Analyzing the Molecular Effects of Endomorphin-2 Degradation on Stabilizing Interactions at the μ-Opioid Receptor

Abstract

Background: Endogenous opioids, such as endomorphin-2, are key regulators of the body’s pain pathways and mediate analgesia by engaging the $\mu$-opioid receptor. This class of opioids are distinguished by their transient activation of the $\mu$-opioid receptor, which is attributed to rapid enzymatic degradation. Methods: To understand how degradation of endomorphin-2 by the enzyme DPP IV affects its interaction with the $\mu$-opioid receptor, we analyzed the ligand–receptor conformational dynamics and interaction patterns of molecular dynamics simulations data of morphine, fentanyl and endomorphin-2 and one degradation product Phe-Phe-NH₂, using molecular fingerprints and the mathematical framework ISOKANN. Results: Our analyses revealed that both the clinically relevant opioids, morphine and fentanyl, as well as the endogenous opioid endomorphin-2, adopt a set of recurring binding conformations within the $\mu$-opioid receptor binding pocket, maintaining overlapping interaction motifs throughout the simulations. In contrast, Phe-Phe-NH₂ failed to maintain a persistent binding mode over the simulated timescale. This instability arises from the dipeptidyl peptidase IV mediated cleavage of endomorphin-2, which generates Phe-Phe-NH₂ and removes critical proline and tyrosine residues, thereby leading to the loss of stabilizing hydrophobic contacts with receptor residues Tyr150^3,33, Val238^5,43 and Val302^6,55. Conclusion: By mapping structural interaction motifs essential for stable $\mu$-opioid receptor binding, this study provides mechanistic insights into how endogenous degradation reshapes ligand–receptor interactions.

1. Introduction

The $\mu$-opioid receptor serves as a primary target receptor for pain relief [1,2,3,4]. Its ligands, collectively referred to as opioids, include a broad range of compounds such as fentanyl, methadone and morphine, all of which effectively bind to the $\mu$-opioid receptor and are widely used as analgesics [5]. Despite their high efficacy in relieving pain, opioids are associated with severe side effects such as respiratory depression, constipation, and dependence, mainly by binding to the $\mu$-opioid receptor, which is ubiquitously expressed in the human body [1,2]. In particular, fentanyl has gained significant attention as a major contributor to the ongoing opioid epidemic [1,6].

These concerns linked to synthetic opioids stress the importance of understanding the body’s own opioid system, where naturally synthesized peptides, known as endogenous opioids, play a vital role in regulating various physiological processes, including pain relief and stress resilience [7].

Prominent endogenous opioids include the class of dynorphins, enkephalins and endomorphins, of which endomorphins are known to have a high selectivity for the $\mu$-opioid receptor compared to other opioid receptors [7,8]. Endomorphin-2, which is a commonly known endogenous opioid belonging to the group of endomorphins, is notable for its high potency, which is comparable to that of fentanyl. However, unlike pharmaceutical opioids that are administered systemically, endomorphin-2 acts as an endogenous ligand whose physiological effects are tightly regulated by localized release and rapid enzymatic degradation, thereby limiting prolonged or widespread $\mu$-opioid receptor activation that is typically associated with adverse effects.

In a previous study, we examined the effects of three pharmaceutical opioids and endomorphin-2 on cAMP levels in enteric neurons using a mathematical modeling approach. Our findings revealed that, following endomorphin-2 administration, side effects such as constipation, which is influenced by reduced cAMP signaling in enteric neurons among other mechanisms, rapidly returned to baseline [9]. This phenomenon is primarily attributed to the rapid enzymatic degradation of endomorphin-2 [10].

While enzymatic degradation governs the stability and lifetime of endogenous opioids at the receptor, the functional outcomes of opioid signaling are dictated primarily by the specific interaction patterns through which opioids first engage the $\mu$-opioid receptor [3,4]. These initial interactions not only influence receptor activation and signaling biases but also ensure a stable ligand–receptor conformation that is essential for receptor signaling [3,4,11,12]. Several $\mu$-opioid receptor residues have been identified as crucial determinants, influencing receptor signaling and stabilizing the ligand within the binding pocket of the receptor. For example, Asp149^3,33 (superscripts indicate Ballesteros–Weinstein numbering [13]) acts as a critical interaction partner by forming a conserved salt bridge with most opioid ligands [3,6,14]. This interaction is essential for ligand stabilization within the binding pocket and downstream receptor signaling [3,6,14]. Furthermore mutations at Tyr328^7,43 and Trp320^7,34 have been shown to influence signaling biases [12,15,16]. In particular, mutations in Tyr328^7,43 increase activity in the $\beta$-arrestin pathway, whereas mutations in Trp320^7,34 lead to complete loss of the $\beta$-arrestin signaling pathway. In addition, several residues within transmembranes 5, 6 and 7 contribute to shaping the hydrophobic properties of the binding pocket that modulate ligand orientation and ensure a stable ligand–receptor conformation. For instance, Val302^6,55 and Ile298^6,51 form a hydrophobic cluster that stabilizes the ligand mainly through van der Waals contacts [5,17]. The residue His299^6,52 often acts as hydrogen bond donor or acceptor, mediating ligand recognition and influencing receptor selectivity [5,17]. Moreover, Trp318^7,32 and Ile322^7,36, located on the upper region of transmembrane 7, support the aromatic ring of many opioids, stabilizing the structure of the receptor in the active state and therefore act as a crucial interaction partner for $\mu$-opioid receptor signaling [3,18].

While these structural insights have significantly advanced our understanding of $\mu$-opioid receptor signaling and the stability of ligand–receptor conformations, little is known about how the degradation of endogenous opioids affects these critical interactions.

One of the enzymes known to be responsible for the degradation of endomorphin-2 is Dipeptidyl Peptidase IV (DPP IV), a membrane-bound serine [10]. The enzyme DPP IV cleaves peptides at the amino terminus, when the proline residue is at the second to last position, as is the case for endomoprhin-2 (sequence: $\mathrm{Tyr}-\mathrm{Pro}-\mathrm{Phe}-\mathrm{Phe}-{\mathrm{NH}}_2$).

Cleavage of endomorphin-2 by DPP IV yields the fragment Phe-Phe-NH₂. While Phe-Phe-NH₂ has been studied in the context of the Neurokinin-1 (NK1) receptor and is established as a candidate antagonist that reduces nociceptive pain, its role in relation to the $\mu$-opioid receptor has not yet been investigated [19,20]. Consequently, it remains unclear which key molecular interactions are lost upon DPP IV-mediated cleavage of endomorphin-2 and how this degradation may affect downstream signaling and the stability of receptor–peptide conformations. To address these mechanistic questions, we employed molecular dynamics (MD) simulations to investigate and compare the interaction profiles of two well-characterized opioids, morphine and fentanyl, alongside those of the endogenous opioid endomorphin-2 and one of its degradation products, Phe-Phe-NH₂. The aim of this analysis was to identify key receptor residues crucial for forming stable $\mu$-opioid receptor–ligand conformations across structurally distinct ligands and to investigate how the loss of specific ligand–receptor interactions upon degradation might alter receptor engagement and with that potential downstream signaling. By contrasting the binding and interaction patterns of stable pharmaceutical opioids with those of the rapidly degraded endogenous ligand endomorphin-2, we sought to identify molecular contacts that are lost or weakened due to enzymatic degradation by DPP IV.

These insights may help clarify how specific molecular interactions govern receptor engagement, shedding light on the mechanistic differences between endogenous opioid signaling, which is rapidly impaired by enzymatic degradation, and the more persistent signaling induced by synthetic or clinically relevant opioids.

2. Methods

2.1. Molecular Dynamics Simulations

Molecular dynamics simulations were conducted using the human $\mu$-opioid receptor structure from the RCSB database (Protein Data Bank (PDB) [21]: 8K9K [22]). Endomorphin-2, fentanyl and morphine were docked into the $\mu$-opioid receptor binding pocket using AutoDock Vina [23,24] (CHARMM-GUI, https://www.charmm-gui.org/ accessed on 1 December 2025). The first docking pose was selected for subsequent MD simulations and ISOKANN.

Each receptor–ligand complex was embedded in a POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) lipid bilayer model using the CHARMM-GUI Membrane Builder [25]. Following the protocols described in [26,27], MD simulations were performed using GROMACS 2024. In brief, the CHARMM36 forcefield was applied to the ligands, receptor and lipids, while solvation was carried out using the CHARMM TIP3 water model [28]. Sodium and chloride ions were added to neutralize the system and achieve a physiological salt concentration of 0.15 M. Energy minimization, equilibration and production simulations were then executed using the GROMACS input files provided by CHARMM-GUI, with production runs extending up to 10 ns [29,30]. The simulations were designed with a 10 ns production timescale to probe early binding events and enable comparison of interaction patterns between ligand–receptor complexes. To improve statistical robustness, three independent MD simulation replicates were performed for each receptor–ligand model. In addition, docking was performed in multiple independent runs, and for each system, the energetically most favorable pose was selected for subsequent simulations.

2.2. Ligand Docking

Molecular docking was performed using the AutoDock Vina within the CHARMM-GUI interface. Potential binding sites in the $\mu$-opioid receptor were identified using the Binding Site Prediction Tool available in CHARMM-GUI. For endomorphin-2, fentanyl and morphine, the top-ranked docking poses, corresponding to the most energetically favorable conformations, were selected for further MD simulations. However, AutoDock Vina failed to produce a docking pose for the degradation byproduct Phe-Phe-NH₂, likely due to its increased flexibility and non-canonical chemical structure. Therefore, DiffDock [31], a deep learning-based docking approach, was employed to generate plausible binding conformations for this compound, as it is designed to handle highly flexible structures compared to conventional docking tools [31].

2.3. Protein–Ligand Interaction Analysis and Visualization

Molecular analysis of protein–ligand interactions was performed using ProLIF version 2.0.3 [32], which enables flexible and extensible interaction fingerprinting for complexes involving proteins and structurally diverse ligands, and has been designed to support integration of molecular dynamics datasets. Interaction fingerprints were generated using a distance threshold of 4 Å, as most relevant interactions, such as hydrogen bonds, van der Waals contacts, and hydrophobic interactions, occur within this range. The snakeplot and helix box representations of the $\mu$-opioid receptor were generated using the GPCRdb webtool [33] (GPCRdb, https://gpcrdb.org/protein/oprm_human/ accessed on 23 December 2025). Identified interactions with ProLIF were cross validated with PLIP [34]. For visualizations of 2D contact maps, the ProteinsPlus web tool was employed [35]. The chemical structures of endomorphin-2 and Phe-Phe-NH₂ were drawn with Marvin JS by Chemaxon [36]. Molecular images of the ligand, receptor and membrane were generated using PyMOL version 3.1.0 Open-Source [37].

2.4. In Silico Preparation of Endomorphin-2 Cleavage Byproducts

The molecular structure of the endomorphin-2 degradation byproduct was obtained from chemicalbook [38]. Minimal modifications were made using the CHARMM-GUI Ligand Reader & Modeler [39] to prepare the molecule for subsequent molecular MD simulations.

2.5. ISOKANN

ISOKANN was used to identify metastable interaction states from the MD simulations through membership functions $\chi$ [40]. As input features, we calculated pairwise distances between heavy ligand atoms and receptor residues, retaining only those atom pairs that approached on average within 4 Å in at least one replica. To reduce redundancy, each residue–ligand atom combination was represented by a single feature corresponding to the smallest observed distance among all possible pairs. A Monte-Carlo estimator of the Koopmann Operator was obtained by pairing each state with its subsequent state along the trajectories. The resulting features were provided to ISOKANN to train a neural network identifying the metastabilities. One neural network was trained per ligand to ensure reproducibility and comparability across the replicas. We characterize the learned slow process using the explainable-AI-based AMORE-MD framework [41]. Specifically, we use ensemble-averaged gradient components per $\mu$-opioid receptor residue to identify binding-pocket rearrangements to which the $\chi$-function is most sensitive. Representative pathways are obtained through a post-processing procedure that connects states that are close both in membership function value and in the Cartesian coordinates used for training [42].

3. Results

3.1. Synthetic and Endogenous Opioids Show Different Interaction Profiles with the MOR in the Fingerprint Space

To compare how morphine, fentanyl, endomorphin-2 and its degradation byproduct Phe-Phe-NH₂ interact with the $\mu$-opioid receptor, taking into account all interactions, we generated molecular interaction fingerprints of each ligand–receptor MD simulation, as detailed in Section 2.3. These fingerprints encode the presence, type, and distance of ligand–receptor interactions over time, providing a compressed representation of how each ligand interacts with the receptor.

We first applied principal component analysis (PCA) to the interaction fingerprints to identify global patterns of ligand–receptor interactions. The first two principal components (PC1 and PC2) accounted for 31.1% and 16.0% of the total variance, respectively (see Figure 1). The PCA analysis highlighted distinct clustering patterns among the ligands. Notably, fentanyl (shown in orange in Figure 1) clustered in the opposite direction along PC1, relative to Phe-Phe-NH₂, indicating a very distinct interaction profile. Meanwhile, morphine (blue data cloud in Figure 1) showed a distinct distribution along PC2, clearly separating it from fentanyl. Endomorphin-2 (green data cloud in Figure 1) clustered in the center of the PCA space. This pattern was consistent in all replicates. To identify the specific receptor–ligand interactions responsible for the observed clustering patterns, we analyzed the residue interaction contributions for each principal component. Along PC1, which distinctly separates fentanyl from Phe-Phe-NH₂, the most negatively contributing interactions are hydrophobic contacts with Lys235^5.62, Ile303^6.54, Leu234^5.61, Val238^5.65 and Tyr150^3.33, indicating that fentanyl formed strong interactions with these residues compared to Phe-Phe-NH₂. Hydrophobic interactions with Gln126^2.60, Ile146^3.29, Val145^3.28, Trp(ECL1) Cys219(ECL2) were the strongest positive contributors and were more profound in Phe-Phe-NH₂. Along PC2, the strongest positive contributors included hydrophobic interactions with residues Tyr328^7.43, Trp295^6.46, Met153^3.36, Leu234^5.61 and Ile303^6.54. Hydrophobic interactions with residues Trp320^7.35 and cationic interactions with Asp149^3.32 contributed most negatively.

Since PC1 and PC2 accounted for only 31.1% and 16% of the variance, respectively, the PCA serves as a broad structural overview into how differently the opioid compounds interact with the $\mu$-opioid receptor. The degradation product Phe-Phe-NH₂ forms a distinct cluster along PC1, indicating a loss of core contacts, while fentanyl clusters narrowly, reflecting a conserved and stable binding mode. To gain deeper insight into the specific interaction patterns that distinguish the studied ligands from each other, we performed a more detailed analysis of the individual temporal interaction fingerprints. In the following section, we highlight key similarities and differences in how each ligand engages with the $\mu$-opioid receptor, with a focus on how these interactions are maintained or altered over time during the MD simulations.

3.2. Temporal Fingerprint Analysis Reveals Distinct Interaction Switches

The analysis of the Tanimoto similarity matrices of receptor–ligand interaction fingerprints revealed distinct clustering patterns among the ligand–receptor complexes (see Figure A2). For all four ligand–receptor models, two main clusters were identified. These clusters represent metastable binding poses in which the ligands adopt comparable orientations within the receptor’s binding pocket, thereby maintaining similar interaction profiles over consecutive simulation frames.

Morphine exhibited two main clusters, spanning frames 0–175 and 176–1000 (see Figure A2A). The first identified cluster of morphine, representing a dominant binding conformation adopted between frames 0 and 175, was characterized by persistent hydrophobic interactions with residues Trp295^6,48, Trp320^7,34 and strong hydrogen bonds with Tyr328^7,42 (see Figure 2A). In the second cluster, morphine primarily formed hydrophobic interactions with Val302^6,55, Phe239^5,44 and Ile324^7,38 (see Figure 2A). Throughout the course of the simulation, morphine maintained a hydrogen bond with Tyr150^3,33 and a salt bridge with Asp149^3,32, as well as hydrophobic interactions with Ile298^6,51 and Val238^5,43 (see Figure 2A).

For fentanyl, the first cluster extended from frame 0 to 404, while the second cluster spanned from frame 405 onward (Figure A2B). Cluster 1 showed prominent hydrophobic interactions with residues Ala119^2,53, Ile303 ^6,56, Phe239^5,44 and His299^6,52, as well as a salt bridge with Asp149^3,32. Cluster 2 revealed hydrophobic contacts with Leu234^5,39, Ile298^6,51, Trp295^6,48 and Val302^6,55. Fentanyl maintained hydrophobic interactions with Met153^3,36, Val238^5,43, Tyr150^3,33, Lys235^5,40 and Tyr328^7,42 throughout the course of the simulation (see Figure 2B). In contrast to morphine, the clustering results for fentanyl were less distinct, suggesting that fentanyl remains in a more consistent position and displays less movement within the binding pocket of the $\mu$-opioid receptor.

For endomorphin-2, two distinct clusters were detected between frames 0–564 and 565–1000 (see Figure A2C). The first cluster showed hydrophobic contacts with Ala119^2,53 and Met153^3,36, alongside hydrogen bonds involving Gln126^2,60 and Val145^3,28 (see Figure 2C, blue residues). In contrast, the second cluster displayed hydrophobic interactions with Ile324^7,38 and Leu221(ECL2), suggesting only a slight repositioning of the peptide within the binding pocket (see Figure 2C). For the course of the simulation, endomorphin-2 maintained interactions with the residues Trp295^6,48, Val302^6,55, Val238^5,43, Ile298^6,51, Trp320^7,34, Ile146^3,29, Tyr150^3,33, Asp149^3,32 and Tyr328^7,42.

The degradation product Phe-Phe-NH₂ formed clusters spanning frames 0–670 and 671–1000 (see Figure A2D). However, the corresponding Tanimoto similarity matrix for Phe-Phe-NH₂ did not display well-defined cluster boundaries. Instead, the trajectory alternated between short periods of similarity and marked dispersion, indicating that Phe-Phe-NH₂ does not adopt a consistently stable binding pose within the receptor during the early binding stages. The interaction profiles of the detected clusters also lacked distinct switching behavior; thus, only one unique interaction was identified for cluster 1, namely a hydrogen bond with Tyr328^7,42 (see Figure 2D). The degradation byproduct Phe-Phe-NH₂ maintained contacts with Gln126 and Asp149^3,32 as well as hydrophobic interactions with the residues Trp295^6,48, Ile298^6,51, Ile324^7,38, Ile146^3,29, Val145^3,28 and Trp135(ECL1).

Interestingly, we observed that all compounds formed some sort of interaction—either a hydrogen bond, a hydrophobic interaction, or a salt bridge—with residues Asp149^3,32, Ile298^6,51, Trp295^6,48, and Tyr328^7,42. In particular, interactions were found with residues Asp149^3,32, Trp295^6,48 and Tyr328^7,42, which have been described as crucial for $\mu$-opioid receptor activation and downstream signaling in previous studies [43,44].

To further understand the effects of endomorphin-2’s degradation on key interactions, we compared the interaction patterns of the $\mu$-opioid receptor agonists morphine, fentanyl and enodmorphin-2 with those of Phe-Phe-NH₂, the degradation product of endomorphin-2. This comparison aimed to identify essential contacts potentially lost after peptide cleavage. Notably, residues Tyr150^3,33, Val238^5,43 or Val302^6,55, consistently contacted by the studied agonists, were absent from the interaction profile of Phe-Phe-NH₂. Of these residues, Tyr150^3,33 is particularly important, as it is known to be critical for $\mu$-opioid signaling and for stabilizing several opioid ligands [43].

In the following section, we focus on a direct structural comparison between endomorphin-2 and Phe-Phe-NH₂, investigating how their conformational differences translate into distinct interaction patterns within the binding pocket of the $\mu$-opioid receptor.

3.3. Endomorphin-2 Drastically Changes Its Interaction Profile upon Degradation

We compared the interaction profiles of endomorphin-2 and its degradation byproduct Phe-Phe-NH₂ to identify similarities and most importantly key differences in their interaction profiles with the $\mu$-opioid receptor. The sequences of endomorphin-2 ($\mathrm{Tyr}-\mathrm{Pro}-\mathrm{Phe}-\mathrm{Phe}-{\mathrm{NH}}_2$) and Phe-Phe-NH₂ have a great overlap; thus, it is not surprising that they share interaction patterns with the $\mu$-opioid receptor (see Figure 3). Thus, both compounds form hydrogen bonds with residues Gln126^2,60, Asp149^3,32 and Ile324^7,38 and hydrophobic interactions with Tyr328^7,42, Trp295^6,48, Ile146^3,29, Ile298^6,51 and Val145^3,28. These residues have been identified to play crucial roles in $\mu$-opioid receptor signaling, suggesting that the compound Phe-Phe-NH₂ could potentially prompt downstream signaling. However, when directly contrasting the two compounds, it becomes clear that Phe-Phe-NH₂ is markedly smaller than endomorphine-2. This is due to the cleavage of the amino acids tyrosine and proline by Dipeptydil Peptidase 4 (DPPIV) at the N-terminius of endomorphin-2 (see Figure 3A).

The amino acid tyrosine, an aromatic residue, is a critical structural element in many endogenous opioids and plays a key role in anchoring the peptide in the binding pocket of the receptor [8,15]. In contrast, the amino acid proline, characterized by its unique structure containing a cyclic ring and a rigid structure, often acts as a helix breaker and contributes to the overall stability of endomorphin-2, helping the compound to be identified as an opioid by the $\mu$-opioid receptor [15].

In the degradation byproduct, both of these crucial structural components are missing, resulting in the fragment not interacting with residues located in transmembrane 4 and 5. Specifically, Phe-Phe-NH₂ failed to form meaningful interactions with residues such as Tyr150^3,33, Met153^3,36 or Trp320^7,34 (see Figure 3D). Endomorphin-2 mainly maintained contacts with these residues via tyrosine and proline (see Figure 3C).

These findings suggest that endomorphin-2 loses key structural features upon degradation, which likely causes a less stable conformation within the early stages of receptor binding. To further investigate the dynamic effects of degradation on the interaction patterns of endomorphin-2 beyond static interaction fingerprints, in the following section, we will mathematically study the metastability of ligand interaction patterns using the ISOKANN framework.

3.4. Metastability Analysis with ISOKANN

Using the ISOKANN [40] framework, we identified the presence of metastable states by analyzing the evolution of the learned $\chi$-function across the MD simulation trajectories. We simulated the ligands within the binding pocket of the activated human $\mu$-opioid receptor to study differences in the initial conformation between the activated receptor and the different ligands. The $\chi$-function separates slow collective variables from fast fluctuations, enabling the detection of long-lived metastable conformations directly from MD simulations [40]. The feature selection was performed as detailed in the Section 2.5.

For the pharmaceutical opioids morphine and fentanyl, as well as for the endogenous opioid endomorphin-2, the resulting $\chi$-functions showed clear transitions between metastable states. In contrast, the degradation byproduct Phe-Phe-NH₂ did not exhibit such behavior. Metastability could be detected for one single replicate and only for a short period of time, which could not be reproduced across replicates.

To gain insight into the structural determinants of metastability, we analyzed the pairwise distances that most strongly contributed to the $\chi$-function for each compound using the recently introduced explainable artificial intelligence framework for reaction coordinates AMORE-MD [41], as seen in Figure 4E.

A comparison of the strongest contributors among all compounds revealed a clear pattern, as seen in Figure 4E). The pairwise distance with residue Asp149^3,32 appeared to be the greatest contributor for most of the compounds, except for morphine, for which the greatest contributor was the distance to residue His299^6,52 (Figure 4).

Furthermore, we noticed an overlap in contributing distances and their magnitude in contributing to the $\chi$-functions of all compounds. These overlaps included distances to residues Tyr328^7,42, Val238^7,42 and Ile324^7,42. We further analyzed the top most contributing residue distances for each compound separately. Morphine’s most important contributors included distances to His299^6,52, Asp149^3,32 and Ala119^7,42. These residues largely overlap with the key interactions identified in the Tanimoto cluster analysis. Fentanyl showed major contributors involving Asp149^3,32, His299^6,52 and Met153^3,36, again corresponding to residues previously identified in the fingerprint analysis. Endomorphin-2’s strongest contributors were distances to Asp149^3,32, Tyr150^3,33 and Trp328^7,42. In contrast, the degradation product Phe-Phe-NH₂ exhibited a substantially reduced set of contributors. Distances to Tyr150^3,33 and His299^6,52 were completely missing, indicating that Phe-Phe-NH₂ did not get into close proximity with these residues during the simulation Figure 4A–D). Notably, we also identified contributors to the $\chi$-function of Phe-Phe-NH₂, which did not seem to have any effect on metastability of the ligand–receptor conformation of the other compounds. These contributors included distances to residues Ile146^3,29, Val145^3,28, Asn129^2,63 and Trp135(ECL2). This suggests that the compound Phe-Phe-NH₂ engaged mostly and strongly with transmembrane 2 and transmembrane 3 in comparison to the other compounds, which maintained interactions with residues of the other transmembranes (see Figure 4A–D).

Interestingly, the profiles of endomorphin-2, morphine and fentanyl showed significant overlap in the contributor profiles. In addition to the compounds previously mentioned (Tyr328^7,42, Val238^5,43 and Asp149^3,32), the residues overlapped in strength and presence of contributors in residues His299^6,52, Tyr150^3,33, Ile298^7,42. The reduced interaction partners and the lack of key interactions with conformation-stabilizing residues such as Tyr150^3,33 and His299^6,52 aligns with the weak metastability observed in both the $\chi$-functions and the fingerprint analysis and suggests that Phe-Phe-NH₂ is unable to adopt a stable receptor-bound conformation in the observed simulation period.

Notably, the analysis of the most strongly contributing distances revealed a substantial overlap among morphine, fentanyl, and endomorphin-2, both in the identity of the residues involved and in the magnitude of their contributions to the $\chi$-function. In addition to the previously identified key residues Asp149^3,32, His299^6,52, and Tyr150^3,33, shared contributions were observed for Val238^5,43, Ile324^7,38, Ile298^6,51, Trp295^6,48, and Met153^3,36. All of these residues have been previously identified as critical interactants governing $\mu$-opioid receptor signaling.

In summary, our ISOKANN analysis therefore independently confirmed our previously stated findings, demonstrating that the pharmaceutical opioids fentanyl and morphine and the endogenous opioid endomorphin-2 adopt stable binding poses within the $\mu$-opioid receptor throughout our simulation periods, whereas Phe-Phe-NH₂ lacks interactions with these core residues.

4. Discussion

In this study, we investigated the interaction profiles of morphine, fentanyl endomorphin-2, and its degradation by-product Phe-Phe-NH₂ to understand their structural differences at the molecular level.

The PCA of the interaction fingerprints provided a broad overview of global trends in ligand–receptor contacts and highlighted clear differences between the compounds. In particular, the analysis revealed that the degradation product Phe-Phe-NH₂ clustered in the opposite direction of fentanyl and the other compounds along the first principle component (PC1), indicating a distinct interaction pattern and a lack of interaction with key stabilizing residues such as Tyr150^3,33, Val238^5,43 and Val302^6,55.

The metastability analysis in the fingerprint space further strengthened these observations. Morphine, fentanyl, and endomorphin-2 exhibited well-defined and stable interaction clusters, while Phe-Phe-NH₂ displayed small and highly dispersed clusters without coherent interaction patterns. This observation for Phe-Phe-NH₂ aligns with the loss of proline and tyrosine after DPPIV cleavage, offering a mechanistic explanation for the loss of stability within the receptor binding pocket.

The ISOKANN analysis independently confirmed these findings, as morphine, fentanyl, and endomorphin-2 showed transitions into metastable states, while Phe-Phe-NH₂ did not. The residues that contributed most strongly to the $\chi$-function matched those highlighted in the fingerprint analysis, underscoring the functional importance of precise interaction networks within the $\mu$-opioid receptor. Furthermore, this demonstrated how subtle differences in the ligand interaction can influence receptor–ligand conformational dynamics.

Endomorphin-2 is highly potent but is rapidly degraded by enzymes such as DDPIV, resulting in the fragment Phe-Phe-NH₂. The degradation product Phe-Phe-NH₂ is missing the key structural features proline and tyrosine, two key residues for endomorphin-2 to adopt a stable conformation within the activated $\mu$-opioid receptor binding pocket. Although the short simulation timescales do not allow definitive conclusions about functional effects, our analyses suggest that degradation of endomorphin-2 by DPPIV and the associated loss of key interactions may contribute to its comparatively reduced analgesic efficacy after a certain time. This highlights that Phe-Phe-NH₂ may exhibit lower stability in the $\mu$-opioid receptor binding pocket, suggesting that Phe-Phe-NH₂ may not act as an efficient $\mu$-opioid receptor agonist. Interestingly, Phe-Phe-NH₂ mitigates pain through alternative pathways, by stably binding the SP1-7 binding site of the Neurokinin-1 (NK1) receptor [19,20]. The SP1-7 binding site is an alternating site on the NK1 receptor, named after a prominent compound that binds it, and is referred to as SP1-7 because it corresponds to the N-terminal fragment of NK1’s primary endogenous ligand, substance P. The NK1 receptor, similar to the $\mu$-opioid receptor, is involved in pain pathways [19,20,46]. Whereas activation of the $\mu$-opioid receptor reduces pain, NK1 receptor signaling transmits nociception and inflammation [19,46,47]. However, compounds such as SP1-7 and Phe-Phe-NH₂ bind to the SP1-7 binding site of the NK1 receptor and counteract substance P-mediated signaling and reduce pain [19,20,48].

Interestingly SP1-7 has also been shown to reverse severe side effects such as tolerance or dependence induced by morphine in mice and rats [48]. The receptor responsible for this effect could not be identified; however, the authors proposed three plausible mechanisms: mitigating the side effects indirectly via the NK1 receptor, binding of SP1-7 to an off-target binding site of the $\mu$-opioid receptor, or by activation of a yet undefined receptor [48]. These findings highlight the complexity of pain modulation and receptor signaling, suggesting that future work must also consider off-target sites of the $\mu$-opioid receptor and potential cross-talk with other receptors, particularly NK1. It is well known that $\mu$-opioid ligands can affect NK1 receptor signaling, and vice versa [46,47]. For example, endomorphin-2 binds the NK1 receptor and reduces inflammation, whereas substance P can reprogram $\mu$-opioid receptor signaling [46,47,48]. This raises the possibility that Phe-Phe-NH₂ and other opioid candidates with low affinity for the main $\mu$-opioid binding pocket may exert analgesic activity via alternative binding sites. To address this, future studies should extend interaction profiling analysis beyond the canonical binding pocket and include off-target receptor sites.

5. Conclusions

This analysis provides an initial molecular-level comparison of pharmaceutical opioids and the endogenous peptide endomorphin-2, focusing on early binding interactions. By examining these early stages, we can begin to identify key differences and overlap in the interaction patterns that may arise from structural features such as degradation susceptibility, which represents a major distinguishing factor between these two classes of opioids. While the relatively short simulation periods limit definitive conclusions regarding stable conformations or functional signaling, these simulations offer a framework to characterize interaction motifs and dynamics. Future studies incorporating longer simulations and experimental validation will be necessary to draw translational conclusions or inform drug discovery efforts. Nevertheless, this work establishes a foundation for critically contrasting endogenous opioids with their pharmaceutical counterparts, highlighting molecular features that could guide subsequent investigations.