Structure-based design of macrocyclic peptides to generate functional antibodies against G protein-coupled receptors

Abstract

G protein-coupled receptors (GPCRs) are critical to a variety of pathophysiological processes, making them attractive targets for the development of drugs or relevant diagnostic tools. Although GPCRs have been successfully targeted with small molecules, the production of reliable anti-GPCR antibodies remains a major challenge. To address this issue, we develop a strategy using macrocyclic peptides designed to mimic the three-dimensional structure of GPCR extracellular loops as immunogens and use the chicken, which is genetically distant from mammals, as an immunization host to produce antigen-specific antibodies. The high-affinity neurotensin receptor type 1 (NTS1), overexpressed in many types of human cancer and associated with poor prognosis, is used as a target. Rational design of macrocyclic epitope mimics and linker selection are achieved using modeling and predictive analysis software tools based on available NTS1 crystal structures. This study particularly highlights the critical role of the linker in peptide macrocyclization, which determines whether antibodies can exert antagonistic activity. Overall, this strategy represents a valuable asset to produce effective spatial anti-GPCR antibodies and holds promise for diagnostic and therapeutic applications.

In this work, functional antibodies against the cancer-related GPCR NTS1 were generated by immunizing chickens with in silico-designed 3D peptides, offering a strategy for GPCR targeting with potential diagnostic and therapeutic applications.

Introduction

G protein-coupled receptors (GPCRs) constitute the largest family of membrane receptors and play a key role in cell physiology and homeostasis^1,2. Upon binding of their endogenous ligands, these receptors activate various signaling pathways and subsequently trigger a variety of cellular responses via G proteins and other intracellular effectors^3–5. Dysregulation of GPCRs and changes in their expression levels have been linked to various pathological conditions, notably tumor growth and cancer^6,7, thereby making them attractive as viable therapeutic targets. Unsurprisingly, over 35% of all drugs approved on the market specifically target this class of receptors⁸.

Given the importance of the GPCR family as drug targets, the development of reliable anti-GPCR antibodies would be beneficial as research tools for receptor characterization and, potentially, as valuable therapeutics, with broad applications in oncology, inflammation, metabolic and infectious diseases^1,9,10. In addition, their growing significance as diagnostic, prognostic and predictive biomarkers cannot be overlooked^9–11. There is therefore an urgent need to develop robust approaches to produce specific, selective and reproducible anti-GPCR antibodies. However, producing antibodies that selectively target GPCRs remains challenging^12–15. To date, only three anti-GPCR antibodies, mogamulizumab, erenumab, and most recently, talquetamab targeting CCR4 (class A), CGRPR (class B), GPCR5D (class C), respectively, have received approval from the U.S. Food and Drug Administration (FDA)^1,16,17.

The production of anti-GPCR antibodies relies on the selection of a suitable antigen and its presentation to trigger antibody generation. The most common method of antibody production involves the immunization of animals, usually mammals, with appropriate antigens. The choice of the antigen format to be injected remains the greatest difficulty, especially when dealing with complex membrane proteins, such as GPCRs that are embedded in the cell membrane¹⁸. To date, most antibodies targeting GPCRs have been generated using the N-terminal domain or extracellular loops (ECL1-3) of these receptors, as immunogens¹. However, the primary challenge lies in identifying correctly folded epitopes that mimic the native conformation of these membrane-spanning proteins^18,19. The high degrees of GPCR sequence and structural homology among species also contribute to the lack of immunogenicity, particularly when the epitope comes from a species closely related to the animal used for antibody production²⁰. In addition, full expression of GPCR proteins is generally weak and of short duration, and the proteins may exhibit misfolding and limited epitope accessibility due to restricted exposure of the extracellular surface^10,21.

To overcome these challenges, we have developed a strategy using macrocyclic peptides designed to mimic the three-dimensional (3D) structure of the extracellular loops of GPCRs as immunogenic agents in chickens. Because chickens are a genetically distant species from mammals, this approach is expected to trigger a strong immunogenic response against mammalian epitopes, resulting in the production of reliable anti-GPCR antibodies^22,23. In this proof-of-concept study, we selected the high-affinity neurotensin receptor type 1 (NTS1) as a target. NTS1 is overexpressed in several types of cancer cells^24,25 and is associated with a poor prognosis^26–28. Its natural ligand, neurotensin (NT), is secreted by tumor cells in an autocrine-dependent manner and acts as a growth factor when bound to NTS1, influencing tumor progression, cell invasion and metastasis^29,30. The crystal structure of NTS1 in its agonist-bound conformation has also been solved previously^31–33, providing valuable insights for predicting the 3D structure of the epitope³⁴.

Herein, we describe a strategy for producing reliable antibodies that recognize both the sequence and 3D structure of the extracellular loops of GPCRs. In particular, this study highlights the critical role of the linker in the macrocyclization of the peptide used as an immunogenic agent. The peptide, which is predicted to adopt the structural conformation corresponding to the second extracellular loop (ECL2) of NTS1 bound to NT, enables the production of an antibody that selectively recognizes the receptor and exerts potent antagonist activity. This functional activity is demonstrated by its ability to block downstream signaling pathways stimulated by NT. A slight modification of the linker used to cyclize the peptide results in antibodies with specific and selective detection properties, which are interesting features for diagnostic applications, but lack antagonist activity. Overall, this proposed technology represents an advance in the development of antibodies targeting GPCRs.

Results

Rational design and synthesis of peptide mimetics of the second extracellular loop of NTS1

To conduct the study, for each extracellular loop of NTS1 (ECL1, ECL2, and ECL3), the design of the macrocyclic antigen was carried out based on two criteria: 1) the size of the cycle, i.e., the number of amino acids in the cycle and the linker enabling macrocyclization. A total of 123 linkers divided into 7 categories (aromatic, biphenyl, indole, piperazine, linear, amino acids and miscellaneous) were evaluated to encompass a broad chemical space and achieve conformational diversity based on the following criteria (Supplementary Fig. 2a): (1) variability in flexibility and chemical structure, (2) ease of synthesis or commercial availability, (3) presence of two functional groups suitable for cyclization, and (4) inclusion of a free thiol group for KLH conjugation (or possibility to insert a cysteine next to the linker). For each loop, between 9 and 12 amino acids were considered, resulting in a total of almost 1500 macrocycles to evaluate in silico. After energy minimization using a model poly-alanine macrocycle and visual inspection, 24 linkers were selected for further molecular dynamics studies, resulting in a total of 288 macrocycles (ECL1, ECL2 and ECL3, from 9 to 12 residues, 24 linkers x 3 loops x 4 sizes) (Supplementary Fig. 2b). Considering the biological results, the molecular modeling, and the selection criteria outlined below, we decided to focus on ECL2.

The peptide sequence selected in this study as the antigenic determinant consists of 11 amino acids (N-R-S-A-D-G-T-H-P-G-G), with 10 to 20 amino acids being the optimal length for an epitope³⁵. The selection and rational design of structurally defined epitope mimetics were based on several criteria, including the relevance of the extracellular domain³⁶, its involvement in receptor activity (as observed with ECL2^31,37), the peptide’s immunogenicity, the availability of the crystal structure of the agonist-bound rat NTS1 receptor (rNTS1)³¹ and the feasibility of peptide synthesis. Multiple algorithms implemented in the Molecular Operating Environment (MOE) software for computational design and simulations were then used to select a suitable linker for peptide cyclization and to predict the flexibility and hydrophilicity of the synthesized macrocyclic peptides³⁸.

The peptide macrocycle design process started with a conformational search to identify initial structural states, followed by molecular dynamics (MD) simulations to explore system dynamics. Next, time-lagged independent component analysis (tICA)³⁹, which identifies slow dynamical processes, was applied, followed by Gaussian Hidden Markov Model (GHMM)⁴⁰ optimization to cluster metastable states and calculate transitions⁴¹. The workflow concludes with full validation and comprehensive analysis of the results, providing insight into the conformational behavior and kinetics of the system (Supplementary Fig. 3). From this process, we identified and synthesized three conformationally constrained peptides mimicking the 3D structure of ECL2 by closing the ring with three different linkers: side-chain S-lysine (3Dpeptide-1), side-chain S-ornithine (3Dpeptide-2) or main-chain S-lysine (3Dpeptide-3). A linear peptide (Lin-peptide) was also synthesized as a control (Supplementary Fig. 1, Supplementary Table 1 and Supplementary Fig. 4b, c). The modeling analysis was conducted on all 3D peptides to predict which of the linkers would generate the best ECL2-mimicking conformation. All the predicted conformational states and their corresponding prevalence (represented by colored spheres as a percentage of occurrence) are shown in Fig. 1. To evaluate the extent to which the spatial peptides mimicked ECL2, root-mean-square deviations (RMSDs) of the atomic positions between each 3D peptide and the crystallized ECL2 of rNTS1 were calculated for each predicted state (Table 1). The RMSD values were calculated from the alpha and beta carbons of the five most exposed amino acids of the loop (D-G-T-H-P) after superimposition on the “Align/Superimpose” panel of the sequence editor module of the MOE software. The conformational states with the lowest RMSD values were considered to represent the most relevant mimetics of ECL2.

Fig. 1. Predicted metastable conformational states of 3D-peptides and ECL2 backbone overlays.

Rows (a–c) correspond to 3D-peptide-1, 3D-peptide-2, and 3D-peptide-3, respectively. In each row, the central panel shows a three-dimensional tICA embedding of the MD data, with Gaussian-HMM-derived metastable states depicted as colored ellipsoids labeled by population percentage. Surrounding the embedding are ribbon representations of 50 randomly sampled conformers per state, overlaid on a translucent density of the full ensemble. The rightmost panel in each row displays a superposition of those two lowest-RMSD centroid structures (sticks) onto the extracellular loop 2 (ECL2) backbone of agonist-bound rNTS1 (cyan ribbon), with stick colors matching the corresponding ellipsoid hues.

Table 1.

Percentage occurrence for each peptide conformational state and their corresponding calculated RMSD values

	3Dpeptide-1		3Dpeptide-2		3Dpeptide−3
	Occurrence (%)	RMSD (Å)	Occurrence (%)	RMSD (Å)	Occurrence (%)	RMSD (Å)
State 1	32.62	1.26	8.21	2.12	16.04	1.09
State 2	17.07	2.39	18.10	1.86	25.55	0.74
State 3	12.92	1.3	16.93	1.79	3.56	1.99
State 4	17.58	2.44	11.39	2.26	11.74	0.62
State 5	13.46	2.01	9.52	1.92	19.43	1.23
State 6	6.35	1.48	12.22	2.16	10.54	1.78
State 7	–	–	7.46	2.63	13.14	1.03
State 8	–	–	16.18	1.31	–	–
% <1.31 Å	45.54% (2 states)		16.18% (1 state)		85.9% (5 states)
Nb of states	6 states		8 states		7 states

An arbitrary threshold of 1.31 Å was determined based on the lowest RMSD of 3Dpeptide-2, and states at or below this threshold are indicated in bold.

For comparison, an arbitrary threshold of 1.31 Å was defined as the highest of the three peptides’ lowest RMSDs (3Dpeptide-1: 1.26 Å; 3Dpeptide-2: 1.31 Å; 3Dpeptide-3: 0.62 Å). Hidden Markov modeling of 3Dpeptide-1 trajectories identified six metastable conformations, two of which (states 1 and 3) exhibited RMSDs ≤ 1.31 Å. Together, these conformations account for a combined occupancy of 45.54 % (Table 1). Notably, state 1 achieved the lowest RMSD (1.26 Å) and the highest stationary probability of all predicted conformers for 3Dpeptide-1. Eight conformational states were predicted for 3Dpeptide-2 and seven states for 3Dpeptide-3 (Fig. 1). Only one state (state 8) met the defined threshold of 1.31 Å, corresponding to only 16.18% probability for 3Dpeptide-2. Regarding 3Dpeptide-3, five states fell below the threshold (states 1, 2, 4, 5, and 7), accounting for 85.9% of the calculated occurrences. Accordingly, this peptide was predicted as the most likely to adopt a 3D structure similar to that of ECL2 in the agonist-bound crystal structure of the rNTS1 receptor. The two states 2 and 4 of 3Dpeptide-3 with the lowest RMSD values (0.62 Å and 0.74 Å) displayed the best backbone superimposition with ECL2 of rNTS1. For each peptide, the superimposition of the two predicted states with the lowest RMSD values onto the ECL2 backbone of agonist-bound rNTS1 are shown in the right-hand panels of Fig. 1. This rank order of native-like occupancy mirrors precisely the assay results, demonstrating that higher residence in the ECL2-mimetic conformations drives stronger binding. The implied-timescale curves plateau at ~ 6 ns, and the CK tests show tight overlap between predicted and observed transition probabilities, confirming that our kinetic models are fully converged and Markovian. These findings thus provide a coherent mechanistic explanation for the experimental potency ranking.

Affinity and selectivity of anti-NTS1 antibodies

Following immunization with the four peptides (Lin-peptide, 3Dpeptides-1, -2 and -3) (Supplementary Table 1; Supplementary Fig. 4a), chicken polyclonal antibodies (IgY) were extracted and immunopurified to generate four distinct anti-NTS1 antibodies: IgY-Lin, IgY-3D-1, IgY-3D-2, and IgY-3D-3. IgY-Lin was expected to primarily detect the linear epitope (N-R-S-A-D-G-T-H-P-G-G), whereas the other three antibodies were designed to recognize the same epitope predominantly in its 3D structure. To validate the ability of these antibodies to bind to their respective peptides, all the antibodies were first tested via an enzyme-linked immunosorbent assay (ELISA) on fixed peptides (Fig. 2a). Inhibition curves were generated by preincubating each antibody with increasing amounts (0 to 10 μM) of its matching free peptide (e.g., IgY-Lin with Lin-peptide) to block detection of the fixed peptide in a dose-dependent manner. The concentration calculated to inhibit 50% of IgY binding to the fixed peptide was similar for all four antibodies, ranging from 2.2 ± 0.6 nM for IgY-Lin to 7.8 ± 2.3 nM for IgY-3D-3 (IC₅₀ values are reported in Fig. 2a). These results confirm that each antibody has a similar detection capacity for its corresponding peptide and validate the effectiveness of the immunization.

Fig. 2. Affinity and specificity of anti-NTS1 antibodies.

a Graph showing the inhibition curves of each IgY (-Lin, -3D-1, −3D-2, and -3D-3) binding to its corresponding bound peptide by increasing the amount of free peptide via an ELISA-based procedure. The absorbance was measured at 450 nm. Control samples containing only the antibody without free peptide were normalized to a total detection of 100%. All samples were then normalized as a percentage of their corresponding control. b Graph illustrating the inhibition of each IgY binding to its corresponding bound peptide in the context of competition with the indicated free peptide at a fixed concentration (10 μM). The positive controls (gray bars) correspond to full detection of the specific peptide bound by its corresponding IgY. All raw data measured at 450 nm were normalized to the control without peptide, which corresponds to 100% detection. Affinities of IgY-3D-1(c) and IgY−3D-3(d) for the different mutated macrocyclic peptides are represented by the inhibition curves generated with increasing concentrations of the five indicated free-mutated peptides (D216A, G217A, T218A, H219A, P220A) and wild-type (WT) 3Dpeptide-1 and − 3. All curves were evaluated via nonlinear regression with three parameters. The curve fits that yielded the IC₅₀ values are shown in the tables under the graphs. Data are shown as means ± S.D. of 2–4 technical replicates per experiment, with each graph representing at least two independent biological experiments (n = 2). Each well corresponds to an independent technical replicate, and the unit of study is a single well in a given experiment. Source data are provided as a Source Data file.

To analyze the selectivity of these anti-NTS1 antibodies, a competition assay was then carried out via ELISA using the four designed peptides as baits and a fixed concentration of 10 μM of each peptide to challenge IgY binding (Fig. 2b). All the antibodies were efficiently blocked by their matching peptides to a similar extent (IgY-Lin: 95% ± 3.5; IgY-3D-1: 100% ± 0.1; IgY-3D-2: 99% ± 1.2; and IgY-3D-3: 97% ± 2.3). The binding of IgY-Lin to its linear peptide was partially blocked by the addition of the three 3D peptides (up to 62% blockage was achieved by 3Dpeptide-1), revealing some cross-reactivity. Interestingly, IgY-3D-1 was effectively blocked by its corresponding peptide (3Dpeptide-1), but was unaffected by the three other peptides, demonstrating high specificity. The IgY-3D-2 and IgY-3D-3 antibodies demonstrated less specificity, compared to the IgY-3D-1 antibody, as the binding of these two antibodies to fixed peptides could be partially displaced by the other peptides, including the linear congener (Fig. 2b). Taken together, these results suggest that the antibodies directed against the 3D peptides display significant specificity toward 3D epitopes compared with a less structured linear epitope.

The importance of the peptide sequence was also evaluated by replacing each amino acid predicted to be predominantly exposed (D-G-T-H-P) with an alanine (A), also known as alanine scanning, which is widely used in epitope mapping and sequence function analyses. This analysis was conducted on the two macrocyclic peptides that best mimicked the 3D structure: 3Dpeptide-1 and 3Dpeptide-3 (Supplementary Table 2). Compared with binding 3Dpeptide-1 WT (IC₅₀ value: 3.2 ± 0.6 nM), IgY-3D-1 showed reduced affinity for all mutated peptides, especially D261A, G217A and T218A, with IC₅₀ values of 2447 ± 852 nM, 1286 ± 354 nM and 708 ± 194 nM, respectively (Fig. 2c). In contrast, IgY-3D-3 binding was less sensitive to these mutations in 3Dpeptide-3, resulting in no significant loss of antibody binding (Fig. 2d). Compared with 3Dpeptide-3 WT (IC50 value: 7.8 ± 2.2 nM), the T218A and H219A mutations even slightly improved IgY-3D-3 binding to these peptides (IC₅₀ values: 0.72 ± 0.04 nM and 0.58 ± 0.07, respectively).

Antibody validation: detection of the NTS1 target protein in cells and brain tissue

We next determined the ability of these anti-NTS1 antibodies to detect the complete target protein in a whole-cell context (Fig. 3). First, immunofluorescence (IF) microscopy was performed using fixed HEK293A cells stably expressing the rNTS1 receptor tagged with hemagglutinin (HA) at its N-terminal extracellular domain (Fig. 3a). A single concentration (5 μg/ml) was used to test the anti-NTS1 antibodies (IgY-Lin, IgY-3D-1, IgY-2, and IgY-3D-3). Colocalization analysis was performed using mouse anti-HA antibodies to validate the receptor subcellular distribution. Signals were detected with dye-conjugated anti-chicken ATTO488 and anti-mouse Alexa Fluor 647 secondary antibodies, which have distinct and nonoverlapping excitation–emission wavelength spectra. For both IgY-3D-1 and IgY-3D-3, z-stack confocal microscopy images revealed strong overlapping of the two fluorescence signals (Fig. 3a). At an identical antibody concentration, IgY-Lin generated a moderate signal, and IgY-3D−2 barely detected NTS1. Immunodetection was fully inhibited in the presence of each matching peptide (Fig. 3a, right panel). To further confirm the effectiveness of these antibodies in recognizing the intended target, a colocalization analysis was performed on all replicates. For each replicate, Pearson and overlap (488/647) coefficients were calculated, and a scatter plot was generated to visualize the correlation between signals (rNTS1 versus HA) (Fig. 3a, middle panel). Scatter plots converging towards the x-axis correspond predominantly to HA detection over rNTS1, as shown with IgY−3D-2. A good correlation, considered here as colocalization, was observed in the scatter plot of IgY-3D-3. The compiled data of the Pearson and overlap coefficients from all the conditions are presented in Fig. 3b, c, respectively. Based on the Pearson coefficient analysis alone, all anti-NTS1 antibodies generated a signal that colocalized with the HA signal, with values between 0.5 and 1.0. This could indicate that none of the antibodies generated nonspecific signals. However, the signal intensity differed significantly, and the overlap coefficient analysis revealed that IgY-3D-2 was not able to correctly detect the rNTS1 receptor (Fig. 3c). Indeed, a value less than 0.5 revealed the absence of colocalization (mean = 0.27 ± 0.08)⁴². In contrast, IgY-3D-1 and IgY-3D-3 generated strong detection signals, almost entirely colocalizing with HA, with overlap coefficients of 0.98 ± 0.19 (3D-1) and 0.97 ± 0.11 (3D-3), respectively. IgY-Lin was less potent at detecting rNTS1 since the overlap coefficient was more variable among replicates (0.89 ± 0.3).

Fig. 3. Antibody validation via IF microscopy and IHC.

a IF microscopy image of fixed HEK293A cells stably expressing HA-rNTS1. The green signal corresponds to the detection of the rat NTS1 receptor by anti-NTS1 antibodies (IgY-Lin, IgY-3D-1/-2/−3), and the red signal corresponds to HA-specific detection. The right panel represents the negative controls generated via preincubation of the antibody with its corresponding peptide at a 1:10 ratio. The Z-stack photos were taken using a confocal microscope with a 60 x objective lens and digitally zoomed by 3 x to better highlight the region of interest. b, c Pearson coefficient and Overlap 488/647 (number of green pixels/number of red pixels) calculated for the images presented in (a). Scatter plots are presented for each antibody, where the y- and x-axes correspond to the green (rNTS1) and red (HA) signals, respectively. Colocalization analysis and quantification of the Pearson coefficient (b) and the overlap coefficient (c) of ten z-stack slices of five randomly chosen fields considered as replicates. The data are presented as Min-to-Max box-and-whisker plots of at least 50 technical replicates, representative of two independent biological experiments (n = 2). Box plots show the median (center line), 25th–75th percentiles (box bounds), and minimum–maximum values (whiskers). Each z-stack slice of a field is considered a technical replicate; the unit of study is one field within a given experiment. One-way ANOVA followed by Tukey’s multiple comparisons test was performed; **** corresponds to a P-value < 0.0001. d Distribution of NTS1 immunoreactivity in fixed tissue sections from adult rat brain with anti-NTS1 antibodies (IgY-Lin, IgY-3D-1/-3). Immunostained NTS1-immunoreactive perikarya and dendrites are evident throughout the substantia nigra, pars compacta (arrows) Scale bar = 200 µm. Within the cerebral cortex, NTS1-expressing cells in layer V represent predominantly pyramidal cells displaying strong immunolabeling of the cell body and apical dendrite (arrows). In addition, labeled neuronal perikarya are evident in layers II and III of the auditory cortex. Scale bar = 150 µm. Source data are provided as a Source Data file.

As suggested previously^43,44, fixing cells or tissues with aldehydes can alter the conformation of protein epitopes. Therefore, antigenic determinants accessible in native proteins may not be exposed in the same way in living and fixed cells. To assess the ability of the isolated antibodies to detect NTS1 in live cells, we performed IF microscopy by incubating living cells with the different anti-NTS1 antibodies and the anti-HA antibody at a single concentration prior to fixation (Supplementary Fig. 5a). Our results revealed that IgY-3D-3 is the only antibody capable of detecting NTS1 in its native conformation in live cells. Accordingly, flow cytometry analysis performed on live cells demonstrated that only IgY-3D-3 achieved signal detection in 89.02% of cells co-labeled with anti-HA antibodies, as observed in the Q1-upper right (UR) quadrant (Supplementary Fig. 5b, c).

To further validate the ability of these anti-NTS1 antibodies to detect the endogenous rNTS1 receptor, immunohistochemistry (IHC) was performed on fixed tissue sections from adult rat brain. Two brain regions previously described as highly expressing rNTS1⁴⁵, namely the substantia nigra and cerebral cortex, are illustrated in Fig. 3d. Within the substantia nigra, pars compacta, IgY-3D-1 and IgY-3D-3 detected intensely stained NTS1-positive nerve cell bodies (arrows). In comparison, although IgY-Lin was able to detect NTS1, rather diffuse staining was observed, making it difficult to identify individual immunoreactive cells. In the auditory cortex, neuronal perikarya immunoreactive to NTS1 were also evident in layers II-III and V. In layer V, NTS1-expressing cells represent mainly pyramidal cells showing strong immunolabeling of the cell body and apical dendrite (arrows).

Antibody Validation: Detection of NTS1 via Biochemical Analyses

We further characterized these anti-NTS1 antibodies via Western blotting (WB) and immunoprecipitation (IP) coupled with mass spectrometry (MS) (Fig. 4). WB is widely used to determine if an antibody detects the denatured antigen (due to the presence of SDS in the lysis buffer), whereas IP is typically conformation sensitive and requires an antibody that recognizes the target protein in its native form⁴⁶. These assays were carried out on whole-cell lysates derived from HEK293A cells expressing or not HA-tagged rNTS1. First, WB analysis with anti-HA antibodies revealed the presence of two bands at 45 and 55 kDa, corresponding to the expected molecular size of two differentially glycosylated forms of NTS1, as reported previously^47,48 (Fig. 4a, red dotted square). Bands of similar molecular weights were also detected with IgY-3D-1 and -3 antibodies, although IgY-3D-3 clearly appeared to be more effective in detecting the NTS1 protein than did IgY-3D-1 (Fig. 4a, red dotted squares). In sharp contrast, IgY-Lin gave rise to nonspecific signals, as proteins of similar molecular weights were detected in HEK293A cells that do not express NTS1, whereas IgY-3D-2 failed to detect NTS1 in WB.

Fig. 4. Antibody validation via WB, IP, and MS.

a Western blots of the rat NTS1 receptor detected with chicken anti-NTS1 (IgY-Lin, IgY-3D-1/2/3) and anti-HA antibodies from lysates of HEK293A cells stably expressing HA-rNTS1 DNA or the empty vector used as a control. The dotted red boxes highlight the signal corresponding to rNTS1. b Immunoprecipitation of NTS1 from cells stably expressing HA-rNTS1 with each chicken antibody conjugated to magnetic beads and blotted with an anti-HA antibody. HEK293A cells transduced with the empty vector were used as a negative control. c Immunoblot of the immunoprecipitates used to isolate rNTS1 for subsequent mass spectrometry identification. The negative control consisted of preincubation of IgY with its corresponding peptide at a 1:10 ratio. d Extracted ion chromatogram (XIC) of the identified rNTS1 receptor from the Rattus norvegicus peptide sequence by mass spectrometry using IgY-3D-1 (red spectrum) and IgY-3D-3 (green spectrum). Negative controls with IgY-matching peptides are represented by gray curves. (insert) Graph presenting the calculated area under the curve (AUC) for each condition (Peakview 2.2 software). All the data are representative of at least two independent experiments. Mass spectrometry analysis was performed at two different research facilities. Source data are provided as a Source Data file.

IP experiments were performed using the HEK293A cell lines described above and anti-NTS1 antibodies conjugated to magnetic beads (Fig. 4b). All 3D-IgY antibodies effectively pulled down the receptor to a level comparable to that of the HA control IP. IgY-Lin also led to receptor pull-down but was less effective than the 3D-antibodies (Fig. 4b). We then validated that the protein pulled down in the IP assay was indeed the NTS1 receptor by combining IP and mass spectrometry^49,50 (Fig. 4c, d). First, the rNTS1 receptor was isolated by IP with magnetic beads conjugated to IgY-3D-1 and IgY-3D−3 from HEK293A cells expressing HA-rNTS1. WB was performed with an anti-HA-HRP antibody (Fig. 4c). The eluted samples were lyophilized and sent for MS analysis at two independent mass spectrometry research facilities. A trypsin-digested peptide of 17 amino acids was clearly identified from the Rattus norvegicus database, corresponding to the C-terminal domain of NTS1. The identified peptide sequence and the resulting spectrum are shown in Fig. 4d (the complete spectrum is presented in Supplementary Fig. 6, and the ion data are presented in Supplementary Table 3). The interaction between anti-NTS1 antibodies and the receptor was significantly blocked by preincubation with the corresponding peptide (Fig. 4d, insert). These compelling results confirm that both the IgY-3D-1 and IgY−3D-3 antibodies interact specifically with the intended NTS1 target protein.

Antibody functional properties

Although antibodies are extensively used as powerful research tools for studying membrane proteins, they are also suitable candidates for the pharmacological modulation of protein function. To this end, we evaluated the ability of these antibodies to modulate NTS1 receptor activity, which would be of great interest from a therapeutic perspective. Since the crystal structure of NTS1 bound to the C-terminal portion of the endogenous agonist NT (i.e., NT(8-13)) reveals that the ligand-binding pocket for NT is partially capped by the ECL2 β-hairpin³¹, we first determined whether these IgY anti-NTS1 antibodies targeting ECL2 competitively displaced radiolabeled NT from its ligand-binding site. Cell membranes from HEK293A cells expressing HA-rNTS1 were thus incubated with radiolabeled NT (¹²⁵I-Tyr³-NT) in the presence of increasing concentrations of IgY-Lin, IgY-3D-1, -2, or − 3 (Fig. 5a). We first observed a dose-dependent displacement of the ¹²⁵I-Tyr³-NT agonist from its ligand-binding site by IgY-3D-3, with an IC₅₀ of 3.2 ± 0.7 μg/ml. IgY-3D-1 and IgY-Lin were significantly less able to displace radiolabeled NT, with IC₅₀ values of 36.5 ± 11.8 μg/ml and 187 ± 94.9 μg/ml, respectively. IgY-3D-2 was not effective at competing with bound NT, which is consistent with its inability to recognize NTS1 in its denatured or native state. Importantly, the IgY-3D-1 and IgY−3D-3 antibodies produced via the proposed strategy exhibited consistent recognition properties across different production lots generated from chickens immunized with 3Dpeptide-1 or 3Dpeptide-3, as demonstrated by the high level of displacement of ¹²⁵I-Tyr³-NT from rNTS1 (Supplementary Fig. 7a).

Fig. 5. Antibody functional properties.

a Displacement curves of radiolabeled NT by chicken antibodies (IgY-Lin, IgY-3D-1/2/3). The curves were evaluated via nonlinear regression with three parameters. The curve fits yielded the IC₅₀ values indicated in the legend. Data represent means ± S.E.M. of technical duplicates, from two independent biological experiments (n = 2 independent plates; each well is considered a technical replicate). b Inhibition of NT-induced cyclic AMP production by IgY antibodies. The negative control corresponds to non stimulated cells (NS), and all the antibodies were tested in the presence of NT (EC₅₀). Data are presented as means ± S.E.M. of at least nine technical replicates per experiment, representative of two independent biological experiments (n = 2). One-way ANOVA followed by Tukey’s multiple comparisons test was performed; **** corresponds to a P- value < 0.0001 (GraphPad reports values < 0.0001 when the exact adjusted p-value is below the software threshold). Capacity of the indicated antibodies to activate the NTS1-dependent Gαq (c) and β-arrestin-2 (e) signaling pathways (agonist mode). Capacity of IgY-Lin and IgY−3D-1/3 to block the NT (EC₅₀)-mediated Gαq (d) and β-arrestin-2 (f) signaling pathways (antagonist mode). SR48692, a nonpeptide NTS1 antagonist, was used as a control. The curves were analyzed by non-linear regression with four parameters. BRET assays were performed in technical duplicates, representative of two independent biological experiments (n = 2). Source data are provided as a Source Data file.

Depending on the cell type, the stimulation of NTS1 by NT leads to the activation of the Gαs, Gαq, Gαi/o, and Gα13 protein signaling pathways as well as to the recruitment of β-arrestins 1 and 2⁵¹ (Supplementary Fig. 7b). To determine whether the anti-NTS1 antibodies were able to modulate NTS1-dependent activity, we first measured Gαs-dependent cyclic adenosine monophosphate (cAMP) accumulation via a homogeneous time-resolved fluorescence (HTRF) cell-based assay. To this end, CHO-K1 cells stably expressing rNTS1 were stimulated with NT (EC₅₀) and simultaneously incubated with IgY-Lin, IgY−3D-1, -2, or −3 (Fig. 5b). Our results revealed that IgY-3D-3 was the only antibody able to significantly block NTS1-mediated cAMP production. We further examined, via bioluminescence resonance energy transfer (BRET) biosensors, the ability of these anti-NTS1 antibodies to modulate the Gαq, Gαi2, Gα13, GαoB, Gαz and β-arrestin 2 signaling pathways following their binding to rNTS1 transiently expressed in HEK293 cells. Our findings demonstrated that IgY-Lin, IgY-3D-1, and IgY-3D-3 failed to activate (agonist mode) the canonical Gαq and β-arrestin 2 receptor signaling pathways (Fig. 5c, e) and to induce NTS1 coupling to Gα13, GαoB, Gαi2 and Gαz (Supplementary Fig. 7). In the antagonist mode, we found that IgY-3D-3 effectively blocked all signaling pathways associated with NTS1 activation in a concentration-dependent manner (Fig. 5d, f and Supplementary Fig. 7). A nonpeptide NTS1-selective antagonist⁵², SR48692, was used as a control in the BRET-based assay and was found to reverse NT signaling activity at the NTS1 receptor site. In the same experimental paradigm, IgY-Lin and IgY-3D-1 did not inhibit the ability of NT to induce the functional coupling of NTS1 to G proteins or β-arrestins.

Antibody cross-reactivity and selectivity

This project exploited the important phylogenetic distance between mammals and chickens to generate two robust anti-NTS1 antibodies, namely, IgY-3D-1 and IgY-3D-3, which were designed against the ECL2 region of rNTS1. The ECL2-based macrocycle designed to immunize chickens was modeled on the rNTS1 X-ray structure, the amino acid sequence of which is highly conserved among species, except for in chickens (Gallus gallus; Fig. 6a). Despite the difference in two amino acids (threonine and proline) between the designed peptide and the human NTS1 protein sequence, superimposition of the designed peptide on the human ECL2 protein resulted in high levels of similarity in their backbone structures (Fig. 6b), thus suggesting a possible conserved 3D conformational structure of NTS1 between rat and human NTS1 species. In this context, we evaluated whether IgY-3D-1 and IgY-3D-3 could detect the human NTS1 receptor (hNTS1) in HEK293 cells expressing HA-tagged hNTS1. IF confocal microscopy (Fig. 6c) and IP experiments using whole-cell lysates (Fig. 6d) revealed that the ability of IgY-3D-1, but not IgY-3D-3, to detect hNTS1 was conserved. Accordingly, IgY-3D-1 was very effective at competing with bound NT, exhibiting comparable IC₅₀ values between hNTS1 and rNTS1 (IC₅₀ values of 44.3 ± 20.6 μg/ml and 36.5 ± 11.8 μg/ml, respectively) (Fig. 6e). Surprisingly, IgY-3D-3 partially retained its ability to displace bound NT from hNTS1, with an IC₅₀ of 181 ± 91 μg/ml (compared with 3.2 ± 0.7 μg/ml for rNTS1). In addition, we measured the Gαq-dependent inositol monophosphate (IP1) accumulation using an HTRF cell-based assay with CHO-K1 cells stably expressing hNTS1. Cells were stimulated with NT at its EC50 concentration and simultaneously incubated with IgY-3D-1 or IgY-3D-3 (Fig. 6f). Both IgY-3D-1 and IgY-3D-3 significantly blocked IP1 production. As in the radiolabeled NT displacement assay, IgY-3D-1 showed more effective inhibition than IgY-3D-3. These findings suggest that IgY-3D-1 has the best ability to recognize the human receptor isoform.

Fig. 6. Anti-NTS1 cross-reactivity and selectivity.

a Alignment of amino acid sequences corresponding to the ECL2 region of NTS1 from different species. The black bars represent the conservation percentage (CLC sequence bio V5). b Overlays of the ECL2 loops of human (green ribbon) and rat NTS1 (white ribbon). The two amino acids (proline and threonine) that differ between rNTS1 and hNTS1 are indicated, as are the N-terminal domain. c Immunofluorescence microscopy of fixed HEK293A cells stably expressing the HA-tagged human NTS1 receptor. The green signal corresponds to the detection of hNTS1 by the tested antibodies (IgY-3D-1 and IgY3D-3), and the red signal corresponds to HA-specific detection (HA-HRP). d Immunoprecipitation of hNTS1 from cells stably expressing HA-hNTS1 with the indicated chicken antibodies conjugated to magnetic beads, followed by blotting with an anti-HA antibody. Cells transduced with the empty vector were used as controls. e Displacement curves of radiolabeled-NT by IgY-3D-1 and -3. The curves were evaluated via nonlinear regression with three parameters. Data represent means ± S.E.M. of three technical replicates per condition, representative of two independent biological experiments (n = 2). f Inhibition of NT-induced IP1 production by IgY antibodies. The negative control corresponds to non-stimulated cells (NS), and all antibodies were tested in the presence of NT (EC₅₀). Data are presented as means ± S.E.M. of at least nine technical replicates per experiment, representative of two independent biological experiments (n = 2). One-way ANOVA followed by Dunnett’s multiple comparisons test was performed. P- values are indicated as follows: **, P = 0.0020; ****, P < 0.0001. (GraphPad reports values < 0.0001 when the exact adjusted p-value is below the software threshold). HEK293A cells transiently expressing HA-tagged rat NTS1, rat NTS2 or rat APJ receptors were processed for Western blotting (g) and native immunoprecipitation (h) via either IgY-3D-1 or IgY−3D-3. Data are representative of at least two independent experiments. Source data are provided as a Source Data file.

Target selectivity is also a crucial element in the development of therapeutic and diagnostic antibodies⁵³. To address this question, IgY-3D-1 and IgY-3D-3 were tested for their ability to detect class A GPCRs closely related to rNTS1, namely, the rat neurotensin receptor type 2 (rNTS2) and the rat apelin receptor (rAPJ). HEK293A cells overexpressing rNTS2, rAPJ or rNTS1 were then processed for WBs (Fig. 6f), native IPs (Fig. 6g) and IF microscopy (Supplementary Fig. 8a, b). Our results revealed that both IgY-3D-1 and IgY-3D-3 selectively and specifically recognized rNTS1 but failed to detect rAPJ and rNTS2. Notably, the amino acid sequence alignments of rNTS2 and rAPJ with the NTS1-ECL2 epitope region presented 36% and 27% similarity with the corresponding rNTS1 region, respectively (Supplementary Fig. 8c), thus limiting the possibility of adopting a similar spatial conformation.

Discussion

GPCRs represent a major opportunity for the development of biotherapeutics, such as antibodies. Yet almost all the approved drugs directed against this family of membrane proteins available today are small molecules or peptides. They do, however, have better PK/PD properties, due to more favorable biodistribution and less frequent dosing, owing to their longer half-life than that of small molecules. Antibody-based therapeutics also offer the advantages of greater specificity and therefore tend to be less prone to off-target toxicity. The main reasons for the delay in anti-GPCR antibody development include their poor immunogenicity, the high degree of sequence conservation across mammalian species and the difficulty of producing conformationally active GPCR antigens. To date, no strategy has been able to reliably generate epitopes that mimic the native structure of targeted GPCRs^9,18.

To overcome these challenges, we developed a strategy using macrocyclic peptides designed to mimic the 3D structure of GPCR extracellular loops as immunogens and used the chicken as an immunization host to produce antigen-specific antibodies. Due to a phylogenetic difference of 200-million years, the use of chickens affords distinct advantages over mammalian systems to enhance the immunogenic response and facilitate the production of elevated titers of high-affinity antibodies against conserved mammalian proteins, such as GPCRs^22,54. This advantage has been demonstrated in other studies involving chickens for the production of antibodies against challenging targets⁵⁵, including GPCRs⁵⁶. Furthermore, chicken antibodies can be easily chimerized or humanized using various strategies, such as CDR grafting in combination with in silico AI prediction tools to select the most appropriate human germline for a successful conversion^57–60. This study also supports the effectiveness of using 3D epitopes instead of traditional N- or C-terminal linear epitopes to target GPCRs, as previously proposed^61–63. Rational design of macrocyclic epitope mimics and linker selection performed with modeling analysis and prediction software tools were critical to design and achieve the production of specific and selective chicken spatial anti-GPCR antibodies. Here, we set out to generate antibodies against NTS1, a well-characterized class A GPCR that is overexpressed in many types of human cancer, including colorectal cancer and non-small cell lung cancer^24,28. Overexpression of NTS1 has also been associated with higher tumor aggressiveness and poor prognosis, notably in pancreatic adenocarcinoma, triple-negative breast cancer and glioblastoma, thereby highlighting its role as a potential prognostic and therapeutic marker^26–28. Finally, the existence of NTS1 crystal structures in complex with NT further strengthened the rational design of the epitope’s 3D structure^31–33. The key features of the 3D antigenic peptides used to generate anti-NTS1 antibodies and the characteristics of these IgY antibodies are summarized in Table 2.

Table 2.

Summary of the main features of the three 3D peptides and the corresponding IgY antibodies

Features	3Dpeptide-1	3Dpeptide-2	3Dpeptide−3
Linker used to produce 3D-peptide	side chain (S)-Lys	side chain (S)-Orn	main chain (S)-Lys
Number of predicted 3D states	6	8	7
Lowest RMSD predicted (Å)	1.26	1.31	0.62
Number of states below 1.31 Å	2	1	5
Incidence of states below 1.31 Å (%)	45.54	16.18	85.9
Predicted alignment with crystal structure	+++	+/−	+++

Target	Features	IgY-3D-1	IgY-3D-2	IgY-3D-3
Peptides	Peptide selectivity	+++	+	+
	ALA scan effect	+++	N/D	+/-
Fixed cells or lysate (rat NTS1)	IF on fixed cells	++	+/−	+++
	IHC	+++	N/D	++
	Western blot	++	–	+++
	IP/MS	+++	N/D	+++
Live cells (rat NTS1)	IF/cytometry	–	–	+++
	NT displacement	+	–	+++
	cAMP production blockade	–	–	+++
	BRET agonist mode	–	N/D	–
	BRET antagonist mode	–	N/D	+++
Human NTS1	IF on fixed cells	+++	N/D	+/-
	Western blot/IP	+++	N/D	+/-
	NT displacement	++	N/D	+
Rat NTS2, APJ	Receptor selectivity	+++	N/D	+++

In the present study, we demonstrated that two of the antibodies generated, IgY-3D-1 and IgY-3D-3, were capable of specifically recognizing NTS1 and its 3D structure. Moreover, IgY-3D-3 exhibited antagonist activity on the NT-bound NTS1 receptor, a very interesting feature from a therapeutic perspective, especially considering that no small molecule has been shown to be an antagonist and selective for NTS1 alone, as there is usually cross-reactivity with NTS2⁶⁴. The results obtained with the anti-NTS1 3D antibodies also correlated with the modeling analysis and the predicted 3D structures of the designed peptides. Compared with the (S)-ornithine linker (3Dpeptide-2), the (S)-lysine linker (3Dpeptide-1 and 3Dpeptide−3) was found to have increased target affinity in hens, as predicted by the ECL2 mimetic analysis of the agonist-bound receptor. Indeed, IgY-3D-2 was unable to detect NTS1, as the RMSD values of nearly all predictive 3D states (7 out of 8) exceeded 1.31 Å, and the only state at this threshold accounted for just 16.18% of the population. The importance of the linker is particularly highlighted by the subtle differences in cyclization between the two macrocyclic peptides (on the side chain for 3Dpeptide-1 or on the main chain for 3Dpeptide-3), resulting in an ability, or lack thereof, to antagonize receptor activity when bound to its natural ligand, NT. The presence of 5 states below the 1.31 Å threshold, resulting in a percentage incidence of 85.9%, compared with 45.54% for 3D-peptide-1, might explain the additional capacities of polyclonal antibodies resulting from immunization with 3D-peptide-3. Importantly, both IgY 3D-1 and IgY-3D-3 were selective for NTS1 and were unable to detect receptors closely related (NTS2) or belonging to the same class of GPCRs (APJ). Overall, this proof-of-concept study demonstrated the effectiveness and value of in silico technologies and highlights the key step of designing macrocycle epitope mimics. Recent structural advances in several GPCRs of therapeutic interest are certainly a major advantage for the design of spatial IgY antibodies⁶⁵.

These chicken anti-NTS1 antibodies were then robustly validated according to guidelines for antibody validation^46,66,67, demonstrating their specificity and selectivity in detecting NTS1 in a variety of experimental molecular and cellular assays, including ELISA, immunofluorescence microscopy on fixed and live cells, immunohistochemistry, cytometry, Western blotting, immunoprecipitation and mass spectrometry. Importantly, using radioligand binding studies and BRET-based biosensors in live cells, we also found that IgY-3D-3 behaved as an antagonist of G-protein and β-arrestin signaling pathways following activation of NTS1 by NT. Finally, with batch-to-batch variability consistently recognized as a major issue¹³, reproducibility was confirmed by testing different lots of antibodies, which retained their ability to specifically detect NTS1. It is, however, important to acknowledge the inherent limitations of studying polyclonal antibodies, as the features observed are likely to arise from multiple immunoglobulin subpopulations, and the data represent an overall estimate rather than individual molecular characteristics. In addition, immunopurification may result in the loss of some high-affinity binders, which could affect the full representation of the immune response. Despite these limitations, polyclonal antibodies remain a valuable tool for identifying key antigenic peptides and functional domains, particularly when combined with computational modeling approaches. This combination improves antigen selection and provides a strategic advantage in the development of functionally relevant antibodies. Importantly, once the optimal antigenic peptide has been identified, the discovery of monoclonal antibody becomes more predictive and less risky, as it can be pursued effectively using advanced technologies such as phage display, yeast display, or single B-cell screening. Initial libraries can be derived from isolated peripheral blood mononuclear cells (PBMCs), cryopreserved spleen tissue, or bone marrow from immunized hens responsible for generating the polyclonal response of interest. This iterative approach ensures the selection of clones with the desired phenotype and functionality, increasing the likelihood of identifying high-affinity and functionally relevant monoclonal antibodies while reducing the risk of selecting non-specific or suboptimal candidates.

This study successfully led to the production and validation of two reliable anti-GPCR antibodies targeting NTS1. Notably, one of these antibodies, IgY-3D-3, demonstrated functional capabilities, making it an asset for therapeutic applications. IgY-3D-1, on the other hand, displayed highly selective features and could be a better candidate for diagnostic use. Alongside their classic use for therapeutic and diagnostic purposes, antibodies targeting GPCRs can also be designed to incorporate effector functions, such as antibody-dependent cellular cytotoxicity (ADCC) to eradicate tumors (such as mogamulizumab). Greater attention is also being paid to the generation of GPCR antibodies conjugated to cytotoxic payloads, leading to antibody-drug conjugates (ADC) or to bispecific antibodies that can simultaneously bind to two different types of antigens or epitopes (such as talquetamab, which binds to both GPRC5D and CD3 receptors), thus offering the potential to regulate GPCR homo- or hetero-dimerization^68–71. Finally, antibodies, interacting with GPCRs, are capable of state-specific recognition and are therefore promising candidates for the identification of therapeutics displaying functional selectivity akin to that of biased ligands. Overall, this strategy of generating spatial antibodies against GPCRs holds promise for diagnostic and therapeutic applications, especially in pathological conditions where GPCRs are of significant clinical relevance, as in cancers, infectious diseases and metabolic disorders.

Methods

Synthesis of linear and macrocyclic peptides

The different steps leading to the synthesis and cyclization of macrocyclic peptides mimicking the second extracellular loop (ECL2) of rNTS1 are described in detail in the Supplementary Methods.

Computational Modeling

All molecular modeling work, comprising conformational sampling, MD simulations, structural superimposition, and downstream kinetic was carried out as follows. Initial conformational searches and model preparation were performed using MOE³⁸. MD simulations were then executed with GROMACS 2024.1 under standard conditions⁷². Resulting MD trajectories were processed with the Python libraries pyEMMA (version 2.5.12) and MSMBuilder (version 3.8.0) to perform time-lagged independent component analysis, construct and interrogate Markov State Models, and carry out estimation, validation, and analysis of molecular-kinetic processes^73,74. MD simulation parameters and protocols are described in the following section³⁸.

Conformational search

The X-ray structure of NT(8-13) bound to rNTS1 (PDB ID 4GRV) was used as the template for modeling the primary sequence of designed peptides. Structures were built via the MOE builder and minimized on all atoms using AMBER 10: EHT forcefield and 0.1 kcal/mol/A2 RMS as gradient and the GB/VI solvation model.

Each of 123 linkers was connected to the sequences of each loop using different, appropriate functional groups compatible with different chemistry (Supplementary Figs. 2a, b). This set of linkers allowed us to create a highly varied initial database of macrocyclic peptides, enabling the identification of conformations that best mimic the exposed and stable regions of the ECL2 domain. This diversity in linker properties played a critical role in conformational variability, as linkers are primarily responsible for inducing large-scale structural changes.

A conformational search was then performed for each structure using the LowModeMD method, which performs molecular dynamics perturbations along low-frequency vibrational modes⁷⁵. The rejection limit was set at 100, the iteration limit at 10000, the RMS gradient at 0.005, and the MM iteration limit at 500. Two conformations were considered equal if the optimal heavy atom RMS superposition distance was less than 0.25 Å. The lowest-energy conformation found was used as the starting point for molecular dynamics simulations.

Molecular dynamics simulations

All MD simulations were carried out with GROMACS 2024.1 using the CHARMM36 all-atom additive protein force field⁷⁶. The simulation system was prepared using CHARMM-GUI⁷⁷ to generate the initial topology. The system was then solved in a cubic box with explicit TIP3P water and neutralized using counterions. Counter-ions were added to neutralize the system, and an initial steepest-descent minimization was carried out. The Verlet cutoff scheme (rlist = 1.2 nm) was used for nonbonded searches, with Particle Mesh Ewald for long-range electrostatics⁷⁸. A second steepest-descent minimization was performed under identical neighbor-list and cutoff settings. All bonds to hydrogen were constrained via LINCS to remove any residual steric clashes⁷⁹. Subsequently, two equilibration steps, including an NVT under constant volume and temperature, which was maintained at 300 K using the velocity-rescaling thermostat⁸⁰, followed by an NPT equilibration under constant pressure, which was controlled via the C-rescale barostat. Finally, 500 ns of production dynamics were run with the Parrinello–Rahman barostat⁸¹. Trajectories were saved every 10 ps for subsequent analysis.

Time-lagged independent component analysis

MD trajectories were analyzed using a custom time-lagged independent component analysis (tICA) pipeline to extract slow collective motions and characterize the underlying conformational landscape³⁹. The backbone-torsion features (ϕ and ψ dihedral angles) were extracted from the resulting MD trajectories using MDtraj and transformed into sine and cosine components for downstream tICA and MSM analyses^82,83_.

All backbone-torsion features were assembled into a numerical matrix and subjected to time-lagged independent component analysis (tICA). The optimal number of independent components was determined by a slope-average test on the implied timescale spectrum to isolate the most significant slow modes. A tICA lag time of 500 frames (5 ns; dt = 0.01 ns) was selected after evaluating convergence of implied timescales across multiple lags and identifying the plateau region^73,84,85. Implied timescales were estimated with Bayesian error bounds to assess statistical uncertainty. As shown in Supplementary Fig. 10, the implied relaxation timescales for 3D-peptide-1, 3D-peptide-2, and 3D-peptide−3 converge between 6–7 ns, confirming the validity of the Markovian assumption and supporting the robustness of the resulting Markov State Models. This analysis highlights the dominant slow collective motions that define each peptide’s conformational landscape and validates the chosen lag times for accurate kinetic modeling.

Clustering and markov state model construction

Structural data projected onto the dominant tICA subspace were discretized into kinetically coherent states using a Gaussian Hidden Markov Model (GHMM) framework⁴⁰. The tICA projection dimension was chosen by a slope-average test of implied timescales to retain only the slowest, most significant modes⁴¹. GHMMs were then trained over tICA dimensions and states, each candidate model was scored by the Bayesian Information Criterion to penalize overfitting while rewarding explanatory power⁸⁶. The optimal GHMM for each peptide was identified as the one minimizing BIC in concert with maximizing the slowest relaxation timescale, thereby ensuring both parsimony and kinetic accuracy.

For 3D-peptide-1, the best model employed three tICA dimensions and six states; for 3D-peptide-2, six dimensions and eight states; and for 3D-peptide−3, three dimensions and seven states (Supplementary Fig. 11). These configurations capture the principal slow conformational transitions and form the basis of robust Markov State Models for each peptide.

The Chapman-Kolmogorov (CK) test was performed to validate the reliability of Markov state models for the three peptides^87–89, which assesses the agreement between predicted and empirically estimated transition probabilities over multiple lag times. For each peptide, Bayesian MSMs constructed at a 6 ns lag time produced self-transition probabilities (solid lines with 95 % credible intervals) that closely matched direct estimates from MD trajectories in Supplementary Fig. 12, thereby confirming the Markovian character and kinetic accuracy of the models.

Superimposition of the macrocycle with the X-ray loop

The X-ray structure of rNTS1 bound to NT(8-13) (pdb ID 4GRV) was loaded to MOE, and the protonated and partial charges were assigned. Prior to alignment, a 500 ns MD simulation of the complex was performed to ensure conformational equilibration and to emulate more physiologically relevant conditions. For each representative conformation of the macrocycles ensemble, structures were loaded into MOE, and amino acid residues were aligned to the corresponding loop region of NTS1 via the sequence alignment module within the sequence editor panel. The five most exposed residues (D-G-T-H-P) were selected and used as reference points for superimposition onto the corresponding residues of the macrocycles. RMSD values were calculated based on the Cα and Cβ atoms of these five amino acids to evaluate the accuracy of alignment and conformational similarity.

Chicken immunization and IgY extraction

Primary immunization of KLH-conjugated peptides (Lin-peptide, 3D-peptide-1, 3D-peptide-2 and 3D-peptide−3) was achieved in complete Freund’s adjuvant, while subsequent boosters were given in incomplete Freund’s adjuvant (Supplementary Fig. 4a). KLH-conjugated peptides were administered subcutaneously to Lohmann Brown (Rhode Island breed) or Lohmann White LSL-LITE (Leghorn) hens (26–28 weeks old), with two chickens assigned per peptide group. After three months of immunization with a monthly antigen boost, all eggs were pooled according to antigen and time of collection. Egg yolk was separated from the egg white, and the proteins extracted, as previously described, with a few modifications⁹⁰. Egg yolk was diluted with two volumes of distilled water, homogenized, and centrifuged. The precipitate was resuspended in 10% (w/v) NaCl prepared in NaOH (0.05 N) to extract the protein fraction, and the homogenate was adjusted to pH 4.0 before centrifugation to remove precipitates. The resulting supernatant was diluted with three volumes of distilled water, centrifuged, and concentrated by ultrafiltration. Immunoglobulins were then precipitated with 20% saturated ammonium sulfate containing 15% (w/v) NaCl, collected by centrifugation, dissolved in distilled water, and desalted using a size-exclusion chromatography column. All extracted IgYs (IgY-Lin, IgY−3D-1, IgY-3D-2 and IgY-3D-3) underwent a second round of immunoaffinity purification using the corresponding peptide conjugated to agarose beads. Briefly, stringent washes with 500 mM NaCl were performed prior to elution to ensure the removal of any nonspecific polyclonal antibody (pAb) binders, allowing only specific NTS1 binders to be collected. Antibodies were then eluted using 100 mM glycine buffer at pH 2.4. The polyclonal antibodies were produced in limited quantities. For further information or to request access to this biological material, please contact Immune Biosolutions.

Enzyme-linked immunosorbent assay (ELISA) of peptides

All the tested peptides were plate-coated at 3.5 μm per well in a clear, Well-Coated^TM Sulfhydryl Binding 8-well strip plate (G-Biosciences, MO 63132-1429, USA) with binding buffer (0.1 M Na₂HPO₄, 0.15 M NaCl, 10 mM EDTA pH 7.2) overnight at 4 °C. For blocking, L-cysteine (1 μM in binding buffer) was added to each well for 1 h at room temperature (RT). The tested antibodies (IgY-Lin, IgY-3D-1, IgY-3D-2 and IgY-3D-3) were diluted to a final concentration of 0.1 μg/ml (0.25% Tween 20 in PBS or 0.3% milk) and preincubated with increasing amounts (from 0.3 nM to 10 µM) of the corresponding free peptide or a fixed concentration (10 μM) of either free peptide. The IgY-peptide mixtures and the positive control (IgY only) were added to individual peptide-coated wells and incubated for 2 h at RT. Four washes were performed with PBS containing 0.25% Tween 20 (PBS-T), followed by incubation with the secondary antibody (alpaca anti-chicken HRP, Immune Biosolutions, Sherbrooke, QC, CA) at a 1:10,000 dilution in PBS-T for 45 min at RT. After four additional washes with PBS-T, 3, 3’, 5, 5’-tetramethylbenzidine (TMB) (Sigma‒Aldrich, St. Louis, Missouri, USA) was added, and the mixture was incubated for 15 min at RT before the reaction was stopped with 2 N HCl. Absorbance at 450 nm was read with a Tecan GENios plate reader (Tecan Austria GmbH, Austria).

Cells and reagents

Human embryonic kidney (HEK) 293 A (Thermofisher, Catalog number: R70507) and 293 T cells (American Type Culture Collection (ATCC), CR-3216) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and antibiotics at 37 °C and 5% CO₂. Chinese hamster ovary (CHO)-K1 cells were obtained from the ATCC (CCL-61) and were maintained in DMEM/F12 supplemented with 10% FBS, L-glutamine, HEPES and antibiotics at 37 °C and 5% CO₂. All cell culture reagents were from HyClone (Thermo Fisher Scientific Inc., Logan, UT, USA).

HEK293A and CHO-K1 cells were transformed with a lentivirus-based system to stably express either the rat or human NTS1 receptor. To generate the lentiviral vectors expressing HA-tagged rat or human NTS1, the 2xHA-rNTS1 (NCBI accession number: NM_001108967.2) and 3xHA-hNTS1 (NCBI accession number: NM_002531.3) coding sequences were amplified from custom plasmids (p2xHA-rNTS1) and pcDNA3.1 + _3xHA-hNTS1 (UMR cDNA resource center, MO, USA) with the following primers: forward for both sequences: 5’-TGCATCTAGAGCCACCATGTACCCATACGA-3’, reverse for rNTS1: 5’-TCACTGCAGAATTCCTAGTACAGGGTCTCC-3’ and reverse for hNTS1: 5’-TCACTGCAGAATTCCTAGTACAGCGTCTCG-3’. The amplified cassettes were then cloned and inserted into pCDH-MCS-Ef1-puro cDNA cloning and expression vectors (System Biosciences, Mountain View, CA, USA) via XbaI and EcoRI restriction enzymes (New England Biolab, Massachusetts, USA). HEK293T cells were then transiently cotransfected with the pCDH lentivectors (containing either rNTS1, hNTS1 or the empty vector), pMD2. G and psPAX2 (gift from Didier Trono, Addgene plasmids #12259 and #12260, respectively) using polyethylenimine as the transfection reagent (MW 25,000; Polysciences, Warrington, PA). The viral particles were then collected 48 h post-transfection, filtered (0.45 μM PVDF syringe filter, Thermo Fisher Scientific Inc., Massachusetts, USA) and used to generate cells stably expressing 2xHA-ratNTS1 or 3xHA-humanNTS1. HEK and CHO-K1 cells expressing pCDH-CMV-NTS1-Ef1-puro were selected with puromycin at concentrations of 1 μg/ml and 8 μg/ml, respectively.

Immunofluorescence microscopy

A total of 120000 HEK293A cells stably expressing HA-rNTS1 or HA-hNTS1 or transiently expressing HA-tagged rNTS1, HA-rNTS2, or HA-rAPJ were seeded on glass microscope coverslips in 6-well plates. The next day, the cells were washed with PBS, fixed for 10 min in 3.7% formaldehyde-PBS at 37 °C, washed three times in PBS and permeabilized with 0.1% Triton X-100 for 2 min on ice. The cells were then washed twice with PBS and blocked with 10% normal goat serum (NGS) in PBS for 1 h at RT. This saturation step was followed by an overnight incubation with a mouse monoclonal HA-probe antibody (1:200, (F-7) sc7392, Santa Cruz Biotechnology, CA, USA) and 5 μg/ml of each chicken polyclonal anti-NTS1 antibody (Lin, 3D-1, 3D-2 and 3D−3) or a mixture containing the antibody and the corresponding peptide at a ratio of 1:10 (IgY: peptide) in 10% PBS. For live cell incubation, all steps were as described above, except for incubation with chicken anti-NTS1 and mouse anti-HA. Live cells were incubated with antibodies at 5 μg/ml in culture medium for 1 h at 37 °C before fixation with 3.7% formaldehyde. Fluorescence staining was performed via incubation with an ATTO488-conjugated goat anti-chicken antibody (Immune Biosolutions, Sherbrooke, QC, CA) and an Alexa Fluor 647-conjugated goat anti-mouse antibody (A-21235, Thermo Fisher Scientific Inc., Massachusetts, USA), each at a 1:200 dilution. The cells were washed 4 times in PBS before being mounted in ProLong Gold Antifade Mountant with DAPI (Thermo Fisher Scientific Inc., Massachusetts, USA). Images were obtained with a confocal Olympus IX81 FV1000 LSM microscope using Fluoview FV10-ASW 3.1 Viewer software. Z-stacks (0.5 μm slices) were acquired from randomly chosen fields with identical imaging parameters across all samples.

Tissue fixation and immunohistochemistry

Adult male rats were anesthetized with 5% isoflurane and transcardially perfused with 4% paraformaldehyde (PFA) in phosphate buffer (0.1 M). The brains were cryoprotected overnight (0.1 M phosphate buffer and 30% sucrose) at 4 °C and frozen in isopentane at − 40 °C for 1 min. Coronal sections were cut on a cryostat and collected in PBS for immunostaining. Free-floating brain sections were washed 4 times with PBS and incubated with PBS-H₂O₂ 3% for 1 h. The sections were subsequently washed 4 additional times with PBS, permeabilized and blocked with a blocking solution (PBS, NGS 3% and Triton-X 0.3%) for 1 h. The sections were incubated for 2 h with the chicken antibodies (IgY-Lin, IgY-3D-1/2 and 3) at a concentration of 5 μg/ml in a dilution solution (PBS, NGS 1% and Triton-X 0.3%). After washing, goat anti-chicken HRP-conjugated antibodies (Immune Biosolutions, Sherbrooke, CA) were added (1:200) to the dilution mixture, which was subsequently incubated for 1 h and washed again. Solutions A (1:100) and B (1:100) were diluted in the dilution solution, added to the slides and incubated for 1 h. After washing, DAB solution (PBS, diaminobenzidine 0.05%, H₂O₂ 0.015%) was added for 3 min, and the reaction was stopped with PBS. All steps were performed at RT under agitation. The brain sections were dried and coverslipped with Permount® Mounting Media (SP15-500, Fisher Thermo Scientific Inc., NY, USA). Images were acquired with a Leica DM4000 microscope with Infinity Capture Software v6.5.6.

Flow cytometry

Live HEK293A cells stably expressing HA-rNTS1 were harvested from a 10 cm petri dish with trypsin. The cells were then washed with PBS and subsequently blocked for 30 min on ice with 8% PBS-NGS. Primary antibodies against both chicken polyclonal anti-NTS1 (Lin, 3D-1, 3D-2 and 3D-3) and mouse anti-HA (SC-7392, clone F-7, Santa Cruz, California, USA) were added to 500,000 cells at a total amount of 1 μg and 0.2 μg, respectively, and incubated for 15 min on ice. The cells were then washed twice and incubated for 15 min on ice with secondary antibodies, Alexa Fluor488-conjugated goat anti-chicken IgY antibody (A-11039, Thermo Fisher Scientific Inc., Massachusetts, USA) and Alexa Fluor 647-conjugated goat anti-mouse antibody IgG (A-21235, Thermo Fisher Scientific Inc., Massachusetts, USA), at a 1:200 dilution. The cells were washed twice and resuspended in PBS. Fluorescence analysis was performed via a Beckman CytoFLEX flow cytometer and CytExpert software v 2.3.0.84.

Western blots

HEK293A cells were transiently transfected with equal amounts (6 μg) of HA-tagged rNTS1, HA-tagged rNTS2, HA-tagged rAPJ or the empty vector as a negative control. Whole cells were lysed in reducing lysis buffer (280 mM NaCl, 50 mM Tris, 0.015 g DTT, 0.5% NP-40, pH 8.0) or native lysis buffer (280 mM NaCl, 50 mM Tris, 0.5% sodium deoxycholate (NaDOC), 1% n-dodecyl β-D-maltoside (DDM)) supplemented with complete protease inhibitor cocktail (Roche, Bale, Switzerland) and sonicated for 30 s 48 h post-transfection. Whole-cell lysates were further processed for immunoprecipitation and/or polyacrylamide gel electrophoresis (PAGE). Finally, the lysates were heated at 37 °C in denaturing/reducing protein sample buffer (60 mM Tris-HCl pH 6.8, 10% glycerol, 0.002% bromophenol blue, 2% SDS, 2% 2-mercaptoethanol) or native protein sample buffer (31.25 mM Tris-HCl pH 6.8, 25% glycerol, 0.005% bromophenol blue). The protein samples (20 μg of total protein per well) were loaded and resolved via SDS‒PAGE or native PAGE. After transfer to nitrocellulose membranes, the blots were probed with different chicken antibodies (IgY-Lin, IgY-3D-1, IgY-3D-2 and IgY-3D-3) at a concentration of 1 μg/ml or a chicken anti-HA antibody (1:1000, Immune Biosolutions Inc., Sherbrooke, QC, CA), followed by a secondary antibody, goat anti-chicken HRP (1:5000, Immune Biosolutions Inc., Sherbrooke, QC, CA). A rat monoclonal anti-HA high-affinity HRP-conjugated antibody (1:1000, Roche, 12013819001, Sigma‒Aldrich, St. Louis, Missouri, USA) was used to detect the protein expression levels of the whole-cell lysates in the selectivity assay. All the blots were visualized via a chemiluminescence detection system (Bio-Rad Clarity Western ECL, Bio-Rad, Hercules, CA, USA). Images were recorded on a ChemiDoc MP using the software Image Lab version 5.0.

Immunoprecipitations

Equivalent molar ratios of all chicken antibodies tested (IgY-Lin, IgY-3D-1, IgY-3D-2, IgY-3D-3 and IgY anti-HA) were conjugated to maleimide magnetic beads using Traut’s reagent (2-iminothiolane) to introduce free thiols (Immune Biosolution, Inc.). The antibody-coupled magnetic beads (10 μl) were added to each sample (200 μl of whole-cell lysate at 1 mg/ml total protein) and to negative controls, which consisted of lysates of cells transfected with the empty vector or samples containing the targeted receptor with a mixture of preincubated IgY peptide at a 1:10 ratio. All the samples were incubated on a rotator at 4 °C overnight. The next day, the protein‒antibody‒magnetic bead complexes were washed with 0.01% PBS-T four times, followed by 4 washes with PBS without Tween in the case of samples subjected to mass spectrometry. Immunoprecipitated proteins were eluted with elution buffer (0.5 M ammonium hydroxide) and processed for Western blotting as described above or snap-frozen and further lyophilized (program: − 40 °C for 6 h; − 21 °C for 10 h; 20 °C for 4 h) via a stoppering tray dryer (Labconco, Kansas City, MO, USA). The samples were sent to two sequencing platforms for mass spectrometry: PhenoSwitch (Sherbrooke, QC, CA) and the proteomic platform of the research center of CHUL (Laval University, QC, CA).

Mass spectrometry and protein identification

Proteins were reconstituted in 100 µl of 50 mM Tris pH 8 + 0.75 M urea, reduced with 10 mM DTT for 15 min at 65 °C and alkylated with 15 mM iodoacetamide for 1 h at RT in the dark. Trypsin/LysC (1 µg) was added to each sample, and the proteins were digested overnight at 37 °C with agitation. The digestion was stopped by the addition of 2% formic acid, and the peptides were purified via reversed-phase SPE. Acquisitions were performed with an ABSciex TripleTOF 5600 (ABSciex, Foster City, CA, USA) and a Thermo Orbitrap Fusion Tribrid (Thermo Fisher Scientific Inc., Logan, UT, USA) by Phenoswitch and the Proteomic Platform of the Université Laval, respectively. Protein identification from Phenoswitch was performed with ProteinPilot V4.5 beta (ABSciex) with the instrument preset for TripleTof5600, with iodoacetamide as a cysteine alkylation as a special factor. A thorough search with false discovery rate analysis was performed with a biological modification emphasis against the rat (Rattus norvegicus) proteome (with custom protein sequences added). For protein identification and data analysis, the global false discovery rate (FDR) was set at 1%, and the local false discovery rate was set at 5%. Peptides for Neurotensin Receptor 1 were reviewed manually to ensure that they met a basic quality standard. The proteomic platform of the Université Laval analyzed all MS/MS samples via Mascot (Matrix Science, London, UK; version 2.5.1). Mascot was set up to search the CP_RattusNorvegicus_10116_CO_20161019 database (unknown version, 31602 entries), assuming the digestion enzyme trypsin. Mascot was searched with a fragment ion mass tolerance of 0.60 Da and a parent ion tolerance of 10.0 PPM. The carbamidomethyl of cysteine was specified in Mascot as a fixed modification. Deamidation of asparagine and glutamine and oxidation of methionine were specified in Mascot as variable modifications. A scaffold (version Scaffold_4.7.5, Proteome Software Inc., Portland, OR) was used to validate MS/MS-based peptide and protein identification. Peptide identifications were accepted if they could be established at greater than 99.0% probability to achieve an FDR less than 1.0% by the scaffold local FDR algorithm. Protein identifications were accepted if they could be established at greater than 99.0% probability to achieve an FDR less than 1.0% and contained at least 1 identified peptide.

Cyclic AMP and IP-One production measurement

CHO-K1 cells stably expressing HA-rNTS1 (cAMP) or HA-hNTS1 (IP-One) were used to perform competitive immunoassay via homogeneous time-resolved fluorescence (HTRF) technology (cAMP-Gs dynamic or IP-One-Gq dynamic kits, Cisbio US, Bedford, MA, USA). The cells (2000/well) were either stimulated with NT(8-13) (6 nM for cAMP or 0.3 nM for IP-One (EC₅₀) and 10 μM (EC_max) alone or with a mixture of NT(8-13) (6 nM or 0.3 nM) and each IgY tested (-Lin, -3D-1, 3D-2 and 3D-3) for 30 min at 37 °C in an Alpha-Plate 384 SW (PerkinElmer, Waltham, MA, USA). Unstimulated cells were used as a negative control. Reagents were added as described in the manufacturer’s protocol. Fluorescence was measured at 620 nm and 665 nm with a Tecan GENios plate reader and Tecan XFluor4 software (Tecan Austria GmbH, Austria).

Binding assay

HEK293 cells expressing HA-tagged rat NTS1 (HA-rNTS1) were frozen when they reached 80% confluency. The cells were scraped off the dish with 10 mM Tris and 1 mM EDTA, pH 7.5, and centrifuged at 15,000 × g for 5 min at 4 °C. The pellet was then resuspended in binding buffer. Competitive radioligand binding experiments were performed by incubating 15 μg of cell membranes expressing the HA-rNTS1 receptor with 45 pM ¹²⁵I-[Tyr³]-NT (2200 Ci/mmol) in binding buffer (50 mM Tris-HCl, pH 7.5, 0.2% BSA) in the presence of increasing concentrations of chicken antibodies (IgY-Lin, IgY-3D-1, IgY-3D-2 and IgY-3D-3) for 60 min at 25 °C. After incubation, the binding reaction mixture was transferred to polyethylenimine-coated 96-well filter plates (glass fiber filters GF/B, Millipore, Billerica, MA). The reaction was terminated by filtration, and the plates were washed three times with 200 μL of ice-cold binding buffer. The glass filters were then counted in a γ-counter (2470 Wizard2 software v 1.0.8, PerkinElmer). Nonspecific binding was measured in the presence of 10⁻⁵ M unlabeled NT(8-13), which represented less than 5% of the total binding. IC₅₀ values were determined from the competition curves as the unlabeled ligand (antibodies) concentration inhibiting half of the ¹²⁵I-[Tyr³]-NT-specific binding.

Bioluminescence resonance energy transfer (BRET)-based biosensor assays

All BRET assays were performed at Domain Therapeutics NA, Inc. (Montreal, QC, Canada). All the experiments were performed in HEK293T cells transiently expressing the HA-rNTS1 receptor and processed as previously described with minor modifications⁹¹. Briefly, 10,000 cells were seeded in an Alpha-Plate 384 SW (PerkinElmer, Waltham, MA, USA) and transfected with the indicated G protein biosensors (Gβγ and Gα subtypes Gq, G13, GoB, Gi2, Gz) or β-arrestin-2 biosensors. Two days post-transfection, the agonist mode was first tested by treating cells with increasing amounts of NT (1-13) (0.1 pM to 0.1 mM), SR48692 (0.1 pM to 0.1 mM) and chicken antibodies, namely, IgY-Lin, IgY−3D-1 and IgY−3D-3 (0.02 µM to 2 µM), for 1 h. The cells were subsequently washed with Tyrode-HEPES buffer (Sigma, Cat # T2145-H9136) and allowed to equilibrate for 60 min at RT. Then, e-Coelenterazine Prolume Purple (1.8 μM, Methoxy e-CTZ; Nanolight, Cat # 369) was added to each well, and a first read was acquired on a Synergy NEO plate reader with BioTek Gen5 software. For the antagonist mode, an EC50 (1 nM) of NT(1-13) was added to the previous wells, which were treated with SR48692 and antibodies, and a second measurement was performed after a 10 min incubation. The BRET signal was calculated as the ratio of acceptor emission to donor emission.

Statistics

All values are expressed as the means ± S.E.M. unless otherwise indicated. GraphPad Prism was used to perform all the statistical tests. One-way ANOVA was generally used to measure the significance of differences between groups, followed by Tukey’s multiple comparisons test. Curves were evaluated via nonlinear regression with three or four parameters, as described in the figure legends.

Ethics satement

All work involving animals and related protocols were approved by the Animal Care Committee of the University of Sherbrooke (protocol n° 2020-2669).

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

M.E.N.T. designed and performed experiments, analyzed data, prepared figures and wrote the first draft of the manuscript. P.L.B. designed experiments, analyzed data and wrote manuscript sections. L.B.C. and M.H. designed experiments and performed and analyzed the molecular dynamics simulations. A.M. and H.T. designed and synthesized the peptides. E.B.O. performed the binding experiments. R.B. performed the HTRF experiments. K.K. performed the IHC experiments. S.M. performed the BRET experiments. J.M.L. provided advice and edited the manuscript. B.B. provided advice and supervised the BRET experiments. F.P.G. provided advice and obtained funding for the study. S.G. developed the concept, designed experiments, supervised all the antibody production and obtained funding for the study. P.S. developed the concept, designed experiments, analyzed data, prepared figures, obtained funding for the study, supervised the study and wrote the manuscript.

Peer review

Peer review information

Nature Communications thanks Andrea Pasquadibisceglie and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

Source data is provided with this paper as a Source data file. The authors declare that the data supporting the findings of this study are provided within the paper, its Supplementary Information, and the Source Data files. Molecular dynamics trajectories have been deposited at 10.5281/zenodo.17246301. Any additional raw data or processed data not included in the Source Data files can be accessed from the corresponding author. Access is unrestricted, and data will remain available for the duration of the archival requirements.

Competing interests

M.E.N.T., P.L.B., H.T., S.G., E.M., and P.S. are patent holders related to this work. S.G. is also a scientific cofounder of Immune Biosolutions. B.B. serves as the Chief Scientific Officer (C.S.O.) of Domain Therapeutics. The remaining authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-025-66030-1.