NMR sample optimization and backbone assignment of a stabilized neurotensin receptor

Abstract

G protein-coupled receptors (GPCRs) are involved in a multitude of cellular signaling cascades and consequently are a prominent target for pharmaceutical drugs. In the past decades, a growing number of high-resolution structures of GPCRs has been solved, providing unprecedented insights into their mode of action. However, knowledge on the dynamical nature of GPCRs is equally important for a better functional understanding, which can be obtained by NMR spectroscopy. Here, we employed a combination of size exclusion chromatography, thermal stability measurements and 2D-NMR experiments for the NMR sample optimization of the stabilized neurotensin receptor type 1 (NTR1) variant HTGH4 bound to the agonist neurotensin. We identified the short-chain lipid di-heptanoyl-glycero-phosphocholine (DH₇PC) as a promising membrane mimetic for high resolution NMR experiments and obtained a partial NMR backbone resonance assignment. However, internal membrane-incorporated parts of the protein were not visible due to lacking amide proton back-exchange. Nevertheless, NMR and hydrogen deuterium exchange (HDX) mass spectrometry experiments could be used to probe structural changes at the orthosteric ligand binding site in the agonist and antagonist bound states. To enhance amide proton exchange we partially unfolded HTGH4 and observed additional NMR signals in the transmembrane region. However, this procedure led to a higher sample heterogeneity, suggesting that other strategies need to be applied to obtain high-quality NMR spectra of the entire protein. In summary, the herein reported NMR characterization is an essential step toward a more complete resonance assignment of NTR1 and for probing its structural and dynamical features in different functional states.

Graphical Abstract

Introduction

G protein-coupled receptors (GPCRs) are one of the largest groups of membrane proteins encoded by the human genome. These essential membrane proteins are involved in a myriad of different cellular and (patho-) physiological processes (Ali et al., 2020; Pierce et al., 2002; Venkatakrishnan et al., 2019). Signal transduction by GPCRs is facilitated by binding of various ligands to the extracellular orthosteric site (Rosenbaum et al., 2009) of the receptor followed by the interaction with their intracellular partner proteins, such as G proteins (Mahoney and Sunahara, 2016) or ß-arrestins (Luttrell and Lefkowitz, 2002). The involvement of GPCRs in diverse signaling processes makes them attractive and promising targets for pharmaceutical drug development (Shimada et al., 2019). In the last two decades, X-ray crystallography (Rasmussen et al., 2011) and cryo-EM (Draper-Joyce et al., 2018) provided atomic-resolution structures of GPCRs in complex with G proteins and ß-arrestins (Huang et al., 2020). Despite these insights, it has been noted that static structures cannot fully explain the activation pathway of these highly dynamic proteins (Hilger, 2021; Weis and Kobilka, 2018).

GPCRs are seven transmembrane helical proteins with a high intrinsic conformational flexibility, enabling the population of inactive and active states (Draper-Joyce and Furness, 2019; Weis and Kobilka, 2018). Nuclear magnetic resonance (NMR) spectroscopy proved to be a versatile tool to decipher GPCR conformational states and dynamics in solution (Ueda et al., 2019) using amino-acid selective (Eddy et al., 2018; Isogai et al., 2016) and fluorine labeling (Liu et al., 2012), or the introduction of methyl probes by chemical modification of lysine (Sounier et al., 2015) or cysteine residues (Goba et al., 2021). Nevertheless, a high-resolution NMR study with a uniformly isotope labeled GPCR has not been reported so far. Production of isotope labeled receptors in different expression hosts (Abiko et al., 2021; Berger et al., 2011; Clark et al., 2017; Werner et al., 2008; Xu et al., 2019) and in a cell-free setup (Shilling et al., 2017) has been achieved and optimized over the past years. In addition, directed evolution strategies made it possible to enhance the expression levels of GPCRs in E. coli (Dodevski and Pluckthun, 2011; Sarkar et al., 2008; Scott et al., 2014) and enabled the production of these challenging membrane proteins for NMR studies (Schuster et al., 2020). Compared to eukaryotic expression hosts, such as insect cells or mammalian cells, E. coli has the major advantage of tolerating high levels of deuterium in the growth medium. The stabilized neurotensin receptor subtype 1 (NTR1) variant HTGH4 can be produced in E. coli at a yield of 2–4 mg/l cell culture and exhibits high thermal stability even in detergent solution (Scott et al., 2014). Crystal structures of HTGH4 and other stabilized NTR1 variants obtained by directed evolution in E. coli (Egloff et al., 2014) bound to the agonist peptide neurotensin-1 (NTS-1) are almost identical to the structure of another stabilized neurotensin receptor obtained from insect cells (White et al., 2012). Structures are known for NTR1 in the active and inactive states, as well as in the G protein bound form (Deluigi et al., 2021; Kato et al., 2019; Krumm et al., 2016; Zhang et al., 2021). Even with the availability of abundant structural information of the inactive and active states, the mechanistic details of the activation pathway of such a peptide receptor by high-resolution NMR methods remain only poorly investigated.

Here, we optimized the sample conditions for the stabilized NTR1 variant HTGH4 and obtained high-resolution NMR spectra of sufficient quality for the collection of multidimensional NMR experiments for sequence-specific NMR backbone resonance assignments. Furthermore, we used NMR chemical shift perturbation and hydrogen deuterium (HDX) mass spectrometry experiments to probe structural changes in the orthosteric binding site of HTGH4 bound to the agonist neurotensin or the antagonist SR142948 (Gully et al., 1997). In addition, we could show that large parts of the transmembrane region of HTGH4 are strongly protected from the solvent, prohibiting the observation of NMR backbone resonances in this region if fully deuterated protein is used. To address this issue, we use partial HTGH4 unfolding by chemical denaturants to enhance amide proton exchange in transmembrane helices and obtain a larger number of amide resonances for further NMR experiments. However, this procedure also led to enhanced sample heterogeneity, suggesting that isotope labeling strategies that omit backbone amide deuteration will be required for further NMR studies of the entire protein. Taken together, these data show that high-resolution NMR studies of GPCRs are facilitated by directed evolution, enabling the investigation of GPCR allostery at atomic resolution.

Material and methods

Protein expression and purification

Rat NTR1 variant HTGH4 was expressed and purified as described previously (Egloff et al., 2015; Egloff et al., 2014). For the production of isotope-labeled protein for 2D-[¹⁵N,¹H]-TROSY NMR experiments, NEB Express I^q E. coli (New England Biolabs) cells were grown in isotope enriched M9 media containing 99 % D₂O, 1 g/L ¹⁵N ammonium chloride and 2 g/L glucose, whereas 2 g/L of ²H ¹³C glucose was used for 3D triple resonance experiments. After induction with 1 mM IPTG, cells were grown at 20 °C for additional 20 hours. Cells were harvested by centrifugation at 4 °C and 6,000 × g. The supernatant was discarded, and cell pellets were stored at −80 °C.

Receptor purification was performed as described previously (Egloff et al., 2015; Egloff et al., 2014), utilizing neurotensin affinity, cation exchange and size exclusion chromatography. Cell pellets were lysed and solubilized in 2x solubilization buffer containing 100 mM HEPES pH 8.0, 20 % glycerol (w/v), 400 mM NaCl, 5 mM MgCl₂, 100 mg lysozyme, one tablet of Complete Protease Inhibitor Cocktail (Roche) and DNaseI (50 μg/ml). For solubilization 0.6/0.12 % CHAPS/CHS and 1.7 % DM were added in a final concentration to the lysate while stirring at 4 °C for 30 min. The lysate was sonicated on ice and then centrifuged at 50,000 × g for 30 min. The supernatant was mixed with the affinity resin pD-NT either wild type (for agonist-bound HTGH4) or mutant Y11A (for antagonist-bound HTGH4) (Egloff et al., 2015) and incubated overnight at 4 °C on a rolling device. The resin was washed first with NT wash buffer 1 (25 mM HEPES pH 8.0, 10 % glycerol (w/v), 600 mM NaCl, 0.5 % DM (w/v)) and then with NT wash buffer 2 (25 mM HEPES pH 7.0, 10 % glycerol (w/v), 150 mM NaCl, 0.3 % DM (w/v) and 2 mM DTT). Cleavage of the receptor from the pD-NT resin was performed with 0.7 mg of HRV 3C protease (produced in house) at 4 °C for 3 hours with rotation. For the purification of antagonist bound HTGH4, mutant pD-NT resin Y11A was mixed additionally with 4.3 mM SR142948 during 3C protease cleavage step. After cleavage, the receptor was eluted with NT wash buffer 2 to give a total of 10 ml elution. Subsequently, DDM was added to a final concentration of 1 % and the eluted NTR1 was diluted threefold with SP binding buffer (10 mM HEPES pH 7.0, 10 % glycerol (w/v), 0.05 % DDM (w/v), 2 mM DTT and 0.3 μM SR142948 or 0.1 μM NT1 peptide). The solution was applied to a SP Sepharose fast flow column and washed first with SP binding buffer, then with SP wash buffer (10 mM HEPES pH 7.7, 10 % glycerol (w/v), 35 mM NaCl, 0.05% DDM (w/v), 2 mM DTT, 0.3 μM SR142948 or 0.1 μM NTS1 peptide, respectively), followed by another wash with SP binding buffer. HTGH4 was eluted by SP elution buffer (10 mM HEPES pH 7.0, 10 % glycerol (w/v), 350 mM NaCl, 2 mM DTT, 0.05 % DDM (w/v), 0.5 μM NTS1 peptide or 0.3 μM SR142948) and further purified by size exclusion chromatography in NMR buffer (20 mM NaPi pH 6.5, 50 mM NaCl, 2 mM DTT, 0.02 % DDM (w/v), 0.1 μM NTS1 peptide or 0.3 μM SR142948).

For purification of HTGH4 in OGNG and DH₇PC, DDM was exchanged on the SP sepharose column. After the second wash with SP binding buffer, a third wash with SP binding buffer containing either 0.2 % OGNG or 0.1 % DH₇PC was included, and the receptor was eluted in elution buffer containing the desired detergent. Insertion of HTGH4 into MSP1D1 and MSP1D1ΔH5 lipid nanodiscs was done as described previously (Hagn et al., 2018) using a 1-palmitoyl-2-oleoyl-glycero-3-phosphocholine (POPC):1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-glycerol (POPG) = 3:1 lipid blend.

Circular dichroism (CD) spectroscopy

CD measurements and thermal stability scans of HTGH4 were recorded with a Jasco J-715 CD spectropolarimeter in a 0.1 cm path length cuvette. Thermal denaturation measurements were obtained with 5 μM of HTGH4 in CD buffer (10 mM NaPi pH 7.0, 2 mM DTT, 0.1 μM NTS1 peptide, 0.02 % DDM (w/v) or 0.3 % DM, 0.2 % OGNG, 0.1 % DH₇PC) at a wavelength of 222 nm from 20 to 100 °C with a heating rate of 1 °C/min. Thermal transition temperatures were obtained by fitting the curve with a Boltzmann equation (Privalov, 1979).

NMR experiments and backbone resonance assignment

2D-[¹⁵N,¹H]-TROSY NMR experiments, were performed at 315 K on a Bruker Avance III HD NMR 800 MHz spectrometer equipped with a cryogenic probe. The concentration of HTGH4 for the 2D experiments was 100–200 μM and 400 μM for 3D triple resonance experiments in NMR buffer containing 5 (v/v) % D₂O and ~100 mM d₂₆-DH₇PC (FB reagents, Sofia, Bulgaria). For obtaining backbone resonance assignments of HTGH4, a set of five TROSY-type triple resonance experiments were recorded (Salzmann et al., 1998) in a non-uniformly sampled (NUS) manner, 3D-HNCA, HNCO, HN(CO)CA, HN(CA)CO, HNCACB, as well as a 3D-¹⁵N-edited [¹H,¹H]-NOESY-TROSY experiment (300 ms mixing time). NMR data were reconstructed with hmsIST (Hyberts et al., 2012), processed with NMRpipe (Delaglio et al., 1995) and analyzed with NMRFAM-Sparky (Lee et al., 2015).

Hydrogen deuterium exchange (HDX) mass spectrometry

For hydrogen/deuterium exchange (HDX) experiments an ACQUITY UPLC M-class system equipped with automated HDX technology (Waters, Milford, MA, USA) was used. HDX kinetics were determined at 20°C taking data points at 0, 10, 60, 600, 1800 and 7200 s in technical duplicates. At the respective data points of the kinetics, 3 μl of a solution of approximately 30 μM protein were diluted 1:20 into 99.9 % D₂O-containing 20 mM sodium phosphate, pH 6.8 (titrated with HCl) or the respective H₂O–containing reference buffer. Quenching of the reaction mixture was achieved by adding 1:1 200 mM KH₂PO₄, 200 mM Na₂HPO₄, pH 2.3 (titrated with HCl), containing 4 M guanidine hydrochloride and 200 mM TCEP at 1 °C. For on-column peptic digest on a Waters Enzymate BEH pepsin column 2.1 × 30 mm at 20 °C, 50 μl of the samples were applied. Peptides were separated by reverse phase chromatography at 0°C applying a gradient increasing the acetonitrile concentration stepwise from 5–35% in 6 min, from 35–40 % in 1 min and from 40–95 % in 1 min. A Waters Acquity UPLC C18 1.7 μm Vangard 2.1 × 5 mm trapping-column and a Waters Aquity UPLC BEH C18 1.7 μm 1 × 100 mm separation column were used. and the eluted peptides were analyzed using an in-line Synapt G2-S QTOF HDMS mass spectrometer (Waters, Milford, MA, USA). All experiments were performed in biological duplicate collecting MS data over an m/z range of 100–2000. Mass accuracy was ensured by calibration with Glu-fibrino peptide B (Waters, Milford, MA, USA). Peptides were identified by MSE ramping the collision energy automatically from 20–50 V. As the automated system handles all samples at identical conditions, deuterium levels were not corrected for back exchange and are therefore reported as relative deuterium levels (Wales and Engen, 2006). Data were analyzed using the PLGS 3.0.3 and DynamX 3.0 software packages (Waters, Milford, MA, USA).

Chemically induced HTGH4 unfolding

Chemical unfolding experiments were conducted with guanidinium chloride (GdmCl) at a concentration range between 0 M and 4 M. 2 μM of HTGH4 in 0.2 % DM, 10 mM HEPES pH 8, 0.5 mM EDTA, 2 mM DTT was incubated at the indicated GdmCl concentrations for 12h at 4°C. The secondary structure content was then monitored at 20°C by the CD signal at 222 nm and the transition was fitted by with a Boltzmann equation (Privalov, 1979). The exact guanidinium chloride concentration in each sample was determined by measurement of the refraction index. Analysis of the unfolding data was done with a Boltzmann equation. For NMR sample preparation, partial unfolding at 1.8 M GdmCl for 12 h at 4°C was done after the SP sepharose purification step with 10 μM ²H,¹⁵N-HTGH4 in 0.2 % DM. After this step, GdmCl was removed by dialysis (6–8 kDa MWCO) for 6 h at 4°C followed by detergent exchange to 0.1 % DH₇PC via SP sepharose and S200 size exclusion chromatography.

Results

Sample optimization and detergent screen

For detergent screening, we selected four detergents, n-dodecyl-β-D-maltoside (DDM), n-decyl-β-D-maltoside (DM), octyl glucose neopentyl glycol (OGNG), and 1,2-diheptanoyl-sn-glycero-3-phosphocholine (DH₇PC). The first two detergents are maltosides with different chain lengths that are commonly used for GPCR extraction and purification, the third is a branched glucoside, and the latter a short-chain lipid. Maltosides are very common detergents for structural biology applications with GPCRs (Isogai et al., 2016; Nygaard et al., 2013). The more novel detergent OGNG has also been reported to be suitable for this receptor class (Lee et al., 2020). Finally, DH₇PC has been successfully used for high-resolution NMR studies of sensory rhodopsin (Gautier et al., 2010).

To evaluate this set of detergents for their use for NMR studies with HTGH4, we first performed size exclusion chromatography (SEC) (Fig. 1A). By this, the size of the complex can be easily assessed by the observed elution volume in SEC. Since the chain length of the detergents is a main factor that dictates the micellar size (Kotov et al., 2019), we could observe a lower elution volume for HTGH4 in DDM micelles than in DM micelles, reflecting molecular weights for the GPCR-micelle complex of 84 and 75 kDa, respectively. In both maltosides, HTGH4 also formed a higher molecular weight component (122 and 113 kDa), that is most likely the dimeric species. The dimer was completely absent in OGNG, owing to its shorter alkyl chain and its slightly harsher property. In addition, due to the formation of a small micelle, the elution volume in OGNG was higher than in the other two detergents, indicating a size of the GPCR-micelle complex of 64 kDa. Finally, HTGH4 in DH₇PC also gave rise to a homogenous SEC peak with a size (72 kDa) that is slightly smaller than DM but larger than OGNG.

Fig. 1. Size exclusion chromatography and CD thermal melting analysis of the stabilized NTR1 variant HTGH4 in complex with neurotensin after purification in the respective detergent.

(A) Analytical SEC (S200 increase 10/300) of HTGH4 after elution from an SP sepharose column in the following detergents, 0.05 % DDM, 0.3 % DM, 0.2 % OGNG and 0.1 % DH₇PC. The molecular weights of each protein species were calculated with reference samples as indicated above the SEC traces. (B) CD thermal melting curves of HTGH4 purified in each detergent. The melting point in each case is indicated. DDM: dodecyl-β-D-maltoside, DM: decyl-β-D-maltoside, OGNG: octyl glucose neopentyl glycol, DH₇PC: 1,2-diheptanoyl-sn-glycero-3-phosphocholine.

To probe the thermal stability of each receptor preparation, we conducted CD-detected melting experiments (Fig. 1B). These data show that HTGH4 purified in the maltoside detergents DDM and DM has thermal melting points of 84.6 °C and 69.8 °C, respectively. The receptor purified in OGNG and DH₇PC is also compactly folded with thermal melting points of 69.5 °C and 72.7 °C, respectively. Considering the combination of a minimal micelle size and a high thermal stability, DH₇PC appeared to be the best candidate for further experiments. Since detergent micelles have a high intrinsic flexibility, larger micelles could still give rise to advanced NMR spectral quality. Therefore, we next conducted 2D-TROSY NMR experiments with all four detergent systems for further assessing their use for more sophisticated NMR experiments.

Effect of detergents on HTGH4 NMR spectral quality

Next, we performed 2D-[¹⁵N,¹H]-TROSY NMR experiments to evaluate the NMR spectral quality of ²H,¹⁵N-isotope-labeled HTGH4 in the above-mentioned set of detergents at temperatures ranging from 303 to 320 K. Since rotational diffusion is enhanced at elevated temperatures, the quality of all recorded NMR spectra improved at higher temperatures (blue to red spectra). However, the sample in OGNG precipitated at 320 K, preventing further NMR experiments. In DDM, the NMR spectrum is indicative of a folded protein with a chemical shift dispersion from 11 to 6.5 ppm in the ¹H dimension. Although the spectral quality in DDM is promising, it shows very intense signals in the random coil chemical shift region around 8 ppm. This indicates a non-homogenous sample of HTGH4 in DDM micelles presumably caused by the large difference in NMR relaxation properties between micelle incorporated and solvent-exposed parts of the receptor. This difference is expected to be less pronounced in the smaller DM micelles. Indeed, the spectral quality in DM is enhanced with more uniform signal intensities in all spectral regions (Fig. 2A). HTGH4 in OGNG shows a very good spectral resolution as well. Yet, the number of observed resonances is much lower than in the other samples, possibly caused by markedly altered receptor dynamics in the rather small OGNG micelle. In DH₇PC, a decent NMR spectral resolution and quality is already observed at 303 K with further improvements up to 320 K. The overall signal pattern is similar to the maltoside samples, confirming the notion that HTGH4 is properly folded in this short-chain lipid. Furthermore, the number of visible NMR signals is the highest in DH₇PC, rendering this system suitable for more sophisticated multidimensional NMR experiments, even though this number is with approximately 150 signals lower than the 323 non-proline amino acid residues in the used HTGH4 construct.

Fig. 2. 2D-[¹⁵N,¹H]-TROSY NMR evaluation of ²H,¹⁵N-HTGH4 in different detergents and lipid nanodiscs.

(A) Neurotensin-bound HTGH4 was purified, and the detergent was exchanged on an SP Sepharose cation exchange column to 0.05 % DDM, 0.3 % DM, 0.2 % OGNG and 0.1% DH₇PC. 100–200 μM of HTGH4 was used to record 2D-[¹⁵N,¹H]-TROSY spectra at the indicated temperatures in 20 mM NaPi pH 6.5, 50 mM NaCl, 0.5 mM EDTA, 2 mM DTT and 5 %(v/v) D₂O at 800 MHz. In OGNG, the GPCR sample precipitated at 320 K. (B) 2D-NMR spectra of ²H,¹⁵N-labeled HTGH4 bound to neurotensin inserted in lipid nanodiscs of different sizes and assembly conditions recorded at 320 K and 700 MHz. The ratio between the GPCR and the membrane scaffold protein as well as the MSP-to-lipid ratio was varied, as indicated above each spectrum.

In addition to the detergent systems described above, we also inserted the receptor into membrane scaffold protein (MSP) lipid nanodiscs (Denisov et al., 2004) that have been described to be suitable for solution-state NMR studies (Gunsel and Hagn, 2022; Hagn et al., 2018; Hagn et al., 2013; Klöpfer and Hagn, 2019). To optimize nanodiscs insertion, we varied the MSP-to-GPCR ratio and the nanodisc size (Fig. 2B). The best spectral quality could be obtained in nanodiscs of 10 nm in diameter and using a large excess of MSP (1:20), which most likely promoted the homogenous insertion of the monomeric form. Despite the improvements achieved by this procedure, the spectral quality was still not sufficient for more challenging NMR experiments. In a recent study where another NTR1 variant was inserted into circularized (cMSP) lipid nanodiscs (Zhang et al., 2021) the NMR spectral quality was slightly better, highlighting the improved NMR spectra of membrane proteins in this advanced nanodisc system (Daniilidis et al., 2022). In summary, the spectra of HTGH4 in DH₇PC micelles are still markedly better.

NMR backbone resonance assignment of HTGH4 bound to the agonist neurotensin

To obtain backbone resonance assignments, we recorded a set of 3D-TROSY-based triple resonance experiments (Salzmann et al., 1998) and a 3D-¹⁵N-edited-[¹H,¹H]-NOESY-TROSY experiment using a ²H,¹³C,¹⁵N-labeled NTS1-bound HTGH4 sample in fatty-acid deuterated (d₂₆) DH₇PC. The quality of the 3D-NMR experiments was sufficiently high to assign 107 of the 154 visible resonances in the 2D-[¹⁵N,¹H]-TROSY spectrum, including 3 tryptophane (trp) side chain signals (Fig. 3A). For instance, the 3D-TROSY-HNCA experiment provided unambiguous sequential connections between Cα resonances of neighboring amino acids (Fig. 3B) and the 3D-NOESY experiment together with the 3D structure of HTGH4 was used to assign the visible trp side chain indole signals (Fig. S1). Since a relatively low number of the expected resonances is visible in the spectrum, we wondered where the assigned resonances are located in the protein. As shown in Fig. 3C, all assigned resonances are in peripheral regions of the receptor, namely the orthosteric ligand binding site, the cytoplasmic loop region and parts of transmembrane helix 1 (α1). The lower degree of assignment in the cytosolic region as compared to the orthosteric binding site is caused by the higher solvent exposure of the latter region, especially during initial stages of the purification where the receptor is not bound to a high-affinity agonist. In addition, the cytosolic region is inherently more dynamic to enable the adoption of the active state that exposes the G-protein binding site (Hilger, 2021). Thus, the NMR signal intensity in this region might be reduced due to motions in the ms to μs time scale, impeding reliable resonance assignments.

Fig. 3. NMR backbone resonance assignment of solvent-exposed residues in HTGH4.

(A) 2D-[¹⁵N,¹H]-TROSY spectrum of NTS1-bound ²H,¹³C,¹⁵N-labeled HTGH4 in d₂₆-DH₇PC at 42 °C with the assigned backbone amide resonances labeled. Around 30 % of the expected signals are visible, presumably due a lack of amide proton back-exchange in the transmembrane helices. (B) Sample strips of a 3D-TROSY HNCA experiment at 800 MHz used for backbone resonance assignment show sequential connections between the amino acids in the receptor (red broken lines). (C) Assignment mapped on the structure of HTGH4. (D) Secondary chemical shift analysis (experimental minus random coil Cα chemical shift values; pos values: α-helix, negative values: β-sheet) of the assigned Cα resonances of neurotensin-bound HTGH4. For comparison, the secondary structure in the HTGH4 crystal structure is indicated (red: α-helix, blue: β-strand).

In contrast, all residues that are located deeper inside the membrane-embedded part of the receptor could not be assigned since they are not visible in the NMR spectra. This behavior suggests a lack of amide proton back-exchange. The amide moieties of the receptor produced in highly deuterated growth medium are initially present in a deuterated form (ND instead of NH). To be visible in proton-detected NMR experiments the amides need to back-exchange to the protonated form, which typically takes place in solvent exposed parts during protein purification in H₂O-based buffers. We analyzed the available NMR data and determined the ¹³Cα secondary chemical shift information in the assigned regions of the receptor, which can be used to determine the secondary structure (Wishart et al., 1992). This analysis shows that the secondary structure detected by NMR in solution coincides very well with the secondary structure seen in the crystal structure of HTGH4 (Egloff et al., 2014) (Fig. 3D). The achieved assignments offers the benefit of being able to selectively observe the structural features of the orthosteric ligand binding site.

Structural impact of agonist and antagonist binding to HTGH4

After the backbone resonance assignment procedure, we further characterized the impact of two orthosteric ligands on the NMR spectra of HTGH4. We performed 2D-[¹⁵N, ¹H]-TROSY experiments of HTGH4 in complex with the peptide agonist neurotensin-1 (NTS1, aa N-GPGGRRPYIL-C) and the antagonist SR142948 (Gully et al., 1997). The NMR spectrum of the antagonist-bound sample showed lower signal intensity than the NTS1-bound form with the remaining stronger NMR resonances predominantly located in the random coil spectral region (around 8 ppm in the ¹H dimension). This spectral signature suggests that the presence of the antagonist leads to conformational exchange on the intermediate NMR timescale, which is in line with the observation of a higher degree of disorder and missing electron density in the crystal structure of NTR1 in complex with SR142948 (Deluigi et al., 2021) than with the agonist NTS1 (Egloff et al., 2015). Furthermore, a recent NMR study with selectively methionine ¹³C-methyl-labeled neurotensin receptor shows that different orthosteric ligands can selectively modulate the receptor dynamics (Bumbak et al., 2023).

Despite a reduction in the NMR spectral quality, CD thermal unfolding experiments indicate properly folded HTGH4 with an increased thermal stability (72.7 °C vs. 81.2 °C), when bound to the antagonist (Fig. S2). This suggests a strong interaction of SR142948 with HTGH4, which is in line with the reported nanomolar affinity of SR142948 for neurotensin receptor (Gully et al., 1997). To explore the impact of SR142948 on the orthosteric binding site of HTGH4, we calculated NMR chemical shift perturbation (CSP) values between the two states (Fig. 4B) and could observe strong spectral changes in the entire binding site (Fig. 4C) but less pronounced effects at the cytosolic side. The lack of structural changes in that region in HTGH4 is caused by a mutation in the functionally essential ionic lock region of NTR1 (R167L) that is preventing efficient G-protein coupling (Goba et al., 2021). To explore the changes in solvent accessibility of HTGH4 in these two states we next employed hydrogen deuterium exchange (HDX) mass spectrometry (Fig. 4D). The detected peptides have more than 80 % coverage of the HTGH4 sequence and show a redundancy of 2.75 (Fig. S3). These data show that the hydrogen exchange rates in the antagonist-bound state are higher than in the agonist bound state, suggesting a structurally less defined conformation (Fig. 4D, Fig. S4). In line with the NMR observations, the cytosolic region does not show pronounced changes in the two states whereas the orthosteric binding site is more heavily affected with a higher degree of protection if bound to NTS1 (Fig. 4E). In addition to structural changes, these data also indicate that the core part of the receptor is strongly protected from hydrogen exchange, corroborating the lack of NMR signals in this region if deuterated protein is used (Fig. 3C).

Fig. 4. Interaction of HTGH4 with different orthosteric ligands.

(A) 2D-[¹⁵N,¹H]-TROSY NMR spectra of HTGH4 bound to the agonist neurotensin (NTS1, red) and the antagonist SR142948 (black) at 320 K. (B) Chemical shift perturbation (CSP) analysis of HTGH4 bound to NTS1 or SR142948. (C) CSP values mapped onto the structure of HTGH4 (Egloff et al., 2014). (D) Hydrogen deuterium exchange (HDX) mass spectrometry to determine the fractional uptake of deuterium in HTGH4 in complex with NTS1 (red) or SR142948 (blue). (E) Data sets from (D), each color-coded onto the structure of HTGH4.

Partial refolding and uniform receptor protonation to visualize the transmembrane region of HTGH4

As suggested by the HDX-MS experiments (Fig. 4D,E), the central part of the GPCR transmembrane helical bundle is highly protected from the solvent, prohibiting hydrogen exchange. Since deuterated protein is beneficial for high-resolution NMR experiments of such a rather large protein-micelle complex, we wondered whether we could utilize HTGH4 unfolding by the addition of the chemical denaturant guanidinium chloride (GdmCl) to facilitate amide proton back-exchange. Full denaturation followed by refolding into detergent micelles was not yet successful for HTGH4 (Fig. S5), which called for a more fine-tuned procedure. Thus, we first conducted CD-detected GdmCl-induced chemical unfolding experiments (Fig. 5A) with HGTH4 in DM micelles. As suggested by the high thermal stability of >70°C (Fig. S2), the midpoint of HTGH4 unfolding is with 2.07 M GdmCl quite high, giving rise to a thermodynamic stability (ΔG) of −21.3 kJ/mol. We adjusted the GdmCl concentration to 1.8 M, which is slightly below the unfolding midpoint, to achieve only partial unfolding of the receptor followed by removal of the denaturant and detection by 2D-NMR (Fig. 5B). By this procedure, we aimed at enhancing hydrogen exchange in the transmembrane regions of the receptor without inducing protein misfolding. Indeed, we could observe marked differences between the 2D-[¹⁵N,¹H]-TROSY experiment of HTGH4 purified with the standard protocol (Fig. 5C, blue) and a sample that was partially unfolded for 12 h at 4°C (Fig. 5C, orange). The number of resonances after partial refolding increased and weak peaks in the initial spectrum became much stronger in the refolded sample, e.g. in the glycine region (105–110 ppm in the ¹⁵N dimension). However, due to the refolding procedure, some resonances in the 2D-spectrum became more heterogeneous, suggesting that even partial HTGH4 unfolding leads to heterogenous conformational states of the receptor. This is most apparent when looking at the tryptophane side chain resonances (marked by boxes in Fig. 5C). HTGH4 contains five Trp residues and only four are visible in the deuterated receptor. After partial refolding, this number increased to ~9 signals, which strongly indicates structural heterogeneity. Thus, this strategy most likely will not lead to a high-quality GPCR sample, even though amide proton back-exchange is facilitated. Nonetheless, these NMR experiments, together with the HDX MS experiments clearly show that a lack of amide proton back-exchange is the main reason why only a minor portion of the backbone resonances in the receptor is visible. To confirm this notion, we additionally produced non-deuterated HTGH4 (¹H,¹⁵N-labeled) and recorded a 2D-TROSY experiment (Fig. 5C, red). In that spectrum, the estimated number of backbone amide resonances is very close to the expected number of non-proline amino acids in the protein, with the correct number of Trp side chain resonances. A 2D-TROSY spectrum of the same sample in D₂O buffer (Fig. 5C, green), where signal loss of every amide located in solvent-exposed regions is expected, contained around 90 signals that are most likely originating from residues in buried transmembrane helices. This data clearly shows that the TMH region can be detected in a protonated sample. It remains to be shown whether resonance assignments can be obtained with a non-deuterated GPCR using multidimensional NMR experiments.

Fig. 5: Partial unfolding of HTGH4 enhances amide proton back-exchange.

(A) Chemically induced unfolding experiments of HTGH4 in DM micelles indicate high thermodynamic stability. At 1.8 M guanidine hydrochloride (GdmCl) the HTGH4 structure is only partially unfolded. (B) To enhance amide proton back-exchange, partial unfolding was done in presence of 1.8 M GdmCl for 12 h at 4°C, followed by GdmCl removal by dialysis and detergent exchange to DH₇PC. (C) 2D-[¹⁵N,¹H]-TROSY spectra of ²H,¹⁵N-labeled HTGH4 obtained from a standard purification (blue), after a partial unfolding (at 1.8 M GdmCl) and refolding cycle (orange), in the ¹H,¹⁵N-labeled form in H₂O (red) or in D₂O buffer (green).

Discussion

In this study we performed high-resolution NMR investigations of the stabilized NTR1 variant HTGH4 by using a robust protocol for the selection of detergents, temperature, and ligands. Our study shows that the short-chain lipid DH₇PC is a suitable membrane mimetic that results in a small GPCR-micelle size and promotes a properly folded protein sample of high stability for NMR. Due to these advantages, DH₇PC has been previously used for the structure determination of the 7-TM helical protein sensory rhodopsin (Gautier et al., 2010). In recent studies, branched maltoside detergents, such as LMNG, have been shown to promote GPCR stability and to be suitable for structural studies and NMR (Lee et al., 2020; Schuster et al., 2020). In addition to detergents, we also evaluated lipid nanodiscs (Denisov et al., 2004; Gunsel and Hagn, 2022; Hagn et al., 2018; Hagn et al., 2013) as a possible membrane mimetic for further NMR studies. Even though we were able to optimize the NMR spectral quality of HTGH4 in lipid nanodiscs the final sample was not yet suitable for high-resolution studies. The use of circularized MSP nanodiscs (Miehling et al., 2018; Nasr et al., 2017) will be beneficial in this attempt (Zhang et al., 2021) and is generally suitable to improve the use of the nanodisc system for NMR studies of complex membrane protein systems (Daniilidis et al., 2022).

We could obtain high-quality 3D NMR experiments which enabled the backbone resonance assignment of most of the visible resonances in the spectrum. The corresponding amino acids mainly cluster to solvent exposed parts of the protein, whereas the signals of amino acids in transmembrane helices are mostly absent. The stable fold of the transmembrane helices and the presence of a detergent micelle provides efficient shielding from the solvent to prevent back-exchange in these regions. This assumption is corroborated by HDX-MS experiments and NMR experiments using protonated receptor. Partial unfolding of HTGH4 by the chemical denaturant GdmCl led to increased signal intensity and to a higher number of peaks in the NMR spectrum. This is due to a destabilization of the helical secondary structure elements in the GPCR leading to enhanced amide exchange. Nevertheless, this procedure also led to higher sample heterogeneity, suggested by a too large number of resonances of e.g. Trp side chains in the 2D-NMR spectrum. A possible strategy to obtain fully back-exchanged protein is to fully refold HTGH4. Even though successful refolding strategies and cell-free production have been reported for other GPCRs (Park et al., 2012; Schrottke et al., 2017) this procedure remains to be optimized for HTGH4.

Another way to avoid refolding is the use of a fully protonated GPCR for NMR. Though, the markedly enhanced transverse relaxation properties of protonated high molecular weight proteins will hamper high-resolution NMR studies (Sattler and Fesik, 1996). Nonetheless, the successful NMR structure determination of sensory rhodopsin in a protonated form demonstrates that this is in principle possible if the protein is compactly folded and tolerates high temperatures of up to 50°C (Gautier et al., 2010). To further improve the spectral quality without the deuteration of amides, other approaches have been suggested, such as the addition of deuterated amino acids to H₂O-based growth media (Chill et al., 2006; Franke et al., 2018). This procedure will still lead to protonation of the acidic α-proton in each amino acid.

Another strategy is to consider the obtained backbone resonance assignments in the deuterated protein as starting points for the structure-based assignment of methyl groups. Methyl group labeling in a highly deuterated background (Goricanec and Hagn, 2019; Goricanec et al., 2016; Hagn and Wagner, 2015; Rosen et al., 1996) does not rely on proton back-exchange. Furthermore, multiple computational approaches exist to perform automated methyl group assignment with a known high-resolution (membrane-) protein structure using a set of 3D- or 4D-NOESY experiments (Kooijman et al., 2020; Nerli et al., 2021; Pritisanac et al., 2019). Methyl groups are valuable probes for investigating the structure and dynamics of high-molecular-weight proteins by NMR (Alderson and Kay, 2021). Thus, in the case of HTGH4, the ability to produce high-level methyl-labeled protein will be highly valuable for the investigation of the conformational landscape of this GPCR bound to different orthosteric ligands or extra-membrane binding partners, such as G-proteins or β-arrestins. Such studies will complement currently available structural information in various functional states. HTGH4 contains 26 mutations that impact its functionality (Goba et al., 2021; Scott et al., 2014). However, only a few back-mutations to amino acids that are present in the wild-type protein suffice to restore its GPCR switching functionality (Goba et al., 2021). Furthermore, incorporation of HTGH4 into lipid nanodiscs will be beneficial for the investigation of its complexes with G-proteins (Goricanec et al., 2016; Zhang et al., 2021) and β-arrestins (Huang et al., 2020) by NMR. In that context, the herein presented optimization of HTGH4 sample conditions and the presented NMR characterization are a promising starting point for obtaining more global resonance assignments and detailed structural and dynamical insights that cannot be captured by other structural methods.

• Show suitability of a GPCR obtained from directed evolution for NMR studies

• Screening procedure for detergents that promote high stability and spectral quality of neutotensin receptor

• Neurotensin receptor responds specifically to agonist and antagonist

• NMR backbone resonance assignment of visible residues in neurotensin receptor

• Comparison of secondary structure in solution and in the crystal

• Lack of amide proton back-exchange as reason for lack of NMR signals