performed on the timescale of microseconds which is unlikely to be adequate for exploration of the entire conformational space of a GPCR, although it appears that the simulations report on the most critical steps of the activation process. Furthermore, even the current simulations revealed significant motional degrees of freedom, in particular in the region of loops and the N- and C-termini. It may be necessary to conduct more extensive conformational averaging of chemical shifts using a representative set of conformations to achieve an even better agreement between experiment and simulations.³⁸ (iii) The CB₂ receptor in the experiment is activated by the full agonist CP-55,940 while activation in the simulation occurred by the endogenous partial agonist 2-AG. (iv) While the estimated purity of the receptor is 90% and the fraction of functionally reconstituted CB₂ receptor (>90%) ²⁷ is rather high, the presence of some unwanted resonances may not be excluded. (v) The subtraction of natural-abundance ¹³C-resonances of lipids from the spectra may not be perfect. With those limitations in mind, the agreement between measured and predicted ¹³C NMR spectra of CB₂ receptor in a membrane is quite reasonable.

Changes in chemical shifts upon activation of CB₂ receptor

The availability of MD simulations of CB₂ receptor showing the effects of ligand binding, breakage of the ionic lock and activation of the rotamer toggle switch enabled us to predict which of the amino-acid residues will experience significant changes of their ¹³C-, and ¹⁵N-chemical shifts upon each of those events. Predicted changes of C_α-chemical shifts upon full activation covering several microseconds of simulation time are plotted in Fig. 6 (bottom panel). The standard deviation of the chemical shift values due to fast motional fluctuations during a sampling time of about 5 ns for each of the inactive and active states was estimated to be ±0.5 ppm on average over all C_α sites (±0.7 ppm for the relatively flexible loops and terminals, and ±0.4 ppm for the helices). Upon activation of the receptor significant changes of C_α-chemical shifts, larger than ±1.5 ppm, were observed only for a limited number of C_α sites that are located primarily in the extracellular loop 2 (ECL2), the second half of intracellular loop 3 (ICL3), as well as in the first half of N- and C-terminal domains (top panel of Fig. 6). A similar trend was observed for C=O-, C_β-, and NH-chemical shifts (Figs. S6–8).

Figure 6.

Predicted ¹³C_α chemical-shift changes (Δδ) upon activation of CB₂ (light green, helices; green, terminals and loops). Amino-acid residues with changes Δδ greater than ±1.5 ppm are marked in the snake plot: 2.0 ppm ≥| Δδ |≥1.5 ppm, light blue; | Δδ |≥2.0 ppm, dark blue. The prediction was made by SHIFTX using structures generated by MD simulations.

In Figs. 7 and 8 the chemical shifts changes for selected regions of CB₂ covering the full activation process (Figs. 7a,8a), the binding of 2-AG and the consecutive breakage of the ionic lock (Figs. 7b,8b), and the activation of the rotamer toggle switch (Figs. 7c,8c) are shown. It is obvious that the critical structural events that are reflected in the chemical shift changes are the ligand binding event and the resulting breakage of the ionic lock, while activation of the toggle switch resulted in minor changes only. It is important to point out that Figs. 7a and 8a cover states of the receptor separated by microseconds of simulation time. The similarities between Figs. 7a and 8a and Figs. 7b and 8b suggest that the biggest structural changes occurred quickly after the ligand binding event and that structural fluctuations of the receptor have a secondary influence on chemical shifts.

Figure 7.

Predicted C_α chemical-shift changes (Δδ/ppm) of amino-acid residues in the extracellular loop 2 (ECL2) (a) upon activation of CB₂ receptor.³⁸ Amino-acid residues with changes Δδ greater than ±1.5 ppm are marked in the snake plot: 2.0 ppm ≥|Δδ |≥1.5 ppm, light blue; |Δδ |≥2.0 ppm, dark blue. Values in parentheses indicate the standard deviation of the chemical shift at each amino-acid residue due to motional fluctuations during the 5-ns periods from inactive- (first value) and active (second value) states of the receptor, respectively. (b) Predicted changes (Δδ/ppm) due to 2-AG binding to the receptor pocket and the following breakage of the ionic lock between helices III and VI during the activation process.³⁸ (c) Predicted changes (Δδ/ppm) due to activation of the rotamer toggle switch at W6.48 that occurred after the breakage of the ionic lock.³⁸ Compared to Δδ from breakage of the ionic lock, Δδ from activation of the toggle switch are minor despite the relatively long time monitored (600 ns).

Figure 8.

Predicted C_α chemical-shift changes (Δδ/ppm) of amino-acid residues in the intracellular loop 3 (ICL3) and in the first half of the C-terminal domain (a) upon activation of CB₂ receptor.³⁸ Δδ for (b) 2-AG binding breakage of the ionic lock, and (c) activation of the rotamer toggle switch as defined in Fig. 7.

ECL2 is known to undergo structural changes upon receptor activation; in rhodopsin ECL2 acts like an opening lid on the retinal-binding site upon photoactivation,⁵⁹ and mutational studies on the dopamine D₂ receptor ⁶⁰ and the M₃ muscarininc acetylcholine receptor ⁶¹ confirmed involvement of ECL2 in ligand binding which explains the predicted changes of ECL2 chemical shifts of CB₂ upon ligand binding (see Fig. 7).

ICL3 constitutes a critical site for binding and activation of heterotrimeric G proteins.²⁶ In the CB₂ simulation, ligand binding was accompanied by breaking of the ionic lock between transmembrane helices III and VI on the cytoplasmic side of the receptor. In turn, transmembrane helix VI bends away from helix III resulting in a large movement of ICL3 which opens up the site for interaction with G protein.³⁸ The structural changes correlate with predicted changes of chemical shifts in ICL3 (see Fig. 8). Mutational and structural studies show that the C-terminal domain, in particular the region near helix VIII, is involved in G protein-activation as well as interactions with other signaling proteins ⁶² explaining predicted changes of chemical shift in the C-terminal region. The limited number of residues undergoing notable changes of chemical shift offers themselves as suitable targets for selective amino-acid labeling to study the mechanism of receptor activation by solid-state NMR.

Selection of NMR resonances according to their expected change of chemical shifts upon receptor activation must not be the sole consideration when planning selective-labeling schemes. It is also of importance to consider what kind of amino-acid labeling would result in the least superposition of resonances to probe a molecular event of interest.^59,63,64 The aromatic amino acid tryptophan (Trp), for example, can be an attractive target for investigations because of its relatively low occurrence (eight) in the CB₂ receptor and its critical role in function such as toggle switch in a ligand-binding pocket.³⁸ Inclusion of ¹⁹F-tryptophan by recombinant expression may also be utilized as a powerful NMR probe.⁶⁵ Fig. 9 shows the predicted two-dimensional ¹³C_α(n)-¹³C=O(n) and ¹³C=O(n)-¹⁵NH(n+1) correlation spectra for a specifically ¹³C-Trp-labeled CB₂ receptor with uniform ¹⁵N-labeling, corresponding to the inactive and active states (Figs. 9a and b, respectively). The better resolution of resonances in the ¹³C=O(n)-¹⁵NH(n+1) correlation spectrum is particularly useful for signal assignment. The presence of Trp-residues around the ligand binding site is also useful for locating an isotope-labeled ligand in the protein. The ¹³C-CP-MAS NMR of ¹³C-labeled CP-55,940 (Fig. 9c) in liposomes (POPC/POPS/CHS (60/15/25, mol/mol/mol)) yielded a well-resolved signal of the branched methyl groups at 28.9 ppm (Fig. 9d). Considering that a method for efficient, selective Trp-labeling of CB₂ receptor has been established,³⁷ and that Trp is likely to be involved in ligand-binding and receptor-activation,³⁸ investigations on the ligand binding pocket by solid-state NMR using isotope-labeled ligands and receptor seem feasible. The NMR experiments on uniformly ¹³C-, ¹⁵N- labeled CB₂ receptor and the spectra predicted from molecular dynamics simulations on receptor activation suggest that mechanisms of receptor activation can be studied by following a limited number of resonances from selectively labeled amino acids.

Figure 9.

Two-dimensional ¹³C_α-¹³C=O dipolar interaction correlation plot for Trp residues of CB₂ receptor predicted for the (a) ligand-free inactive and (b) agonist-bound active states, along with ¹³C=O(n)-¹⁵NH(n+1) dipolar interaction correlation plot for Trp(n) and the next amino-acid residues (n+1). The spectra were predicted by SHIFTX using structures generated by MD simulations. The circles at each cross peak indicate spectral ranges corresponding to the diameters of 0.5, 1.0, and 1.5 ppm. (c) ¹³C-labeled cannabinoid agonist CP-55,940 at the branched methyl groups in the nonyl tail. (d) ¹³C-CP-MAS spectra of the ¹³C-labeled CP-55,940 in liposomes (POPC/POPS/CHS (60/15/25, mol/mol/mol) (blue), and of a liposome sample without ligand (black). The measurements were performed at 17°C and a MAS frequency of 15 kHz. The sample was prepared at a ligand/lipid ratio of 1/10 (w/w). The peak marked with an asterisk is the ligand signal. Assignments for major lipid resonances are shown.⁶⁶