Binding preference at the μ-opioid receptor underlies distinct pharmacology of cyclopropyl versus valeryl analogs of fentanyl

Abstract

Illicitly manufactured fentanyl is driving the current opioid crisis, and various fentanyl analogs are appearing in recreational drug markets worldwide. To assess the potential health risks posed by fentanyl analogs, it is necessary to understand structure-activity relationships for these compounds. Here we compared the pharmacology of two structurally related fentanyl analogs implicated in opioid overdose: cyclopropylfentanyl and valerylfentanyl. Cyclopropylfentanyl has a three-carbon ring attached to the carbonyl group on the fentanyl scaffold, whereas valerylfentanyl has a four-carbon chain at the same position. In vitro assays examining μ-opioid receptor (MOR) coupling to G proteins in CHO cells showed that cyclopropylfentanyl is a full agonist (EC₅₀=8.6 nM, %Emax=113%), with potency and efficacy similar to fentanyl (EC₅₀=10.3 nM, %Emax=113%). By contrast, valerylfentanyl is a partial agonist at MOR (EC₅₀=179.8 nM, %Emax=60%). Similar results were found in assays assessing MOR-mediated β-arrestin recruitment in HEK cells. In vivo studies in male CD-1 mice demonstrated that both fentanyl analogs induce naloxone-reversible antinociception and respiratory suppression, but cyclopropylfentanyl is 100-times more potent as an antinociceptive agent (ED₅₀=0.04 mg/kg, s.c.) than valerylfentanyl (ED₅₀=4.0 mg/kg, s.c.). Molecular simulation results revealed that the alkyl chain of valerylfentanyl cannot be well accommodated by the active state of MOR and may transition the receptor toward an inactive state, converting the fentanyl scaffold to a partial agonist. Taken together, our results suggest that cyclopropylfentanyl presents much greater risk of adverse effects when compared to valerylfentanyl. Moreover, the summed findings may provide clues to the design of therapeutic opioids with reduced adverse side effects.

1. Introduction

The United States of America is in the midst of an unprecedented opioid epidemic, with more than 80,000 overdose fatalities recorded during 2021 (Ahmad et al., 2022). Illicitly manufactured fentanyl is a major driving force in the current opioid crisis, and a number of fentanyl analogs have also appeared on recreational drug markets as standalone products, heroin adulterants, or constituents of counterfeit pain medications (Han et al., 2019; O’Donnell et al., 2021). As a means to assess the public health risks posed by fentanyl and its analogs, it is essential to investigate structure-activity relationships (SAR) for these compounds (Baumann et al., 2018). Fentanyl is a potent and selective agonist at the μ-opioid receptor (MOR), which activates both the G_i/o protein and arrestin signaling pathways (Stanley, 2014).

Cyclopropylfentanyl and valerylfentanyl are two fentanyl analogs associated with opioid intoxications and overdose deaths (Fagiola et al., 2019; Fogarty et al., 2018; Walsh et al., 2021). Cyclopropylfentanyl has been implicated in many overdose fatalities since 2017 (Fogarty et al., 2018), whereas valerylfentanyl has appeared more recently (Walsh et al., 2021). In most of the death cases involving either cyclopropylfentanyl or valerylfentanyl, other illicit drugs, including opioids, were present. From a chemical structure perspective, cyclopropylfentanyl has a three-carbon ring attached to the carbonyl moiety of the fentanyl scaffold, while valerylfentanyl has a four-carbon chain at the same position (see Figure 1). Despite the subtle structural difference between cyclopropyl and valeryl analogs, these compounds are reported to have striking differences in their affinity, potency, and efficacy at MOR (Astrand et al., 2020; Baumann et al., 2018; Eshleman et al., 2020; Vasudevan et al., 2020).

Figure 1.

Chemical structures of fentanyl, cyclopropylfentanyl, and valerylfentanyl.

Radioligand binding experiments carried out in rat brain tissue demonstrate that cyclopropylfentanyl and valerylfentanyl have greater affinity for MOR over the δ-opioid receptor (DOR) and the κ-opioid receptor (KOR) (Baumann et al., 2018). However, the MOR affinity of cyclopropylfentanyl (K_i=2.11 nM) is similar to that of fentanyl (K_i=2.76 nM), while the MOR affinity of valerylfentanyl is much weaker (K_i=53 nM) (Baumann et al., 2018). Opioid receptor binding assays carried out in Chinese Hamster Ovary (CHO) cells transfected with rat MOR confirm that cyclopropylfentanyl displays about 20-fold higher affinity for MOR when compared to valerylfentanyl (Eshleman et al., 2020). Functional studies carried out in CHO cells show that fentanyl and cyclopropylfentanyl are full agonists in the GTPγS assay (i.e., 88–89% of DAMGO efficacy) but valerylfentanyl is a weak partial agonist (i.e., 30% of DAMGO efficacy) (Eshleman et al., 2020). Vasudevan et al. compared the effects of various fentanyl analogs in Human Embryonic Kidney (HEK) cells co-transfected with human MOR and mini-Gi or β-arrestin to assess possible ligand bias (Vasudevan et al., 2020). Their findings reveal that fentanyl and cyclopropylfentanyl are full agonists in both the mini-Gi and β-arrestin 2 assays, whereas valerylfentanyl is a partial agonist, with efficacies ranging from 29–44%.

No studies to date have directly compared the in vivo pharmacological effects of cyclopropylfentanyl and valerylfentanyl, but the available evidence seems to support the in vitro findings that cyclopropylfentanyl is much more potent (Bergh et al., 2021; Varshneya et al., 2019; Walentiny et al., 2019). For example, a recent study by Bergh et al. (2021) found that subcutaneous (s.c.) administration of cyclopropylfentanyl to male rats produces antinociception in the hot plate test with an ED₅₀=48 μg/kg, close to the potency of fentanyl in the same test (ED₅₀=21 μg/kg) (Bergh et al., 2021; Vandeputte et al., 2022). By contrast, studies in mice show the potency of valerylfentanyl to induce antinociception in the warm water tail flick test (ED₅₀=6.43 mg/kg) is nearly 80-fold weaker than that of fentanyl (e.g., ED₅₀=21 μg/kg), close to the potency of morphine (ED₅₀=7.82 mg/kg) (Varshneya et al., 2019).

High-resolution structures of MOR in both inactive (Manglik et al., 2012) and active (Huang et al., 2015; Koehl et al., 2018) states can help to elucidate mechanistic details of ligand binding at MOR, as well as the coupling between MOR and its signaling proteins. Based on these structures, the binding mode of fentanyl at MOR has been extensively studied, but no consensus has been reached about its precise orientation in the binding pocket of MOR (de Waal et al., 2020; Ellis et al., 2018; Eshleman et al., 2020; Lipinski et al., 2019; Podlewska et al., 2020; Ricarte et al., 2021; Vo et al., 2021). Therefore, we previously evaluated the possible binding modes of fentanyl at MOR by integrating data obtained from simulations of either the fentanyl-alone or the fentanyl-MOR complex situation. Our results indicated that the most preferred fentanyl binding pose in the MOR binding pocket is with the amide and aniline groups pointing towards the orthosteric binding site and located near the middle portions of TMs 3 and 6 (termed as the APF pose, see the definition described in Figure S1), while an alternative pose with the opposite orientation (the FPA pose) is also possible (Xie et al., 2022).

Given the information discussed thus far, we wished to further investigate the pharmacology, as well as the action mechanisms, for cyclopropyl and valeryl analogs of fentanyl. To this end, the present study had two major aims: 1] to directly compare the in vitro and in vivo pharmacological effects of cyclopropylfentanyl and valerylfentanyl in mice and, 2] to use in silico methods to characterize the binding modes of cyclopropylfentanyl and valerylfentanyl at MOR and to shed light on the structural basis of their observed differences in potency and efficacy. A major strength of our study relies on the side-by-side assessment of the pharmacological effects of cyclopropyl and valeryl analogs using in vitro, in vivo, and in silico methods.

2. Materials and Methods

2.1. Chemicals and reagents

Fentanyl, cyclopropylfentanyl, valerylfentanyl, morphine, naloxone, [d-Ala²,MePhe⁴,Gly(ol)⁵]enkephalin (DAMGO), [D-Pen²,D-Pen⁵]enkephalin (DPDPE), and U50,488H were generously provided by the National Institute on Drug Abuse, Drug Supply Program (Rockville, MD, USA). Na¹²⁵I and [³⁵S]GTPγS were purchased from PerkinElmer (Waltham, MA, USA). Other chemicals were obtained from Fischer Scientific (Pittsburgh, PA, USA) or MilliporeSigma (St Louis, MO, USA) unless otherwise noted.

2.2. Animals

Male CD-1 mice weighing 25 – 35 g (Charles River Laboratories, Wilmington, MA, USA) were maintained on a 12 h light/dark cycle with food and water available ad libitum. Mice were housed in groups of five until testing. All animal experiments were reviewed and approved by the Institutional Animal Care and Use Committee of Memorial Sloan-Kettering Cancer Center.

2.3. Competition binding assay with [¹²⁵I]IBNtxA

Membranes were isolated from CHO cells stably transfected with murine MOR (mMOR), mDOR, or mKOR as described elsewhere (Pan et al., 1999). Membrane protein concentrations were determined using the Lowry method, with BSA as the standard (Lowry et al., 1951). The high affinity opioid receptor ligand, [¹²⁵I]IBNtxA, was synthesized by radiolabeling the benzoyl derivative of 6β-naltrexamine with Na¹²⁵I (Majumdar et al., 2011a). Competition assays were performed with [¹²⁵I]IBNtxA, as previously described (Majumdar et al., 2011b; Majumdar et al., 2012). Briefly, 0.1 nM [¹²⁵I]IBNtxA was incubated with 20 μg of membranes and various concentrations of fentanyl, cyclopropylfentanyl, valerylfentanyl, or morphine for 90 min at 30°C in binding buffer (50 mM KPO₄ and 5 mM MgSO₄, pH 7.4). Samples were then filtered through glass fiber filters under vacuum and washed three times with ice-cold filtering buffer (50 mM Tris-HCl, pH 7.5). Bound radioactivity was measured on a gamma counter (Wizard Gamma Counter 1470, PerkinElmer). Specific binding was defined as the difference between total binding and nonspecific binding, determined in the presence of levallorphan (8 μM). The IC₅₀ values were calculated by nonlinear regression analysis (GraphPad Prism 8, San Diego, CA, USA) and converted to the Ki values using the following formula: K_i = IC_50/(1+[[125I]IBNtxA]/K_D) where K_D is the dissociation constant for MOR (K_D = 0.11 ± 0.02 nM), KOR (K_D = 0.027 ± 0.001 nM), or DOR (K_D = 0.24 ± 0.05 nM), respectively, which was determined by saturation studies with [¹²⁵I]IBNtxA in the same CHO cells stably transfected MOR, KOR, or DOR, as described previously (Majumdar et al., 2011a). The B_max values obtained from the same saturation studies were 3.81 ± 0.42, 0.96 ± 0.2 and 0.57 ± 0.02 pmol/mg protein for MOR, KOR, and DOR, respectively. The K_D and B_max values were calculated by nonlinear regression analysis (GraphPad Prism 8). Data were analyzed by one-way ANOVA.

2.4. [³⁵S]GTPγS binding assay

The functional coupling of opioid receptors to G proteins was examined using [³⁵S]GTPγS binding assays in CHO cells transfected with mMOR, mDOR, or mKOR as previously described (Bolan et al., 2004). Fifty μg of membranes from the stable CHO transfectants were incubated in binding buffer (50 mM Tris-HCl, pH 7.7, 3 mM MgCl₂, 0.2 mM EGTA, 100 mM NaCl) containing 0.05 nM [³⁵S]GTPγS, 30 μM GDP and a protease inhibitor cocktail (leupeptin, bestatin, aprotinin, and pepstatin) for 60 min at 30°C in the presence of various concentrations of fentanyl, cyclopropylfentanyl, valerylfentanyl, or morphine. The incubation was terminated by rapid filtration through glass fiber filters under vacuum, followed by three washes with 3 mL of ice-cold 50 mM Tris-HCl (pH 7.4). Bound radioactivity was determined by liquid scintillation spectrophotometry after overnight extraction in 5 mL Liquiscint scintillation fluid (Fisher Scientific). Basal binding was assessed in the absence of drug. One μM DAMGO, DPDPE, or U50488H was used as the full agonist control for MOR, DOR, or KOR, respectively. Percent of maximum stimulation for each drug was calculated by normalizing the stimulation values to DAMGO for the MOR, or DPDPE for the DOR or U50,488H for the KOR. EC₅₀ and % maximum (%Emax) stimulation was calculated by nonlinear regression analysis. The EC₅₀ and %Emax values at MOR were analyzed by one-way ANOVA to determine significant differences among the compounds (GraphPad Prism 8).

2.5. β-arrestin 2 recruitment assay in HEK cells

G protein coupled receptor (GPCR) recruitment of β-arrestin 2 and AP2 was measured using the CISBIO β-arrestin recruitment assay (PerkinElmer). The Fluorescence Resonance Energy Transfer (FRET) signal in the CISBIO assay is triggered when 2 dyes - donor and acceptor - are in close proximity, and the intensity of fluorescence is directly proportional to the number of β-arrestin 2/AP2 complexes formed in the sample. For the assay, FLP-FRT-HEK cells stably expressing the human MOR (Cai et al., 2019) were washed with PBS buffer, dissociated from cell culture dishes with 0.05% trypsin-EDTA, and centrifuged at 1000 rpm for 5 min. The cell pellet was suspended in DMEM with 10% fetal bovine serum and counted; 24,000 cells/well (100 μL) were transferred to a 96-well plate coated with poly-D-lysine (to achieve optimal cell attachment), and was incubated overnight (at least 20 h) at 37°C. The next day, medium was removed, and cells were incubated at room temperature with 100 μL of compounds for 30 min and concentrations indicated, followed by addition of 30 μL of stabilization buffer (15 min incubation), and 3 washes with 100 μL wash buffer. Finally, overnight incubation at room temperature with 100 μL of Europium cryptate (donor fluorophore) and d2-labeled CISBIO β-arrestin 2 antibody (acceptor) pre-mixture in detection buffer (Cisbio Assays, Perkin Elmer). The levels of FRET signal were assessed by calculating the fluorescence ratio at 665 nm and 620 nm emissions, which were measured with a PHERAstar FSX plate reader (BMG Labtech, Cary, NC, USA).

2.6. Radiant heat tail flick assay in mice

Antinociception was measured at 10 min (fentanyl and its analogs) or 30 min (morphine) after subcutaneous (s.c.) administration of drugs using a mouse radiant-heat tail-flick assay, with a maximal latency of 10 sec to minimize tissue damage, as described elsewhere (Pan et al., 2009; Xu et al., 2015). Results were calculated as the percentage of maximum possible effect (% MPE) as follows: [(latency after drug – baseline latency)/(10 - baseline latency) × 100]. ED₅₀ values were determined using nonlinear regression analysis (GraphPad Prism 8). The opioid specificity of antinociceptive effects was determined by pretreatment with MOR antagonist naloxone (1 mg/kg, s.c.), 5 min prior to agonist drug administration. The effects of naloxone on drug-induced antinociceptive effects were evaluated using unpaired t tests (GraphPad Prism 8).

2.7. Breath rate test in mice

Breath rate was measured in freely moving mice using the MouseOx pulse oximeter system (Starr Life Sciences, Oakmont, PA, USA) as previously described (Majumdar et al., 2011b). Briefly, mice were shaved around the neck 24 h before testing. Mice were habituated to a CollarClip with the sensor clipped on the shaved neck for at least 1 h before testing. A 5 sec average breath rate was recorded at 5 min intervals. A baseline for each animal was recorded over a 25-min period before drug administration. Mice (n = 5 per group) received s.c. saline or drug administration, and testing began at 15 min postinjection and continued for 50 min. Data are reported as the percentage of baseline readings. The opioid specificity of respiratory effects was determined by pretreatment with MOR antagonist naloxone (1 mg/kg, s.c.), 5 min prior to agonist drug administration. Time-course data were analyzed by two-way analysis of variance (drug treatment × time) followed by Tukey’s post hoc test (GraphPad Prism 8).

2.8. Molecular modeling of human MOR in both active and inactive conformations

The construction of the human MOR (hMOR) models in the active state has been previously described (Xie et al., 2022). Briefly, the cryogenic electron microscopy (cryo-EM) structure of the mMOR in complex with the Gi protein (PDB 6DDF) was used as the main template to build our hMOR-Gi models. Eight missing N-terminal residues (residues 59–66 in hMOR numbering) and 7 missing H8 residues (residues 348–354 in hMOR numbering) in the structure 6DDF were added to the template based on two other MOR structures with PDB codes, 5CM1 and 4DKL, respectively, using homology modeling with Modeller (version 9.24) (John and Sali, 2003). The inactive structure of mMOR in complex with β-FNA (PDB 4DKL) was used as the main template to build our inactive hMOR model. Five missing N-terminal residues (62–66 in hMOR numbering) were added to the template based on the structure 5C1M. Intracellular loop 3 (residues 264–273 in hMOR numbering) was built with the loop module of Modeller. Val68^1.30, Val189^4.45, and Ile308^EL3 residues in mMOR were mutated to the aligned human residues in our hMOR models. In each of the Modeller runs for either the active or inactive hMOR modeling, the model with the lowest Discrete Optimized Protein Energy (DOPE) score was selected.

The selected active and inactive hMOR models were then processed through the Protein Preparation Wizard in Maestro of Schrodinger (version 2020–2). Using PROPKA (Olsson et al., 2011) at pH 7.0, the protonation states of the titratable residues were predicted, while the hydrogen bond assignment was optimized. See Table S4 for the protonation states of histidine residues in the MOR models. Energy minimization of the structure was conducted with the default constraint of 0.3 Å heavy atoms root-mean-square deviation (RMSD).

Fentanyl was docked into the prepared active and inactive receptor structures via the induced-fit docking (IFD) protocol (Sherman et al., 2006) implemented in Schrodinger suite. The center of the docking boxes was determined by the center of mass of the bound ligand in each structure. A hydrogen-bond constraint was applied to filter the poses that form an ionic interaction with D149^3.32. Both APF and FPA poses were identified from the docking results as described in our previous study (Xie et al., 2022). The cyclopropylfentanyl and valerylfentanyl bound hMOR models were established by modifying the bound fentanyl molecule in the equilibrated and representative hMOR models after prolonged MD simulations for each indicated condition (Table S1).

2.9. Molecular dynamics (MD) simulation protocol

The active and inactive state hMOR models were further processed to build the simulation systems with the Desmond System Builder of Schrodinger suites (version 2021–2 with the OPLS4 force field and version 2020–3 with the OPLS3e force field). In addition to proteins and lipids, the OPLS3e and OPLS4 force fields also include the parameters for small compounds, which have been extensively optimized and cover a wide chemical space (Lu et al., 2021; Roos et al., 2019). For the ligands modeled and simulated in this study, further force field parameter optimization was not required. Residues D116^2.50 and D166^3.49 in the active model were protonated to their neutral forms as assumed in the active state of rhodopsin-like GPCRs (Asher et al., 2022), and His299^6.52 was adjusted to the HID protonation state based on the findings from previous studies (Mahinthichaichan et al., 2021; Xie et al., 2022). A Na⁺ ion interacting with D116^2.50 was added to the inactive model as assumed in the inactive conformation of the receptor. The prepared models were immersed in explicit 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine lipid bilayer (POPC) and solvated with water layers on both sides of the lipid bilayer, defined by the simple point charge (SPC) water model. The net charge of the system was neutralized by Cl⁻ ions, and additional Na⁺ and Cl⁻ ions were added to achieve 150 mM NaCl ion strength. The process resulted in a system in an active state model with dimensions of 106×117×151 ų and approximately 190,000 atoms. The dimensions of the inactive system were 98×94×97 ų with approximately 92,000 atoms.

MD simulations were carried out through Desmond MD systems (D. E. Shaw Research, New York, NY), using the NPT ensemble. The constant temperature 310 K was maintained by Langevin dynamics, while 1 atm constant pressure was achieved with the hybrid Nose-Hoover Langevin piston method on an anisotropic flexible periodic cell. The system was initially minimized and equilibrated with restraints on the ligand heavy atoms and protein backbone atoms, similar to our previous simulation protocols for GPCRs (Lane et al., 2020). In the production runs, all restraints on both active and inactive states of hMOR were released; however, to retain the integrity of the Gi protein while allowing it adequate flexibility to interact with the receptor, the heavy atoms of residues 46–55, 182–189, and 230–242 of Gα, and the entire Gβ and Gγ subunits, were restrained with a force constant of 1 kcal/mol/Å.

At least three trajectories starting from different random number seeds for each condition were collected.

2.10. Conformational analysis

We used the Protein Interaction Analyzer (Michino et al., 2017; Stolzenberg et al., 2016) in analyzing the MD simulation results of the MOR. Note that this analysis of pairwise distances and their differences do not require frame alignment (and thus are not biased by how the frames are aligned). For analysis of coarse-grained interaction network of the hMOR, we defined the following subsegments of the transmembrane domain: TM1e (the extracellular section (e) of TM1, residues 68–74), TM1m (the middle section (m) of TM1, residues 75–84),TM1i (the intracellular section (i) of TM1, residues 85–97), TM2i (residues 105–117), TM2m (residues 118–126), TM2e (residues 127–131), TM3e (residues 140–48), TM3m (residues 149–157), TM3i (residues 158–172), TM4i (residues 183–193), TM4m (residues 194–200), TM4e (residues 201–207), TM5e (residues 229–240), TM5m (residues 241–248), TM5i (residues 249–259), TM6i (residues 275–293), TM6m (residues 294–301), TM6e (residues 302–307), TM7e (residues 314–323), TM7m (residues 324–331), and TM7i (residues 332–341).

We assembled the representative ensembles of frames for analysis by randomly selecting 5,000 frames with replacement (bootstrapping) for each condition from all the trajectories of that condition. The same datasets were used for all the geometric calculations and analyses. Pairwise Root Mean Square Deviations (RMSDs) can avoid the bias of a single reference. To evaluate the stability of ligand binding, we aligned all possible pairs of the representative MD frames for a given condition according to the Cα atoms of the binding site residues of hMOR: Thr122^2.56, Phe125^2.59, Gln126^2.60, Asn129^2.63, Trp135^EL1.50, Val1453.28, Ile146^3.29, Asp149^3.32, Tyr150^3.33, Met153^3.36, Asp218^EL2.49, Cys219^EL2.50, Trp295^6.48, Ile298^6.51, His299^6.52, Trp320^7.35, His321^7.36, Ile324^7.39, Gly327^7.42, and Tyr328^7.43, then calculated the RMSD based on the ligand heavy atoms.

2.11. MM/GBSA calculations

Molecular mechanics/generalized Born surface area calculations (MM/GBSA) analysis was carried out using the Prime of Schrodinger (version 2021–3) to estimate the binding free energies between bound ligands and the hMOR. VSGB2.1 solvation model with solvent dielectric constants of 80.0 was adopted. The binding free energies were calculated for every 3 ns frame from the production runs and averaged later.

3. Results

3.1. Cyclopropylfentanyl has much greater MOR affinity, potency, and efficacy in vitro when compared to valerylfentanyl

To characterize the pharmacology of the fentanyl analogs, we first carried out in vitro opioid receptor binding using [¹²⁵I]IBNtxA (Figure S2 and Table 1). IBNtxA is an atypical opioid analgesic that has sub-nM binding affinities for MOR (K_D, 0.11 nM), DOR (K_D, 0.24 nM), and KOR (K_D, 0.03 nM) (Majumdar et al., 2012). Results revealed that fentanyl had much higher affinity for MOR (K_i=7.6 nM) when compared to affinity for DOR (K_i=248 nM) and KOR (K_i=786 nM), indicating MOR selectivity. A similar scenario was seen for morphine binding. The results with fentanyl and morphine are similar to those observed in competition binding assays using more traditional tritiated opioid ligands (Bolan et al., 2004; Majumdar et al., 2011a; Obeng et al., 2021; Schmid et al., 2017). Cyclopropylfentanyl displayed MOR affinity (K_i=2.8 nM) that was slightly greater than fentanyl, whereas valerylfentanyl had much weaker MOR affinity (K_i=49.7 nM). A one-way ANOVA demonstrated that valerylfentanyl had significantly lower affinity for MOR when compared to the other drugs tested (F[3, 8] = 63.39, p<0.0001]. Both cyclopropyl and valeryl analogs exhibited much weaker affinity for DOR and KOR when compared to MOR (Figure S2 and Table 1).

Table 1.

Summary of in vitro and in vivo assay results.

Assay	Fentanyl	Cyclopropylfentanyl	Valerylfentanyl	Morphine
MOR [125I]IBNtxA Ki (nM)	7.6±3.0	2.8±1.0	49.7±4.5	6.8±0.7
DOR [125I]IBNtxA Ki (nM)	248±69	119±14	1443±121	242±24
KOR [125I]IBNtxA Ki (nM)	786±293	517±115	731±67	356±21
MOR [35S]GTPγS EC50 (nM) [%Emax]	10.3±1.0 [113±1 %]	8.6±1.5 [113±1%]	179.8±18.3 [60±1 %]	15.5±1.8 [121±3%]
DOR [35S]GTPγS EC50 (nM) [%Emax]	1078±255 [66±8%]	1062±260 [68±2%]	n.d.	1052±291 [66±5%]
KOR [35S]GTPγS EC50 (nM) [%Emax]	1165±273 [58±2%]	844±63 [73±4%]	n.d.	715±48 [92±5%]
MOR β-arrestin 2 EC50 (nM) [%Emax]	172.8±17.0 [70±4%]	85.7±7.2 [76±5%]	77.6±10.0 [24±7%]	DAMGO 267.9±21.8 [100%]
Antinociception ED50 (mg/kg, sc)	0.03±0.01	0.04±0.01	4.0±0.6	1.8±0.4

[¹²⁵I]IBNtxA (0.1 nM) binding assays for mMOR, mDOR, and mKOR were performed as described in Materials and Methods. Ki values are mean±SEM for N=3 experiments. [³⁵S]GTPγS (0.05 nM) functional assays for mMOR, mDOR, and mKOR were performed as described in Materials and Methods. EC₅₀ values are mean±SEM for N=3 experiments. %Emax reflects efficacy with respect to maximal stimulation induced by 1 μM DAMGO, DPDPE, or U50,488H for MOR, DOR, or KOR, respectively. β-Arrestin 2 assays for hMOR were performed as described in Materials and Methods. EC₅₀ values are mean±SEM for N≥4 experiments. %Emax reflects efficacy with respect to the efficacy of DAMGO. Antinociception was determined using the mouse radiant heat tail flick assay as described in Materials and Methods. ED₅₀ values are mean±SEM for N=6 mice per group. n.d. = not determined.

To evaluate the capability of the fentanyl analogs to stimulate G protein coupling, we tested the drugs in a [³⁵S]GTPγS binding assay, and determined maximal stimulation (%Emax) by normalizing the stimulation levels for each drug with those of the full agonists, DAMGO for MOR, DPDPE for DOR, or U50,488H for KOR (Figure S3 and Table 1). In the MOR [³⁵S]GTPγS assay, fentanyl, cyclopropylfentanyl, and morphine displayed similar low nM potencies and full-agonist efficacies, whereas valerylfentanyl was more than 10-fold less potent and only exerted 60% efficacy. A one-way ANOVA demonstrated that valerylfentanyl displayed significantly lower EC₅₀ (F[3, 8]=82.69, p<0.0001) and %Emax (F[3,8]=346.8, p<0.0001) at MOR, when compared to the other drugs tested. In the DOR assay, fentanyl, cyclopropylfentanyl, and morphine had similar low μM potencies, but valerylfentanyl did not stimulate [³⁵S]GTPγS binding and acted as a DOR antagonist. In the KOR assay, fentanyl and cyclopropylfentanyl showed lower efficacy than morphine, while valerylfentanyl appeared to inhibit [³⁵S]GTPγS binding and functioned as a weak partial or inverse agonist (Figure S3 and Table 1).

We also characterized the effects of the fentanyl analogs using a HTRF-based β-arrestin 2 recruitment assay. EC₅₀ and %Emax values from the β-arrestin 2 assays are summarized in Table 1. We found that morphine acts as a partial agonist at MOR-mediated β-arrestin 2 recruitment (data not shown), similar to the finding of others (Kelly, 2013; Narayan et al., 2021; Xu et al., 2017), where DAMGO was used as the reference full efficacy agonist. As shown in Table 1, cyclopropylfentanyl and fentanyl had somewhat lower efficacies than DAMGO, while cyclopropylfentanyl displayed slightly greater potency than fentanyl. By contrast, valerylfentanyl acted as very weak partial agonist. When the potencies of all compounds are compared in the [³⁵S]GTPγS and β-arrestin 2 assays, the potencies are reduced in the latter assay (see Table 1), which is consistent with previously reported trends (Kelly, 2013).

3.2. Cyclopropylfentanyl has much greater in vivo potency in mice when compared to valerylfentanyl

As a means to examine the effects of the compounds in vivo, we tested the antinociceptive effects of the fentanyl analogs in CD-1 mice using a radiant-heat tail-flick assay. The left panel of Figure 2 depicts the dose-response effects of fentanyl, cyclopropylfentanyl, valerylfentanyl, and morphine in the radiant heat tail flick assay, whereas ED₅₀ potency values are summarized in Table 1. In the tail flick assay, fentanyl (ED₅₀=0.03 mg/kg, s.c.) was 60-fold more potent than morphine (ED₅₀=1.80 mg/kg, s.c.). Cyclopropylfentanyl had antinociceptive potency (ED₅₀=0.04 mg/kg, s.c.) that was nearly equivalent to fentanyl, while valerylfentanyl was 100-fold less potent (ED₅₀=4.0 mg/kg, s.c.). Despite the variable potencies across the drugs, all of the compounds induced maximal antinociceptive effects at sufficient doses (Figure 2 and Table 1). To confirm the opioid specificity of antinociceptive effects, we administered 1 mg/kg naloxone prior to administration of high drug doses (i.e., 5-times the ED₅₀ for each drug). The naloxone antagonist findings are shown in the right panel of Figure 2. Naloxone pretreatment significantly reduced the antinociceptive effects of fentanyl (t=16.68, df=8, p<0.0001), cyclopropylfentanyl (t=26.59, df=8, p<0.0001), valerylfentanyl (t=9.264, df=8, p<0.0001), and morphine (t=14.12, df=8, p<0.001), confirming a critical role of MOR in mediating effects of these drugs (Figure 2).

Figure 2. Antinociception of fentanyl, cyclopropylfentanyl, valerylfentanyl, and morphine in CD-1 mice using a radiant tail flick assay.

For the dose-response studies, male CD-1 mice received s.c. doses of fentanyl, cyclopropylfentanyl, valerylfentanyl, and morphine. Antinociception was measured using a radiant-heat tail-flick assay 10 min (fentanyl or fentanyl analogs) or 30 min (morphine) after agonist administration. Dose-response data represent % maximal possible effect (% MPE), expressed as mean±SEM for N=6–10 mice per group. Antagonist data represent % MPE, expressed as mean±SEM for N=5 mice per group; For antagonist studies, s.c. fentanyl (0.2 mg/kg), cyclopropylfentanyl (0.2 mg/kg), valerylfentanyl (20 mg/kg), or morphine (10 mg/kg) was administered 5 min after s.c. pretreatment with saline (Sal) or naloxone (Nal; 1 mg/kg). Asterisks represent significant effects compared to the appropriate saline+agonist treatment (p<0.05 Tukey’s post hoc test).

We also investigated the effects of cyclopropylfentanyl and valerylfentanyl on breath rate in mice using an oximeter-based assay. When drugs were administered at doses 5-times greater than their ED₅₀ doses for inducing antinociception (i.e., 0.2 mg/kg for cyclopropylfentanyl and 20 mg/kg for valerylfentanyl), we found that breath rate was significantly reduced by both cyclopropyl (F[2,12]=45.24, p<0.0001) and valeryl (F[2,12]=13.65, p<0.0001) analogs. Maximal reductions in breath rate of 46% and 51% of basal levels were observed for cyclopropylfentanyl and valerylfentanyl at 15–20 min after drug administration. The breath rate suppression slowly recovered to 75% and 86% of basal levels for cyclopropylfentanyl and valerylfentanyl by 50 min post-injection (Figure 3). Naloxone pretreatment fully antagonized the effects of cyclopropylfentanyl and valerylfentanyl on breath rate, and naloxone plus agonist effects were not significantly different from saline vehicle treatment.

Figure 3. Time-course effects of cyclopropylfentanyl or valerylfentanyl on breath rate in mice.

Male CD-1 mice received s.c. cyclopropylfentanyl (0.2 mg/kg), valerylfentanyl (20 mg/kg), or saline vehicle, and breath rate was measured in 5-min bins for 50 min post-injection using the MouseOx pulse oximeter system (Starr Life Sciences, Oakmont, PA, USA). For the antagonist groups, naloxone (1 mg/kg, s.c.) was administered 5 min prior to agonists. Data represent % of baseline breath rate (% Basal), expressed as mean±SEM for N=5 mice per group. Filled symbols represent significant effects compared to the saline control group at a specific time point (p<0.05 Tukey’s post hoc test).

3.3. The APF pose of valerylfentanyl cannot be well accommodated in either the active or inactive states of the MOR.

From a chemical structure perspective, cyclopropylfentanyl has a 3-carbon ring attached to the carbonyl moiety of fentanyl, while valerylfentanyl has a 4-carbon chain at the same position. Despite the subtle structural difference between the compounds, cyclopropylfentanyl displays much greater MOR affinity when compared to valerylfentanyl (Table 1). Likewise, we showed that cyclopropylfentanyl is a potent full-efficacy MOR agonist in functional assays, whereas valerylfentanyl is a weak partial agonist (Table 1). To explore the structural basis for the differences in potency and efficacy between cyclopropyl and valeryl analogs, we carried out molecular dynamics (MD) simulations using MOR models in complex with fentanyl, cyclopropylfentanyl, or valerylfentanyl, and compared the resulting conformations stabilized by these different ligands.

Previously, we found that fentanyl can be stable in two opposite orientations, the APF and FPA poses (Figure S1) (Xie et al., 2022). In this study, we chose the most favored condition for each orientation, either APF_DC or FPA_DT, as the starting points to build ligand bound MOR models in both the active and inactive states (see Methods). Note the subscripts “C” and “T” represent cis- and trans-amide configurations of the fentanyl scaffold, respectively, while “D” indicates that His299^6.52 (superscripts denote Ballesteros-Weinstein numbering (Ballesteros and Weinstein, 1995)) of MOR is in the HID protonation state.

In our analysis of the APF_DC simulation results, we first focused on the small cavity in the MOR binding pocket with which the divergent alkyl moieties of fentanyl analogs interact, specifically, Lys235^5.39, Val238^5.42, Phe239^5.43, His299^6.52, and Val302^6.55 at the interface between the extracellular portions of TM5 and TM6 (TM5e and TM6e) (Figure 4). In the control simulations of the MOR models in complex with the ligands from the original structures, i.e., DAMGO for the active structure 6DDF, and β-FNA for the inactive structure 4DKL (Table S1), the Cα-Cα distance of Val238^5.42 and Val302^6.55 remained ~10 Å, very close to the corresponding distances in the original structures (Table S2). This consistency demonstrates that the lipid bilayer environment and our simulation protocols do not affect the conformations revealed by cryo-EM and crystallography.

Figure 4. Valerylfentanyl in the APF pose disrupts the binding pocket of the MOR in both the active and inactive states.

(A) the APF binding poses of fentanyl, cyclopropylfentanyl and valerylfentanyl in the active state of the MOR. The common parts of fentanyl and its analogs were colored as light grey. The residues forming a hydrophobic cavity were colored in green. The cyclopropyl moiety of cyclopropylfentanyl (yellow) can be accommodated by this cavity as the ethyl tail of fentanyl, while the butyl tail of valerylfentanyl (orange) protrudes to interact with K235^5.39 and F239^5.43 and disrupts the cavity. The Cα atoms of V238^5.42 and V302^6.55 were shown in spheres. (B) Binding poses of fentanyl (dark grey) and valerylfentanyl (orange) in the inactive conformation. The residues forming the hydrophobic cavity were colored in cyan. A similar disruption of the cavity by the butyl tail of valerylfentanyl was observed as well. The black dotted lines are the Cα-Cα distances of V238^5.42 and V302^6.55 (see Table S2). The poses shown are the representatives from the pose ensembles of the simulated conditions (Figure S8).

In the MD simulations of the active MOR models with ligands in the APF orientation, the butyl tail of valerylfentanyl protrudes further into the cavity and interacts with Lys235^5.39 and Phe239^5.43, while fentanyl and cyclopropylfentanyl are not in contact with these two residues (Figure 4A, Table 2). Compared to fentanyl, cyclopropylfentanyl has a higher frequency of interacting with Tyr150^3.33 and a lower frequency in contacting Asn152^3.35 (Table 2), but these interactions do not induce significant conformational changes in MOR (see below). The crowded interactions between the butyl tail of valerylfentanyl and all the residues forming the cavity result in a longer Val238^5.42–Val302^6.55 distance, compared to those in the active MOR models bound with either fentanyl or cyclopropylfentanyl, and that of the DAMGO bound model (Table S2). A similar trend is observed in the simulations of the inactive MOR model bound with the ligands in the APF orientation (Figure 4B, Tables S2 and S3). This local disruption induced by the bound valerylfentanyl induced an outward rearrangement of TM6e, which is demonstrated by a longer TM5e-TM6e distance of the valerylfentanyl bound MOR models in both active and inactive states compared to those of the fentanyl or cyclopropyl bound models (Figure 5A,B, Table S2).

Table 2.

The contact frequencies of the residues interacting with fentanyl and its analogs in the active state of hMOR.

Residue	fentanyl		cyclopropylfentanyl				valerylfentanyl
	APFDC	FPADT	APFDC	FPADT	−fentanyl		APFDC	FPADT	−fentanyl
					APFdc	FPAdt			APFDC	FPADT
Thr1222.56	0.28	0.98	0.13	0.93	−0.15	−0.05	0.23	1.00	−0.05	0.02
Phe1252.59	0.05	0.97	0.05	0.92	0.00	−0.05	0.05	0.99	0.00	0.02
Gln1262.60	1.00	0.99	1.00	1.00	0.00	0.01	1.00	1.00	0.00	0.01
Asn1292.63	0.98	0.89	0.99	0.87	0.01	−0.02	0.98	0.95	0.00	0.06
Tyr1302.64	0.93	0.14	0.95	0.04	0.02	−0.10	0.94	0.83	0.01	0.69
Trp1353.18	0.99	1.00	1.00	0.99	0.01	−0.01	0.99	1.00	0.00	0.00
Val1453.28	0.97	0.99	0.94	0.97	−0.03	−0.02	0.96	0.99	−0.01	0.00
Ile1463.29	0.99	1.00	0.99	1.00	0.00	0.00	0.99	1.00	0.00	0.00
Asp1493.32	1.00	1.00	1.00	1.00	0.00	0.00	1.00	1.00	0.00	0.00
Tyr1503.33	0.32	0.97	0.86	0.96	0.54	−0.01	0.45	1.00	0.13	0.03
Asn1523.35	0.70	0.00	0.13	0.00	−0.57	0.00	0.54	0.00	−0.16	0.00
Met1533.36	1.00	0.99	1.00	0.99	0.00	0.00	1.00	1.00	0.00	0.01
Asp218EL2.49	0.13	0.54	0.21	0.94	0.08	0.40	0.20	0.89	0.07	0.35
Cys219EL2.50	1.00	1.00	1.00	1.00	0.00	0.00	1.00	1.00	0.00	0.00
Lys2355.39	0.02	0.01	0.00	0.01	−0.02	0.00	0.77	0.00	0.75	−0.01
Val2385.42	1.00	0.12	1.00	0.09	0.00	−0.03	0.99	0.02	−0.01	−0.10
Phe2395.43	0.00	0.00	0.00	0.00	0.00	0.00	0.62	0.00	0.62	0.00
Trp2956.48	0.91	0.59	0.89	0.66	−0.02	0.07	0.94	0.34	0.03	−0.25
Ile2986.51	1.00	0.98	1.00	0.99	0.00	0.01	1.00	0.99	0.00	0.01
His2996.52	0.99	0.68	1.00	0.70	0.01	0.02	0.95	0.34	−0.04	−0.34
Val3026.55	0.97	0.11	0.98	0.17	0.01	0.06	0.98	0.02	0.01	−0.09
Thr3177.32	0.00	0.00	0.00	0.00	0.00	0.00	0.00	0.56	0.00	0.56
Trp3207.35	0.32	0.73	0.29	0.75	−0.03	0.02	0.42	0.78	0.10	0.05
His3217.36	0.00	0.86	0.00	0.58	0.00	−0.28	0.00	0.94	0.00	0.08
Ile3247.39	1.00	1.00	1.00	1.00	0.00	0.00	1.00	1.00	0.00	0.00
Gly3277.42	0.98	0.84	0.97	0.91	−0.01	0.07	0.91	0.97	−0.07	0.13
Tyr3287.43	1.00	1.00	1.00	1.00	0.00	0.00	0.95	1.00	−0.05	0.00

In a molecular dynamics (MD) simulation frame, if the shortest heavy-atom distance between the ligand and any given residue of the MOR was within 5 Å, we defined that the ligand forms an interaction with this residue. The residues that have at least one contact frequency > 0.5 in any indicated condition are included in this table. The results shown here are based on the simulations with the OPLS4 force field. The differences between the analogs and fentanyl in their APF_DC conditions are shown in the “- fentanyl” columns.

Figure 5. The local conformational disruption by valerylfentanyl in APF propagates to distort TM6e.

In both the inactive (A) and active (B) states of the MOR, the superimposed conformations of the receptor bound valerylfentanyl (orange) and fentanyl (grey) shows that TM6e moves outwards in the presence of valerylfentanyl. This trend is confirmed by a quantitative analysis of the pairwise distances of the TMs subsegments (TM2e to TM7e) in which TM6e has longer distances (red pixels in the heatmaps) to many other subsegments in the presence of valerylfentanyl in both the inactive (C) and active (D) states, while these distances are comparable between the fentanyl bound and cyclopropylfentanyl bound conditions (D).

To comprehensively characterize the differential impacts of the bound ligands on the receptor conformation, we evaluated the pairwise-distances of all the extracellular sub-segments using the Protein Interaction Analyzer (see Methods). The results confirm that the most prominent impact is on TM6e, which has longer distances to other sub-segments in the presence of valerylfentanyl, compared the fentanyl-bound situation (Figure 5D). These same pairwise distances only have minor differences between the fentanyl- and cyclopropyl-bound situations (Figure 5D). A similar trend is observed in the inactive state (Figure 5C).

3.4. Valerylfentanyl can be reasonably accommodated in the FPA_DT condition

In the FPA_DT condition, fentanyl, cyclopropyl, and valerylfentanyl are bound in the MOR binding pocket in an orientation that is opposite to the APF_DC condition. More specifically, in the FPA_DT condition, the divergent alkyl modifications of the phenylamide point extracellularly into a relatively more open space and can be easily accommodated by the MOR in both the active and inactive states. As we have previously described, when fentanyl is bound in the FPA_DT pose in the active MOR model, its carbonyl group can form a hydrogen-bond (H-bond) with the side chain of either Gln126^2.60 or Asn129^2.63 (Xie et al., 2022). Interestingly, the bound valerylfentanyl in the FPA_DT pose has a stronger tendency of forming this H-bond, which is reflected in the narrower and higher peak of the distance distribution between valerylfentanyl and Asn129^2.63 when compared to those of cyclopropylfentanyl and fentanyl (Figure 6A,C). In addition, the butyl tail of valerylfentanyl has a higher probability of forming hydrophobic interactions with Tyr130^2.64, Thr317^7.32, and His321^7.36 (Figure 6A,D, Table 2). Overall, in the active MOR model, valerylfentanyl appears to have a more stable FPA_T pose than fentanyl and cyclopropylfentanyl. On the other hand, the Val238^5.42–Val302^6.55 and TM5e–TM6e distances are similar among the FPA_DT conditions of fentanyl, cyclopropylfentanyl, and valerylfentanyl (Table S2). Importantly, our estimations of the binding free energy with the MM/GBSA approach shows that valerylfentanyl in the FPA_DT condition has 7.8 kcal/mol lower energy than in the APF_DC condition in the active MOR model, while this trend is not observed for cyclopropylfentanyl. Thus, valerylfentanyl may favor FPA_DT rather than APF_DC.

Figure 6. Valerylfentanyl in FPA affects the TM2e orientation in the active state of MOR.

In the FPA_DT pose, the butyl tail of valerylfentanyl (orange) protrudes into the interface between TM2e and TM7e and forms interactions with Tyr130^2.64, T317^7.32, and His321^7.36 (A), which results in an outward movement of TM2e compared to the fentanyl-bound condition (B). In addition, valerylfentanyl appears to have a more defined FPA_T pose than fentanyl (grey) and cyclopropylfentanyl (yellow), which is reflected in more persistent interactions with Asn129^2.63 (the distance between the asparagine side chain nitrogen and the ligand amide oxygen) (C) and His321^7.36 (the closest heavy atom-heavy atom distance between the histidine and the ligand) (D). The PIA analysis for extracellular subsegments shows that the fentanyl- and cyclopropylfentanyl-bound conditions have similar conformations (E), while TM2e moves away from TM6e in the presence of valerylfentanyl (F), as shown in panel B.

In the pairwise distance analysis of extracellular subsegments, a stronger interaction of valerylfentanyl with TM2e results in a longer distance between TM2e and TM6e than that in the fentanyl or cyclopropylfentanyl bound condition (Figure 6B,E,F, and Figure S7), which is reminiscent of the trend observed between the inactive and active MOR structures (PDB 4DKL and 6DDF, respectively), i.e., the same distance, which is on the longer side of the binding pocket, is longer in the inactive state (Figure S4). Thus, it is tempting to speculate that when valerylfentanyl is bound in the FPA pose, it may shift the receptor towards to the inactive state.

In our simulations of the inactive MOR model bound with either valerylfentanyl or fentanyl in the FPA_DT condition, both ligands adopted more favored linear poses (Figure S5A,B) rather than bent poses in the active state (Figure 6A). This divergence likely resulted from the different shapes of the ligand binding pocket in the active and inactive states, which were associated with the different side chain rotamers of the toggle switch Trp295^6.48 (Shi et al., 2002) (Figure S6). Similar to what we observed in the active MOR model simulations in the FPA_DT condition, valerylfentanyl forms a persistent H-bond with Asn129^2.63 and has a tighter interaction with His321^7.36 than fentanyl in the inactive MOR model (Figure S5C,D), while the Val238–Val302 and TM5e-TM6e distances are similar between fentanyl and valerylfentanyl bound models (Table S2).

4. Discussion

The main goal of the present study was to directly compare the pharmacological effects of cyclopropylfentanyl and valerylfentanyl, two fentanyl analogs associated with opioid intoxications and overdose deaths in human users (Bergh et al., 2019; Fagiola et al., 2019; Fogarty et al., 2018; Ton et al., 2021; Walsh et al., 2021). To this end, we performed the side-by-side assessment of the pharmacological effects of the drugs using in vitro, in vivo, and in silico methods. Our in vitro findings confirm that cyclopropylfentanyl is a potent full-efficacy agonist at MOR, similar to the parent compound fentanyl, whereas valerylfentanyl is a weak partial agonist at MOR (Astrand et al., 2020; Eshleman et al., 2020; Vasudevan et al., 2020). The in vivo findings reported here corroborate that cyclopropylfentanyl is about 100-times more potent than valerylfentanyl in assays measuring antinociception and respiratory depression in mice. Importantly, despite the weak partial agonist effects of valerylfentanyl in vitro, this compound is capable of inducing maximal antinociceptive effects and substantial decreases in breath rate, when administered at sufficient doses. Our results from molecular dynamic simulations reveal that cyclopropylfentanyl and fentanyl are more stable in the APF pose, whereas valerylfentanyl may prefer the FPA pose. In addition, the extended alkyl chain of valerylfentanyl seems to favor binding to an inactive state of MOR. Thus, the weak partial agonist effects of valerylfentanyl could be related to its unique binding interactions at MOR, which differ substantially when compared to fentanyl and cyclopropylfentanyl. From a public health perspective, our findings suggest that cyclopropylfentanyl poses much greater risks than valerylfentanyl to human drug users who are inadvertently exposed to the compounds.

Our results with [¹²⁵I]IBNtxA binding to mMOR agree with previous studies showing that cyclopropylfentanyl and fentanyl have high affinities for rat MOR labeled with [³H]DAMGO, and valerylfentanyl displays about 20-fold lower affinity for rat MOR under the same binding conditions (Baumann et al., 2018; Eshleman et al., 2020). Consistent with our findings in functional assays, a recent study investigating the pharmacology of alicyclic fentanyl analogs showed that smaller cyclopropyl, cyclobutyl, and cyclopentyl analogs of fentanyl are full agonists at MOR, while larger cyclohexyl, and 2,2,3,3-tetramethylcyclopropyl analogs are weak partial agonists (Astrand et al., 2021). Moreover, previous work demonstrates that valerylfentanyl is a partial agonist in the MOR-mediated GTPγS binding assay in transfected CHO cells (Eshleman et al., 2020). Taken together, the available data indicate that increasing alkyl chain length or bulk at the carbonyl moiety of fentanyl might decrease drug potency and efficacy at MOR by preventing optimal drug interactions in a binding cavity with defined size.

As far as we are aware, the present mouse studies represent the first side-by-side comparison of the in vivo pharmacological effects of cyclopropylfentanyl and valerylfentanyl. We show that both drugs can induce antinociception and respiratory depression, but cyclopropylfentanyl is about 100-times more potent than valerylfentanyl. In a previous rat study, cyclopropylfentanyl produced dose-related antinociception, catalepsy, and hypothermia after s.c. administration (Bergh et al., 2021). The reported ED₅₀ for cyclopropylfentanyl to induce antinociceptive effects in the rat hotplate test is 0.05 mg/kg, s.c. (Bergh et al., 2021), nearly identical to the ED₅₀ of 0.04 mg/kg, s.c. that we observed here for cyclopropylfentanyl in the mouse radiant-heat tail-flick assay. Varshneya et al. demonstrated that valerylfentanyl is a low-potency agonist (ED₅₀=6.4 mg/kg, s.c.) in the warm-water tail-withdrawal test in mice, but had no locomotor stimulant activity (Varshneya et al., 2019). This latter observation differs from most full-efficacy MOR agonists, which are known to produce robust locomotor stimulant effects in mice at high doses (see (Varshneya et al., 2019)). However, it is noteworthy that valerylfentanyl was able to induce maximal antinociceptive effects in vivo when administered at high enough doses, even though the drug functioned as a partial agonist in the GTPγS binding and β-arrestin 2 assays in vitro. One possible explanation for the full efficacy agonist effects of valerylfentanyl in vivo might be related to different “efficacy requirements” for the in vitro versus in vivo assays used in our experiments. For example, Selley et al. demonstrated that the efficacy requirement for MOR-mediated stimulation of GTPγS binding is much higher than the efficacy requirement for induction of antinociception in the mouse (Selley et al., 2021). Such findings suggest that partial agonist effects determined in vitro might not always translate to in vivo responsiveness. Future studies should evaluate the effects of opioid agonists using both in vitro and in vivo assays to more fully examine the concepts of efficacy requirement and “receptor reserve”, which is the fraction of the receptor population not required to induce a maximal pharmacological response (Black and Leff, 1983; Gillis et al., 2020).

We carried out in-depth molecular modeling and simulation studies to examined how drug interactions at MOR might explain the stark differences in affinity, potency, and efficacy between cyclopropyl and valeryl analogs of fentanyl. In the APF pose, alkyl extensions from the carbonyl moiety of fentanyl are accommodated in a small cavity within the binding pocket of MOR. Importantly, this cavity is selective with respect to length and shape of the alkyl extensions that are tolerated. In particular, a cyclopropyl extension can comfortably fit in the cavity, while the butyl tail cannot be properly accommodated and induces a local conformational shift that propagates to cause an outward movement of TM6e in both the active and inactive states of MOR. This perturbation by valerylfentanyl in APF may be responsible for its reduced binding affinity, potency, and efficacy at MOR. In one of the valerylfentanyl APF_DC trajectories when MOR is in the active state, the critical ionic interaction between the positively charged piperidine nitrogen and Asp^3.32 was lost, suggesting weakening of this interaction, which might also contribute to the reduced binding affinity and potency.

In the FPA_T pose, the butyl tail of valerylfentanyl can be easily accommodated and forms more distinct interactions with TM2 and TM7 residues when compared to the effects of fentanyl and cyclopropylfentanyl. The binding of valerylfentanyl in FPA_T results in conformational changes in MOR that are reminiscent of the transition from the active to inactive state, whereas the binding of fentanyl and cyclopropylfentanyl do not have these effects. The summed findings prompt us to propose that valerylfentanyl may favor the FPA pose, which is the most likely structural basis of its partial agonism. Regardless, in both the APF and FPA poses, the binding of valerylfentanyl results in different MOR conformations compared to those stabilized by fentanyl or cyclopropylfentanyl.

Prior studies suggest that MOR-mediated recruitment of β-arrestin 2 is involved with respiratory depressant effects of opioid agonists (Schmid et al., 2017). We found that valerylfentanyl was a partial agonist in MOR-mediated recruitment of β-arrestin 2, yet the drug still substantially suppressed breathing in mice. Our data cannot evaluate whether valerylfentanyl displays bias towards β-arrestin 2 versus G protein transduction mechanisms because it is impossible to accurately calculate bias factors from the different cell systems used in our β-arrestin 2 and [³⁵S]GTPγS assays. Nevertheless, our in vivo results with fentanyl analogs in mice are not consistent with the notion that β-arrestin 2 recruitment, rather than G protein activation, is associated with respiratory depression (Schmid et al., 2017). A recent study using mini-Gi and β-arrestin 2 recruitment assays did not find significant bias for many fentanyl analogs, including cyclopropylfentanyl and valerylfentanyl (Vasudevan et al., 2020). Furthermore, fentanyl and morphine can elicit robust respiratory depression in mutant mouse models where MOR-mediated β-arrestin 2 recruitment is impaired (Kliewer et al., 2019). We found that valerylfentanyl has similar potency to fentanyl and cyclopropylfentanyl in the β-arrestin 2 recruitment assay, while the valeryl analog had significantly lower potency than the other compounds in the [³⁵S]GTPγS binding assay. Thus, the potencies in the [³⁵S]GTPγS assay appear to correlate more with in vivo potency of fentanyl analogs.

Determining the potential clinical risks associated with cyclopropylfentanyl and valerylfentanyl was a major impetus for investigating the preclinical pharmacology of the compounds. The findings reported here clearly show that cyclopropylfentanyl will pose much greater risk than valerylfentanyl to humans who are inadvertently exposed to the drugs via recreational opioid use. Indeed, cyclopropylfentanyl would be expected to induce clinical symptoms and adverse effects (e.g., respiratory depression) that are identical to illicitly manufactured fentanyl, which is a principle driving force in the current opioid overdose epidemic. Interestingly, cyclopropylfentanyl and valerylfentanyl have both been associated with opioid overdose fatalities (Bergh et al., 2019; Fagiola et al., 2019; Fogarty et al., 2018; Ton et al., 2021; Walsh et al., 2021). This apparent paradox can be explained by the fact that nearly all post-mortem cases involving valerylfentanyl were polydrug cases that had illicit opioids or other drugs of abuse present. Two recent forensic reports, one from New York City and another from Michigan, identified several cases where valerylfentanyl was detected in post-mortem blood samples (Ton et al., 2021; Walsh et al., 2021). In both reports, post-mortem concentrations of fentanyl were higher than concentrations of valerylfentanyl in nearly every case. Given the much greater potency and efficacy of fentanyl compared to valerylfentanyl shown here, it seems likely that fentanyl was the cause of death in the cases examined. The toxicological findings illustrate the complexity of relating specific opioid drugs with overdose death in polydrug cases and reinforce the need for preclinical studies investigating the pharmacology of new opioid drugs as they emerge in clandestine markets.

In summary, the in vitro data reported here confirm that cyclopropylfentanyl is a potent and efficacious MOR agonist, similar to the parent compound fentanyl, while valerylfentanyl is a weak partial agonist at MOR. Our in vivo studies in mice verify that cyclopropylfentanyl is 100-times more potent than valerylfentanyl as an antinociceptive agent, which suggests that cyclopropylfentanyl poses much greater risk than valerylfentanyl in human users who are exposed to the drugs. Importantly, findings from our molecular dynamic simulations provide an atomistic basis for the low potency and partial agonist effects of valerylfentanyl at MOR. The summed findings shed light on the structural basis of converting fentanyl to a partial agonist at MOR, which may provide new insights for the development of safer analgesics. Finally, the multidisciplinary approach for profiling the pharmacology of fentanyl analogs described here might be useful for rapidly examining the potential risks of new opioid drugs emerging on clandestine markets.

Highlights.

• Cyclopropylfentanyl is more efficacious than valerylfentanyl at MOR in vitro

• Both analogs induce naloxone-reversible antinociception and respiratory suppression

• Bulky carbonyl modifications to fentanyl decrease the potency and efficacy at MOR

• The butyl modification prevents optimal ligand binding in the active state of MOR

• Cyclopropylfentanyl presents greater risk for adverse effects than valerylfentanyl