Stabilization versus flexibility: Detergent‐dependent trade‐offs in neurotensin receptor 1 GPCR ensembles

Abstract

Detergents provide essential membrane‐mimetic environments for studying G protein‐coupled receptors (GPCRs), but their molecular impact on receptor energetics remains incompletely understood. We combined ligand binding, thermostability measurements, and atomistic molecular dynamics to dissect detergent‐ versus ligand‐driven stabilization in a thermostabilized neurotensin receptor 1 (enNTS1). Circular dichroism and ligand binding assays revealed that apo enNTS1 becomes progressively more stable in decyl maltoside (DM), dodecyl maltoside (DDM), and lauryl maltose neopentyl glycol (LMNG). However, this gain in baseline stability was accompanied by an initially counterintuitive observation: LMNG, the most stabilizing detergent, supported the weakest neurotensin agonist binding affinity. Thermodynamic analysis shows that this behavior arises naturally from partitioning stability between detergent‐driven conformational rigidity (ΔG _conf) and ligand‐induced stabilization (ΔG _ligand). In DM, ΔG _ligand contributions were large, consistent with the receptor's engineered background. In contrast, LMNG maximized ΔG _conf, constraining conformational flexibility and reducing ΔG _ligand. Molecular dynamics simulations corroborated these results, showing that LMNG formed denser, less mobile detergent shells around the receptor, enhancing protein–detergent interaction energies while limiting conformational flexibility. Redistribution of ligand contacts, particularly at neurotensin residue Y11, further underscored detergent‐dependent modulation of the binding pocket. The results in this thermostabilized neurotensin receptor illustrate a fundamental trade‐off: LMNG provides exceptional receptor stabilization, supporting structural studies, but may mask conformational states relevant to signaling. In contrast, less rigid detergents preserve ligand‐induced transitions at the expense of stability. We therefore propose this system as a case study of how detergent chemistry can redistribute stability between conformational rigidity and ligand‐induced effects, with implications for guiding detergent choice depending on whether the goal is structural resolution or dynamic characterization.

1. INTRODUCTION

G protein‐coupled receptors (GPCRs) constitute the largest family of plasma membrane receptors in eukaryotes, functioning as versatile sensors of hormones, neurotransmitters, peptides, and metabolites (Rosenbaum et al., 2009; Venkatakrishnan et al., ²⁰¹³). Their pharmacological importance is underscored by the fact that ~34% of approved drugs target GPCRs, spanning indications from cardiovascular to neurological disorders (Hauser et al., 2017; Zalewska et al., ²⁰¹⁴). Despite this therapeutic relevance, high‐resolution structures exist for fewer than 200 unique receptors out of the more than 800 encoded in the human genome (Isberg et al., 2016; Yang et al., ²⁰²¹). A major challenge remains the purification and stabilization of GPCRs in membrane‐mimetic environments that maintain native conformational equilibria (Tate, 2010).

Although nanodiscs, bicelles, and liposomes have expanded the membrane‐mimetic toolkit (Majeed et al., 2021; Ratkeviciute et al., ²⁰²¹), detergent micelles remain indispensable for initial solubilization and purification. Nonionic maltoside detergents such as n‐decyl‐β‐D‐maltoside (DM) and n‐dodecyl‐β‐D‐maltoside (DDM) have been widely used owing to their relatively mild properties (Seddon et al., 2004). More recently, branched maltose‐neopentylglycol (MNG) detergents were introduced to enhance receptor stability and crystallization success (Chae et al., 2010). Among these, lauryl maltose neopentyl glycol (LMNG) has emerged as a workhorse, used to solve more than half of all GPCR structures (Lee et al., 2022), and is widely regarded as a “gold standard” detergent that simultaneously enhances receptor stability and supports high‐affinity ligand binding in structural and biophysical studies. LMNG improves thermostability in multiple receptors, including adenosine A2A receptor (A2AR), β2‐adrenergic receptor (β2AR), opioid receptors, and muscarinic receptors (Caffrey & Cherezov, 2009; Granier et al., ²⁰¹²; Haga et al., ²⁰¹²; Kruse et al., ²⁰¹²; Kruse et al., ²⁰¹³; Manglik et al., ²⁰¹²; Miller & Aricescu, ²⁰¹⁴; Rasmussen, Choi, et al., ²⁰¹¹; Ring et al., ²⁰¹³; Rosenbaum et al., ²⁰¹¹; White et al., ²⁰¹²). LMNG's branched design yields tighter micelle packing, reduced detergent mobility, and greater thermal stabilization (Lee et al., 2016; Lee et al., ²⁰²⁰). However, the energetic consequences of this stabilization are incompletely understood, particularly regarding how detergents redistribute the balance between conformational rigidity and ligand‐induced effects.

Ligand binding is known to enhance GPCR stability by increasing global rigidity (Zhang et al., 2015), yet this contribution (ΔG _ligand) depends on the baseline conformational ensemble established by the membrane‐mimetic environment. A detergent that rigidifies the apo state may maximize conformational stability (ΔG _conf) but simultaneously reduce the conformational freedom available for ligand‐induced stabilization. Such trade‐offs have direct implications for receptor activation, as conformational shifts between apo and ligand‐bound states underlie GPCR signaling (Bumbak et al., 2020; Cong et al., ²⁰¹⁸).

Here, we address this question by dissecting detergent versus ligand contributions in enNTS1, a thermostabilized variant of the rat neurotensin receptor 1. enNTS1 was originally selected for stability and ligand binding in DM detergent (Scott & Pluckthun, 2013), so the trends we observe across DM, DDM, and LMNG should be interpreted in the context of this engineered background rather than as a direct survey of wild‐type GPCR behavior. Using CD spectroscopy, ligand‐binding thermostability assays, and atomistic molecular dynamics, we show that LMNG maximizes ΔG _conf at the expense of ΔG _ligand. This apparently counterintuitive combination—highest baseline stability but weakest agonist affinity—is naturally explained within a near‐additive free energy framework by LMNG's ability to restrict conformational heterogeneity and remodel ligand contact profiles (notably at NT(8–13) residue Y11) (Asadollahi et al., 2023; Bumbak et al., ²⁰²⁰; Cong et al., ²⁰¹⁸). This conceptual thermodynamic partitioning framework may provide a generalizable lens for understanding how membrane mimetics alter GPCR energetics, ligand recognition, and activation pathways (Chung et al., 2012; Rasmussen, DeVree, et al., ²⁰¹¹; Thomas et al., ¹⁹⁹⁹).

2. RESULTS

2.1. Detergent environments differentially influence agonist binding affinity of enNTS1

Detergents are well known to influence GPCR stability, yet their effect on ligand binding affinity has received far less attention. To directly compare micelle environments, we measured the affinity of NT(8–13), the minimal activating peptide of neurotensin, for enNTS1 solubilized in DM, DDM, or LMNG micelles using biolayer interferometry (BLI). NT(8–13) was immobilized via an N‐terminal hexaHis tag on the BLI probe, and binding was measured against increasing receptor concentrations. As illustrated in Figure 1, steady‐state analysis yielded K _d values of 13.7 ± 3.3 nM (DM), 35.9 ± 5.1 nM (DDM), and 94.9 ± 26 nM (LMNG). The highest affinity in DM is expected, given that the enNTS1 construct was thermostabilized for improved NT(8–13) binding in this detergent environment (Scott & Pluckthun, 2013). More surprising, however, is that LMNG, which is predicted to provide greater overall receptor stability than DDM, supports weaker binding affinity. This observation suggests that detergent environments shape not only overall receptor thermostability but also the conformational ensembles that determine ligand binding competence. To explore this possibility, we next examined the thermostability of apo and ligand‐bound enNTS1 across the same detergent series.

FIGURE 1.

Detergent‐dependent affinity and thermostability of enNTS1. (a) NT(8–13) was immobilized through an N‐terminal hexaHis tag on the biolayer interferometry (BLI) probe, and binding was measured against increasing receptor concentrations solubilized in decyl maltoside (DM, top), dodecyl maltoside (DDM, middle), or lauryl maltose neopentyl glycol (LMNG, bottom) micelles. Equilibrium dissociation constants (K _d ) were derived by fitting steady‐state kinetics to a quadratic binding model. 95% confidence bands are illustrated as dashed lines above/below the fitted curve. (b) Far‐UV CD spectra were collected at increased temperatures for 5 μM enNTS1 in DM (top), DDM (middle), and LMNG (bottom). Apo and NT(8–13)‐bound enNTS1 are colored darker and lighter, respectively. The absorption at 222 nm was normalized and fit with a six‐parameter Boltzmann equation to yield apparent secondary structure melting temperature (^2° T _m ) and slope (Dubois et al., 2009). (c) Apparent tertiary structure melting temperature (^3° T _m ) curves monitor the ability of enNTS1 to bind Alexa Fluor 647 dye‐labeled NT(8–13) at increasing temperatures. Measurements were collected with 5 μM enNTS1 in DM (top), DDM (middle), and LMNG (bottom) in both apo (darker color) and NT(8–13)‐bound (lighter color) states. The data were normalized and fit with a six‐parameter Boltzmann equation to yield ^3° T _m and slope.

2.2. Detergent environments differentially stabilize apo and ligand‐bound enNTS1

Circular dichroism (CD) spectroscopy was used to assess how detergents alter the thermostability of enNTS1 secondary structure. Far‐UV CD reports on α‐helical content through minima at ~210 and 222 nm. Monitoring ellipticity at 222 nm as a function of temperature provided melting curves from which we determined the apparent secondary structure melting temperature (^2° T _m ). Measurements were performed in DM, DDM, and LMNG micelles for both apo enNTS1 and the agonist‐bound state with NT(8–13). Apo enNTS1 displayed the lowest stability in DM micelles (^2° T _m  = 50.4 ± 0.07°C), which increased to 71.6 ± 0.2°C in DDM and 82.4 ± 0.1°C in LMNG (Figure 1b). NT(8–13) binding increased the ^2° T _m in all detergents, yielding values of 76.4 ± 0.06°C, 91.2 ± 0.2°C, and 97.6 ± 0.2°C for DM, DDM, and LMNG, respectively (Figure 1b). Ligand‐induced stability is not unusual within the GPCR superfamily (Zhang et al., 2015) and it was shown previously that NT(8–13) binding increases the global rigidity of DDM reconstituted enNTS1 (Bumbak et al., 2023). Because enNTS1 was mutationally thermostabilized for high‐affinity NT(8–13) binding in DM, the large change in thermostability between apo and holo conditions (Δ^2° T _m ) observed in DM (26.0°C) is expected. More revealing is the trend across detergents: Δ^2° T _m decreased to 19.6°C in DDM and 15.2°C in LMNG, despite LMNG conferring the highest overall stability. Melting slopes further emphasized this pattern. In DM, the slope steepened substantially upon ligand binding (apo = 3.9°C, holo = 2.0°C), consistent with enhanced unfolding cooperativity. DDM showed a similar but less pronounced effect (apo = 4.4°C, holo = 2.7°C). In contrast, LMNG exhibited steep slopes that changed little with ligand addition (apo = 2.1°C, holo = 1.6°C), suggesting that detergent‐driven rigidity minimized ligand‐induced contributions.

The apparent tertiary structure melting temperature (^3° T _m ) was evaluated by measuring loss of NT(8–13) binding as a function of temperature using Alexa Fluor 647‐labeled peptide (Figure 1c). Apo enNTS1 was again least stable in DM micelles (^3° T _m  = 35.9 ± 0.2°C), increasing to 57.7 ± 0.4°C in DDM and 74.4 ± 0.4°C in LMNG. NT(8–13) binding increased ^3° T _m in all detergents, with values of 50.8 ± 0.2°C (DM), 64.9 ± 0.2°C (DDM), and 76.5 ± 0.3°C (LMNG). As expected, the enNTS1 tertiary structure is less stable than the individual helices (Vogt & Argos, 1997); yet, the relative increases in ligand‐mediated stabilization mirrored the secondary structure results: Δ^3° T _m  = 14.9°C in DM, 7.2°C in DDM, and only 2.1°C in LMNG. Analysis of tertiary unfolding slopes reinforced these findings. In DM, the slope decreased from 3.4°C (apo) to 1.5°C (holo), and in DDM from 3.3°C (apo) to 1.6°C (holo), indicating that ligand binding sharpened unfolding cooperativity in both environments. In LMNG, however, slopes were nearly unchanged (apo = 2.6°C, holo = 2.5°C), again pointing to detergent‐imposed rigidity.

Plotting secondary versus tertiary structure melting temperatures revealed a strong correlation for both apo (Slope = 1.18, R ² = 0.99) and holo (Slope = 1.17, R ² = 0.97) states (Figure S1), consistent with a shared structural basis for unfolding across detergents. Taken together, these results indicate that the large ligand‐induced stabilization in DM reflects the engineered background of enNTS1, while the progressively smaller T _m and slope changes in DDM and LMNG emphasize how detergent environments shift the balance of stability contributions. In LMNG, where detergent interactions (ΔG _conf) dominate, ligand contributions (ΔG _ligand) are minimized, consistent with largely separable free‐energy contributions. To provide a simple numerical estimate of these contributions, we treated the apo tertiary melt temperature (^3° T _m ) as a proxy for the detergent‐driven component and the ligand‐induced change in tertiary stability (Δ^3° T _m ) as a proxy for the ligand‐driven component. Normalizing to the DM values (i.e., set to 1.0), the relative detergent‐driven contributions (ΔG _conf,rel) are approximately 1.0, 1.6, and 2.1 for DM, DDM, and LMNG, respectively (using apo ^3° T _m values of 35.9, 57.7, and 74.4°C), whereas the corresponding ligand‐induced contributions (ΔG _ligand,rel) are 1.0, 0.5, and 0.1 (using Δ^3° T _m values of 14.9, 7.2, and 2.1°C). These dimensionless ratios are not absolute free energies, but they summarize the experimental trends and make explicit that stabilization shifts from a more balanced partitioning between detergent and ligand in DM to a predominantly detergent‐driven component in LMNG, with only a small residual ligand‐induced contribution.

2.3. Molecular dynamics models of enNTS1 in detergent micelles

To investigate the mechanisms of enNTS1 stability in different detergents, we performed all atom molecular dynamics (MD) simulations of the receptor in DM, DDM, and LMNG detergents (Table 1). Initial models of the apo and NT(8–13)‐bound states were built upon crystal structures of a similar, thermostabilized rat NTS1 (PDB:6Z66 and PDB:6YVR, respectively) (Deluigi et al., 2021). Intracellular loop 3 (ICL3), which was truncated in the crystal constructs, was reintroduced, while the C‐terminal DARPin fusion was removed. Helix 8 was modeled from another thermostabilized rat NTS1 structure (PDB: 4BWB) (Egloff et al., 2014). Finally, the primary sequence of these homology models was modified to match the receptor construct used in the thermostability experiments. Receptor–detergent complexes were generated using the CHARMM‐GUI Micelle Builder (Cheng et al., 2013; Jo et al., ²⁰⁰⁸). Each of the six systems (apo and NT(8–13)‐bound states in DM, DDM, and LMNG) was simulated in five independent 1 μs trajectories with randomized initial velocities.

TABLE 1.

Molecular dynamics (MD) parameters for enNTS1 in different detergents.

Simulation	Base structure (PDB)	Helix 8 modeling (PDB)	Detergent molecules	Water molecules	NaCl (mM)
DM Apo	6Z66	4BWB	192	35,445	150
DDM Apo	6Z66	4BWB	192	35,206	150
LMNG Apo	6Z66	4BWB	88	49,957	150
DM NT(8–13)	6YVR	4BWB	192	39,586	150
DDM NT(8–13)	6YVR	4BWB	192	39,364	150
LMNG NT(8–13)	6YVR	4BWB	88	54,745	150

Note: Starting conditions for the CHARMM‐based enNTS1 models in decyl maltoside (DM), dodecyl maltoside (DDM), and lauryl maltose neopentyl glycol (LMNG) detergent micelles.

Abbreviation: PDB, Protein Data Bank.

Structural stability was assessed by RMSD (root‐mean‐square deviation) from the starting structure (Figure S2). Percent helicity was then plotted as a function of TM backbone RMSD from the initial structure (Figure S3). Apo models maintained ~2.5%–5% higher helicity than their starting crystal structures, with fluctuations most pronounced in TM5. In the apo state, LMNG‐solubilized enNTS1 showed a slightly lower average TM backbone RMSD compared to DM and DDM, but with a broader distribution, suggesting greater conformational heterogeneity. NT(8–13) binding reduced helicity, consistent with crystal structures, and decreased average TM backbone RMSD across all detergent conditions. Backbone root‐mean‐square fluctuations (RMSF) mapped onto representative structures revealed similar dynamic regions across detergents, with apo receptors particularly flexible in the extracellular region of TM7 and the intracellular portions of TM5/6 (Figure S3). Ligand binding suppressed extracellular fluctuations around the orthosteric pocket, indicating that agonist engagement quenches local flexibility regardless of detergent environment.

2.4. Molecular dynamics reveal detergent‐dependent energetic partitioning in enNTS1

To assess how detergent environments shape receptor energetics, we calculated total receptor–detergent interaction energies at 50, 200, 400, and 1000 ns for each MD trajectory. These values included Lennard‐Jones and Coulomb nonbonded energies between the enNTS1 transmembrane backbone and surrounding detergent molecules (Figure S4). In the apo state, total interaction energies increased progressively from DM to DDM to LMNG. At 1000 ns, the average interaction energy rose by ~30 kJ/mol from DM to DDM and by an additional ~38 kJ/mol from DDM to LMNG (Figure S4a). This trend parallels the experimental thermostability data, consistent with LMNG forming the most stabilizing detergent–protein interactions. In the NT(8–13)‐bound state, DM again had the lowest interaction energy, but DDM exceeded LMNG (Figure S4b). Relative to apo, the increase in interaction energy was ~48 kJ/mol in DM, ~83 kJ/mol in DDM, and only ~28 kJ/mol in LMNG. Thus, the apo‐to‐holo energetic shift was largest in DDM and smallest in LMNG.

This outcome mirrors the thermostability measurements: ΔT _m values were greatest in DM, intermediate in DDM, and minimal in LMNG. Importantly, the large ΔT _m in DM is expected given that the enNTS1 construct was thermostabilized based on NT(8–13) binding in DM, and this optimization likely accentuates the magnitude of ligand‐induced stabilization we observe in DM relative to DDM and LMNG. Plotting the calculated total internal energies against the experimentally determined thermostability values produces linear correlations with R ² = 0.88 and 0.65 for secondary and tertiary structure thermostabilities, respectively (Figure 2). Interestingly, removing the outlying DDM NT(8–13) condition increases the correlation to an R ² value of 0.97 for the CD and 0.85 for the functional thermostability data; this datapoint was not removed as there is no justification for its elimination. This trade‐off emerges most clearly when comparing LMNG and DDM: LMNG provides greater baseline stability, yet ligand‐induced energetic stabilization is smaller. In terms of the thermodynamic framework, LMNG favors detergent‐driven stabilization (ΔG _conf), while DDM permits a larger contribution from ligand binding (ΔG _ligand). We emphasize that we did not perform explicit free‐energy calculations (e.g., FEP, TI, or enhanced‐sampling methods); ΔG _conf and ΔG _ligand are used here as qualitative descriptors of detergent‐ and ligand‐associated contributions to stability inferred from experimental thermostability and relative interaction energies.

FIGURE 2.

Detergent‐dependent interaction energies reveal energetic partitioning of enNTS1 stability. Plots of the molecular dynamics (MD) total interaction energy against (a) circular dichroism and (b) functional thermostability measurements yields linear correlations. enNTS1 in decyl maltoside (DM, blue), dodecyl maltoside (DDM, orange), and lauryl maltose neopentyl glycol (LMNG, purple) detergent conditions are plotted for both the apo (open shape) and NT(8–13)‐bound (filled shape) states. The number of interhelical (c) hydrogen bonds and (d) van der Waals forces for apo and NT(8–13)‐bound enNTS1 in DM, DDM, and LMNG. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

To further dissect these effects, we enumerated persistent interhelical hydrogen bonds and van der Waals (vdW) contacts within the enNTS1 transmembrane region (>40% contact frequency; Figure 2c,d). In apo simulations, hydrogen bonds increased modestly from DM (321) to DDM (325) to LMNG (330). Upon NT(8–13) binding, bond counts rose to 340 in DM and 354 in both DDM and LMNG. For vdW contacts, apo enNTS1 exhibited 939 (DM), 938 (DDM), and 969 (LMNG). With ligand bound, these increased to 993 (DM), 1002 (DDM), but only 979 (LMNG). Together, these data suggest that ligand‐induced stabilization arises primarily from vdW contacts in DM and DDM, but is dominated by hydrogen bond formation in LMNG. Thus, detergent environments not only modulate receptor baseline stability but also reshape the mechanism of ligand‐induced stabilization. In LMNG, extensive detergent packing reduces the conformational flexibility available for ligand binding to stabilize, leading to smaller ΔG _ligand contributions compared to DM or DDM.

2.5. Micelle shape reflects detergent headgroup chemistry

To examine how detergent chemistry influences the global geometry of enNTS1 proteomicelles, we analyzed micelle eccentricity from the most populated conformation (i.e., the largest cluster from cluster analysis performed for the whole proteomicelle) of our MD simulation (Figure 3a,b) (Lipfert et al., 2007; Oliver et al., ²⁰¹³). The receptor–detergent complexes adopted spheroidal shapes whose eccentricity varied with detergent composition. In the apo state, LMNG micelles were less oblate than DM or DDM micelles (Figure 3c). This is consistent with micelle packing models showing that larger headgroups disfavor oblate shapes (Dupuy et al., 1996; Iyer & Blankschtein, ²⁰¹²). LMNG's branched headgroup likely introduces electrostatic repulsion that promotes a more spherical geometry. Upon NT(8–13) binding, LMNG proteomicelles shifted toward more oblate conformations, resembling DM and DDM, and displayed a broader population distribution (Figure 3d). These results suggest that ligand binding perturbs micelle geometry most strongly in LMNG, potentially reflecting rearrangements in the transmembrane helical bundle that challenge detergent packing. Thus, micelle shape analysis highlights how detergent headgroup chemistry sets the baseline organization of the proteomicelle, while ligand binding induces compensatory geometric changes.

FIGURE 3.

Detergent headgroup chemistry shapes enNTS1 proteomicelles and ligand‐induced geometry changes. Representative structures of (a) apo and (b) NT(8–13)‐bound enNTS1–detergent proteomicelles from molecular dynamics (MD) simulations. Calculated eccentricity of (c) apo and (d) NT(8–13)‐bound enNTS1 in decyl maltoside (DM, blue), dodecyl maltoside (DDM, orange), and lauryl maltose neopentyl glycol (LMNG, purple). Eccentricity is defined as the ratio of the short axis over the long axis (a/b). Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

2.6. Detergent packing dynamics reveal a trade‐off between baseline stability and ligand‐induced effects

To evaluate how detergent molecules interact dynamically with the receptor, we examined packing behavior in MD simulations. RMSF measurements showed that LMNG molecules were less flexible than DM or DDM, indicating tighter packing in both apo and NT(8–13)‐bound states (Figure S5). Spatial distribution analysis of single detergent molecules further illustrated this effect: DM and DDM diffused more freely, whereas LMNG molecules remained comparatively fixed over time (Figure 4). Radial distribution function (RDF) analysis quantified detergent density within 2 nm of the receptor core (Figure 5a). LMNG exhibited the highest headgroup and tail densities in both apo and holo states, with ~60% (head) and ~39% (tail) greater density than DDM in the apo state and ~51%E (head) and ~39% (tail) increases in the holo state. DM consistently displayed the lowest densities. These results indicate that LMNG forms a more compact and continuous detergent shell around the receptor.

FIGURE 4.

Detergent mobility differs across micelles with lauryl maltose neopentyl glycol (LMNG) exhibiting reduced dynamics. The spatial distribution function for single (a, d) decyl maltoside (DM), (b, e) dodecyl maltoside (DDM), and (c, f) LMNG molecules are illustrated on representative (a–c) apo and (d–f) holo enNTS1 models. Initial positions of the detergent molecules are shown as stick representations. Spatial distributions are shown as blue (carbon atoms) and red (oxygen atoms) dots. Analysis was performed on the 5 μs (N = 5 × 1 μs) molecular dynamics (MD) simulation trajectories.

FIGURE 5.

Detergent packing density and hydrophobic mismatch vary between decyl maltoside (DM), dodecyl maltoside (DDM), and lauryl maltose neopentyl glycol (LMNG) micelles. (a) Plot of the total areas under the radial distribution function (RDF) curves within 2 nm of the receptor core for both the head (H) and tail (T) regions of DM (blue), DDM (orange), and LMNG (purple). The open bars represent the apo state and the filled bars represent the NT(8–13)‐bound state. (b) Residual hydrophobic mismatch displaying total surface area of the TM regions exposed to solvent for enNTS1 in DM (blue square), DDM (orange circle), and LMNG (purple triangle) both apo and NT(8–13)‐bound. Analysis was performed on the 5 μs (N = 5 × 1 μs) molecular dynamics (MD) simulation trajectories. Analysis was done with two‐way ANOVA with multiple comparison using Tukey's test. Nonsignificant (ns) p > 0.05, *p ≤ 0.05, ***p ≤ 0.001, ****p ≤ 0.0001.

Residual hydrophobic mismatch, which occurs when the hydrophobic thickness of the membrane mimetic does not match the hydrophobic thickness of the membrane protein, supported this interpretation. When taking the non‐persistent contacts with less than 20% frequency, DM and DDM exhibited ~3–4 Ų greater solvent‐exposed transmembrane surface area than LMNG, consistent with looser packing (Figure 5b). NT(8–13) binding increased mismatch across all three detergents, but the effect was most pronounced in LMNG, where tighter apo packing was disrupted by ligand‐induced conformational changes. Together, these findings demonstrate that LMNG provides superior hydrophobic coverage and the most rigid packing of enNTS1 in the apo state. However, ligand binding destabilizes this optimized packing, leading to smaller relative gains in stability compared to DM or DDM. In thermodynamic terms, LMNG maximizes detergent‐driven stabilization (ΔG _conf) while constraining additional ligand‐induced stabilization (ΔG _ligand).

2.7. Detergent environments redistribute NT(8–13) contacts, with Y11 as a sensitive reporter

MD simulations of agonist‐bound enNTS1 in DM, DDM, and LMNG provided insight into how detergent environments reshape ligand dynamics. Analysis of NT(8–13) backbone RMSD revealed overall stability of ~0.1–0.2 nm across all detergents, but LMNG trajectories showed intermittent fluctuations up to 0.3 nm, including early instability (<200 ns) and later spikes at ~300, 550, and 700 ns (Figure S7). These results suggest that LMNG alters the conformational ensemble of the bound peptide relative to DM and DDM. Residue‐level contact frequency analysis highlighted detergent‐dependent differences (Figure S8a). Most strikingly, NT(8–13) residue Y11 exhibited the largest variance in contacts across detergents (Figure S8b), with more modest differences observed for R9 and I12. By contrast, P10 maintained consistent contact frequencies regardless of detergent environment. R9 displayed a slight preference for DM and DDM over LMNG, I12 favored Y347 in LMNG and DM over DDM, and L13 showed enhanced contact with Y351 in LMNG.

The Y11 contact profile was particularly sensitive to detergent chemistry (Figure 6). DM and DDM yielded broadly similar patterns (Figure 6a,d): DM‐preferred contacts were located in the N‐terminus and ECL1, while DDM‐preferred contacts clustered in ECL2. LMNG, however, produced a distinct shift in Y11 interactions. Contacts with L55 (N‐terminus), H132 (ECL1), and H348 (helix 7) were reduced, while interactions with S214 in ECL2 became strongly favored (Figure 6b,e,f). Together, these results demonstrate that detergent environments remodel the binding microenvironment of NT(8–13), with Y11 acting as a sensitive reporter of these changes. The similarity of DM and DDM underscores their comparable packing properties, whereas LMNG produces a distinct redistribution of contacts, consistent with its denser detergent shell and reduced conformational heterogeneity. Functionally, this implies that LMNG stabilization (ΔG _conf) constrains the range of ligand‐induced conformational adjustments (ΔG _ligand), altering how key residues such as Y11 engage the receptor.

FIGURE 6.

Detergent‐dependent redistribution of NT(8–13) Y11 contacts highlights altered binding pocket interactions. Plots of Y11 contact frequency with enNTS1 residues 55, 132, 213, 214, 224, 226, and 348 in (a) decyl maltoside (DM) versus dodecyl maltoside (DDM), (b) DM versus lauryl maltose neopentyl glycol (LMNG), and (c) DDM versus LMNG. Plotted deltas of Y11 contact frequencies with enNTS1 residues displaying calculated differences between (d) DM and DDM, (e) DM and LMNG, and (f) DDM and LMNG. Analysis was performed on the 5 μs (N = 5 × 1 μs) molecular dynamics (MD) simulation trajectories.

3. DISCUSSION

Detergents remain indispensable for GPCR structural biology, but their molecular influence extends beyond solubilization to shaping receptor energetics. Our work interprets the experimental and simulation data in terms of two thermodynamic components: detergent‐driven conformational rigidity (ΔG _conf) and ligand‐induced stabilization (ΔG _ligand). These quantities are introduced as a phenomenological decomposition of stability rather than as directly computed absolute free energies. This framework rationalizes the otherwise counterintuitive finding that LMNG, while conferring the greatest baseline stability to enNTS1, simultaneously supports the weakest agonist binding affinity, as a natural consequence of strong detergent‐induced stabilization leaving less scope for additional ligand‐induced gains. Given that LMNG is widely used as the default detergent in GPCR structural work based on its reputation for simultaneously enhancing receptor stability and preserving high‐affinity ligand binding, recognizing this trade‐off is important when selecting detergents for studies that aim to characterize activation energetics and conformational landscapes.

In DM, the detergent environment permits broad conformational heterogeneity of the apo receptor, which ligand binding resolves into a more rigid, cooperative state. The large ΔT _m and energetic shifts observed in DM thus reflect substantial ΔG _ligand contributions, consistent with the receptor's engineered background in this detergent (Sarkar et al., 2008; Shibata et al., ²⁰⁰⁹). This redistribution is evident in simple relative metrics derived from the tertiary melting data: using apo ^3° T _m and Δ^3° T _m as proxies, the detergent‐driven and ligand‐driven components (ΔG _conf,rel/ΔG _ligand,rel) are approximately 1.0/1.0 in DM, 1.6/0.5 in DDM, and 2.1/0.1 in LMNG, making explicit the shift toward a predominantly detergent‐driven contribution in LMNG. DDM provides an intermediate case, where detergent interactions contribute more strongly to stability but still allow significant ligand‐induced gains. In contrast, LMNG maximizes ΔG _conf as MD simulations revealed dense detergent packing, reduced detergent mobility, and minimized hydrophobic mismatch. As a result, the apo ensemble is rigidified, and ligand binding produces only modest additional stabilization. This thermodynamic redistribution may have functional consequences for receptor signaling. Analysis of NT(8–13) dynamics showed that residue Y11, a known reporter of receptor activation (Asadollahi et al., 2023; Bumbak et al., ²⁰²⁰; Cong et al., ²⁰¹⁸), displayed detergent‐specific contact patterns. In DM and DDM, Y11 engaged the N‐terminus and ECL1, while in LMNG it preferentially contacted ECL2. Such detergent‐dependent remodeling of the binding pocket is consistent with the hypothesis that LMNG stabilization may constrain ligand‐induced conformational adjustments, narrowing the energetic landscape available for activation.

Our findings echo broader observations that detergents and membrane mimetics can bias GPCR ensembles (Chung et al., 2012; Rasmussen, DeVree, et al., ²⁰¹¹; Thomas et al., ¹⁹⁹⁹; Zhang et al., ²⁰¹⁵). While LMNG excels at stabilizing receptors for crystallography and cryo‐EM, our data lead us to propose that it may underrepresent conformational states important for signaling or pharmacology. Conversely, less rigid detergents may better preserve ligand‐induced transitions but at the cost of reduced overall stability. This balance highlights the need for careful detergent selection depending on whether the experimental goal is to maximize structural resolution or to capture conformational dynamics. We note, however, that we did not measure G protein or arrestin coupling in this study, so these implications for signaling should be viewed as testable hypotheses rather than demonstrated outcomes.

By dissecting detergent‐ versus ligand‐driven stabilization, our study reframes detergent effects as an energetic trade‐off rather than a uniform stabilizing influence in this engineered neurotensin receptor. We anticipate that the same conceptual framework will be relevant for many GPCRs and other membrane proteins, but its generality remains to be tested systematically across additional receptors and membrane mimetics. Ultimately, understanding how ΔG _conf and ΔG _ligand are partitioned will be critical for aligning biophysical characterization with the functional roles of membrane proteins in their native environments. Our conclusions are nonetheless constrained by the scope of the present work. All experiments and simulations were performed on a single thermostabilized GPCR construct, enNTS1, which was originally selected for enhanced stability and ligand binding in DM detergent. This engineered background likely contributes to the particularly large ligand‐induced stabilization observed in DM and may influence the relative behavior we observe in DDM and LMNG. As such, our study should be viewed as an instructive case study illustrating how detergent chemistry can redistribute stability between ΔG _conf and ΔG _ligand, rather than as a demonstration that all GPCRs will exhibit the same quantitative stabilization–flexibility trade‐off.

To our knowledge, there are currently no structural studies directly comparing a given GPCR construct in different membrane mimetics. A recent cryo‐EM study of the ABC transporter MsbA provides a rare side‐by‐side comparison of a membrane protein in 12 different membrane mimetics (Hoffmann et al., 2025), revealing pronounced environment‐dependent conformational bias. ABC transporters undergo large‐scale structural rearrangements between outward‐facing (OF) and inward‐facing (IF) conformers to transport substrates through the lipid bilayer. In this study, detergent micelles generally favored a wide inward‐facing (IF_wide) conformation, whereas nanodiscs and peptidiscs predominantly stabilized a narrow inward‐facing (IF_narrow) state; only LMNG, GDN, and the largest MSP2N2 nanodiscs supported both IF_wide and IF_narrow conformations. A double electron–electron resonance (DEER) spectroscopy study showed that apo MsbA indeed samples both IF_narrow and IF_wide in a native membrane environment (Galazzo et al., 2022), suggesting close resemblance to the conformational distribution observed for GDN, LMNG, and MSP2N2. Despite these differences, ATPase activity was uniformly high across all nanodisc preparations and uniformly lower in detergents, peptidisc, and amphipol, indicating that function does not map in a simple way onto a single conformation (Hoffmann et al., 2025). Together with our results, this example highlights that membrane mimetics can funnel proteins into distinct subsets of their accessible conformational landscape. In the context of GPCRs, LMNG‐stabilized structures used for crystallography or cryo‐EM are therefore likely to emphasize LMNG‐favored, rigidified states and may underrepresent more flexible or ligand‐responsive conformations that would be populated in native lipids or large nanodiscs. Our energetic partitioning framework (ΔG _conf + ΔG _ligand) should extend to such lipidic environments, but with ΔG _conf dominated by protein‐lipid and scaffold interactions rather than detergent packing, and with the balance between baseline rigidity and ligand‐induced stabilization depending sensitively on the chosen membrane mimic.

4. METHODS

4.1. Reagents

The construct used for expression of enNTS1 in Escherichia coli BL21(DE3) cells was generated from a previously published plasmid (Egloff et al., 2014). Thermo Scientific™ HisPur™ Cobalt Resin used in the IMAC purification step was purchased from Fisher Scientific (#89964). All pre‐packed affinity resin columns used in the purification of enNTS1 (IMAC, IEX, SEC) were purchased from GE (#45‐000‐192, #17‐5247‐01, #29‐0452‐69). The “MyOne Streptavidin T1” Dynabeads used for functional thermostability assays were purchased from Thermo Fisher (#65601). The fluorescent NT(8–13)^AF647 was made in house using the dye AF647 from AAT Bioquest (#1833).

4.2. enNTS1 plasmid construct and protein expression

The previously characterized functional variant enNTS1 (Bumbak et al., 2018) was used in an expression vector (pDS170) containing an open reading frame encoding an N‐terminal maltose‐binding protein signal sequence (MBPss), followed by a 10x‐Histidine tag, a maltose‐binding protein (MBP), a NNNNNNNNNNG linker and a HRV 3C protease site (LEVLFQGP) which were linked via a BamHI restriction site (resulting in additional residues GS) to residue T42 of the receptor. At the C‐terminus, T416 of the receptor was linked via a NheI restriction site (resulting in additional AS residues) to an Avi‐tag allowing biotinylation, another HRV 3C protease site, a GGSGGS linker, a monomeric ultra‐stable green fluorescent protein (muGFP) (Scott et al., 2018), and an additional 10×‐Histidine tag. The enNTS1 plasmid was transformed into BL21(DE3) E. coli cells and plated on LB agar supplemented with 100 μg/mL carbenicillin and 1% (w/v) glucose at 37°C overnight. Liquid LB media starter cultures were supplemented with 100 μg/mL carbenicillin and 1% (w/v) glucose then inoculated with several colonies and incubated overnight at 37°C and 220 RPM. Six liters of 2xYT media were each supplemented with 100 μg/mL carbenicillin and 0.2% (w/v) glucose, inoculated with 10 mL/L of overnight LB starter culture and incubated at 37°C and 220 RPM to an OD600 ≅ 0.3. The cultures were then cooled to 16°C. Once each culture reached an OD600 ≅ 0.7, they were induced with 0.3 mM IPTG and incubated for ~18 h at 16°C and 220 RPM. The cultures were harvested via centrifugation at 5000 rcf and cell pellets stored at −80°C. The final amino acid sequence for the purified construct was:

GPGSTSESDTAGPNSDLDVNTDIYSKVLVTAIYLALFVVGTVGNGVTLFTLARKKSLQSLQSRVDYYLGSLALSSLLILLFALPVDVYNFIWVHHPWAFGDAGCKGYYFLREACTYATALNVVSLSVERYLAICHPFKAKTLMSRSRTKKFISAIWLASALLSLPMLFTMGLQNLSGDGTHPGGLVCTPIVDTATLRVVIQLNTFMSFLFPMLVASILNTVIARRLTVMVHQAAEQARVSTVGTHNGLEHSTFNMTIEPGRVQALRRGVLVLRAVVIAFVVCWLPYHVRRLMFVYISDEQWTTALFDFYHYFYMLSNALVYVSAAINPILYNLVSANFRQVFLSTLASLSPGWRHRRKKRPTFSRKPNSMSSNHAFSTASGLNDIFEAQKIEWHEGSGLEVLFQ.

4.3. enNTS1 protein purification

Cell pellets were solubilized in a solubilization buffer (100 mM HEPES, 500 mM NaCl, 20% [v/v] glycerol, 10 mM MgCl₂, 10 mM imidazole, pH 8.0) supplemented with 100 mg lysozyme, 1 unit DNAse, 0.2 mM PMSF, and one Roche cOmplete EDTA‐free protease inhibitor cocktail tablet. The solution was stirred on ice for 30 min before being subjected to sonication on ice: 3 min processing time (10 s on, 20 s off) at 35% maximum amplitude. The receptor was then solubilized in detergent with the addition of final concentrations of 1% (w/v) DM detergent, 0.12% (w/v) (CHS), and 0.6% (w/v) CHAPS. The solution was incubated at 4°C for 2 h with stirring. The solution was then centrifuged at 24,424 RCF for 45 min. The supernatant containing detergent‐solubilized enNTS1 was then collected and incubated with TALON resin equilibrated with equilibration buffer (25 mM HEPES, 10% [v/v] glycerol, 400 mM NaCl, 0.15% [w/v] DM, pH 8.00) at 4°C for 45 min. Following TALON incubation, the receptor solution was placed into a gravity column to allow flow through of unbound protein. The TALON resin was then washed with TALON wash #1 (25 mM HEPES, 10% [v/v] glycerol, 500 mM NaCl, 0.15% [w/v] DM, 10 mM Imidazole, 4 mM ATP, 10 mM MgCl₂, pH 8.0) followed by TALON wash #2 (25 mM HEPES, 10% [v/v] glycerol, 350 mM NaCl, 0.05% [w/v] DDM, 10 mM Imidazole, pH 8.0). This second wash step also served as a detergent exchange step from DM to DDM. Following detergent exchange in TALON wash #2, enNTS1 was eluted with TALON elution buffer (25 mM HEPES, 10% [v/v] glycerol, 500 mM NaCl, 0.05% [w/v] DDM, 350 mM Imidazole, pH 8.0) and incubated with 3 mg of HRV 3C precision protease for 16 h or overnight at 4°C to cleave MBP and muGFP expression tags. The cleaved enNTS1 was concentrated in a 30 kDa MWCO concentrator via centrifugation at 3800 RCF and then diluted 10‐fold in SP equilibration buffer (20 mM HEPES, 10% [v/v] glycerol, 0.05% [w/v] DDM, pH 7.4). This dilution was then split into equal thirds with one batch remaining in DDM, one batch exchanging back into DM, and one batch exchanging into LMNG. Each batch was loaded onto an equilibrated 5 mL SP cation‐exchange (CEX) column via a GE AKTA Pure system run at 4 mL/min flow rate. The SP CEX column was washed with SP wash buffer (20 mM HEPES, 10% [v/v] glycerol, 300 mM NaCl, 0.03% [w/v] DDM, pH 7.4) until the A280 reading stabilized. The DM batch was then subjected to 30 mL of DM exchange buffer (20 mM HEPES, 10% [v/v] glycerol, 100 mM NaCl, 1% [w/v] DM, pH 7.4) at 1 mL/min flow rate. The LMNG batch was treated the same way with a LMNG exchange buffer (20 mM HEPES, 10% [v/v] glycerol, 100 mM NaCl, 0.1% [w/v] LMNG, pH 7.4). For elution an equilibrated 1 mL Ni2+‐NTA column was attached in‐tandem after the 5 mL SP CEX column, and the receptor batches were eluted with the appropriate SP elution buffer (20 mM HEPES, 10% [v/v] glycerol, 1 M NaCl, [0.03% [w/v] DDM] or [0.15% (w/v) DM] or [0.01% (w/v) LMNG], 15 mM Imidazole, pH 7.4). The enNTS1 elutions were each concentrated in a separate 30 kDa MWCO concentrator via centrifugation at 3800 RCF and individually injected onto a GE S200 Increase SEC column equilibrated in the appropriate SEC buffer (20 mM HEPES, 150 mM NaCl, [0.03% (w/v) DDM] or [0.15% (w/v) DM] or [0.01% (w/v) LMNG], pH 7.4). Following SEC, the desired enNTS1 fractions for each run were pooled, concentrated to 100–300 μM, and flash‐frozen via liquid nitrogen and stored at −80°C.

4.4. Biolayer interferometry

BLI measurements were performed using a Gator Prime 8‐channel system (GatorBio) in 96‐well plate format. All experiments were carried out at ambient temperature. Ni‐NTA biosensors were pre‐hydrated for 20 min in assay buffer (20 mM HEPES, 300 mM NaCl, [0.03% (w/v) DDM], [0.15% (w/v) DM], or [0.00025% (w/v) LMNG]). Binding was measured by dipping ligand‐coated probes into a dilution series of enNTS1 concentrations. For DDM, NT(8–13) was loaded at 50 nM onto Ni‐NTA sensors, and enNTS1 ranged from 0 to 50 nM. For DM, NT(8–13) was loaded at 5 nM, and enNTS1 concentrations ranged from 0 to 200 nM. For LMNG, NT(8–13) was loaded at 5 nM, and enNTS1 concentrations ranged from 0 to 120 nM.

BLI traces were collected by acquiring a baseline in assay buffer (200 s), followed by ligand immobilization (3600 s). A second baseline was acquired post‐immobilization (200 s), followed by the association phase by dipping the sensor into wells containing enNTS1 (1500 s). Dissociation was monitored by transferring the sensor back into buffer‐only wells (1500 s). Double referencing was used to minimize background noise. Each protein concentration was paired with a buffer‐only reference probe, and its signal was subtracted from its corresponding sample. In all cases, a reference well was included where the ligand‐coated sensor was dipped into buffer alone during the association step. Raw data were processed by subtracting the reference well and buffer‐only reference probe. The binding curves were analyzed using the quadratic equation that accounts for ligand depletion, single‐site binding, and non‐specific binding in GraphPad Prism (Version 10.4.0).

4.5. Circular dichroism

For CD measurements, purified enNTS1 was diluted to a final concentration of 5 μM in a CD buffer (20 mM HEPES, 150 mM NaCl, [0.03% (w/v) DDM] or [0.3% (w/v) DM] or [0.003% (w/v) LMNG]). For agonist‐bound enNTS1, a final concentration of 50 μM NT(8–13) peptide was added to the CD buffer. Temperature melting curves were collected on a Jasco J‐715 Spectropolarimeter using a 1 mm quartz cuvette, monitoring at 222 nm wavelength, a temperature range of 5 to 100°C, a temperature interval of 0.2°C, and a response time of 1 s. Data were collected for enNTS1 in each detergent (DDM, DM, and LMNG) in both apo and NT(8–13) bound states. All data were collected in triplicate. The data were normalized between the highest temperature and the lowest temperature points and fit to a 6‐parameter Boltzmann equation consisting of a low‐temperature asymptote, high‐temperature asymptote, low‐temperature slope, high‐temperature slope, T _m , and T _m slope. For the LMNG holo condition, the high temperature asymptote was constrained to 0.07 based upon fitted values for the other five conditions.

4.6. Functional thermostability assay

Functional thermostability of enNTS1 was tested by measuring the receptor's ability to bind a fluorescent NT(8–13)^AF647 ligand over a thermal gradient. The enNTS1 construct was diluted to 25 nM in reaction buffer (20 mM HEPES, 150 mM NaCl, [0.03% (w/v) DDM] or [0.3% (w/v) DM] or [0.003% (w/v) LMNG]). Each detergent condition, as apo and NT(8–13) bound consisted of 36 samples which were then subjected to temperatures ranging from 22 to 95°C (2°C intervals). For the NT(8–13) bound samples, 250 nM of fluorescent NT(8–13)^AF647 was added to each sample prior to the thermal melt. After the thermal melt, 250 nM fluorescent NT(8–13)^AF647 was added to both the apo and holo samples. The enNTS1 construct possesses a C‐terminal Avi‐tag which was used to immobilize the purified receptor on magnetic Dynabeads with a Streptavidin tag. 3.5 μL of 10 mg/mL Dynabead stock pre‐equilibrated with reaction buffer was added to the samples after the thermal melt and a Promega (#V8351) 96‐well magnetic side strip plate was used to capture the Dynabead‐enNTS1 complex for at least 5 min. Unbound fluorescent peptide was washed away with reaction buffer and the remaining Dynabead‐enNTS1‐NT(8–13)^AF647 complex was taken to a plate reader (BioTek Synergy Neo2) to measure final fluorescence intensity. All data were collected in triplicate. The data were normalized between the highest and lowest temperature points and fit to a 6‐parameter Boltzmann equation consisting of an upper asymptote, lower asymptote, upper slope, lower slope, T _m , and T _m slope. The upper and lower slopes were constrained to 0. For the semi‐quantitative partitioning analysis in the Results and Discussion, apo ^3° T _m and Δ^3° T _m were used as simple proxies for ΔG _conf and ΔG _ligand, respectively, and normalized to the DM values to yield dimensionless relative contributions; these quantities are intended only as indicators of the experimental trends and not as absolute free energies.

4.7. Molecular dynamics simulations

MD simulations were performed by utilizing the GROMACS package (version 2021/2022) with the Chemistry HARvard Molecular Mechanics (CHARMM) 36 force field for proteins, n‐Decyl‐β‐d‐Maltopyranoside (DM), n‐Dodecyl‐β‐D‐Maltopyranoside (DDMs) or Lauryl maltose neopentyl glycol (LMNG) detergent molecules, Na + CL− ions, and using CHARMM Transferable Intermolecular Potential with 3 Points (TIP3P) water as solvent. The enNTS1 models were generated from the ligand‐free state (Protein Data Bank [PDB] ID: 6Z66) or the NT(8–13)‐bound rat NTS1 (PDB ID: 6YVR) crystal structures. First, the DARPin crystallization chaperone was removed from both initial crystal structures after which Helix 8 was transposed by aligning the template structures with the NT(8–13)‐bound NTS1 mutant without Lysozyme (PDB ID: 4BWB) using Maestro (Schrödinger). Next, the sequences of the generated homology models were adapted to mimic the sequence of the experimental enNTS1 setup, and ICL3 was generated using Prime and further optimized with the refine loops module in Maestro. Lastly, missing side chains and hydrogen atoms were added, protein chain termini were capped with neutral acetyl and methyl amide groups, and histidine‐protonated states were assigned. We created the simulation box using the CHARMM–Graphical User Interface (CHARMM‐GUI) and positioned the receptor in the micelle using the PPM2.0 web server of the OPM database (orientation of proteins in membranes) using the structure inputs of PDB IDs: 6Z66 and 6YVR for alignment of the TM helices of the protein structure and inserted a pre‐equilibrated (B)DM, (B)DDM or (B)LMNG micelle. Final system dimensions of the ligand‐free and NT(8–13)‐bound enNTS1 (B)DM micelle were respectively: 110 Ȧ by 110 Ȧ by 110 Ȧ, including 192 (B)DM detergent molecules, 35,445 water molecules, 150 mM NaCl; and 113 Ȧ by 113 Ȧ by 113 Ȧ, including 192 (B)DM detergent molecules, 39,586 water molecules, 150 mM NaCl. Final system dimensions of the ligand‐free and NT(8–13)‐bound enNTS1 (B)DDM micelle were: 110 Ȧ by 110 Ȧ by 110 Ȧ, including 192 (B)DDM detergent molecules, 35,206 water molecules, 150 mM NaCl; and 113 Ȧ by 113 Ȧ by 113 Ȧ, including 192 (B)DDM detergent molecules, 39,364 water molecules, 150 mM NaCl. Final system dimensions of the ligand‐free and NT(8–13)‐bound enNTS1 (B)LMNG micelle were respectively: 122 Ȧ by 122 Ȧ by 122 Ȧ, including 88 (B)LMNG detergent molecules, 49,957 water molecules, 150 mM NaCl; and 125 Ȧ by 125 Ȧ by 125 Ȧ, including 88 (B)LMNG detergent molecules, 54,745 water molecules, 150 mM NaCl. Next, we minimized all six systems, and equilibrated them using a 1 ns long NVT (constant temperature, constant volume) ensemble and consequently with an NPT (constant temperature, constant pressure) ensemble where we gradually reduced the position restraints from 5 to 0 kcal/mol/Ȧ² with each step being 5 ns long. Since Helix 8 was transposed from a different crystal structure, we performed an extra 50 ns 1 kcal/mol/Ȧ² restraint NPT ensemble on the whole protein except for Helix 8, allowing it to adjust to its new orientation. The final part of equilibration involved a 200 ns long unrestrained NPT simulation before running a total of five independent production MD simulations, each assigned with random velocities and a 1000 ns long. Snapshots were captured every 20 ps, and the entire 1000 ns × 5 runs amounting to 5000 ns of simulation time was used for analysis. We visualized all trajectories using PyMOL (Molecular Graphics System, version 2.0, Schrödinger) and Visual Molecular Dynamics (VMD). The trajectories were analyzed using the GROMACS package (version 2021/2022). The convergence of the MD simulations was confirmed by calculating the change in root‐mean‐square deviation of the backbone atom coordinates of the residues in the TM region (Apo: TM1 61–91; TM2 100–130; TM3 139–172; TM4 183–206; TM5 231–271; TM6 299–333; TM7 341–370; NT(8–13)‐bound: TM1 61–91; TM2 97–130; TM3 139–172; TM4 183–207; TM5 231–271; TM6 301–333; TM7 341–373) and ligand. All data were analyzed using GraphPad Prism 10 (GraphPad Software, San Diego, California, USA).

4.8. Contact analysis of NT(8–13) with enNTS1

To establish contacts between NT(8–13) and enNTS1 in DM, DDM, and LMNG, we used Get_contacts (https://getcontacts.github.io). Contact frequencies were calculated from the aggregated trajectories, and a contact frequency cut‐off of 40% was applied (i.e., counting for at least two MD simulation runs). The frequencies were displayed in a contact heatmap using GraphPad Prism 10 (GraphPad Software, San Diego, California, USA).

4.9. Interaction energy of TM backbone of enNTS1

Interaction energies were calculated on the TM backbone of enNTS1 using the gmx energy function of the GROMACS package. Both Lennard‐Jones short‐range and Coulomb short‐range energies were calculated on all nonbonded interactions between atoms and summed as the total interaction energies. These interaction energies are not binding or folding free energies, and no free‐energy perturbation, thermodynamic integration, or enhanced‐sampling calculations were performed. Instead, we use these interaction energies as relative energetic indicators within our qualitative ΔG _conf/ΔG _ligand partitioning framework.

4.10. Radial distribution function of detergent molecules

For analyses of how DM, DDM, and LMNG molecules organize around the enNTS1, we calculated the radial distribution function for the head and tail of the detergent molecules from the center of mass of the receptor using the gmx rdf function of the GROMACS package.

4.11. Spatial distribution function of detergent molecules

For the localization of DM, DDM, and LMNG molecules around the enNTS1 during the MD simulations, we captured the position of one detergent molecule at the beginning and end of each MD run (i.e., 0 and 1000 ns) and these were displayed as a mesh representation around enNTS1 using PyMOL. The initial position of the detergent molecules was shown as a stick representation.

4.12. Eccentricity of micelle

The eccentricity of the ellipse of the DM, DDM, and LMNG detergents around enNTS1 is defined as the ratio of the long axes to the short axes and was calculated using the gmx gyrate function of the GROMACS package with the option ‐moi.

4.13. Root‐mean‐square fluctuation of detergent molecules and TM backbone of enNTS1

The RMSF of the TM‐backbone atoms of enNTS1, as well as the detergent molecules DM, DDM, and LMNG, was calculated using the gmx rmsf function in GROMACS. To render the extent of flexibility on the enNTS1 structure as a heat map, we converted the RMSF values to thermal B‐factor using loadbfacts in PYMOL.

4.14. Root‐mean‐square deviation versus α‐helicity of TM backbone

The average percent α‐helicity of TM backbone residues and RMSD correlation of enNTS1 was obtained by first calculating the α‐helicity of every residue in the TM region using the helicity.tcl VMD script, which uses STRIDE to identify if a residue is in helical conformation, and subsequently gmx rms on all TM‐backbone residues (see above). The distributions were plotted using Python packages numpy, matplotlib, and seaborn.

4.15. Calculation of hydrophobic mismatch surface area

For calculating unfavorable hydrophobic interactions between enNTS1 and DM, DDM, or LMNG, we first identified the hydrophobic residues in the TM that do not make persistent contacts (i.e., <20% of the total simulation snapshots) with the detergents'tail or head groups using Get_contacts (https://getcontacts.github.io). The hydrophobic mismatch surface area between receptor and micelle was obtained by taking the sum of the solvent accessible surface area of TM hydrophobic residues (glycine, alanine, proline, valine, methionine, cysteine, isoleucine, tryptophan, phenylalanine, tyrosine, and leucine) that show less than 20% persistence contact with DM, DDM, or LMNG (20). The solvent accessible surface area per residue was calculated using the gmx sasa function of the GROMACS package.

AUTHOR CONTRIBUTIONS

James B. Bower: Investigation; writing – original draft; writing – review and editing; formal analysis; methodology. Wijnand J. C. van der Velden: Investigation; writing – original draft; writing – review and editing; methodology; formal analysis. Karen P. Gomez: Investigation; formal analysis. Mingzhe Pan: Formal analysis; investigation. Fabian Bumbak: Conceptualization; writing – review and editing; writing – original draft; investigation; formal analysis; methodology. Nagarajan Vaidehi: Conceptualization; funding acquisition; writing – review and editing; methodology; supervision; formal analysis; writing – original draft; resources. Joshua J. Ziarek: Conceptualization; funding acquisition; writing – original draft; writing – review and editing; formal analysis; methodology; supervision; resources.

Supporting information

Figure S1. Correlation between secondary and tertiary thermostability highlights detergent‐ and ligand‐dependent stabilization. Correlation plot comparing CD thermostability (^2° T _m ) and functional thermostability (^3° T _m ) measurements for enNTS1 in DM (blue square), DDM (orange circle), and LMNG (purple triangle) detergent conditions. Linear fits were applied to both the apo (open shapes, dashed line) and NT(8–13)‐bound (filled shapes, solid line) states. Linear fit equation and R² values are displayed next to each fit. Error bars are smaller than the shapes.

Figure S2. Detergent environments differentially stabilize enNTS1 transmembrane helices. TM backbone RMSD values plotted as a function of simulation time for apo enNTS1 model in (a) DM, (b) DDM, and (c) LMNG; and the NT(8–13)‐bound enNTS1 model in (d) DM, (e) DDM, and (f) LMNG. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S3. Conformational heterogeneity of enNTS1 ensembles reveals detergent‐ and ligand‐dependent differences. Distributions of the average percent helicity of TM region residues and RMSD correlation for each MD snapshot of enNTS1 in DM (a) apo and (b) NT(8–13)‐bound, DDM (c) apo and (d) NT(8–13)‐bound, and LMNG (e) apo and (f) NT(8–13)‐bound. The dashed lines in each plot show the average helicity of the crystal structure used for apo (6Z66) and NT(8–13) (6YVR) enNTS1. Representative structures for each condition are shown with a RMSF heat map of the TM residues. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S4. Detergent‐dependent receptor–micelle interaction energies mirror experimental stability trends. Plots of MD calculated total interaction energies for enNTS1 in DM (blue square), DDM (red circle), and LMNG (green triangle) across the simulation trajectory at 50, 200, 400, and the full 1000 ns. Calculations were done in both the (a) apo and (b) NT(8–13)‐bound states. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S5. Detergent mobility differs across micelles, with LMNG forming the most rigid shell. RMSF of each detergent molecule for DM (blue), DDM (orange), and LMNG (purple) around (a) apo and (b) NT(8–13)‐bound enNTS1. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S6. Radial distribution functions show denser LMNG packing around enNTS1 compared to DM and DDM. Radial distribution function (RDF) plots displaying the density for either the tail group (red) or head group (blue) of the detergents as a function of distance from apo enNTS1 in (a) DM, (b) DDM, and (c) LMNG; and NT(8–13)‐bound enNTS1 in (d) DM, (e) DDM, and (f) LMNG. The gray shaded region represents the first 2 nm from enNTS1. The dashed line is set at an RDF value of 6 for clarity. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S7. Ligand backbone dynamics differ by detergent, with LMNG showing intermittent fluctuations. RMSD values plotted as a function of simulation time for NT(8–13) bound to the enNTS1 model in (a) DM, (b) DDM, and (c) LMNG. Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Figure S8. Detergent environments redistribute NT(8–13) contacts, with Y11 as a sensitive reporter of binding pocket changes. (a) Contact frequency measurements for NT(8–13) residues with enNTS1 residues in DM, DDM, and LMNG. (b) Comparison of individual NT(8–13) contact frequencies with enNTS1 residues in DM (blue square), DDM (orange circle), and LMNG (purple triangle). Analysis was performed on the 5 μs (N = 5 × 1 μs) MD simulation trajectories.

Bower JB, van der Velden WJC, Gomez KP, Pan M, Bumbak F, Vaidehi N, et al. Stabilization versus flexibility: Detergent‐dependent trade‐offs in neurotensin receptor 1 GPCR ensembles. Protein Science. 2026;35(2):e70475. 10.1002/pro.70475

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.