Effect of Ligands and Transducers on the Neurotensin Receptor 1 Conformational Ensemble

Abstract

Using a discrete, intracellular ¹⁹F nuclear magnetic resonance (NMR) probe on transmembrane helix 6 of the neurotensin receptor 1 (NTS1), we aim to understand how ligands and transducers modulate the receptor’s structural ensemble in a solution. For apo NTS1, ¹⁹F NMR spectra reveal an ensemble of at least three conformational substates (one inactive and two active-like) in equilibrium that exchange on the millisecond to second timescale. Dynamic NMR experiments reveal that these substates follow a linear three-site exchange process that is both thermodynamically and kinetically remodeled by orthosteric ligands. As previously observed in other G protein-coupled receptors (GPCRs), the full agonist is insufficient to completely stabilize the active-like state. The inactive substate is abolished upon coupling to β-arrestin-1 (βArr1) or the C-terminal helix of Gα_q, which comprises ≳60% of the GPCR/G protein interface surface area. Whereas βArr1 exclusively selects for pre-existing active-like substates, the Gα_q peptide induces a new substate. Both transducer molecules promote substantial line broadening of active-like states, suggesting contributions from additional microsecond to millisecond exchange processes. Together, our study suggests that (i) the NTS1 allosteric activation mechanism may be alternatively dominated by induced fit or conformational selection depending on the coupled transducer, and (ii) the available static structures do not represent the entire conformational ensemble observed in a solution.

Graphical Abstract

INTRODUCTION

G protein-coupled receptors (GPCRs) serve as the primary hubs to relay changes in extracellular environments across the eukaryotic cell membrane.¹ The more than 800 members of this protein superfamily share a conserved seven transmembrane helix (TM) bundle architecture that recognizes a large variety of ligands, comprising small molecules, hormones, peptides, and photons.² As such, it is no surprise that they encompass over 30% of the drug market.³ Although atomic models are still relatively scarce compared to other protein classes, there are currently 121 unique receptor structures, or ~14% of the total GPCR superfamily.⁴ The difficulty of GPCR structural studies primarily reflects inherent protein instability and low recombinant expression. Through the use of detergent membrane mimetics and creative receptor engineering, the rate at which new receptor structures are determined has increased in recent years.⁵ These atomic models have revealed conserved, long-range allosteric activation networks that link the receptor orthosteric pocket to the intracellular bundle across the cell membrane. Most notably, the DRY, PIF, CWxP, and NPxxY motifs serve as internal molecular “switches” of class A GPCRs, connecting ligand binding to downstream effector molecule complexation and activation events, spanning a distance of nearly 50 Å.⁶ The modeling of allosteric switches across numerous receptors has led to a putatively conserved structural activation profile.^7,8

Neurotensin receptor 1 (NTS1) has quickly become one of the most well-characterized GPCRs with structures of the apo state, complexes with various pharmacological ligands, and ternary complexes with both the heterotrimeric G_i protein and β-arrestin-1 (βArr1) transducers.^9–17 NTS1 is a class A, β group receptor that is expressed throughout the central nervous system and the gastrointestinal (GI) tract.¹⁸ Activation by its endogenous tridecapeptide ligand neurotensin (NT) mediates a variety of physiological processes including low blood pressure, high blood sugar, low body temperature, mood, and GI motility.¹⁹ It is also a long-standing therapeutic target for Parkinson’s disease, schizophrenia, obesity, hypotension, psychostimulant substance use disorders, and cancer.²⁰

Current atomic models derived from either X-ray crystallography or cryo-electron microscopy (EM) capture NTS1 in different stages of activation, mediated by bound ligands and transducer proteins. A hallmark of GPCR activation is the outward movement of transmembrane helix 6 (TM6) to accommodate G protein and arrestin complexation.²¹ In NTS1, ligand binding at the extracellular orthosteric pocket allosterically induces a ~13 Å lateral displacement at the intracellular tip of TM6.¹⁰ Ultimately, these models remain static. This has left a void in the literature detailing the NTS1 conformational ensemble and the pleiotropic effects ligands and transducers have on individual substates. This inspired us to pursue solution nuclear magnetic resonance (NMR) spectroscopy as a complementary approach to better characterize the allosteric activation mechanism in NTS1.

In this study, we ¹⁹F-label TM6 of a thermostabilized NTS1 construct solubilized in 2,2-didecylpropane-1,3-bis-β-D-maltopyranoside (LMNG) detergent micelles. Trifluoromethyl NMR probes are an optimal choice for site-selective isotopic labeling due to their low background signal and high spin-1/2 natural abundance.²⁵ Their observed chemical shift value is dominated by solvent polarity and the local electronic environment, which makes them very sensitive to large conformational rearrangements observed in GPCRs. For example, as the intracellular tip of TM6 moves outward to accommodate transducer proteins, we anticipate an upfield chemical shift perturbation reflecting increased solvent exposure.²⁶ NMR can provide both qualitative and quantitative information regarding the timescale of structural motions.²⁷ Although very fast rotation about the methyl axis averages any local fluctuations into a single peak, slower “biologically relevant” motions on approximately the microsecond to millisecond timescale affect both the resonance linewidth and chemical shift.²⁸ As conformational exchange slows further into the millisecond to second regime, the averaged resonance will split into distinct peaks with characteristic linewidths and chemical shifts.²⁹ Herein, we employ the G_q C-terminal α5-helix peptide and a pre-activated βArr1 to recapitulate responses to the heterotrimeric G_q protein and βArr1.^30,31 Together, this enables us to develop a dynamic model of NTS1 activation in which ligands and transducers are allosterically coupled.³²

RESULTS

Thermostabilized NTS1 Retains Signaling Activity.

The well-characterized structure of NTS1 in a variety of pharmacologically relevant states creates an ideal system for exploring the allosteric mechanisms of GPCR activation. Yet, wildtype NTS1 structural characterization remains challenging due to poor receptor stability following isolation from native membranes.³³ All published NTS1 structures to date incorporate some combination of thermostabilizing mutations, lysozyme fusions, DARPin fusions, or conformationally selective antibodies.^9,12,34,35 Here, we employed a functional, thermostabilized rat (r)NTS1 variant (termed enNTS1) for solution NMR spectroscopy.³²

To further validate enNTS1’s functional integrity, we performed a cell-based alkaline phosphatase (AP) reporter assay for G protein activation. The stimulation of Gα_q and Gα_12/13 leads to ectodomain shedding of an AP-fused transforming growth factor-α (TGFα), which is then quantified using a colorimetric reporter.²² HEK293A cells were transfected with AP–TGFα and a NTS1 plasmid construct. A hexapeptide corresponding to NT residues 8–13 (NT8–13) is sufficient to generate a full agonist response in wildtype rNTS1;³⁶ NT8–13 stimulates robust, concentration-dependent G protein-coupling to enNTS1 in the TGFα shedding assay, although with reduced efficacy compared to human (h)NTS1 (Figures 1A and S1). Both enNTS1 and hNTS1 were equally expressed on the cell surface (Figure S1C). βArr1 recruitment was also measured using a NanoBiT enzyme complementation system.²³ The large and small fragments of the split luciferase were fused to the N-terminus of βArr1 and the C-terminus of NTS1, respectively, and these constructs were expressed in HEK293A cells. As a negative control, we used the vasopressin V2 receptor (V2R) C-terminally fused with the small luciferase fragment. enNTS1 exhibited weak basal βArr1 recruitment that did not increase upon agonist addition (Figures 1B and S1B). Nonetheless, the addition of the βArr1-biased allosteric modulator (SBI-553) dose dependently potentiates NT8–13-mediated βArr1 recruitment (Figure S1D).^37,38 As SBI-553 alone is unable to substantially stimulate βArr1 recruitment to enNTS1 at the same concentration, we conclude that enNTS1 recruits using the same molecular mechanism as wildtype NTS1, although with reduced potency (Figure S1E).

Figure 1.

Orthosteric ligands modulate the enNTS1 conformational ensemble. (A) G protein activation was assessed using a TGFα shedding assay on HEK293A cells transiently transfected with vasopressin receptor 2 (V2R; Mock), human (h)NTS1, or enNTS1.²² Cells were stimulated with vehicle (brown) or 1 μM NT8–13 (gray). Error bars represent SEM from three independent experiments. (B) βArr1 recruitment to V2R (Mock), hNTS1, and enNTS1 was measured using a NanoBiT-based assay.²³ Cells were stimulated with vehicle (brown) or 1 μM NT8–13 (gray). Luminescence counts recorded from 5 to 10 min following stimulation were averaged and normalized to the initial counts. Error bars represent SEM from four independent experiments. (C) Deconvoluted ¹⁹F NMR spectra of enNTS1[Q301C^BTFMA] in various liganded states. All the ligands are added to the receptor at 10 Meq. The relative population and LWHH are indicated for each substate. (D) Chemical shift value of each deconvoluted resonance was confirmed by monitoring the residual error while constraining peak height and LWHH. The chemical shift was constrained to a new value and the procedure was repeated. The lowest residual error value for each substate represents the chemical shift used in deconvolution.²⁴

It is unclear which enNTS1 thermostabilizing mutations are responsible for attenuating G protein activation and βArr1 recruitment. We reverted stabilizing mutations adjacent to the connector region (V358F^7.42) and within the sodium binding site (S113D^2.50/A362S^7.46), which are considered critical for activity, but neither backmutation recovered signaling in the TGFα shedding assay (data not shown).^16,39

¹⁹F NMR Probe Does Not Affect enNTS1 Function.

To characterize enNTS1’s structural ensemble in solution, we developed protocols to selectively incorporate cysteine-reactive ¹⁹F NMR probes onto TM6. Many previous ¹⁹F NMR studies of GPCRs target position 6.27 (Ballesteros–Weinstein nomenclature), but coupling the ¹⁹F-2-bromo-N-(4-(trifluoromethyl)phenyl)acetamide (BTFMA) probe at this site reduced enNTS1 expression yields and stability (data not shown).^24,40,41 MtsslWizard was used to model cysteine-conjugated BTFMA probes at various alternative positions along TM6 of the apo (PDB 6Z66), agonist NT8–13-bound (PDB 4BWB), antagonist SR142948-bound (PDB 6Z4Q), Gα_iβγ protein ternary (PDB 6OS9), and βArr1 ternary NTS1 complex structures (PDB 6UP7 and 6PWC).^{9,10,12,13,34,42} MtsslWizard rapidly screened 200 randomly generated BTFMA rotamers and enumerated all conformers that did not clash with the receptor to a tolerance of 3.4 Å. Although position 6.27 is unrestricted in antagonist and transducer-bound models, the tight TM5/TM6 packing in the apo and agonist-bound structures sterically restricted BTFMA to 18 and 110 potential rotamers, respectively, suggesting a mechanism for its observed instability (Figure S2A). In contrast, the neighboring residue Q301C^6.28 presented completely unhindered mobility in all six structural models (Figure S2A and Table S1). BTFMA labeling at position 6.28 had no effect on receptor thermostability or yield.

In the final construct, herein enNTS1[Q301C^BTFMA], solvent-exposed C172^3.55 was mutated to serine to prevent off-site labeling. Site-specific BTFMA labeling was confirmed by liquid chromatography–mass spectroscopy and NMR with estimated efficiencies of >95 and >80%, respectively (Figure S2B,C and Table S2). enNTS1[Q301C^BTFMA] showed no appreciable difference in affinity for agonist NT8–13 in saturation binding experiments compared to unlabeled enNTS1, indicative of proper receptor folding (Figure S2D). Dynamic NMR experiments require the sample to be stable throughout multiday data acquisition. To confirm enNTS1[Q301C^BTFMA] would remain viable during extended periods of data collection, we measured its ability to bind fluorescently labeled NT8–13 as a function of time. After 10 days at 37 °C, 55.0 ± 5.7% apo and 82.1 ± 17.1% agonist-bound enNTS1[Q301C^BTFMA] preserved binding competency (Figure S3).

enNTS1’s Conformational Ensemble is Sensitive to Orthosteric Ligands.

We collected one-dimensional (1D) ¹⁹F NMR spectra of enNTS1[Q301C^BTFMA] in the absence and presence of saturating (10:1 Meq) orthosteric ligand concentrations to investigate the conformational ensemble; all the spectra possessed S/N ratios ranging from 95.5 to 199.1. The spectral deconvolution of ligand-free enNTS1[Q301C^BTFMA] best fits three Lorentzian curves, which qualitatively indicates a three-state equilibrium in slow (ms to s) exchange on the NMR timescale (Figures 1C and S4). The area, chemical shift, and linewidth at half-height (LWHH) for each deconvoluted resonance serve as direct reporters of the relative population, chemical environment, and flexibility of each conformer, respectively.⁴³ Following the approach established by Prosser and colleagues, best-fit values were identified by individually constraining a given substate’s chemical shift over a range of frequencies and then globally fitting the remaining parameters (Figures 1D and S4).²⁴ A chemical exchange saturation transfer (¹⁹F-CEST) experiment was performed on apo enNTS1[Q301C^BTFMA] to further validate the existence of three substates. The region from −800 to +600 Hz (16.67–11.81 ppm), relative to the S₂ substate, was scanned in 100 Hz increments with 1 s saturation pulses (Figure 2A). Frequency-dependent changes in peak height confirm the existence of three substates undergoing slow conformational exchange on the NMR timescale (Figure 2A).

Figure 2.

enNTS1 substate exchange and transverse relaxation line broadening. (A) ¹⁹F-CEST experiments using apo enNTS1[Q301C^BTFMA] confirm the presence of three substates qualitatively interconverting on the millisecond to second timescale. A series of 1D spectra were collected, varying the offset frequency of a 1 s saturation pulse at 100 Hz intervals. Spectra were deconvoluted and the height of each substate was normalized to its respective height in the presence of a far off-resonance 1 s control saturation. (B) Series of deconvoluted CPMG T₂ spectra collected with 1 ms CPMG spin-echo and total relaxation delay varied from 1 to 6 ms. (C) Fitting the normalized peak heights of the deconvoluted CPMG T₂ spectra to a monoexponential model. CEST and CPMG T₂ error bars represent the standard deviation as calculated at a single offset frequency across the entire spectral series. (D) TM4-TM6 Cα distance between NTS1 S182^4.38 and Q301^6.28, plotted for all NTS1 atomic models deposited in the Protein Data Bank, correlates with their putative activation state.

In the apo state, the three resonances (labeled S₁, S₂, and S₃) were populated at 5, 80, and 15%, respectively, with LWHH ranging from 146 to 313 Hz (Figure 1C). For the deconvoluted ¹⁹F-1D spectra, LWHH = 1/πT₂*, where T₂* = T_2,homogenous + T_{2,inhomogenous}. T_2,homogenous is the natural linewidth modulated by microsecond to millisecond chemical exchange, whereas T_{2,inhomogenous} results from magnetic field inhomogeneities and qualitatively slower chemical exchange processes.^44,45 A Carr–Purcell–Meiboom–Gill (CPMG) T₂ experiment is capable of refocusing T_{2,inhomogenous}, and thus, to first approximation, reports only the T_2,homogenous contribution.^44,45 Using a train of ~1 ms CPMG spin-echo periods over a 1–6 ms total relaxation delay, all three substates exhibited a monoexponential decay in peak height (Figure 2B). Both the LWHH and T₂ were directly fitted for each substate and compared to the linewidths and T₂* derived from the 1D deconvolution (Figure 2C and Table S3). We report T_{2,inhomogenous} contributions on the order of 51–62% T₂* for each substate; assuming a homogenous sample preparation in a well-shimmed modern spectrometer, T_{2,inhomogenous} should be negligible.⁴⁶ Thus, our results suggest that conformational exchange on the order of the millisecond CPMG delay is also being partially refocused, although rigorous determination would require relaxation dispersion-type CPMG experiments.⁴⁶

The same three substates were also present in agonist- and antagonist-bound spectra; agonist reduced the S₁ population while increasing S₂, whereas antagonist had the opposite effect (Figure 1C and Figure S4). Both ligands similarly decreased the S₁ LWHH ~20 Hz, suggesting a slight stabilizing effect. The S₂ substate exhibited subtle ligand-dependent frequency perturbations—shifting approximately 0.01 ppm downfield and 0.04 ppm upfield in response to SR142948 antagonist and NT8–13 agonist, respectively (Figure 1C). The simplest explanation for this behavior is that the metastable S₂ substate is in fast exchange between two high energy microstates, such as local stereoisomers, where the S₂ chemical shift reflects the relative population of each microstate.⁴⁶ These modest chemical shift perturbations were accompanied by ~20 Hz line broadening. Similarly, the S₃ linewidth reported on ligand efficacy with the agonist decreasing, and the antagonist increasing, the LWHH by 20 Hz (Figure 1C).

Orthosteric Ligands Modulate Distinct Conformational Kinetics.

The simultaneous observation of three distinct enNTS1[Q301C^BTFMA] resonances defines an upper limit of approximately 10⁻³ s⁻¹ to the exchange rates. We undertook saturation transfer difference (STD) dynamic NMR experiments to quantify the exchange kinetics between substates. ¹⁹F-STD experiments employ a low power pulse to selectively saturate (i.e., reduce the intensity) a single substate frequency, ν_s. When a saturated substate exchanges, it decreases the signal at the other site(s). A series of ¹⁹F-1D spectra were collected with the saturation pulse duration varied from 50 to 1000 ms. To account for off-resonance saturation effects, a second series of ¹⁹F-1D spectra were collected with a control saturation pulse set at an equal, but opposite, offset (ν_c) from the substate of interest (Figure S5). The difference in peak height between on- and off-resonance experiments (ν_s,eff), as a function of saturation pulse length, can be fitted to the Bloch–McConnell equations to yield the exchange rate constant (k) with the irradiated resonance, and by extension, the lifetimes (τ_s = 1/k) of each conformer.⁴⁷ Judicious selection of irradiation frequencies is paramount for the minimization of off-resonance artifacts and incomplete saturation, but the limited spectral dispersion of enNTS1[Q301C^BTFMA] substates presents an insurmountable experimental challenge that underlies ambiguity in the accuracy of fitted exchange rates (Figure S6 and Table S4).^48–50 Nonetheless, we were unable to observe direct exchange between S₁ and S₃ under any condition which supports a linear activation trajectory (S₁ → S₂ → S₃) from the inactive conformer to the most solvent exposed position (Figure S7 and Table S4). Such a sequential transition was also observed for ¹⁹F-TM6^6.31 of the adenosine A_2A receptor in LMNG micelles.⁵¹ Although subsequent studies in nanodiscs resolved these resonances into two distinct nucleotide-exchange competent states and an activation intermediate, which complicates this comparison with enNTS1[Q301C^BTFMA].⁵² As an orthogonal estimate of exchange rates, we collected a two-dimensional [¹⁹F,¹⁹F]-exchange spectroscopy (EXSY) spectrum on apo enNTS1[Q301C^BTFMA] with a 100 ms mixing time. Weak diagonal peaks were only observed for S₂ and S₃ substates, but no exchange cross peaks were visible, likely due to the overall poor S/N ratio (Figure S8).

Altogether, we hypothesize that S₁ is an inactive conformation, whereas S₂ and S₃ reflect active-intermediate and active-like states, respectively. This is based upon several similar observations with other ¹⁹F-TM6 labeled receptors: (i) the comparatively broad S₁ linewidth, which is consistent with microsecond to millisecond timescale motions such as DRY ionic-lock flickering between formation/disruption reported for the β₂-, β₁-adrenergic, and A_2A adenosine receptors;^53–55 (ii) the near disappearance of the S₁ resonance and concurrent increase of the S₂ population upon agonist addition; and (iii) the S₁, S₂, and S₃ substate chemical shifts are increasingly upfield, which is consistent with increased solvent exposure as the cytoplasmic cavity expands for transducer association.^41,56 Although likely an oversimplification, we speculate that the S₁, S₂, and S₃ substates represent the three conformations that all 24 NTS1 atomic models can be organized based on the intracellular TM4–TM6 distance (Figure 2D).

G Protein Mimetic Stabilizes a Novel Conformation.

Next, we investigated the interaction of enNTS1 with a synthetic peptide (herein Gα_q peptide) corresponding to residues 333–359 of the Gα_q C-terminus (a.k.a. α5-helix). The α5-helix is conserved across Gα protein subunits as a random coil that adopts a helical structure upon receptor recognition that comprises 55–69% of the GPCR/G protein interface surface area.^57–59 We first characterized the efficacy of the enNTS1/Gα_q peptide interaction using an affinity pulldown approach.³⁰ The Gα_q peptide was N-terminally fused to a biotin tag and enNTS1 contained a C-terminal monomeric, ultra-stabilized green fluorescent protein (muGFP) fusion (enNTS1-muGFP).⁶⁰ Binding efficacy was quantified as the fluorescence ratio of streptavidin-captured enNTS1-muGFP versus total (i.e., streptavidin-captured plus unbound) fluorescence.

In the absence of a ligand, the Gα_q peptide captured 26.2 ± 6.4% apo enNTS1-muGFP (Figure 3A). Repeating the pulldown in the presence of a saturating NT8–13 agonist increased the enNTS1-muGFP capture efficiency to 36.4 ± 6.6%, whereas the SR142948 antagonist had no significant effect. Performing the experiment with enNTS1[Q301C^BTFMA] showed no appreciable differences from enNTS1, indicating that the TM6 ¹⁹F-BTFMA label does not influence Gα_q peptide interaction (Figure 3A). Assuming a quadratic binding model, our results suggest an enNTS1[Q301C^BTFMA]/Gα_q peptide complex K_d ~ 225 μM, or a higher basal-state association as reflected in the TGFα shedding assay (Figure 1A).

Figure 3.

Gα_q peptide and βArr1[ΔCT] stabilize distinct enNTS1 substates. (A) Binding of the Gα_q peptide to enNTS1 and enNTS1[Q301C^BTFMA] was measured using a Dynabead sequestration assay.³⁰ Ligands were incubated at 10 Meq and Gα_q-bound receptor was calculated as the ratio of input/bound receptor. Bars represent the average bound percentage from both experimental and instrumental triplicates; error bars represent the standard deviation. Statistical significance between conditions was calculated at P = 0.05 using a one-way ANOVA test; “ns” denotes no significance between two conditions as determined via calculated F-ratio at P = 0.05. (B) Deconvoluted ¹⁹F NMR spectra of enNTS1[Q301C^BTFMA] in the presence of 10 Meq NT8–13 and 5 Meq Gα_q peptide. Two Lorentzian lineshapes (S₂ and S₄) provided the best fit to the experimental data. (C) Overlay of NTS1 receptor from NT8–13/heterotrimeric G_i protein (PDB 6OS9; magenta) and NT8–13/βArr1 (PDB 6UP7; blue) complex structures. Transducer and agonist were removed for clarity. The distance between Q301 in the two structures is 0.8 Å and the all-atom rmsd = 0.68 Å as calculated using PYMOL. (D) Affinity (K_d) of NT8–13-bound enNTS1[Q301C^BTFMA] for βArr1[ΔCT] was measured using MST. Fluorescent NTA-labeled βArr1[ΔCT] (25 nM), NT8–13 (22.5 μM), and PIP₂ (10 Meq) were incubated with increasing concentrations of enNTS1[Q301C^BTFMA]. Data points represent the average normalized MST signal from data collected in both experimental and instrumental triplicates; error bars represent the standard deviation. Equilibrium dissociation constants were calculated from a global fit of experimental data using the quadratic binding model. (E) Deconvoluted ¹⁹F NMR spectra of enNTS1[Q301C^BTFMA] in the presence of 10 Meq NT8–13, 10 Meq PIP₂, and 5 Meq βArr1[ΔCT]. Two Lorentzian lineshapes (S₂ and S₃) provided the best fit to the experimental data.

The Gα_q peptide modified the NT8–13-bound enNTS1[Q301C^BTFMA] ¹⁹F NMR spectra by replacing the inactive S₁ substate with a unique peak at 13.26 ppm (S₄) alongside S₂ (Figures 3B, S9, and S10). Relative to NT8–13 alone, the formation of the Gα_q peptide ternary complex increases the S₂ population from 84 to 88%, broadens the linewidth to 57 Hz, and perturbs the chemical shift an additional 0.11 ppm upfield. If we assume the frequency difference between the two pure S₂ microstates ≲100 Hz (0.18 ppm), the exchange process is likely on the low millisecond timescale, although relaxation dispersion-type CPMG experiments are required for quantification.⁴⁶ The Gα_q peptide-induced S₄ substate is upfield of any ligand-only conformer, consistent with previous studies that agonist alone is unable to completely stabilize the fully active conformation.²¹ It is possible that the S₄ substate exists as a broad underlying resonance that is undetectable in the absence of Gα_q peptide. If so, we would anticipate observation in the CEST experiment, although S/N could be limited (Figure 2A). The S₄ resonance linewidth is 372 Hz and constitutes 12% of the observed populations (Figure 3B,E). It is also possible that the broad linewidth is a result of multiple overlapping resonances, but there is insufficient evidence to deconvolute additional conformers.

A similar distribution of TM6 G protein-bound conformers has also been observed for the adenosine A_2A receptor in complex with Gα_s peptide via ¹⁹F NMR.⁵¹ Once bound to the stimulatory G protein peptide and cognate agonist, TM6 populated two distinct conformers at the expense of all inactive substates. A concurrent population increase for the upfield-most chemical shift occurred, similar to S₄ for enNTS1[Q301C^BTFMA]. It is possible that for a majority of class-A GPCRs, complexation with G proteins induces microsecond to millisecond timescale chemical exchange of TM6, reflecting pre-coupling conformations prior to full receptor stimulation.

Arrestin Stabilizes Pre-existing Conformations.

Recent cryo-EM structures of the hNTS1/βArr1 and hNTS1/G_i proteins reveal a remarkably conserved receptor architecture with a 0.67 Å all-atom root-mean-square deviation (rmsd) (Figure 3C).^10,12 We next wanted to test if βArr1 modified the enNTS1 intracellular landscape similar to the Gα_q peptide. βArr1 recruitment is physiologically dependent on receptor phosphorylation, primarily on intracellular loop 3 (ICL3) and the C-terminus, but the number and location of sites necessary and sufficient to promote coupling is relatively unknown.⁶¹ To reduce system complexity and facilitate a high-affinity interaction, we employed a pre-activated human βArr1 variant truncated at N382 (herein βArr1[ΔCT]).³¹ Microscale thermophoresis (MST) was used to determine the apparent equilibrium dissociation constants (K_d) of enNTS1[Q301C^BTFMA]/βArr1[ΔCT] complexes. The N-terminal His-tag of βArr1[ΔCT] was site-specifically labeled with the RED-tris-NTA 2^nd generation (monolith) fluorescent dye. RED-βArr1[ΔCT] was then incubated with increasing enNTS1[Q301C^BTFMA] concentrations in the presence or absence of a saturating NT8–13 agonist. The interactions followed a sigmoidal dose–response and affinities were calculated using the quadratic binding model. Apo enNTS1[Q301C^BTFMA] bound RED-βArr1[ΔCT] with a K_d ≥ 506.2 ± 48 nM, consistent with the high affinity reported for pre-activated arrestin variants (Figure 3D).^10,62 The NT8–13 agonist increased affinity to 90.6 ± 5.8 nM, which is similar to the NTS1/Gα_iβγ ternary complex in phospholipid nanodiscs.¹⁴

Analogous to the Gα_q peptide, βArr1[ΔCT] abolished the inactive S₁ substate (Figures 3E, S9 and S10). However, rather than inducing a new substate resonance, βArr1[ΔCT] selectively restructured the existing conformational landscape by increasing the S₃ population to 28% and decreasing S₂ to 72% (Figure 3D). The linewidths of S₂ and S₃ increased to 237 and 376 Hz, respectively, suggesting additional contributions from microsecond to millisecond timescale chemical exchange (Figure 3E).

DISCUSSION

Provided that the distance between the TM4 and TM6 intracellular tips approximates transducer binding competency, Figure 2D illustrates that all NTS1 structures determined to date can be organized into three functional categories (inactive, active-intermediate, and active). Our spectroscopic results demonstrate that enNTS1 dynamically populates an ensemble of at least four conformers that are allosterically tuned by the orthosteric pocket. It is likely that some of these ¹⁹F-TM6 substates correspond directly to the static structures; future experiments could validate interhelical distances using electron paramagnetic resonance or fluorescence spectroscopy.

As shown for other class-A GPCRs, agonist binding does not stabilize a single enNTS1 active-like conformation but rather tunes the energetic landscape (Figure 4).²¹ Congruently, βArr1[ΔCT] selects from pre-existing active-like NTS1 substates (Figure 3D). As observed in published atomic models of NTS1/βArr1, steric clash between TM6 and the βArr1 finger-loop eliminates the inactive conformer.^10,13 The Gα_q peptide also reduces the inactive substate while inducing one active-like substate distinct from those observed in the presence of orthosteric ligands or βArr1 (Figure 3B). The Gα_i α5-helix peptide had a similar effect on the conformational ensemble of a related thermostabilized NTS1 variant.⁶³ Both transducer mimetics increased the active-intermediate and active-like linewidths by 130% relative to NT8–13. This line broadening likely reflects additional dynamics associated with encounter complex formation and the established inherent dynamics of both molecules.^{14,52,64–66}

Figure 4.

Model summarizing the effect of orthosteric ligands and transducers on each enNTS1 substate’s lifetime, population, and dynamics.

Although both transducers dock helical motifs into the receptor’s cytosolic core, the orientation of each segment is structurally unique; the inserted βArr1 finger-loop is 90° relative to TM6, whereas the G protein α5-helix is parallel.^10,12 These subtly distinct interfaces may promote structural fluctuations of TM6 that are not easily captured in static models. Together, this suggests that the enNTS1 allosteric activation mechanism may alternate between induced fit (Gα_q) and conformational selection (βArr1) depending on the coupled transducer. Employing a nucleotide-free heterotrimeric G protein and a phosphorylation-mediated βArr1 ternary complex, which represent “end point” conformations, should enable path delineation. It is important to note the inherent challenge of extrapolating observations at a single probe site to global conformational dynamics. Yet, the sensitivity of ¹⁹F NMR to distinct ternary complexes, despite highly similar architecture, suggests that varying the probe location around the helical bundle will provide unprecedented insight into the lowly populated states of the NTS1 ensemble.^10,12

An ensemble view in which transducers and ligands modulate the thermodynamic populations and exchange kinetics of enNTS1 substates provides a foundation for designing molecules that select discrete transducer pathways. The relatively recent recognition of this so-called biased signaling, in which ligands preferentially activate either the G protein or arrestin pathway, offers a new mechanism for reducing drug side effects.^67–69 There are several promising biased agents in preclinical and clinical trials; most notably, the opioid receptor G protein-biased ligand oliceridine (TRV130), which was approved for pain management in August 2020.^70,71 A class of βArr1-biased agents has also been developed that target NTS1 and attenuate methamphetamine and cocaine abuse while limiting G protein-mediated on-target side effects.⁷² Our results suggest that ¹⁹F NMR may serve as a powerful discovery platform to delineate biased agonists and biased allosteric modulators.

An important limitation of our model system is the employment of a thermostabilized receptor and pre-activated transducer mimetics. Although commonly employed to stabilize a single conformational state for structure determination, these mutations rigidify receptor motions and alter specific inter-residue and receptor/solvent interactions.^73,74 Nonetheless, enNTS1 responds appropriately, although with reduced efficacy, to ligands and allosteric modulators in functional assays. Given that mutations are more commonly loss-of-function rather than gain-of-function, we hypothesize that thermostabilization does not result in an entirely distinct activation landscape. Although more experiments are required to further the dynamic NTS1 model presented here, this study illustrates the importance of orthogonal structural techniques in understanding the complete mechanism of GPCR activation.