Introduction

Post-translational modifications (PTMs) expand the functional and chemical diversity of neuropeptides, many of which rely on such modifications for efficient activation of G protein-coupled receptors (GPCRs) (1–4), thereby regulating diverse physiological processes (5–13). Canonically, PTMs are thought to modulate neuropeptide activity through direct ligand-receptor interactions (14–19), enhancing receptor affinity or signaling efficacy. However, the mechanistic basis by which certain PTMs, such as the N-terminal pyroglutamylation (conversion to pyroglutamic acid, pQ), can exert subtype-selective modulation without directly engaging receptor residues remain unclear. Here, we investigate how this N-terminal PTM may reconfigure ligand-receptor interaction dynamics via distal conformational control.

pQ introduces a rigid five-membered lactam ring and is prevalent among neuropeptides (20–23). It is best known for conferring resistance to aminopeptidase degradation, as observed in GnRH, TRH and Tachykinin (22, 24, 25). Beyond this stabilization role, pQ has been proposed to influence receptor signaling through structural effects rather than forming direct ligand-receptor contacts (20, 26–31). For example, pQ may stabilize preferred ligand configurations (26, 28), potentially through restriction of conformational flexibility as demonstrated in solution-based studies (32, 33), or maintenance of an optimal binding orientation to enhance GPCR affinity, as inferred from studies where its substitution reversed binding pose (30, 34). Despite these proposals, no neuropeptide system involving pyroglutamylation has yet provided direct molecular-level evidence demonstrating how such conformational effects might dictate ligand-receptor interactions and signaling outcomes. In most cases, pQ either enhances receptor activation (23–26, 28) or has negligible impact upon removal (27, 29, 31), complicating efforts to disentangle its structural or mechanistic roles from stability-related effects.

This limitation is overcome by the Aplysia PRXamide model system. The PRXamide families (referred to as the Neuromedin U (NmU) family in vertebrates), defined by a C-terminal PRXamide motif, is conserved across protostomes and deuterostomes (35–39). Here, we described two newly identified Aplysia PRXamide receptors, ApPRXa-R1 and ApPRXa-R2, and discovered that they respond oppositely to the same pQ-modified ligand, MMG2-pDPb (pQPPLPRYamide) (40). ApPRXa-R2 follows the canonical pattern in which pQ enhances ligand potency, similar to many other pQ-containing peptides (24–26, 28) including Aplysia Tachykinins (23). In striking contrast, ApPRXa-R1 exhibits reduced responsiveness upon pQ modification. This bidirectional, subtype-specific effect of pQ is unprecedented in neuropeptide signaling. More importantly, such opposite regulation suggests that the outcome is determined not solely by the presence of pQ, but by receptor subtype-specific properties, such as distinct interaction networks and/or differences in binding-pocket environments.

In the Aplysia PRXamide system, an integrated strategy combining AI-assisted peptide-GPCR tools (41, 42) and residue mutagenesis reveals that N-terminal pQ remains solvent-exposed and does not directly contact receptor residues. Instead, its lactam ring constrains the peptide conformational ensemble, reweighting interaction strengths across a largely shared set of receptor contacts. Extending this molecular logic to mammals, canine endogenous pQ-NmU-7 (pQFLFRPRNamide) (43) biases activation of human NmU receptor subtypes through an analogous non-contacting mechanism. Both static pocket and molecular dynamics analyses based on recent high-resolution cryogenic electron microscopy (cryo-EM) structures of human NmU receptors (17, 18) further show that the direction of this distal modulation is determined by a key molecular variable that is largely overlooked in the prevalent “induced-fit” framework (44–48): the intrinsic properties of receptor orthosteric binding pocket. Specifically, the more permissive and looser hNmU-R1 pocket better accommodates the rigid pQ-NmU-7, whereas the more compact hNmU-R2 pocket preferentially favors the flexible non-pyroglutamylated NmU-8, which may also explain the bidirectional pQ regulation observed in the Aplysia PRXamide system.

Overall, our work reveals an indirect “PTM distal steering” mechanism in neuropeptide-GPCR signaling, where a non-contacting N-terminal lactam governs receptor activation through conformational biasing and pocket-dependent accommodation. This mechanism refines classical “lock-key” (49) and “induced-fit” (44) models by introducing an intermediate mode of molecular recognition. In this mode, ligand-borne conformational constraints are differently interpreted by receptor pockets with distinct permissiveness. More broadly, this work establishes pyroglutamylation as a versatile N-terminal chemical modification for tuning peptide-GPCR recognition and provides a rational design principle for optimizing stable, conformation-biased neuropeptide analogs.

Results

Establishment of an endogenous PRXamide system for probing N-terminal pyroglutamylation

To establish a native PRXamide system for probing the role of N-terminal pyroglutamylation (pQ), we first employed liquid chromatography-mass spectrometry (LC-MS/MS) (50, 51) to confirm the endogenous coexistence of MMG2-DPb and its pyroglutamylated counterpart MMG2-pDPb in Aplysia central ganglia (Figs. S1, S2 and Table S1). In parallel, bioinformatics and phylogenetic analyses identified four candidate PRXamide/Neuromedin U-like Class A GPCRs as their potential cognate receptors (Figs. S3-S5 and Table S2). Broad functional screening against a panel of all known Aplysia PRXamide-related peptides, including MMG2-DPs, pleurins, SCPs and D-type pleurins, deorphanized ApPRXa-R1 and ApPRXa-R2 as cognate PRXamide receptors (Fig. 1A and Figs. S6-S8). D-type pleurins did not activate any of the candidate receptors under these conditions, and therefore were not tested further. Detailed identification and screening procedures are provided in the Supporting Results and Discussion.

Figure 1.

A single N-terminal lactam enables bidirectional GPCR subtype tuning

We next generated full concentration-response curves to quantify ligand potency at ApPRXa-R1 and ApPRXa-R2 (Fig. 1 C, D). ApPRXa-R1 showed broad responsiveness across ApPRXamide classes, with the highest potency toward MMG2-DPs (EC₅₀ = 0.21 nM) (Fig. 1 C, G). In contrast, ApPRXa-R2 displayed more selective activation, responding strongly only to MMG2-DPb and MMG2-pDPb, with EC₅₀ values of 260 nM and 64 nM, respectively (Fig. 1 D, G). Together, these data identify ApPRXa-R1 and ApPRXa-R2 as functional receptors for MMG2-DPs, with additional receptor analyses provided in the Supporting Results and Discussion and Figs. S9-S12. The two receptors also differed in maximal efficacy (E_max), with ApPRXa-R2 exhibiting higher E_max values than ApPRXa-R1 despite lower apparent potency (Fig. 1F), possibly reflecting difference in receptor regulation (52) or desensitization kinetics (23, 53–55).

Strikingly, comparison of MMG2-DPb with its pyroglutamylated counterpart MMG2-pDPb revealed opposite effects of the same N-terminal pQ. At ApPRXa-R2, pQ enhanced ligand potency, consistent with its commonly observed potentiating role (24–26, 28). At ApPRXa-R1, however, pQ reduced ligand potency relative to MMG2-DPb (Fig. 1 E, G). Thus, a single N-terminal lactam modification converts a minimal chemical difference into bidirectional GPCR subtype outputs. More importantly, such a bidirectional activation pattern indicates that pQ does not function as a universal stabilizing or affinity-enhancing group, but instead acts in a receptor-dependent manner, suggesting that the outcome must depend on subtype-specific interaction networks and possibly on how receptor-specific binding pockets accommodate the pQ-imposed ligand conformation.

Docking reveals a non-contacting role of N-terminal pQ in reweighting receptor-contact networks

To obtain the structural basis of this bidirectional regulation, we modeled ApPRXa-R1 and ApPRXa-R2, and docked MMG2-pDPb and MMG2-DPb into the receptor binding pockets using HPEPDOCK (56), with model quality evaluation and diffdock (57) cross-validation provided in Figs. S13-S17 and Supporting Results and Discussion.

In all four ligand-receptor complexes, the C-terminal PRXamide pharmacophore inserted into the predicted orthosteric pockets (58), whereas the N-terminal Q or pQ remained solvent-exposed. The overall conformations thus point to the C-terminal PRXamide motif as the principal active determinant involved in direct contact with the receptors, while the peptide’s extended, linear backbone effectively projects the N-terminus outside the pocket, so on purely positional grounds, the N-terminal pQ/Q is unlikely to form direct contacts with receptor residues.

Despite the absence of direct pQ/Q-receptor interactions, MMG2-pDPb and MMG2-DPb adopted distinct conformations within both ApPRXa-Rs (Fig. 2 A, B, Fig. 3 A, B and Fig. S18 A, B). Mapping of receptor-contact residues showed strong overlap with experimentally defined interaction networks in human homologous Neuromedin U receptors (17, 18), supporting the relevance of the modeled binding interface (Figs. S19, S20 and Table S3). Notably, MMG2-pDPb and MMG2-DPb engaged largely shared receptor-contact sites, but differed in the strength and geometry of interactions formed by central and C-terminal residues (Fig. 2 C, D, Fig. 3 C, D, and Table S4). These observations suggest that the N-terminal pQ may act as a distal conformational constraint that reweights activation-related interactions at core sites and ultimately affects receptor activity, providing an initial structural basis for its opposite effects on ApPRXa-R1 and ApPRXa-R2.

Figure 2.

Figure 3.

Receptor mutagenesis validates N-terminal pQ-dependent reweighting of interaction networks

To experimentally test this model, we performed site-directed mutagenesis of receptor residues predicted to interact with MMG2-DPb or MMG2-pDPb. Hydrophobic residues were generally substituted with glutamine (Q) to disrupt nonpolar contacts, whereas other residues were substituted with alanine (A). In ApPRXa-R1, mutants E229^2.61A, Y237^ECL1Q, L332^ECL2Q, N334^ECL2A, and W432^ECL3Q significantly impaired receptor activity, while Y308^4.60A abolished activation. In ApPRXa-R2, Y179^4.60A, Y186^ECL2Q, M203^ECL2Q, F210^ECL2Q, and T302^6.58A markedly reduced activity, whereas E100^2.61A led to complete inactivation (Table S4).

For ApPRXa-R1, docking predicted three ligand interaction regions with notable differences between MMG2-pDPb and MMG2-DPb (Fig. 2 E-G and Fig. 4). The most critical site was W432^ECL3, which engaged in hydrophobic interactions with P² of MMG2-pDPb but with L⁴ of MMG2-DPb (Fig. 2 G). Given that leucine is more hydrophobic than proline (59), W432^ECL3Q was expected to preferentially disrupt MMG2-DPb recognition. Consistent with this prediction, W432^ECL3Q caused a larger potency loss for MMG2-DPb, even eliminating the potency difference between MMG2-pDPb and MMG2-DPb (EC₅₀ = 45.5 and 36.1 nM, respectively) (Fig. 2 G, J, K). These data demonstrate that W432^ECL3-mediated hydrophobic packing is central to the N-terminal pQ-driven conformational disfavoring in receptor activation of ApPRXa-R1. Additionally, R⁶ of both ligands formed hydrogen bonds with T424^6.58 and salt bridges with E229^2.61. Docking suggested that the R⁶-E229^2.61 salt bridge was weaker for MMG2-pDPb than for MMG2-DPb (Fig. 2 E), making MMG2-pDPb more dependent on the T424^6.58-R⁶ hydrogen bond. In agreement with this model, T424^6.58A sharply decreased the potency of MMG2-pDPb (EC_50: from 4.79 to 204 nM), whereas the potency of MMG2-DPb remained largely unaffected (EC₅₀: from 0.43 to 0.79 nM) (Fig. 2 E, J, K). Furthermore, Y237^ECL1 formed hydrophobic contacts with P² of both ligands, with an additional CH-pi interaction unique to MMG2-DPb (Fig. 2 F), theoretically enhancing its potency. We anticipated a relatively smaller impact of Y237^ECL1Q on MMG2-pDPb compared to MMG2-DPb; however, Y237^ECL1Q significantly increased EC₅₀ values for both ligands (Fig. 2 F, J, K) without clear differentiation. These findings suggest that hydrophobic interactions mediated by Y237^ECL1 critically influence receptor activation for both ligands, regardless of the presence of the additional CH-pi interaction, providing valuable clues for subsequent ligand-focused studies (see the next section).

Figure 4.

For ApPRXa-R2, three interaction regions modulated by pQ were also identified (Fig. 3 E-G and Fig. 4). R⁶ of both ligands interacted with S103^2.64 and E100^2.61 via hydrogen bonds and salt bridges, with the salt bridge being stronger for MMG2-pDPb (Fig. 3 E). As expected, S103^2.64A slightly affected MMG2-pDPb potency (EC₅₀: from 119 to 178 nM), but significantly impaired MMG2-DPb activation (EC₅₀: from 334 to 1700 nM) (Fig. 3 E, J, K). Additionally, F210^ECL2formed a hydrophobic network with P³ and P⁵ and pi-pi stacking with Y⁷ in MMG2-pDPb, whereas its interaction with MMG2-DPb was limited mainly to P³ (Fig. 3 G). Despite predicting a more pronounced effect on MMG2-pDPb, F210^ECL2Q comparatively increased EC₅₀ for both ligands (Fig. 3 G, J, K), supporting that the conserved hydrophobic interactions involved in F210^ECL2 and P³ play an indispensable role in maintaining the receptor binding interface. Similarly, Y186^ECL2 formed dual hydrophobic and CH-pi contacts with P² of MMG2-pDPb but only hydrophobic contacts with MMG2-DPb (Fig. 3 F). Although Y186^ECL2Q had no selective effects on the two ligands (Fig. 3 F, J, K), the strong loss of potency for both ligands supports its role in stabilizing the ApPRXa-R2 binding pocket.

Beyond individual contacts, the two receptors differ in overall pocket architecture. In ApPRXa-R2, MMG2-pDPb is cradled within a hydrophobic pocket defined by Y186^ECL2, F210^ECL2 and M203^ECL2, which engaged multiple ligand positions, including P², P³, L⁴ and P⁵ (Fig. 3 C, Fig. S20 and Table S4). This distributed hydrophobic interface may provide a relatively loose and permissive environment for pQ-mediated potentiation. Consistently, mutations at S103^2.64, M203^ECL2 and T302^6.58 only moderately impair receptor activation (Fig. 3 J-K and Fig. S20 E). By contrast, in ApPRXa-R1, Y237^ECL1, L332^ECL2, N334^ECL2 and W432^ECL3 contact a narrow set of ligand positions, mainly P² and P³, generating a more compact and tightly optimized recognition network (Fig. 2 C, Fig. S20 and Table S4). In this compact space, individual mutations dramatically impaired receptor activation (Fig. 2 J-K and Fig. S20 E), indicating reduced tolerance to ligand conformational perturbation. Structural alignments further supported that pQ amplifies subtype-specific conformational divergence between ligand-bound states (Fig. S13 C and Fig. S18 C, D). These features also align with the broader ligand tolerance and generally lower EC₅₀ values of ApPRXa-R1 compared with ApPRXa-R2 when activated by either MMG2-pDPb or MMG2-DPb (Fig. 1 G and Fig. S21). Together, at the molecular level, pQ remodels the ligand conformation and redistributes the contribution of pre-existing receptor-contact networks (Figs. 2-4). We therefore propose that the distinct pocket properties of ApPRXa-R1 and ApPRXa-R2 may differentially interpret the pQ-imposed regulation, providing a plausible basis for the observed subtype-selective activation.

Cell surface receptor expression analyses of FLAG-tagged mutants confirmed comparable expression levels to wild-type receptors (Figs. S22-S24), indicating that EC₅₀ differences reflect binding alterations rather than trafficking variations. Moreover, the EC₅₀ values for FLAG-tagged receptors exhibited a similar trend to those of untagged receptors (Figs. S23, S24).

Activation of MMG2-pDPb analogs defines pQ as a distal conformational modulator

We tested a series of MMG2-pDPb analogs at both ApPRXa-R1 and ApPRXa-R2, and modeled their receptor-bound conformations to further define the ligand-side determinants of pQ-dependent subtype tuning. Replacing the N-terminal pQ with alanine generated [Ala¹]MMG2-pDPb, which resembled MMG2-DPb in both docking pose, recognition sites (Fig. 5 B, C and Table S5) and ligand potency (Fig. 5 D-F, J, K) rather than MMG2-pDPb. This further supports the idea that the pQ effects arise primarily from conformational modulation rather than direct receptor contacts. Next, we examined an acetylated analog, [Ac-Ala¹]MMG2-pDPb, to test whether N-terminal capping with charge neutralization alone could account for the pQ-dependent effects. This analog partially reproduced the inhibitory effect of MMG2-pDPb at ApPRXa-R1, but failed to fully restore MMG2-pDPb-like potency at ApPRXa-R2 (Fig. 5 D-F, J, K), indicating that the pQ lactam provides a more specific structural constraint for productive ApPRXa-R2 activation.

Figure 5.

For C-terminal amidation, its removal (MMG2-pDPb-OH) sharply increased EC₅₀ values, reaching 6.2 μM for ApPRXa-R1 and exceeding 10 μM for ApPRXa-R2, which could not be fully activated (Fig. 5 D-F, J, K). Structurally, amidation loss resulted in the disappearance of hydrogen bonds and affected the conformation of R⁶ and Y⁷ (Figs. S25, S26 and Table S5), highlighting its important role as a primary activation pharmacophore.

In addition to PTMs, alanine-scanning of residues 2-7 ([Ala²]-[Ala⁷]MMG2-pDPb) revealed a gradient of functional contributions (Fig. 5 G-K, Figs. S25, S26 and Table S5). For ApPRXa-R1 system, Ala² substitution merely shifted the hydrophobic interaction between Y237^ECL1 and P² to P³. Similarly, in ApPRXa-R2, A² maintained hydrophobic contact with Y186^ECL2. Potency of [Ala²]MMG2-pDPb was nearly identical to the native peptide in both receptors, indicating no critical interaction loss in either case. In contrast, [Ala³]MMG2-pDPb abolished the hydrophobic interaction between F210^ECL2 and P³ in ApPRXa-R2, leading to a marked reduction in receptor activation, while the interaction pattern and efficacy remained largely unchanged in ApPRXa-R1. This demonstrates the dual role of the hydrophobic “anchor” interactions in peptide binding, while secondary contacts modulate fine conformational tuning. Notably, [Ala⁶]MMG2-pDPb exhibited a complete loss of activity in both receptors. Structural modeling and mutagenesis data (Fig. 2 J, Fig. 3 J, Figs. S25, S26 and Table S5) indicated the absence of key hydrogen bonds and salt bridges originally formed with R⁶, particularly the salt bridges involving E229/E100^2.61, which proved essential for activation (EC₅₀-E229^2.61A: 9120; E100^2.61A: inactive). This likely reflects the strong electrostatic attraction between the guanidino group of arginine and the carboxyl group of glutamic acid (60), which is consistent with findings from other PRXamide/Neuromedin U structure-activity relationship studies (17, 18, 61). A general trend was observed whereby alanine substitutions closer to the C-terminal motif led to a progressive loss of potency (EC₅₀ fold change: 3.2-265 in ApPRXa-R1; 12-156 in ApPRXa-R2). Interestingly, [Ala⁴]MMG2-pDPb was less disruptive than [Ala⁵] MMG2-pDPb in ApPRXa-R2, possibly due to the boundary location of L⁴ and the conformational flexibility imparted by its side chain, suggesting its involvement in receptor selectivity. [Ala⁷] MMG2-pDPb partially activated ApPRXa-R2 (Emax ∼65% to wild-type), consistent with loss of T302^6.58 hydrogen bonding or destabilized C-terminal amidation interactions.

pQ-driven bidirectional steering logic extends to human NmU receptor subtypes

The subtype-selective regulation observed in the Aplysia PRXamide signaling (Fig. 1 E and Fig. 4) prompted us to ask whether an analogous distal steering logic might operate in mammalian NmU signaling. Although most mammalian NmUs lack an N-terminal Q, canine pQ-NmU-7 (pQFLFRPRNamide) (Fig. S2 D) uniquely carries an N-terminal pQ (43), raising the intriguing possibility that pQ may act as conserved molecular switch. To evaluate ligand-receptor interaction networks, we leveraged available cryo-EM structures of human receptor subtypes (hNmU-R1 and hNmU-R2) bound to NmU-8 (17, 18), and performed comparable docking analyses with NmU-8 and its two analogs: pQ-NmU-7 (Fig. S27) (43) and Q-NmU-7 (QFLFRPRNamide) (A similar modeling analysis of the insect RPCH signaling system with a pQ ligand is provided in Supporting Results and Discussion and Fig. S28).

Docking of NmU-8 recapitulated cryo-EM observations (17, 18), showing that N-terminal residue (Y) does not directly contact receptors (Figs. S29 A, S30 A), and that the identified interaction residues fully overlapped with experimentally resolved contacts (Fig. S31). Consistent with this, the N-terminal residue (pQ, or Q) in both pQ-NmU-7 and Q-NmU-7 also remains solvent-exposed and does not directly engage receptor residues (Figs. S29 B-C, S30 B-C).

Despite this lack of direct contact, pQ markedly alters interaction patterns. In hNmU-R1, the rigidity imparted by pQ (pQ-NmU-7) appears to optimize the side-chain positioning of L³, reinforcing hydrophobic packing against W320^ECL3 and F334^7.31, while R⁵ approached C219^45.50 to form a polar contact (Fig. S29 B, E, G). Conversely, in hNmU-R2, the same pQ constraint hinders the conformational rearrangements required to accommodate bulky side chains within a narrower binding pocket, causing a steric mismatch that diminishes the L³-centered hydrophobic network (involving A302^7.28 and F305^7.31) and disrupts polar contacts near R⁵ (with T203^ECL2 and T205^ECL2) (Fig. S30 B, E, G). Thus, our analyses support pQ as an indirect conformational switch, changing the number or strength of bonds formed by a largely shared set of interacting receptor residues to bias subtype activation—enhancing it in hNmU-R1 while attenuating it in hNmU-R2.

To experimentally validate these predictions, we synthesized NmU-8, pQ-NmU-7 and Q-NmU-7 (Fig. 6), and evaluated their potencies at hNmU-1 and hNmU-2. In hNmU-R1, pQ-NmU-7 significantly potentiated the receptor activation compared with both NmU-8 and Q-NmU-7, yielding approximately a 10-fold decrease in EC₅₀ (Fig. 6 A, C). In contrast, Q-NmU-7 had no obvious impact on hNmU-R1 activation (Fig. 6 A, C). Intriguingly, in hNmU-R2, the regulatory pattern differed. pQ-NmU-7 modestly attenuated receptor activation relative to NmU-8, increasing EC₅₀ values from 3.9 to 19.6 (Fig. 6 B, C). A similar weakening trend was observed compared to Q-NmU-7, though not statistically significant (Fig. 6 B, C). As observed for hNmU-R1, Q-NmU-7 did not significantly modulate hNmU-R2 activation (Fig. 6 A-C). Thus, the overall activation patterns are consistent with our prediction.

Figure 6.

Pocket permissiveness determines whether the pQ-constrained ligand is favored or penalized

The bidirectional regulation observed for canine pQ-NmU-7 further prompted us to examine whether receptor pocket properties could explain why the same N-terminal lactam constraint produced opposite effects at human NmU receptor subtypes. Previous structural study demonstrated that L³, F⁴ or R⁵ of NmU-8 are accommodated in a more extensive residue environment in hNmU-R1, whereas the corresponding recognition region in hNmU-R2 is more compact (17). This distinction suggested that the more conformationally rigid pQ-NmU-7 might be differentially accommodated by the binding pockets of the two receptors, thereby leading to opposite subtype outputs.

To test this hypothesis, we performed quantitative analyses using cryo-EM structures of both hNmU-R1 and hNmU-R2 (17). First, we analyzed the orthosteric pocket volumes of the ligand-bound active-state hNmU receptor structures. CASTp analysis (62) showed that the hNmU-R1 binding pocket had a larger molecular-surface (MS) volume than hNmU-R2 (2622.2 ų versus 2425.0 ų) (Fig. 6 D), consistent with a looser receptor environment for ligand recognition.

Second, we performed 300-ns all-atom molecular dynamics (MD) for NmU-8 bound to hNmU-R1 and hNmU-R2. Notably, radius of gyration (Rg) values were consistently higher for the hNmU-R1 complex than for the hNmU-R2 complex (Fig. 6 E), suggesting a less restricted and more open overall receptor-ligand architecture. Root-mean-square fluctuation (RMSF) analysis further showed higher residue fluctuations in hNmU-R1 than in hNmU-R2, particularly around residues 200-300, including major binding regions (Fig. 6 F). Parallel 300-ns MD simulations based on the pQ-NmU-7 docking models showed the same Rg and RMSF trends (Fig. 6 G, H), indicating that this pocket-property difference is maintained in both NmU-8-and pQ-NmU-7-bound complexes. Root-mean-square deviation (RMSD) trajectories and free-energy landscapes (FEL) indicated that both complexes remained structurally stable during the simulations (Fig. S32).

Together, these static and dynamic analyses indicated that hNmU-R1 provides a spacious and dynamically permissive pocket environment, whereas hNmU-R2 forms a more compact and restrictive architecture. This difference explains why the conformationally rigid pQ-NmU-7 is favored by hNmU-R1 but penalized by hNmU-R2, converting the same non-contacting N-terminal modification into opposite subtype-selective activation (Fig. 6 I).

Discussion

This study establishes how a minimal N-terminal chemical modification can remotely control peptide-GPCR recognition. Using the Aplysia PRXamide system as a molecularly tractable structure-activity relationship (SAR) model, we show that two receptor subtypes (ApPRXa-R1 and ApPRXa-R2) respond in opposite directions to the same pyroglutamylated ligand, MMG2-pDPb. This bidirectional activity cannot be explained by a simple stabilizing or affinity-enhancing role of pQ. Instead, multiple SAR analyses support an indirect mechanism in which the N-terminal pQ lactam constrains ligand conformation and reweights a shared receptor-contact network. Extension to human NmU receptors further explain how the direction of such distal modulation is determined: the loose and permissive hNmU-R1 pocket better accommodates the constrained pQ ligand, whereas the compact hNmU-R2 pocket penalizes it. Thus, pQ converts a minimal chemical difference into subtype-specific GPCR signaling through receptor pocket-dependent conformational steering.

Binding pocket permissiveness converts pQ constraint into subtype-selective signaling

Docking and mutagenesis analyses demonstrate that N-terminal pQ confers receptor subtype-selectivity via an indirect, conformation-based mechanism. Although MMG2-pDPb and MMG2-DPb engage largely overlapping receptor-contact residues, pQ reshapes the ligand’s bound configuration (Fig. 2 C, D, Fig. 3 C, D, Fig. S13 C, and Fig. S18 A, B), thereby altering the topology or strength of key interactions (Fig. 2 E-G, Fig. 3 E-G and Fig. 4). Notably, the overall binding mode, in which C-terminus inserts into the orthosteric pocket while N-terminus remains solvent-exposed, closely parallels that observed in cryo-EM structures of human NmU-8 bound to its receptors (17, 18) (Fig. S33), supporting a conserved directional binding architecture across phylogenetically distant neuropeptide-GPCR systems.

Classical “lock-key” model (49) is insufficient to capture the dynamic complexity of neuropeptide-GPCR interactions, as it neglects the intrinsic conformational flexibility of both ligands and receptors (47) (Fig. 6 I). This limitation motivated the development of “induced fit” paradigm, which emphasizes receptor rearrangement following ligand engagement (44–47). However, given the diversity of neuropeptides and GPCR architectures, it remains unclear whether induced-fit applies universally. Importantly, “induced fit” models generally presume that optimization arises from direct ligand-receptor contacts (45, 63) and from a relatively extensive receptor binding pocket. Consequently, they do not readily explain how ligand features that do not directly contact the receptor can modulate signaling, nor why outcomes diverge across closely related receptor subtypes (Fig.6 I).

Our findings address these issues by demonstrating that an N-terminal pQ moiety acts as a modulator to reshape the ligand core conformation, effectively reducing ligand flexibility (Fig. 6 I). Importantly, the functional outcome of this constraint depends on receptor pocket permissiveness, where static pocket analysis and MD simulations further reveal that the underlying logic is governed by receptor pocket properties: the looser pocket of hNmU-R1 can accommodate the pQ-defined interaction pattern more productively than the non-pQ ligand, resulting in enhanced activation, resembling the canonical induced-fit response (Fig. 4 and Fig. 6 I). Given the well-documented role of pQ in enhancing receptor activation across neuropeptide systems (23–26, 28), our results suggest that conformationally constrained ligands may be particularly effective in receptors that undergo induced-fit adaptations. In contrast, the more compact hNmU-R2 pocket cannot readily accommodate the pQ-imposed pattern relative to the non-pQ ligand, leading to conformational mismatch and suppressed activation (Fig. 4 and Fig. 6 I). This represents an intermediate mode of recognition between static “lock-key” complementarity and the adaptive “induced-fit” accommodation, in which a ligand-borne conformational constraint is interpreted differently by receptor pockets of distinct permissiveness. This mode has not been formally described in neuropeptide-GPCR signaling and further implies that the receptors with compact pockets favor conformationally flexible ligands. Notably, these mechanisms likely explain the bidirectional regulation observed in the Aplysia PRXamide system.

We therefore define an indirect “PTM distal steering” mechanism, in which a non-contacting N-terminal PTM functions as a molecular tuner that reshapes the conserved ligand core to differentially engage receptor pockets of distinct architecture. Through this mechanism, a single peptide scaffold can generate opposed functional outputs across receptor subtypes via a minimal chemical change. In this framework, the Aplysia model reveals how pQ remotely reweights interaction networks at a molecular level (Fig. 4), whereas the human NmU system further demonstrates how receptor pocket properties determine whether the pQ-constrained ligand is favored or penalized (Fig. 6). This principle provides a mechanistic foundation for conformation-guided design of stable, subtype-selective peptide ligands and suggests that analogous distal conformational control may be broadly exploitable in small-molecule and macrocyclic ligand design to encode receptor selectivity without requiring direct contact modifications.

Broader implications for PTM-mediated conformational control

Our findings position N-terminal pyroglutamylation within a broader class of PTMs that regulate GPCR signaling indirectly through conformational tuning. C-terminal amidation is well established to enhance receptor affinity through direct ligand-receptor interaction (16–19, 64). However, our prior work in Aplysia leucokinin system showed that the C-terminal amide does not directly penetrate the receptor pocket (65) (facing the solvent, as also observed in TRH (66)), yet its removal still abolishes receptor activation. This behavior parallels the conformational role delineated for pQ here, although precise mechanisms remained unsolved in that study. A related example is provided by Aplysia ATRP, in which L-to D-amino acid isomerization produces diastereomeric peptides with opposing effects on two receptor subtypes (67, 68), a phenomenon that may likewise arise from modification-induced conformational biasing. Similarly, N-terminal glycosylation in NPY and glucagon families may modulate GPCR signaling by altering peptide secondary structure or steric bulk (69). Disulfide bonds in oxytocin and vasopressin similarly confer ligand selectivity and stabilize productive ring conformations (3, 64, 70) without necessarily forming new receptor contacts (64), instead positioning internal residues for optimal interactions within a spatially constrained binding pocket.

Despite their chemical diversity, these PTMs converge on a shared functional logic: they impose topological constraints that shape ligand conformational landscapes and thereby tune interaction networks within the receptor pocket, rather than through direct contact alone.

Conservation of pQ-dependent subtype-specific regulation and translational implications

The conservation of pQ-dependent modulation from Aplysia PRXamide to mammalian NmU signaling highlights its broader functional relevance. Although pQ-containing NmU peptides are rare, canine pQ-NmU-7 (pQFLFRPRNa) demonstrates that this modification can bias subtype preference in human NmU-Rs, consistent with the high sequence similarity between canine and human NmU receptors (Fig. S34). The retained sensitivity of hNmU-R1 to pQ-mediated potentiation (Fig. 6 A) further supports the transferability of this chemical logic across species.

From a translational perspective, NmU receptors remain promising yet challenging therapeutic targets. hNmU-R2 has been prioritized for anti-obesity applications due to its metabolic efficacy with fewer side effects (71–74), whereas hNmU-R1 remains less explored but has been implicated in glucose-lowering effects (74, 75), immune regulation (76), and even potential application in treating bladder dysfunctions, including stress urinary incontinence or detrusor underactivity (77–80). However, clinical development of NmU-based ligands has been limited by peptide stability.

Although NmU peptides exhibit robust activity in receptor assays, their in vivo efficacy is constrained by rapid degradation (81–83). In this context, pQ provides a dual advantage: its known ability to enhance peptide stability (22, 24, 25) can be complemented by our finding that pQ can also tune receptor engagement through distal conformational control. Compared to synthetic N-terminal modifications or linker-based stabilization strategies (73, 84), pQ offers a chemically minimal, evolutionarily validated solution for improving both pharmacokinetic and pharmacodynamic properties.

Overall, our work defines an indirect “PTM distal steering” mechanism in which a non-contacting N-terminal pQ modulates subtype-specific GPCR activation by remodeling ligand conformation ensembles rather than by forming direct receptor contacts (Fig. 4). The functional direction of this steering is dictated by receptor pocket mechanics: compact pockets favor more flexible ligands, whereas loose pockets better accommodate rigid pQ-constrained ligands (Fig. 6 I). The extension of this molecular logic from Aplysia PRXamide to mammalian NmU signaling suggests that pQ can act as an evolutionarily conserved conformational switch capable of fine-tuning neuropeptide-GPCR communication. Given the widespread occurrence of N-terminal and cyclization PTMs, this mechanism may be relevant to other neuropeptide families and suggests that ligand-receptor interactions can operate through multiple regulatory modes beyond the canonical “induced-fit” paradigm. By translating this principle into a design framework, our study links fundamental mechanistic insight to new opportunities for the rational optimization of neuropeptide therapeutics, and may further inform the design of small-molecule and macrocyclic scaffolds.

Materials and Methods

Subjects and reagents

Experiments were performed on Aplysia californica (100-350 g) obtained from Marinus, California, USA. Aplysia are hermaphroditic (i.e., each animal has reproductive organs normally associated with both male and female sexes). Animals were maintained in circulating artificial seawater at 14°C-16°C and the animal room was equipped with a 24 h light cycle with the light period from 7:00 a.m. to 7:00 p.m. Before dissection, animals were anesthetized by injection of isotonic 333 mm MgCl₂ (about 50% of body weight) into the body cavity. All reagents were purchased from Sigma-Aldrich unless otherwise stated. Peptides were synthesized by Sangon Biotech Pharmaceutical Co., Ltd, Guoping Pharmaceutical Co., Ltd., or Synpeptide Co., Ltd and the purity of the peptides was >95%. Peptides were aliquoted into 50 nmol per micro-centrifuge tube and stored at -20 °C until use.

Peptide measurements by LC-MS/MS

To structurally characterize PRXamide peptides derived from the actual MMG2 precursor gene in the Aplysia CNS, peptide extracts from all central ganglia (abdominal, buccal, cerebral, pleural and pedal) of the CNS were sequenced de novo by LC-MS/MS (85). Isolated and desheathed ganglia were homogenized in acidified methanol (80:18:2 MeOH: water: acetic acid) using Precellys bead beater set to the “Soft” method (3 × 15 s, 5800 rpm protocol). Homogenates were collected and centrifuged at 14,000 x g for 15 min to precipitate tissue debris. Resulting supernatants were transferred into another set of vials, dried using a speed-vac. Dry crude extracts were reconstituted in 0.1% formic acid, mixed thoroughly, centrifuged, and the resulting supernatant was collected for further processing; the pellet was discarded.

Aplysia PRXamide peptides extracts were characterized using the Bruker nanoElute LC hyphenated to Bruker timsTOF Pro via CaptiveSpray source with external column oven (Bruker Daltonics, Billerica, MA). Samples were first injected onto Acclaim™ PepMap™ 100 C18 pre-column trap for additional desalting and preconcentration, followed by separation on a Bruker PepSep column (75 μm ID, 1.9 μm particle size, 250 mm length) at a flow rate of 400 nL/min and a column temperature of 35°C. The mobile phase consisted of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in acetonitrile). A gradient of solvent B was applied, starting from 2% and increasing to 10% in 5 min and then to 50% over next 80 min, followed by a rapid ramp to 95% for wash, and finally equilibration with starting conditions. The positive ion mass spectra were acquired in Parallel Accumulation Serial Fragmentation (PASEF) mode with a cycle time of 1.1 s and 1/k0 range 0.6-1.6 V s/cm².

Identification of PRXamide receptor sequences in Aplysia

We primarily conducted sequence search and BLAST using the NCBI (https://www.ncbi.nlm.nih.gov/) and AplysiaTools (http://aplysiatools.org) databases. These latter databases (Dr Thomas Abrams, University of Maryland) include data for Aplysia transcriptome and genome (86). The ORFs of the putative receptor full-length cDNA sequence were obtained using ORF Finder (https://www.ncbi.nlm.nih.gov/orffinder/). BioEdit software (https://bioedit.software.informer.com/7.2/) was selected to compare the PRXa/NmU peptide and receptor sequences with those of other species. For the putative PRXamide receptors, TM domains and conserved sites were predicted using TMHMM Server, v.2.0 (87) (http://www.cbs.dtu.dk/services/TMHMM/) and NCBI Conserved Domain Database (88) (http://www.ncbi.nlm.nih.gov/Structure/cdd/cdd.shtml).

Phylogenetic analysis and molecular evolution

Multiple sequence alignments of receptor sequences were performed by ClustalW and manually corrected. Phylogenetic trees were constructed by maximum likelihood (ML) methods of MEGA X and Phylosuite software. For Fig. S4, the LG+G+F model and 1000 times bootstrap were set. For Fig. S9, the LG+G model and 1000 times bootstrap were set. TBtools (89) and website Chiplot (90) (https://www.chiplot.online/tvbot.html) were used to combine and visualize the phylogenetic trees.

We used MUSCLE (codon) in MEGA X to compare the CDS sequences of PRXa/NmU receptors, and the aligned CDS sequences were converted into pml format for the construction of ML phylogenetic tree by Phylosuite (91) software. The ratio of non-synonymous/synonymous (ω) in the branch and specific positively selected sites was inferred using the Branch Model and Branch-Site Model in EasycodeML_v1.0 software (92), while Likelihood Ratio Test (LTR) was used to determine site differences (93).

Generation of receptor plasmids

Aplysia PRXamide receptors were derived by RNA extraction and cDNA cloning. For RNA extraction, animals were anesthetized by injection of 333 mM MgCl₂ (30-50% of body weight). Cerebral, pleural-pedal, buccal, and abdominal ganglia were then dissected and maintained in artificial seawater (460 mM NaCl, 10 mM KCl, 55 mM MgCl₂, 11 mM CaCl₂, 10 mM HEPES, pH 7.6) within Sylgard-lined dishes (Dow Corning). For RNA isolation, ganglia were transferred to 200 μL Trizol reagent (Sigma, T9424) and flash-frozen at -80°C.

Frozen samples were thawed and homogenized using a plastic pestle. Trizol volume was adjusted to 1 mL, followed by 10 min incubation at room temperature. Subsequent chloroform extraction (200 μL) involved vigorous vortexing and 15 min incubation on ice. Phase separation was achieved by centrifugation (12,000×g, 4°C, 15 min). The aqueous phase was combined with an equal volume of isopropanol, mixed by inversion, and precipitated at -20°C for 2 h. RNA pellets were obtained by centrifugation (12,000×g, 4°C, 15 min), washed twice with 75% ethanol, and air-dried for 5-10 min. Purified RNA was resuspended in 30 μL nuclease-free water, with concentration quantified via ND-1000 spectrophotometer (Thermo Fisher Scientific).

First-strand cDNA synthesis was performed using 1 μg total RNA with PrimeScript RT Master Mix (Takara, RR036A) under manufacturer-specified conditions. Synthesized cDNA was stored at -20°C until subsequent PCR.

The synthesized cDNA above was used as a polymerase chain reaction (PCR) template. Specific primers were designed based on protein-coding sequences for the putative ApPRXa-Rs. The PCR reaction was performed with 98 °C/2 min predenaturing, 98 °C/10 s denaturing, ∼64 °C/15 s annealing, 72 °C/30 s extension, and 72 °C/5 min re-extension for 35 cycles. The PCR products were subcloned into vector pcDNA3.1(+) or pcDNA3.1(+)-FLAG (for ELISA experiments only, see “Cell surface expression analysis by ELISA” section) and sequenced to ensure the sequences were correct.

Human Neuromedin U receptors (hNmU-R1 and hNmU-R2) were synthesized by Sangon Biotech Pharmaceutical Co., Ltd., based on sequences derived from the Genebank database (accession numbers: AF272362.1 for hNmU-R1 and AF272363.1 for hNmU-R2). The synthesized genes were subcloned into pcDNA3.1(+) and sequenced to ensure the sequences were correct.

Cell culture and transfection

The CHO-K1 cells (Procell, CL-0062) were cultured in F-12K medium (Gibco, 21127-022) with 10% fetal bovine serum (FBS) (Genial, G11-70500) at 37 °C in a humidified incubator containing 5% CO₂. Cell culture dishes (Corning, 430166) were used for culture until cells reached the logarithmic growth phase. Then, the medium was discarded, and cells were washed twice with PBS and digested with trypsin. Transfection experiments were performed when the cells were grown to 70−90% confluence and using jetPRIME (Polyplus Transfection, PT-114-15) transfection reagent, according to the manufacturer’s instructions. Specifically, in all experiments in this study, for each 35mm dish, 2 μg of the putative receptor plasmids [in pcDNA3.1(+) or pcDNA3.1(+)-FLAG] and 2 μg Gαq plasmids [in pcDNA3.1(+)] were mixed with 200 μL jetPRIME Buffer (Lot# 0000000396), followed by the addition of 6 μL jetPRIME. Gαq plasmids were added to ensure that the ligand receptor binding would generate an IP1 response (see “IP1 accumulation assay section”) no matter what Gα protein a putative receptor might associate because Gαq is a promiscuous protein that will bind to most GPCRs. The CHO cells with the reagents added above were mixed gently and incubated at room temperature for 15 min. The DNA/jetPRIME mixture dropwise was then added to the dish, and the cells were incubated at 37 °C in 5% CO₂ overnight.

IP1 accumulation assay

Inositol monophosphate (IP1) accumulation assay measures the concentration of IP1, which is hydrolyzed from the second messenger, inositol trisphosphate (IP3), generated by the Gαq pathway (94, 95) when a G-protein coupled receptor (GPCR) expressed in CHO-K1 cells is activated by an appropriate ligand. On the third day, transfected cells were trypsinized and reseeded in 384-well tissue culture-treated plates (Corning, 3570) at a density of 20,000 cells/well in F-12K and 10% FBS at 37 °C in 5% CO₂ overnight. The next day, the medium in wells was discarded and exposed to potential peptide agonists in 14 μL stimulation buffer (Lot No. 621P1FDC) for 1h. Peptide agonists, ranging in concentrations from 10⁻¹² to 10⁻⁴ M, were used to construct dose-response curves. The activation of the putative receptors was detected by monitoring intracellular IP1 concentration using an IP-One Gq kit (Cisbio, 62IPAPEB) in Tecan Spark following the manufacturer’s instructions with minor modifications (for using 0.5×reagents for the anti-IP1-cryptate and IP1-d2 reagents). Specifically, after adding the detection solution to each well, the sealed plate was incubated for a minimum of 1h at RT protected from light, and each plate was then read with Spark (TECAN, USA) with an HTRF compatible setup: excitation at 320 nm and emitted light at 620 and 665 nm.

Molecular modeling and docking

Since pyroglutamic acid (pQ) is a non-canonical amino acid, MMG2-pDPb/DPb and other ligands were constructed using ChemDraw (96) and exported in sdf format for subsequent molecular docking analysis. No manual energy minimization was applied, as all docking protocols used in this study perform internal conformational sampling and scoring. Three-dimensional models of all receptors were generated independently using both the website Swiss model (97) (https://swissmodel.expasy.org/) and the Robetta server (98) (http://robetta.bakerlab.org/), the resulting models were compared for overall structural consistency. Model qualities were assessed using the SAVES v6.1 server (https://saves.mbi.ucla.edu/), and stereochemical properties were evaluated by Ramachandran plot analysis with MolProbity (99) (https://molprobity.biochem.duke.edu/).

Peptide-protein docking was performed primarily using HPEPDOCK 2.0 (56) (http://huanglab.phys.hust.edu.cn/hpepdock/) and diffdock (57) (https://huggingface.co/spaces/simonduerr/diffdock). HPEPDOCK supports the input of peptide sequences containing non-standard amino acids; accordingly, ligand sequences were submitted directly with the pre-modeled receptor structures. In contrast, diffdock requires explicit structural input, including ligand sdf files and receptor pdb files. For each docking output, blind docking was performed, and the top 10 ranked models with plausible binding pocket engagement were selected for interaction analysis. Recurrent interacting residues and representative docking poses were visualized and analyzed using PyMOL and ChimeraX (100, 101). Additional detailed docking workflow and complementary peptide-protein docking are provided in the Supplementary Results and Discussion section.

Molecular dynamics simulations

The CHARMM-GUI platform (102) was employed to integrate the structures into a hexagon pure 1-palmitoyl-2-oleoyl-sn-glycerol-3-phosphocholine (POPC) lipid bilayer, solvated with a 22.5 Å water layer and 150 mM Na⁺/Cl^-. Given that MMG2-pDPb contains an N-terminal pQ residue that is not included in the PTM library of CHARMM-GUI, we excluded the predefined amino acid annotation of the ligand peptide and performed atomistic modeling by MOE (103) to retain and simulate the pQ moiety. All systems were described using the CHARMM36m (104) force field, and all MD simulations were carried out using NAMD 3 (105). Each system was first subjected to 5000 steps of energy minimization, followed by the standard six-step equilibration protocol as recommended by CHARMM-GUI to gradually relax the membrane, solvent and protein components. Production simulations were then performed for 300 ns under constant temperature (303.15 K) and pressure (1 bar) (Fig. 5 E). The stability of the simulated systems was evaluated by calculating the root-mean-square deviation (RMSD), root-mean-square fluctuation (RMSF), free energy surface (FES), and radius of gyration (Rg) based on all heavy atoms of the receptor-peptide complexes. Trajectory analysis and all calculations were conducted using VMD (106) and MDAnalysis (107).

Binding-pocket volume analysis

The orthosteric binding pockets of human NmU receptors were defined and quantified using CASTp with a probe radius of 1.4 Å (62). Ligand-bound active-state structures of hNmU-R1 and hNmU-R2 were used for pocket analysis (17, 18). Before CASTp calculation, non-receptor components, including NmU-8, G proteins, lipids, and solvent molecules, were removed, while the ligand-bound receptor conformations were retained. Molecular surface volume (MS volume) was used to compare pocket size between receptor subtypes. Structural visualization of receptor pockets and CASTp-defined pocket surfaces was performed in UCSF Chimera (108).

Mutagenesis of the ApPRXa-Rs

In a first set of experiments, mutagenesis of the receptors was performed without a FLAG tag. Specifically, construction of the mutants was performed employing the full-length ApPRXa-Rs cDNA cloned into the pcDNA3.1(+) plasmid. Site-directed mutagenesis was performed using the site-directed mutagenesis kit (Sangon Biotech), following the manufacturer’s instructions. Briefly, forward and reverse primers containing the expected mutation were mixed with kit components, and 10 ng of pcDNA3.1(+)-receptors was used as the mutation template. After 14 to 18 rounds of PCR amplification, 1 μl of DpnI was added and incubated at 37 °C for 1 h to digest the template. The primers used to obtain the mutants were designed based on the cDNA sequences as listed in Supplementary Table 2. Mutants were confirmed via DNA sequencing. IP1 accumulation assay with ApPRXa-Rs and mutants without FLAG tags was performed using procedures described in “IP1 accumulation assay” section.

Cell surface expression analysis by ELISA

Cell surface expression determination of ApPRXa-Rs and mutant receptors, all with FLAG tags, was performed using a procedure modified from previous work (23, 65, 109–111). Specifically, CHO-K1 cells were maintained in F-12K medium with 10% FBS at 37 °C in 5% CO₂. When the cells were grown to 70 to 90% confluence, transfection experiments were performed. Briefly, in each well, 2 μg of the receptor plasmids [in pcDNA3.1(+)-FLAG] were mixed with 200 μl of jetPRIME buffer, followed by the addition of 6 μl of jetPRIME and incubated at room temperature for 15 min. The DNA/jetPRIME mixture dropwise was then added to the dish. The cells were incubated at 37 °C in 5% CO₂ overnight. After 24 h, transfected cells were plated in 96-well white clear-bottom cell culture plates (Corning, 3610) at a density of 20,000 cells in 100 μl per well and incubated overnight. The following day, culture media was aspirated and cells were washed twice with 200 μlof1×PBS (Procell, PB180327). Then, 100 μl of 1× PBS containing 5% (w/v) bovine serum albumin was added to each well and incubated at room temperature. After 30 min, 100 μl of 1:10,000 anti-FLAG M2-HRP conjugate (Sigma-Aldrich, Cat A8592) was added to each well and incubated for 30 min at 37 °C. Cells were washed twice with 200 μl of 1× PBS and then incubated with 200 μl of TMB Chromogen Solution (Sangon Biotech, E661007) for 30 min at 37 °C in the dark. Finally, 50 μl of ELISA Stopping Solution (Sangon Biotech, E661006) was added to each well to stop the reaction. The absorbance at 450 nm was measured using a Tecan Spark microplate reader. In each experiment, expression of the mutant was assessed and compared with that measured for the corresponding wildtype receptor in the same experiment.

Statistical analysis

Dose-response curves and bar graphs for experimental data were plotted using Prism software (GraphPad, https://www.graphpad.com/). Data are presented as mean ± SEM of at least three independent experiments. Statistical tests were performed using Prism software. They included Student’s t-test, and one-way or two-way ANOVA, as appropriate. Statistical significance was set at P < 0.05.

Data availability

The original contributions presented in the study are included in the article and Supplementary information.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (grants 32471073, 32171011, 31861143036, 31671097 and 31371104 to J. J., grant 32100816 to G.Z.), the Department of Chemistry (UIUC) Discovery Fund to J. V. S., and the Natural Science Foundation of Henan Province (grant 262300421110 to G. Z.). We thank the Collaborative Innovation Center of Advanced Microstructures, Nanjing University, for providing the computational resources and technical support required for the molecular dynamics simulations. We thank Drs. Hillel J. Chiel and Chao Yan for their useful comments on the work.

Additional information

Author contributions

J.-h. C., W.-j. L., and J. J. writing–original draft; J.-h. C., W.-j. L., G. Z., S.-q. W., C.-p. L., J. V. S., and E. V. R. validation; J.-h. C., W.-j. L., S.-q. W., W.-j. L., C.-p. L., F. L., X.-y. D., H.-y. W., J.-p. X., P. F., Y.-l. Z., Q.-c. J., C.-y. L., R.-t. M., Y.-c.-f. Z., G. Z., J. V. S., E. V. R. and J. J. investigation; J.-h. C., W.-j. L., S.-q. W., C.-p. L., J. V. S., E. V. R. data curation; J.-h. C., W.-j. L., and J. J. conceptualization; J.-h. C., W.-j. L., G. Z., J. V. S., E. V. R., and J. J. writing– review and editing; G. Z., J. V. S., and J. J. funding acquisition; J. J. supervision; J. J. project administration; J. J. methodology.

Funding

MOST | National Natural Science Foundation of China (NSFC) (32471073)

• Jian Jing

MOST | National Natural Science Foundation of China (NSFC) (32171011)

• Jian Jing

MOST | National Natural Science Foundation of China (NSFC) (31861143036)

• Jian Jing

MOST | National Natural Science Foundation of China (NSFC) (31671097)

• Jian Jing

MOST | National Natural Science Foundation of China (NSFC) (31371104)

• Jian Jing

MOST | National Natural Science Foundation of China (NSFC) (32100816)

• Guo Zhang

Department of Chemistry (UIUC) Discovery Fund

• Jonathan V Sweedler

Natural Science Foundation of Henan Province (Henan Natural Science Foundation) (262300421110)

• Guo Zhang

Additional files

Supporting information

Supplementary table 1

Supplementary table 2

Supplementary table 3

Supplementary table 4

Supplementary table 5

Supplementary table 6

Supplementary table 7

Supplementary table 8

Supplementary table 9