Characterization of recent non-fentanyl synthetic opioids via three different in vitro μ-opioid receptor activation assays

Abstract

New synthetic opioids (NSOs) are one of the fastest growing groups of new psychoactive substances. Amid this dynamic landscape, insight into the pharmacology of NSOs is important to estimate the harm potential of newly emerging drugs. In this work, we determined the μ-opioid receptor (MOR) affinity and activation potential of seven poorly characterized non-fentanyl NSOs (N-ethyl-U-47700, 3,4-difluoro-U-47700, U-47931E/bromadoline, 2,4-difluoro-U-48800, U-62066/spiradoline, 2F-viminol, ketobemidone) and a panel of nine reference opioids. MOR affinity was determined via [³H]-DAMGO binding in rat brain tissue homogenates, and was found to correlate well with different functional parameters. MOR activation potential was studied at different levels of receptor signaling using three distinct assays (NanoBiT^® MOR-β-arrestin2/mini-G_αi and AequoScreen^®). The most active compounds were ketobemidone (EC₅₀ 32.8-528 nM; E_max 105-271%, relative to hydromorphone) and N-ethyl-U-47700 (EC₅₀ 241-767 nM; E_max 139-247%). The same opioids showed the strongest MOR affinity. As most of the other NSOs only weakly activated MOR in the three assays (EC₅₀ values in the high nM-μM range), they likely do not pose a high overdose risk. 2F-viminol (EC₅₀ 2.2-4.5 μM; E_max 21.2-61.5%) and U-47931E/bromadoline (EC₅₀ 0.55-2.9 μM; E_max 52.8-85.9%) were partial agonists compared to hydromorphone, and maximum receptor activation was not reached for 2,4-difluoro-U-48800 (EC₅₀ >22 μM). We further highlight the importance of considering specific assay characteristics upon interpretation of potencies, efficacies and biased agonism. As absolute values may greatly differ between assays with varying experimental set-ups, a comparison of functional parameters to those of well-characterized reference agonists is considered the most informative.

Introduction

Over the past few years, there has been a dramatic rise in the number of new synthetic opioids (NSOs) appearing on non-medical (i.e., recreational) drug markets worldwide. NSOs include various fentanyl analogs and non-fentanyl compounds that function as μ-opioid receptor (MOR) agonists. The total share of NSOs as a proportion of all synthetic new psychoactive substances (NPS) has quadrupled between 2015 and 2019 (UNODC 2020), rendering NSOs one of the fastest growing groups of NPS. Given the high potency of many NSOs, combined with the substantial risk of fatal overdoses due to respiratory depression, this rapid growth and increasingly widespread availability are coupled to significant public health risks (Prekupec et al. 2017; Baumann and Pasternak 2018; UNODC 2020). Adding an extra layer of complexity, the nature of the opioids entering recreational markets has largely diversified in recent years (UNODC 2020; EMCDDA 2021). Whereas fentanyl derivatives were predominant between 2013 and 2018, the introduction of different class-wide scheduling efforts targeting fentanyl analogues (DEA 2018; Bao et al. 2019) has facilitated the emergence of a new generation of non-fentanyl synthetic opioids (UNODC 2020; EMCDDA 2021). A plethora of compounds with opioid effects have been described in the scientific and patent literature, whereby the compounds were evaluated as potential analgesics with widely varying potencies and side effects. Nowadays, illicit drug manufacturers capitalize on this early or more recent research, synthesizing “novel” opioids that are specifically engineered to bypass existing drug control laws (Griffiths et al. 2013; Sharma et al. 2019; Vandeputte et al. 2021a). These opioids of the ‘post-fentanyl analogue era’ form a structurally diverse group, dissimilar to fentanyl and most other known opioids. Yet, each new member carries a substantial and often unpredictable risk of toxicity (Sharma et al. 2019; Verougstraete et al. 2020; Krotulski et al. 2021; Blanckaert et al. 2021).

Insight into the pharmacology of opioids has become an important strategy to estimate the harm potential of these rapidly emerging NPS (Baumann et al. 2018a). Specifically, in vitro functional assays of MOR activation can give an important first indication of the activity of a newly emerging opioid, as MOR is considered the main molecular target for both therapeutic (e.g., analgesic) as well as adverse (e.g., respiratory depression) opioid effects (Charbogne et al. 2014). MOR is a G protein-coupled receptor (GPCR). GPCRs are integral membrane proteins containing 7 transmembrane domains that signal via heterotrimeric G proteins composed of α, β, and γ subunits (Lefkowitz 2000; Niedernberg et al. 2003). Several comprehensive reviews have detailed the mechanisms involved in MOR signaling (Law et al. 2000; Waldhoer et al. 2004; Al-Hasani and Bruchas 2011; Stein 2016; Valentino and Volkow 2018). In short, receptor activation promotes the exchange of guanosine 5’-diphosphate (GDP) for the corresponding 5’-triphosphate (GTP) on the G_α subunit, leading to G_α-G_βγ dissociation. Both subunits (G_α and G_βγ) subsequently interact with a wide range of cellular effector systems, eventually leading to the known opioid effects. One of these effects is the G_αi/o-GTP-mediated inhibition of adenylyl cyclase, which leads to a decrease in cyclic adenosine 5’-monophosphate (cAMP) levels within the cell. While it is known that MOR is predominantly coupled to the G_αi/o protein subtype, interactions with other G proteins have also been described, further diversifying intracellular signaling (Connor and Christie 1999; Waldhoer et al. 2004). G_α16, for example, has been shown to interact with recombinant MOR (Offermanns and Simon 1995; Connor and Christie 1999) and subsequently activate phospholipase C (PLC), an enzyme that catalyzes the formation of inositoltriphosphate (IP₃). IP₃ is involved in the mobilization of calcium from intracellular stores, leading to an increase in intracellular Ca²⁺-levels upon GPCR activation (Niedernberg et al. 2003). The coupling of G_α16 to downstream Ca²⁺ mobilization allows to measure G protein-dependent MOR activation. Apart from signaling via G protein-dependent pathways, upon receptor activation, different protein kinases phosphorylate the intracellular domains of the GPCR. Distinct phosphorylation patterns induce differential binding of β-arrestins (βarr, e.g., βarr2) to the receptor (Latorraca et al. 2020), initiating G protein-independent response pathways and desensitization of receptor activation (Lefkowitz 2005; Latorraca et al. 2020; Smith et al. 2021).

The often fleeting appearance of a new compound on the recreational market, combined with the increasing structural diversity of many non-fentanyl opioids, makes it difficult for clinicians, toxicology labs and law enforcement agencies to keep up with current trends. To aid in better understanding the potential harms of this new generation of opioids, we recently reported on the in vitro functional characterization of a panel of 11 non-fentanyl NSOs using two MOR activation assays based on two receptor-proximal (upstream) events, i.e. the recruitment of either βarr2 or mini-G_αi to the activated receptor (Vandeputte et al. 2020). As the activity of several opioids of this panel was comparable to or even exceeded that of fentanyl (e.g. isotonitazene, bromadol), many of these somewhat ‘obscure’ opioids have the potential to pose a significant threat to public health. Therefore, in the current work, we evaluated a new panel of non-fentanyl NSOs (Figure 1, panel B), of which little is known regarding their harm potential. As a second layer of inquiry, and to obtain a more comprehensive view of the in vitro mechanism of action of these NSOs, we added a third assay (AequoScreen^®) to our in vitro platform to also allow further insight into downstream events dependent on G protein activation. By doing so, we effectively profiled a new panel of seven NSOs at three different levels of MOR signaling: (1) βarr2 recruitment (as a proxy for activation of the G protein-independent pathway); (2) mini-G_αi recruitment (as a proxy for G_αi activation); and (3) downstream measurement of G protein activation by intracellular calcium release. In addition, five previously evaluated NSOs (Vasudevan et al. 2020; Vandeputte et al. 2020) (Figure 1, panel C) were also analyzed in the AequoScreen^® assay to allow a more in-depth interpretation of the results obtained with the three distinct assays. Last, as several of the newly investigated compounds were only weakly active at MOR, we also evaluated their MOR binding affinities and these compounds’ abilities to block receptor activation caused by fentanyl. Insight into binding affinities and potential antagonizing or modulating properties may further increase our understanding of the potential risks the use of these opioids may bring along.

Figure 1.

Structures of the studied opioids. Panel A consists of four reference compounds, providing a framework for interpretation of the results. The newly evaluated NSOs, tested in three different MOR activation assays, are included in Panel B. Panel C consists of diverse opioids for which new insights into MOR activation are studied by means of the AequoScreen^® assay. Panel C further helps in the interpretation of the results obtained in the three different assays applied in this work.

Materials & Methods

a. Chemicals and reagents

The compound hydromorphone (A) was from Fagron (Nazareth, Belgium) and fentanyl (B) was obtained from LGC Chemicals (Wesel, Germany). DAMGO (C) was from Sigma-Aldrich (Overijse, Belgium) and Cayman Chemical Company (Ann Arbor, Michigan, USA). Morphine (D), N-ethyl-U-47700 HCl (trans-3,4-dichloro-N-[2-(dimethylamino)cyclohexyl]-N-ethylbenzamide monohydrochloride) (1), 3,4-difluoro-U-47700 (trans-N-2-(dimethylamino)cyclohexyl)-3,4-difluoro-N-methylbenzamide) (2), U-47931E/bromadoline (trans-4-bromo-N-2-(dimethylamino)cyclohexyl]-benzamide) (3), 2,4-difluoro-U-48800 HCl (trans-2-(2,4-difluorophenyl)-N-(2-(dimethylamino)cyclohexyl)-N-methylacetamide monohydrochloride) (4), U-62066/spiradoline mesylate (rel-3,4-dichloro-N-methyl-N-[(5S,7R,8R)-7-(1-pyrrolidinyl)-1-oxaspiro[4.5]dec-8-yl]-benzeneacetamide monomethanesulfonate) (5), 2F-viminol (α-[[bis(1-methylpropyl)amino]methyl]-1-[(2-fluorophenyl)methyl]-1H-pyrrole-2-methanol) (6), bromadol (trans-4-(4-bromophenyl)-4-(dimethylamino)-1-(2-phenylethyl)-cyclohexanol) (8), brorphine HCl (1-[1-[1-(4-bromophenyl)ethyl]-4-piperidinyl]-1,3-dihydro-2H-benzimidazol-2-one monohydrochloride) (9) and 7-OH-mitragynine ((αE,2S,3S,7aS,12bS)-3-ethyl-1,2,3,4,6,7,7a,12b-octahydro-7a-hydroxy-8-methoxy-α-(methoxymethylene)-indolo[2,3-a]quinolizine-2-acetic acid methyl ester) (11) were obtained from Cayman Chemical Company (Ann Arbor, Michigan, USA). Ketobemidone (1-(4-(m-hydroxyphenyl)-1-methyl-4-piperidyl)-1-propanone) (7) and (−)-mitragynine ((αE,2S,3S,12bS)-3-ethyl-1,2,3,4,6,7,12,12b-octahydro-8-methoxy-α-(methoxymethylene)-indolo[2,3-a]quinolizine-2-acetic acid methyl ester) (10) were procured from Chiron AS (Trondheim, Norway). Ketobemidone (7) was also purchased from Toronto Research Company (Northyork, Canada). TRV-130/oliceridine (N-[(3-methoxythiophen-2-yl)methyl]-2-[(9R)-9-pyridin-2-yl-6-oxaspiro[4.5]decan-9-yl]ethanamine) (12) was purchased from AdooQ Bioscience (Irvine, California, USA).

The human embryonic kidney (HEK) 293T cells (passage 20) were a kind gift from Prof. O. De Wever from Ghent University Hospital, Belgium. Dulbecco’s Modified Eagle’s Medium (DMEM; GlutaMAX^™), Opti-MEM I Reduced Serum Medium, penicillin-streptomycin (5000 U/mL), and amphotericin B (250 ^μg/mL) were supplied by Thermo Fisher Scientific (Pittsburgh, Pennsylvania, USA). Sigma-Aldrich (Overijse, Belgium) supplied poly-D-lysine and fetal bovine serum (FBS). The reagents needed for the NanoBiT^® assays, i.e. the Nano-Glo Live Cell Assay System (containing the Nano-Glo Live Cell Substrate and Nano-Glo LCS Dilution Buffer), were purchased from Promega (Madison, Wisconsin, USA). The AequoScreen^® recombinant CHO-K1 cell line (ES-542-AV) was purchased from Perkin Elmer (Groningen, Netherlands). DMEM/Ham’s F12 without phenol red supplemented with 15mM HEPES, 2.5 mM L-glutamine, FBS, trypsin and genetecin were supplied from Thermo Fisher Scientific (Gothenburg, Sweden). Protease-free bovine serum albumin (BSA), digitonin, DMSO, and ATP were supplied from Sigma-Aldrich (Darmstadt, Germany). Coelenterazine was supplied from Nanolight Tech (Pinetop, Arizona, USA) and zeocin was from Invitrogen (Invitrogen, Carlsbad, California, USA).

b. Determination of in vitro biological activity at the μ-opioid receptor (MOR)

i. Cell culture

For the NanoBiT^® assays, HEK-293T cells were routinely maintained at 37°C in a humidified atmosphere containing 5% CO₂. Cells were cultured in DMEM (GlutaMAX^™) supplemented with 10% heat-inactivated FBS, 100 IU/mL penicillin, 100 μg/mL streptomycin, and 0.25 μg/mL amphotericin B. The cells stably expressed either the MOR-βarr2-GRK2 or MOR-mini-G_αi system (see below), and expression levels were routinely monitored via flow cytometric analysis of co-expressed markers (Vasudevan et al. 2020).

For the AequoScreen^® assay, CHO-K1 cells were routinely maintained at 37°C in a humidified atmosphere containing 5% CO_2. Cells were cultured in Ham’s F12 medium supplemented with 10% FBS, 0.4 mg/mL geneticin and 0.25 mg/mL zeocin. The cells were passaged every 3-4 days and not cultured beyond 30 passages.

ii. NanoBiT^® MOR-βarr2/mini-G_αi recruitment assays

Two previously described MOR activation assays, monitoring recruitment of either βarr2 (Cannaert et al. 2019; Vasudevan et al. 2020) or mini-G_αi (GTPase domain of the G_αi subunit) (Vasudevan et al. 2020) were used as a proxy for assessing initiation of the G protein-independent and -dependent signaling pathways, respectively, following MOR activation. Briefly, human MOR is fused to one part (LargeBiT, LgBiT) of a split nanoluciferase enzyme (NanoLuc Binary Technology^®, Promega). Receptor activation leads to recruitment of either βarr2 (in the presence of co-expressed G protein-coupled receptor kinase 2, GRK2) or mini-G_αi, which is fused to the complementing part (SmallBiT, SmBiT) of the split nanoluciferase. Recruitment to the activated receptor results in functional complementation of the enzyme, restoring its luciferase activity. Upon addition of the substrate furimazine, a measurable bioluminescent signal is generated (Figure 2A–B).

Figure 2.

Schematic representation of the three MOR activation assays applied in this study. 2a and 2b represent the employed NanoBiT^® assays, in which receptor activation leads to recruitment of either βarr2 (2a) or mini-G_αi (2b), allowing functional complementation of a split nanoluciferase enzyme. Upon addition of the substrate furimazine, a bioluminescent signal is generated. LgBiT, Large BiT; SmBiT, Small BiT. Figure 2c represents the AequoScreen^® assay. Upon addition of the substrate coelenterazine, apoaequorin is converted to the calcium-binding photoprotein aequorin. Following receptor activation, calcium is released from intracellular stores, allowing binding to aequorin. Coelenterazine is then oxidized by calcium-bound aequorin with the emission of flash luminescence. PLC, phospholipase C; IP₃, inositol triphosphate.

On day one, HEK-293T cells expressing either MOR-βarr2-GRK2 (further referred to as MOR-βarr2) or MOR-mini-G_αi were seeded in 96-well plates (5 x 10⁴ cells/well) pre-coated with poly-D-lysine. After overnight incubation, the cells were washed twice with Opti-MEM I Reduced Serum Medium, followed by addition of 90 μL Opti-MEM. Next, the Nano-Glo Live Cell Reagent was prepared by 20-fold dilution of Nano-Glo Live Cell Substrate with Nano-Glo LCS Dilution Buffer, and 25 μL was added to each well. The plate was then placed into a TriStar² LB 942 multimode plate reader (Berthold Technologies GmbH & Co., Bad Wildbad, Germany), and luminescence was continuously monitored for an equilibration period of 10-15 minutes, until the signal stabilized. Next, 20 μL of a 6.75-fold concentrated stock solution of each test compound (in Opti-MEM/MeOH, Opti-MEM/ACN or Opti-MEM/DMSO) was added per well and luminescence was monitored for 2 hours. Appropriate solvent controls were included on each plate.

iii. AequoScreen^® assay

The AequoScreen^® assay was used to measure G protein signaling following MOR activation, using an increase in levels of intracellular Ca²⁺. In short, AequoScreen^® CHO-K1 cells expressing the human MOR, the promiscuous G_α16 subunit (a universal G protein adaptor coupling receptor activation to Ca²⁺ release (Offermanns and Simon 1995)), and apoaequorin were used (Perkin Elmer) (Figure 2C). Upon addition of the substrate coelenterazine, the calcium-binding photoprotein aequorin is formed. Following receptor activation, calcium is released from intracellular stores, allowing it to bind to aequorin. Upon exposure to Ca²⁺, aequorin oxidizes coelenterazine with the emission of flash luminescence (Figure 2C).

The CHO-K1 cells were passaged once without antibiotics before experiments, cultivated to a confluency of 70-90% and trypsinized (10 min), after which they were centrifuged (150 x g for 5 min at room temperature) and resuspended in pre-warmed DMEM/Ham’s F12 medium without phenol red, supplemented with 15mM HEPES, 2.5 mM L-glutamine and 0.1% protease-free BSA, at a concentration of 3 x 10⁵ cells/mL. Cells were pre-incubated with coelenterazine (2.5 μM) on a rotating wheel at room temperature for 3 hours, and were protected from light. Test compounds were serially diluted in duplicate in opaque 96-well plates. As negative controls, wells containing medium and cells were used. As positive controls for coelenterazine loading, digitonin (67 μM) and ATP (7 μM) were used. The receptor activation at each concentration was determined by dispensing 50 μL of cells into each well and registering 200 luminescense readings for 25 s using a TECAN Spark 10 (Tecan, Switzerland). A baseline for 10 cycles was set.

iv. Data and statistical analysis

All compounds were tested at minimum 7 concentration levels ranging between 10 pM and 100 μM in minimum three independent experiments (n ≥ 3), with duplicates run for each concentration within an experiment. For the NanoBiT^® assays, absolute time-luminescence profiles obtained during the 2h readout were corrected for inter-well variability and used for calculation of the area under the curve (AUC) via the trapezoidal method (Pottie et al. 2020b). In all instances, solvent-correction was applied by subtracting the calculated AUC of the corresponding blank. For the AequoScreen^® assays, AUC values were calculated in the same way, starting from the time point at which the cells were injected (between 2.4 and 2.5 s), for a total of 25 s. Linear regression was performed (based on signals for digitonin) in Microsoft Excel 16 to compensate for potential signal drift attributable to settling out of cells stored in the syringe of the instrument. Solvent controls consisted of subtraction of the average AUC value (adjusted for plate drift) of the negative control (blank) wells. From this point onwards, data analysis was performed identically for all three activation assays to ensure maximal comparability of the results. Concentration-response curves were generated in GraphPad Prism 9 (San Diego, California, USA) via three-parametric logistic regression with a Hill slope of 1 (the latter is required for implementation of the relative intrinsic activity calculation, as outlined below). To allow maximal comparison with data from previous studies (Vasudevan et al. 2020; Vandeputte et al. 2020), data were normalized to the maximum response of hydromorphone (arbitrarily set at 100%) in each experiment. Fentanyl, DAMGO, and morphine were included to facilitate interpretation of the data (Panel A). In line with previous studies using the NanoBiT^® system (Vandeputte et al. 2020, 2021b; Pottie et al. 2020a), AUC values from the highest concentration(s) were excluded in case of a reduction of ≥ 20% compared to the AUC of the next dilution. Inclusion of such data points may inadvertently skew the obtained concentration-response curve, as the rapid drop in AUC at high concentrations may be the result of cell toxicity and/or solubility issues. The complete data set (NanoBiT^® and AequoScreen^® data) was tested for outliers via the standard Grubbs’ test. In total, 17 (0.69%) outliers were identified and excluded from further analysis. For each compound, the normalized data from minimum three independent experiments were used to obtain combined concentration-response graphs, generating EC₅₀ (potency) and E_max (efficacy, relative to hydromorphone) values for each opioid. NanoBiT^® data for bromadol (8), brorphine (9), mitragynine (10), 7-OH-mitragynine (11), and fentanyl (B) were previously reported in (Vandeputte et al. 2020), and were included here to allow an in-depth comparison with the novel results obtained from the AequoScreen^® assay. A part of the NanoBiT^® data for TRV-130/oliceridine (12) (n=3 out of a total of n=4 (mini-G_αi) and n=5 (βarr2) for this compound) are from (Vasudevan et al. 2020). NanoBiT^® data for morphine (n=5) were also used in (Vandeputte et al. 2021b).

The previously described relative intrinsic activity (RA_i) approach was used to evaluate the activity of each compound in the three different assays, including assessment of potential bias via calculation of a bias factor (Ehlert et al. 1999; Griffin et al. 2007; Ehlert 2008; Rajagopal et al. 2011). Application of the RA_i method requires a Hill slope fixed at 1, as obtained here via three-parameter logistic regression (cfr. supra). The RA_i was calculated for each compound (excl. (A) and (4)) in all three assays (Equation 1; i = test compound, HM = hydromorphone, the (unbiased) reference agonist):

$${RA}_{i,referenceagonist}^{assay}=\raisebox{1ex}{$\frac{E_{max,i}}{{EC}_{50,i}}$}\!\left/ \!\raisebox{-1ex}{$\frac{E_{max,HM}}{{EC}_{50,HM}}$}\right.=\raisebox{1ex}{${EC}_{50,HM}x{E}_{max,i}$}\!\left/ \!\raisebox{-1ex}{${EC}_{50,i}x{E}_{max,HM}$}\right.$$ Equation 1

Given that the employed functional assays can be considered as a proxy for different signaling pathways, the evaluated compounds could also be assessed for their potential pathway bias. To this end, the calculated RA_i values per pathway were combined into a bias factor (β_i) according to Equation 2 (i = test compound, X and Y = assay type, i.e. βarr2 (X) vs. mini-G_αi (Y), βarr2 (X) vs. AequoScreen^® (Y) or mini-G_αi (X) vs. AequoScreen^® (Y)):

$${\beta }_i=\mathrm{l}\mathrm{o}\mathrm{g}(\raisebox{1ex}{${RA}_{i,HM}^X$}\!\left/ \!\raisebox{-1ex}{${RA}_{i,HM}^Y$}\right.)$$ Equation 2

The bias factor was calculated for each compound (Panels A-C, excl. (A) and (4)) from three independent experiments, each performed in duplicate. Statistical analysis was carried out in GraphPad Prism 9 by non-parametric one-way ANOVA (Kruskal-Wallis), followed by a post hoc Dunn’s test. Differences compared to the unbiased reference agonist hydromorphone (β = 0) were considered statistically significant when p < 0.05.

c. Fentanyl competition experiments

To evaluate whether the newly studied NSOs (Panel B) could block MOR activation induced by fentanyl, the MOR-βarr2 assay was used to perform competition experiments. For completeness, compounds from Panels A and C were also tested in these experiments. The assay protocol was carried out as described above. After the equilibration phase, 10 μL of solvent (MeOH, ACN or DMSO) or of a 12.5x-concentrated stock solution of the different compounds (in MeOH, ACN or DMSO; final in-well concentration of 10 μM) was added to the cells. Naloxone (final in-well concentration of 1 μM) was included as known MOR antagonist, and the high-affinity partial MOR agonist buprenorphine (final in-well concentration of 10 μM) was added as an additional control. Receptor activation was measured for approximately 5 minutes, after which 10 μL of solvent (Opti-MEM/MeOH) or of a 13.5x-concentrated fentanyl solution (in Opti-MEM/MeOH) was injected to each well. Prior to injection of solvent or fentanyl solution, the injector was washed and primed with the appropriate solutions (Opti-MEM/MeOH or fentanyl/Opti-MEM/MeOH). The final fentanyl concentration in each well after injection was 0.05 μM. Wells containing only fentanyl (0.05 μM, no injection) were also included. Every condition was measured in duplicate and luminescence was monitored for a total of two hours using a Tristar² LB 942 multimode plate reader. The resulting MOR activation profiles were corrected for inter-well variability and AUC calculations were performed as previously described (Pottie et al. 2020b). In each experiment (n = 3), the solvent-corrected AUC values were then normalized to the average AUC originating from 0.05 μM fentanyl (final in-well concentration, no injection).

d. μ-opioid receptor (MOR) binding experiments

To determine the MOR affinities of NSOs and reference compounds, radioligand binding assays were carried out. Whole rat brains from male Sprague-Dawley rats were purchased from a commercial vendor (BIOIVT, Westbury, New York, USA). Membranes were prepared by homogenizing 0.5 g of brain tissue in 20 mL of ice-cold 50 mM Tris (pH adjusted to 7.4) using a polytron (Kinematica Inc., New York, NY, USA) on setting 6 for 20 sec. Homogenates were centrifuged at 19000 rpm for 15 min at 4°C. Supernatants were discarded and the pellets resuspended in 20 mL ice-cold 50 mM Tris pH 7.4. Tissue was centrifuged at 19000 rpm for 15 min at 4°C. Supernatants were discarded and the pellets were resuspended in cold 50 mM Tris pH 7.4 to yield a final concentration of 100 mg wet tissue weight per mL of buffer. Assays were carried out in test tubes containing 0.5 mL 50 mM Tris pH 7.4 at room temperature for 60 min. Binding assays were initiated by adding 100 μL of tissue homogenate to tubes containing 1.0 nM [³H]-DAMGO (Perkin Elmer, Waltham, Massachusetts, USA) and opioid test compounds. Concentration-response curves were generated by testing 8 different concentrations of each opioid compound in triplicate. Nonspecific binding was determined in the presence of 10 μM naloxone. The binding reaction was stopped by rapid filtration through Whatman GF/B filters (Whatman Inc., Clifton, New Jersey, USA) presoaked in 50 mM Tris pH 7.4 and 0.1% polyethyleneimine solution. The filters were washed with 50 mM Tris pH 7.4 (3 x 4 mL) and the bound radioactivity measured by liquid scintillation spectrometry. K_i values were calculated by nonlinear regression analysis using commercially available software (GraphPad Prism 7, San Diego, California, USA). All drugs were stored at −80°C as 10 mM stock solutions, which were diluted to appropriate working concentrations immediately prior to the binding experiments.

Results

Figure 3 shows the MOR activation profiles obtained for each compound in the three assays. The calculated efficacy (E_max, relative to hydromorphone) and potency (EC₅₀) values are summarized in Table 1. Being known MOR agonists, the four reference opioids (Panel A, compounds A-D) generated concentration-response curves as expected in all three assays. The opioids in Panel A could be considered full agonists compared to hydromorphone, though absolute efficacy varied across this group depending on the assay system. For hydromorphone (A), used as a reference compound for normalization, three overlapping curves were obtained in the different assays. When focusing on the newly evaluated NSOs (Panel B, compounds 1-7), it was interesting to note that, in general, most of these only weakly activated MOR. For 2,4-difluoro-U-48800 (4), the least active compound of the panel, the maximal signal could not be reached in the AequoScreen^® and MOR-βarr2 assays at 100 μM. This compound did not generate a reproducible signal in the MOR-mini-G_αi assay. Similarly, also for the weakly potent 2F-viminol (6) - a partial agonist in the AequoScreen^® and MOR-βarr2 assays - a reproducible signal was not obtained in the MOR-mini-G_αi assay at a concentration of 100 μM. Bromadoline (3) was a partial agonist compared to hydromorphone in the three assay systems. The most active compound of Panel B was ketobemidone (7), followed by N-ethyl-U-47700 (1). Importantly, in the AequoScreen^® assay, aspecific signals (not mediated by MOR activation) seemed to be picked up for N-ethyl-U-47700 (1) and U-62066/spiradoline (5), as could be demonstrated by some extent of off-target activation of the cannabinoid receptor CB₁ at the highest tested concentration(s) (Supplementary Information S1). Overall, the results obtained with the AequoScreen^® assay for the opioids from Panel C (compounds 8-12) were largely in line with those obtained previously in the two NanoBiT^® assays (Vasudevan et al. 2020; Vandeputte et al. 2020). Bromadol (8) was the most potent opioid in this panel, followed by brorphine (9) or TRV-130 (12). 7-OH-mitragynine (11) was generally more potent than the parent compound mitragynine (10), although 95% confidence intervals were overlapping. Mitragynine (10), 7-OH-mitragynine (11), and TRV-130 (12) were partial agonists compared to hydromorphone.

Figure 3.

MOR activation profiles of the three panels of studied opioids (mean receptor activation ± standard error of the mean, normalized to the maximum response of hydromorphone (= 100%)). Panel B depicts the newly evaluated non-fentanyl synthetic opioids. Note the differences in the scales of the y-axes. Blue dots, MOR-βarr2 data; black triangles, MOR-mini-G_αi data; red diamonds, AequoScreen^® data. ^a NanoBiT^® data from (Vandeputte et al. 2020). ^b NanoBiT^® data also used in (Vandeputte et al. 2021). ^c NanoBiT^® data (partially) from (Vasudevan et al. 2020).

Table 1.

Overview of potency (EC₅₀) and efficacy (E_max, relative to hydromorphone) data obtained in the three different MOR activation assays. 95% confidence intervals are shown between parentheses.

	NanoBiT® MOR-βarr2		NanoBiT® MOR-mini-Gαi		AequoScreen®
	EC50 (nM)	Emax (% relative to HM)	EC50 (nM)	Emax (% relative to HM)	EC50 (nM)	Emax (% relative to HM)
A. Hydromorphone	17.9 (13.4-24.1)	99.9 (95.7-104)	24.1 (16.2-36.4)	99.8 (94.4-105)	15.8 (9.84-26.2)	97.0 (90.7-104)
B. Fentanyl a	14.3 (11.3-18.1)	163 (157-169)	32.7 (23.2-45.8)	284 (269-300)	1.40 (1.15-1.71)	95.9 (93.4-98.3)
C. DAMGO	90.2 (40.7-199)	199 (176-222)	419 (323-540)	322 (308-335)	4.59 (3.23-6.48)	110 (105-115)
D. Morphine b	338 (239-478)	117 (111-123)	385 (247-593)	121 (115-127)	624 (349-1074)	109 (100-117)
1. N-ethyl-U-47700	451 (239-821)	161 (144-178)	767 (599-975)	247 (237-258)	241 (148-405)	139 (131-147)
2. 3,4-difluoro-U-47700	> 4000	132 (119-147)	> 9000	169 (154-186)	1817 (1553-2132)	103 (101-106)
3. U-47931E/bromadoline	2856 (1953-4226)	52.8 (49.3-56.5)	1703 (1072-2786)	64.1 (58.6-70.2)	554 (423-723)	85.9 (82.4-89.4)
4. 2,4-difluoro-U-48800	> 22000	3.85 *	-	-	> 28000	45.9 *
5. U-62066/spiradoline	1904 (1369-2678)	97.0 (90.3-104)	1788 (1294-2503)	126 (118-133)	598 (420-846)	134 (128-141)
6. 2F-viminol	2151 (486-11920)	21.2 (15.1-34.9)	-	-	4520 (2601-7819)	61.5 (55.1-68.4)
7. Ketobemidone	297 (149-591)	196 (177-215)	528 (367-743)	271 (259-284)	32.8 (24.3-44.3)	105 (101-109)
8. Bromadol a	1.89 (1.23-2.93)	182 (172-192)	3.04 (1.48-6.28)	462 (414-512)	0.478 (0.235-0.931)	120 (112-128)
9. Brorphine a	31.1 (20.7-47.0)	226 (207-246)	106 (84.1-134)	385 (363-408)	5.64 (4.21-7.53)	118 (113-122)
10. Mitragynine a	773 (513-1148)	19.2 (17.4-21.1)	1089 (280-4442)	29.7 (21.6-41.8)	254 (66.8-912)	25.1 (19.5-31.0)
11. 7-OH-Mitragynine a	369 (220-612)	43.5 (39.9-47.3)	191 (92.4-415)	52.2 (46.1-58.9)	55.9 (39.3-78.9)	86.5 (82.3-90.9)
12. TRV-130/oliceridine c	19.5 (8.06-48.9)	49.4 (42.7-56.5)	22.6 (14.0-37.4)	67.2 (61.1-73.7)	15.0 (6.68-36.4)	72.7 (64.3-81.4)

^*Maximum receptor activation seen at the highest concentration evaluated. EC₅₀ values to be interpreted with caution.

^aNanoBiT^® data from (Vandeputte et al. 2020).

^bNanoBiT^® data also used in (Vandeputte et al. 2021).

^cNanoBiT^® data (partially) from (Vasudevan et al. 2020).

Figures 4 and 5 provide a comparative assessment of the potencies (pEC₅₀ values) and efficacies (E_max values, relative to hydromorphone), obtained for compounds B-D and 1-12 in the three different assays. Overall, slightly higher potencies (i.e., lower EC₅₀ or higher pEC₅₀ values) were obtained in the AequoScreen^® assay compared to the NanoBiT^® systems (Figure 4A), although there was some overlap in the 95% confidence intervals obtained for some compounds in the different assays. The very weak MOR activation potential of 2F-viminol (6) resulted in wide, overlapping 95% confidence intervals, hampering comparison of the outcomes of the different assays. Potencies obtained in the two NanoBiT^® assays were mostly comparable or slightly lower in the mini-G_αi assay. Despite differences in absolute values, ranking of the different opioids based on their potencies was largely similar for the three assays, as the same compounds were included at the top (most potent) or the bottom (least potent) of the ranking (Figure 4B).

Figure 4.

Comparison of potencies (pEC₅₀) of the different compounds (1-12, B-D), a. between assays, and b. within one assay, to obtain a rank order of the different compounds based on their potencies. Data are depicted as means with corresponding 95% confidence intervals (n ≥ 3). At large, taking into account (overlapping) confidence intervals, the same compounds are included at the top (most potent) and bottom (least potent) of the ranking in the three assays, with only small differences in the exact order of the compounds.

Figure 5.

Comparison of efficacies (E_max, relative to hydromorphone) of the different compounds (1-12, B-D), a. between assays, and b. within one assay, to obtain a rank order of the different compounds based on their efficacies. Data are depicted as means with corresponding 95% confidence intervals (n ≥ 3). Note the differences in the scales of the y-axes. In the AequoScreen^® assay, the efficacies of full agonists are largely clustered around 100%. For N-ethyl-U-47700 (1) and U-62066/spiradoline (5) (indicated with an open circle), non-MOR mediated signals were picked up for some concentrations, potentially contributing to the efficacy exceeding 100% in this assay (Supplementary Information S1). In the NanoBiT^® assays, the same compounds were identified as being the most efficacious, albeit in a different order. Partial agonists were consistent in the different assays.

The range of efficacies obtained in the different assays varied widely. The widest range of E_max values was clearly seen in the MOR-mini-G_αi assay, followed by MOR-βarr2 (Figure 5A). Efficacies for many analytes obtained in the AequoScreen^® assay were close to 100%, which is to be expected for full agonists in this assay (cfr. infra). For N-ethyl-U-47700 (1) and U-62066/spiradoline (5), non-specific activation was observed at the 50 and/or 100 μM concentrations in the AequoScreen^® assay (Supplementary Figure S1), which may have contributed to the efficacy exceeding 100% for these compounds. In the NanoBiT^® assays, the same compounds were identified as being among the most and least efficacious, albeit in a different order (Figure 5B). All compounds that were partial agonists compared to hydromorphone in the NanoBiT^® assays, were also identified as such in the AequoScreen^® assay.

Calculation of the relative intrinsic activity (RA_i) value and subsequent calculation of bias factors for each compound in the different assays allowed insight into potential functional selectivity of the compounds. Theoretically, the MOR-mini-G_αi and AequoScreen^® assays both monitor the same pathway, i.e., that of G protein activation. The MOR-βarr2 assay monitors an upstream event following activation of the G protein-independent pathway. Figure 6 includes scatterplots of RA_i values (6A) and quantitative bias plots (6B). The RA_i values for the different compounds correlated well in the three employed assays (Figure 6A). For the calculation of bias (Figure 6B), N-ethyl-U-47700 (1) and U-62066/spiradoline (5) were excluded from analysis in the AequoScreen^® assay due to their off-target receptor activation (Supplementary Information S1). 2,4-difluoro-U-48800 (4) was excluded as no maximum was reached. Note that, as a result, statistical analysis was performed on a different number of opioids in Figure 6B. In addition, while 2,4-difluoro-U-48800 (4) and 2F-viminol (6) were very weak agonists in two assays, no reproducible signal could be picked up in the NanoBiT^®-mini-G_αi assay, theoretically indicating some extent of “qualitative bias” towards the NanoBiT^®-βarr2 and AequoScreen^® assays. When comparing the two NanoBiT^® assays, no statistically significant bias (compared to hydromorphone) could be detected. Comparison of the MOR-βarr2 and MOR-mini-G_αi assays with the AequoScreen^® system resulted in a negative bias factor for all compounds, except for morphine. The same trend is also reflected in Figure 6A, where the linear trendline is more shifted towards the AequoScreen^® assay in the respective scatterplots. As a consequence, also more deviating bias factors were obtained, with a statistically significant bias being obtained for U-47931E/bromadoline (3) and 7-OH-mitragynine (11), and for fentanyl (B) and DAMGO (C) in the MOR-βarr2-AequoScreen^® and MOR-mini-G_αi-AequoScreen^® comparisons, respectively.

Figure 6.

a Scatterplots of calculated relative intrinsic activity (log(RA_i)) values of the test compounds (B-D, 1-3, 5-12; 2,4-difluoro-U-48800 (4) was excluded as no maximum was reached) obtained in three independent experiments using the different MOR activation assays. RA_i values were calculated as described in Materials & Methods (Equation 1) and the graphs were plotted using Microsoft Excel 16.16.27. Pearson correlation coefficients (r) were calculated via Graphpad Prism 9. Figure 6b Quantitative MOR bias plots. Bias factors (β) ± SEM (calculated from three independent experiments) are plotted for all compounds (B-D, 1-3, 5-12). Hydromorphone was used as unbiased reference agonist (β = 0, not shown). For comparisons with the AequoScreen^® assay, compounds (1) and (5) were excluded due to the occurrence of aspecific signals in the AequoScreen^® system, potentially affecting the obtained E_max values in this assay. Depending on the assays, statistically significant (p < 0.05) biased agonism was detected for U-47931E/bromadoline (3) and 7-OH-mitragynine (11) (MOR-βarr2 vs AequoScreen^®) or fentanyl (B) and DAMGO (C) (MOR-mini-G_αi vs AequoScreen^®). No statistically significant biased agonism was found when comparing data from the MOR-βarr2 and MOR-mini-G_αi assays. 2,4-difluoro-U-48800 (4) and 2F-viminol (6) could be considered qualitatively biased towards the AequoScreen^® and MOR-βarr2 assays, as no reproducible signal could be detected in the MOR-mini-G_αi assay.

Given the relatively weak agonism of several of the newly evaluated NSOs in Panel B, we wondered whether some of these compounds would be capable of blocking fentanyl-induced MOR activation. We assessed this in the MOR-βarr2 assay, by pre-incubating the cells with a high concentration of agonists B-D and 1-12, and then adding fentanyl. Naloxone and buprenorphine were included as controls: as expected, being a known MOR antagonist, naloxone fully blocked receptor activation by fentanyl. Injection of fentanyl onto the known high-affinity partial agonist/antagonist buprenorphine (present in a 200x higher concentration), did not further increase the extent of buprenorphine-induced receptor activation. This was in line with the obtained K_i values, which were comparable for fentanyl and buprenorphine (Supplementary Table S2) (i.e., having an equally high affinity as buprenorphine, a lower concentration of fentanyl was not capable of displacing buprenorphine from the receptor in this scenario). Overall, the results for the different test compounds (shown in Figure 7A) could be categorized into two groups. In group 1, injection of fentanyl led to a signal increase (Figure 7A, orange versus green bars). For 3,4-difluoro-U-47700 (2), the signal rose to 100% that of fentanyl alone (blue bar). For the other compounds in this group (3, 4, 6 and 10), the extent of receptor activation generated by fentanyl alone (blue bar) was not reached. The observation that the signal further increased upon addition of a 200-fold lower concentration of fentanyl (as compared to the concentration of the test compound), is in line with the relatively low activity and MOR binding affinity (K_i values between 140-3931 nM, whereas the K_i of fentanyl = 8.43 nM) found for these compounds (Figure 7B, Supplementary Table S2). In other words, high-affinity fentanyl was capable of at least partially displacing these compounds from the receptor and further inducing MOR activation. Conversely, fentanyl injection did not further increase receptor activation for the compounds in group 2 (Figure 7A, compounds B-D, 1, 5, 7-9, 11, 12). Even for morphine (D), U-62066/spiradoline (5), 7-OH-mitragynine (11), and TRV-130/oliceridine (12), whose receptor activation at 10 μM was below that of fentanyl (orange vs. blue bar in Figure 7A), addition of fentanyl did not considerably change the signal (hence, these situations are comparable to what was seen with buprenorphine). As the binding affinities of these compounds were largely comparable to that of fentanyl (and generally higher than those found for the compounds in group 1; see also Supplementary Table S2), it can be hypothesized that a 200-fold lower concentration of fentanyl was not capable of displacing these compounds from the receptor and, hence, of further increasing the signal.

Figure 7.

a Results of competition experiments between different opioids (B-D, 1-12, naloxone and buprenorphine; final in-well concentration 10 µM) and fentanyl (final in-well concentration 0.05 μM) using the MOR-βarr2 assay. In short, NanoBiT^®-βarr2-expressing cells were pre-incubated with 10 μM of opioid agonist, followed by injection of solvent (orange bars) or fentanyl (green bars). The average normalized area under the curve (AUC) (± SEM) from three independent experiments is shown. Naloxone (a known MOR antagonist) and buprenorphine (a high-affinity partial MOR agonist/antagonist) were added as controls, and the signal generated by fentanyl alone (0.05 μM) is shown for comparison (blue bar). The tested opioid agonists can be categorized into two distinct groups (highlighted). In the first group (compounds 2, 3, 4, 6, 10), addition of fentanyl caused an increase in the agonist signal. In the second group, the signals did not further increase upon injection of fentanyl (compounds B-D, 1, 5, 7-9, 11, 12). Figure 7b Concentration-response curves from the MOR binding experiments of the different studied MOR agonists, categorized according to the classification in Figure 7a. In general, compounds in group 1 have lower binding affinities as compared to those in group 2. K_i values of the different compounds are included in Supplementary Information S2.

The herein presented data also allowed the assessment of relationships between MOR binding affinities and functional parameters (potency, efficacy and the combined relative intrinsic activity) obtained in the different assays (Figure 8). The correlation plots show that K_i values correlate well with potency (EC₅₀) and relative intrinsic activity (RA_i) values obtained in the three functional assays. A weaker or no significant correlation was found between binding affinity and efficacy values.

Figure 8.

Correlation plots investigating the relationship between MOR binding affinity (K_i values) and different functional parameters (EC₅₀, E_max and RA_i) as obtained in the employed MOR activation assays. Relative intrinsic activity of each compound is depicted as the log-transformed average RA_i of three independent experiments. Pearson correlation coefficients (r) were calculated with Graphpad Prism 9.

Discussion

The focus of the current study was threefold. First, we evaluated the MOR activation potential of a poorly characterized panel of non-fentanyl opioids, including several NSOs that have recently appeared on recreational drug markets or are being discussed on online drug user fora. Second, the use of three distinct in vitro assays allowed a more in-depth characterization of the different compounds and demonstrated the influence of different experimental methods on the obtained parameters. The latter highlights the importance of carefully considering an assay’s characteristics when interpreting potency and efficacy values. Third, given the relatively weak MOR agonism of most newly evaluated opioids, we assessed their MOR binding affinity and their ability to block receptor activation induced by fentanyl. The fentanyl interaction experiments may help to shed some light onto the effects of these opioids in the presence of fentanyl and their potential agonist/antagonist properties.

MOR activity profiling of a new panel of recent non-fentanyl NSOs

A well-known characteristic of the NPS market is its rapid evolution over time, with new drugs entering and leaving circulation continuously (UNODC 2021; Vandeputte et al. 2021a). In many cases, this constant flux and transient presence of a new drug on the market, makes it difficult to obtain a comprehensive view about its effects and harm potential. In vitro assays monitoring receptor activation have become important tools to obtain a first estimate of the potential in vivo effects of a compound (Baumann et al. 2018a; Vandeputte et al. 2021b). Importantly, it should be stressed that the results from in vitro assays primarily reflect a compound’s intrinsic receptor activation potential and, hence, are a mere estimate of the (extent of) potential effects experienced after consumption of a drug. Direct translation to the in vivo situation is complicated by many other factors (e.g., bioavailability, metabolic stability, off-target toxicity, tolerance) that are typically not considered in the in vitro setting. Importantly, MOR binding affinity has been reported not to predict the in vivo analgesic potency of opioid compounds (Volpe et al. 2011; Baumann et al. 2018b). Nevertheless, it has been shown that in vitro functional parameters offer a good estimate of the activity and potential harm a new drug may evoke in vivo (Benredjem et al. 2019; Gillis et al. 2020a; Vandeputte et al. 2021b). For example, in vitro efficacy has been shown to inversely correlate with the in vivo therapeutic window in mice (Gillis et al. 2020a). In addition, while functional assays are typically not performed in native tissues, a high degree of linear correlation was found here between MOR binding affinities in native tissue and different functional parameters obtained in MOR activation assays performed in cells ectopically expressing MOR (Figure 8). This indicates that there is a link between in vitro receptor activation data from bioassays and MOR affinity in native tissue.

N-ethyl-U-47700 (1), 3,4-difluoro-U-47700 (2), U-47931E/bromadoline (3), 2,4-difluoro-U-48800 (4), and U-62066/spiradoline (5) are compounds of the U-series, a group of drugs first synthesized by the Upjohn Company in the 1970-80s (Sharma et al. 2019; Kyei-Baffour and Lindsley 2020; Baumann et al. 2020). Based on the absence or presence of a methylene spacer between the aromatic ring and the carbonyl moiety of these compounds, U-drugs can be broadly categorized into two groups: i) compounds related to the well-known NSO U-47700 (e.g. (1), (2) and (3)), and ii) compounds related to U-50488, a known kappa-opioid receptor-selective agonist (e.g. (4) and (5)) (Baumann et al. 2020).

In the present study, three analogues of U-47700 were included. First, substitution of the N-methyl group for an ethyl function yields N-ethyl-U-47700 (1). To the best of our knowledge, N-ethyl-U-47700 was not described in early research by the Upjohn Company. Its first known emergence on the recreational drug market dates from March 2020, when N-ethyl-U-47700 was identified in seized material in the United States (U.S.) (Krotulski et al. 2020a). While it is less potent than U-47700 (EC₅₀ = 167 nM in the MOR-βarr2 assay (Vasudevan et al. 2020)), it is unclear how substituting the N-methyl for an N-ethyl group will translate into altered bioavailability (e.g., blood-brain barrier penetration or metabolism) or opioid activity in vivo. With an efficacy comparable to that of fentanyl in the two NanoBiT^® assays (and off-target activity detected in the AequoScreen^® assay), N-ethyl-U-47700 (1) is one of the most active compounds of Panel B. A second U-47700 analogue is 3,4-difluoro-U-47700 (2), for which difluorination drastically decreased potency. This reduced potency is in line with the results from a Catchpoint^® cAMP assay (Hsu et al. 2019), where no detectable MOR activity was found for 1 μM 3,4-difluoro-U-47700 (compound U06 in the study by Hsu et al.). While we obtained an efficacy slightly lower than that of fentanyl, the very high EC₅₀ values (μM) that were consistently obtained in our study, imply that very high doses of 3,4-difluoro-U-47700 (2) would need to be ingested before resulting in significant opioid effects. In March 2020, 3,4-difluoro-U-47700 (2) was first identified in seized material in the U.S. (Krotulski et al. 2020b). Lastly, the third U-47700 analogue included in this study is U-47931E/bromadoline (3). Bromadoline was first synthesized in 1978 (Szmuszkovicz 1978). In our in vitro assays, this compound was a partial agonist compared to hydromorphone, with a potency in the low μM range. Our results are in line with those of a [³⁵S]GTPɣS assay, where U-47931E/bromadoline was found to be a partial agonist compared to DAMGO (Janowsky 2017; UNODC). The potency in the [³⁵S]GTPɣS assay (447 ± 12 nM) was higher than that obtained via the NanoBiT^® assays, and more comparable to that found via the AequoScreen^® assay (Janowsky 2017; UNODC). The decrease in MOR activity seen for bromadoline (3) versus U-47700 (Vasudevan et al. 2020) is in line with the results of early studies investigating structure-activity relationships of various U-compounds in guinea pig brains, where an important role for halogens in both positions 3 and 4 was found (Szmuszkovicz 1978; Loew et al. 1988). In bromadoline, no halogen is present in position 3, potentially leading to a decreased MOR activity. On the other hand, it has been hypothesized that the less electronegative bromine atom may increase MOR affinity compared to a chlorine atom, but a direct comparison of bromadoline and its chlorinated analogue has not been made (Fujimoto et al. 1989; Sharma et al. 2019; Baumann et al. 2020). The analgesic activity of U-47931E/bromadoline (3) has been evaluated in vivo in mice and rats, where it was found to be around 6-7 and >800 times less potent as an antinociceptive agent when compared to morphine and fentanyl, respectively (Hayes et al. 1987). In a different study evaluating the antinociceptive effects of bromadoline using the warm water tail withdrawal assay, the drug was found to be 1.8 and > 300 times less potent than morphine and fentanyl, respectively (Flanigan et al. 2018). Corresponding quite well with the in vivo data, our in vitro assay platform indicated that bromadoline was about equal (AequoScreen^®) or 4-8 times less potent (MOR-mini-G_αi and -βarr2, respectively) than morphine. Compared to fentanyl, it was between 52-429 times less potent, depending on the assay used. Bromadoline appears to have been evaluated for its potential analgesic activity in humans and canines (Peng et al. 1985), but details are lacking (Sharma et al. 2019). The drug was first identified on the recreational market in Slovenia in December 2017 (EMCDDA 2017). In the U.S., the first identification of bromadoline was reported in seized material from August 2018 (Krotulski et al. 2018). As far as we are aware, no fatalities have been reported in literature for the three above-mentioned U-47700 analogues, although it should be noted that likely very few labs actively monitor these compounds.

2,4-difluoro-U-48800 (4) and U-62066/spiradoline (5) contain a methylene spacer between the aromatic ring and the carbonyl moiety, a structural characteristic that is known to increase the binding selectivity for the kappa-opioid receptor (KOR) over MOR (Szmuszkovicz 1999). Given that the focus of this study was the evaluation of opioids as potential drugs of abuse, which relies on MOR activation, evaluation of these compounds at KOR was beyond the scope of the present work. Furthermore, the abuse potential of kappa-opioid agonists could be limited by adverse effects (i.e., dysphoria) (Bruijnzeel 2009; Brust et al. 2016). Our results indicate a very low MOR activation potential of compounds (4) and (5). To the best of our knowledge, we are the first to report on the lack of MOR activation by 2,4-difluoro-U-48800 (4). Given this very low MOR activation, it is unlikely that 2,4-difluoro-U-48800 will generate substantial MOR-mediated effects at concentrations that are typically encountered in vivo. Secondly, we evaluated U-62066/spiradoline (5), a known kappa-selective opioid with weak MOR activity attributable to the (+)-enantiomer (VonVoigtlander and Lewis 1988; Meecham et al. 1989). The weak MOR activity is confirmed in the three in vitro functional assays used in this study. While spiradoline has not been formally identified on the recreational market, it is being offered online (Lamy et al. 2021).

Structurally different from U-compounds or other known NSOs, 2F-viminol (6) was first described in a patent from 1979 (Conti 1979). It is a fluorinated analogue of viminol or viminolo, an analgesic and antitussive agent marketed in some countries under the name Dividol^®. The 1979 patent states that the R,R-stereoisomer of 2F-viminol has a greater therapeutic index than both racemic viminol and its most active isomer (coined “R2” in the original patent (Della Bella et al. 1974) and in (Della Bella et al. 1976)), with a lower ED₅₀ and higher LD₅₀ in mouse assays (Conti 1979). However, 2F-viminol was never approved for medical use. In our in vitro assays, racemic 2F-viminol was a weak partial agonist at MOR. Importantly, it cannot be excluded that optically pure stereoisomers of 2F-viminol may display different effects at MOR, as has been shown for the different stereoisomers of viminol/viminolo. For example, the analgesic activity of the “R2”-stereoisomer of viminolo was shown to be nearly eight times higher than that of the racemic mixture, while some of the S,S- and S,R-isomers were found to have opioid antagonist activity (Della Bella et al. 1974, 1976). 2F-viminol was first notified to the European Early Warning System in July 2019 on behalf of Sweden (EMCDDA 2020). It was identified in seized material in the U.S. later that year, in October (Krotulski et al. 2019). No fatalities have been reported in the scientific literature for either viminol or 2F-viminol, although a case with dependence on viminol, requiring methadone substititution, has been reported (Turkiewicz and Baltieri 2007).

The last opioid in Panel B was ketobemidone (7). Ketobemidone, a known MOR agonist and NMDA receptor antagonist (Andersen et al. 1996), is the active ingredient of a prescription drug sold as Ketogan^® or Ketorax^® in mostly Scandinavian countries. Synthesis routes for ketobemidone and related piperidyl ketones were first described in patents dating from the 1940s (CIBA 1948). Several studies in humans show that ketobemidone and morphine display similar analgesic potency (Peltola 1958; Tamsen et al. 1982; Öhqvist et al. 1991; Jylli et al. 2004; Wolff et al. 2009). In the NanoBiT^® assays, the potency of ketobemidone was largely comparable to that of morphine, whereas ketobemidone was around 19 times more potent than morphine in the AequoScreen^® assay. In the NanoBiT^® assays, ketobemidone was more efficacious than morphine, potentially indicating stronger opioid effects with ketobemidone at higher concentrations. Apart from potential misuse of ketobemidone itself, analogues of the drug (e.g. acetoxymethylketobemidone or O-AMKD (Verougstraete et al. 2020; Lamy et al. 2021)) have started appearing on the recreational drug market.

Comparison of the three distinct MOR activation assays: potency and efficacy

In this study, a structurally diverse set of opioids was characterized in terms of their MOR activation potential using three different in vitro assays. The two employed NanoBiT^® assays (Figure 2A–B) monitor MOR activation in real-time over a period of two hours, generating a dynamic receptor activation profile of the studied opioid. Compared to the linear βarr2/mini-G_αi-receptor interaction, signals in the AequoScreen^® assay (Figure 2C) are substantially amplified: activation of one receptor will lead to multiple active G proteins, which eventually causes release of a multitude of calcium ions via a cascade of events. The result is a rapid and transient (10-15s) calcium flux inside the cell. Given the transient nature of the calcium flux, bioluminescence is short-lived and allows real-time measurement of receptor activation over a short period of time (approximately 25s in total).

Some key differences in the experimental set-up of the different assay types are summarized in Table 2. In theory, each of these aspects may contribute to the differences in absolute values obtained for the distinct opioids evaluated in this study.

Table 2.

Overview of selected key characteristics of the NanoBiT^® and AequoScreen^® assays used for the monitoring of MOR activation potential.

	NanoBiT® (MOR-βarr2 and MOR-mini-Gαi)	AequoScreen®
Cellular background	HEK-293T	CHO-K1
Tagging of receptor/transducer	Yes (SmBiT/LgBiT)	No
Bioluminescence catalyst	NanoLuc (deep sea shrimp-derived)	Aequorin (jellyfish-derived)
Substrate	Furimazine	Coelenterazine
Measurement level	Receptor-proximal(βarr2/mini-Gαi recruitment)	Downstream (Gα16-mediated release of intracellular Ca2+)
Measurement	2 hours, real-time	25 seconds, real-time
Signal amplification	Minimal	High

Our data clearly show the importance of considering the specific assay characteristics upon interpretation of potency (EC₅₀) and efficacy (E_max) values. In terms of potency, generally higher potencies (i.e., lower EC₅₀ values) were found in the AequoScreen^® assay, which may be the result of the high degree of signal amplification, known to cause a left-shift of concentration-response curves (Rajagopal et al. 2011). This underscores the importance of interpreting potency values relative to well-characterized reference agonists, such as those in Panel A of this study, rather than focusing on absolute EC₅₀ values obtained in different assays. This is also reflected in Figure 4A, where the rank order of potency is largely the same across the three assays.

The obtained range of efficacies was different for the three assays. In line with findings from previous studies using the NanoBiT^® MOR-βarr2 and MOR-mini-G_αi assays, the most dynamic range of efficacies was found in the latter assay (Vasudevan et al. 2020; Vandeputte et al. 2020, 2021b). Interestingly, the opposite trend was observed in a highly similar NanoBiT^® assay platform monitoring activation of the cannabinoid receptor CB₁ instead of MOR (Wouters et al. 2020), rendering it unlikely that an inherent assay characteristic can explain this observation for opioids. In the AequoScreen^® assay, a system ceiling is reached around 100% efficacy compared to hydromorphone. A ceiling effect is typically the result of saturation of downstream effectors, limiting the system response. For the AequoScreen^® assay, it has been suggested that regeneration of the aequorin enzyme requires low calcium concentrations. Hence, binding of calcium to aequorin is effectively irreversible during the rapid calcium flux following receptor activation, resulting in a progressive depletion of the biosensor (Charlton and Vauquelin 2010). This may translate to the observed assay ceiling. In the NanoBiT^® system, functional complementation of NanoLuc is reversible and no obvious ceiling effect was observed in these assays. Another interesting observation in this context lies in the consistent identification of partial agonists across the three assays. For highly amplifying systems, the difficulty in distinguishing partial and full agonists due to ‘receptor reserve’ is a well-known limitation (Wouters et al. 2019; Gillis et al. 2020b; Hill and Canals 2021). Receptor reserve refers to the phenomenon whereby only submaximal receptor occupancy is needed for the system to reach its maximal response, resulting in a comparable E_max for both full and partial agonists (Rajagopal et al. 2011; Wouters et al. 2020). Apparently, the signals of the different designated partial agonists in this study were below the system maximum of the AequoScreen^® assay, allowing their identification as partial agonists, regardless of the signal-amplifying nature of this assay.

Evaluation of biased agonism and relative intrinsic activity of the test compounds

Biased agonism or functional selectivity refers to the ability of a ligand to preferentially activate one signaling pathway over another, potentially leading to a different spectrum of in vivo effects (Wootten et al. 2018; Gillis et al. 2020b). For opioids, G protein-biased agonists have long been put forward as improved analgesics with a reduced side effect profile (Bohn 1999; Bohn et al. 2000; Raehal et al. 2005). While clear strides have been made in this research area (e.g., the first purported G protein-biased MOR agonist, TRV-130/oliceridine (12), was approved in the U.S. for use in acute pain in August 2020 (Lambert and Calo 2020)), the ‘G protein-bias hypothesis’ has recently been challenged (Gillis et al. 2020b, a; Stahl and Bohn 2020). Importantly, also the assessment of the ‘biased’ nature of compounds strongly depends on the deployed assays, as exemplified by our earlier work, in which we did not find a significant bias for TRV-130/oliceridine (12) using the MOR-βarr2 and MOR-mini-G_αi assays that were also used in the present study (Vasudevan et al. 2020). In this context, from a methodological point of view, the set-up of the current study was ideally suited to illustrate the impact of different experimental methods on the assessment of bias. The MOR-βarr2 assay can be considered a proxy for initiation of the G protein-independent pathway, whereas the MOR-mini-G_αi and AequoScreen^® assays reflect activation of the G protein-dependent pathway at two different levels (upstream and downstream, respectively). With the caveat that two different G_α protein subunits are monitored in the MOR-mini-G_αi and AequoScreen^® assays, it is interesting to note that even upon comparison of these assays, bias factors quite distinct from zero were found for most opioids, particularly when related to those obtained in the MOR-βarr2/mini-G_αi assay comparison. This underscores the caution required when interpreting bias factors: even upon comparison of two assays that theoretically monitor the same pathway, some extent of ‘bias’ can be detected, likely owing to the different assay set-ups and kinetics of the assays. This is also exemplified by 2,4-difluoro-U-48800 (4) and 2F-viminol (6), which were weak agonists in the AequoScreen^® and MOR-βarr2 assays, but did not lead to a reproducible signal in the MOR-mini-G_αi assay. Last, although one might expect that the interpretation of bias factors is complicated by the amplifying character and ceiling effect of the Aequoscreen^® assay, in contrast to the widely varying E_max values found in the NanoBiT^® assays, the obtained RA_i values correlated well in all three assays (Figure 6A).

Other factors that complicate the interpretation of bias factors in this study include (but are not limited to) the differences in expression levels of the utilized constructs across the three assays, and the differing cellular backgrounds (HEK-293T vs. CHO-K1) (Wouters et al. 2019). Hence, the different patterns of ‘bias’ seen here for the same opioids across different assays, may be the result of substantial ‘system bias’, rather than reflecting the actual (biased/balanced) nature of the ligand (Wouters et al. 2019). In addition, while both assay formats allow real-time measurement of receptor activation, potential dynamic aspects of signaling and/or biased agonism (coined ‘temporal bias’) (Grundmann and Kostenis 2017; Wouters et al. 2019) are likely differentially captured in the 25-second read-out used in the AequoScreen^® assay versus the 2h read-out in the NanoBiT^® assays. Finally, it should be noted that it is not clear if and to what extent findings related to in vitro bias may translate to the in vivo situation (Luttrell et al. 2015; Vandeputte et al. 2021b).

Evaluation of agonist/antagonist properties of the different opioids

Given the relatively weak MOR activation potential of most opioids in Panel B, we performed competition experiments with fentanyl to gain insight into potential agonist/antagonist properties of the opioids included in this study. To allow better interpretation of these results and to further deepen our in vitro characterization, MOR binding experiments were performed to determine receptor affinities. For the opioids in group 1 (2, 3, 4, 6, 10), all binding affinities were lower than that of fentanyl, providing a possible explanation as to why addition of fentanyl further increased the total extent of receptor activation, even in the presence of a 200-fold molar excess of the test compounds (i.e., having a higher MOR affinity, fentanyl was likely capable of displacing the test agonists from the receptor). Group 2 (B-D, 1, 5, 7-9, 11, 12) consisted of opioids with a binding affinity greater than or comparable to that of fentanyl. Combined with the generally higher intrinsic MOR activation potential of these opioids as compared to those in group 1, this likely explains why addition of fentanyl did not further increase the signal in this group. Notably, the obtained results for morphine (D), U-62066/spiradoline (5), 7-OH-mitragynine (11) and TRV-130/oliceridine (12) were comparable to that of buprenorphine, the latter being a partial agonist/antagonist known to produce comparable analgesia as full opioid agonists, but with a reduced likelihood of severe respiratory depression (Pergolizzi et al. 2010; Gillis et al. 2020a).

Conclusion

The increasing presence of synthetic opioids on the recreational drug market continues to pose a substantial threat to public health. Here, we evaluated a panel of poorly characterized non-fentanyl opioids for their potential to activate MOR, the primary molecular target for both clinically applied and abused opioids. Several of the studied opioids have recently emerged on the recreational drug market or have been the subject of discussions among drug users on online platforms. By comparing the results obtained via three in vitro MOR activation assays, we highlight the influence of distinct assay characteristics on the obtained efficacy and potency values, including the assessment of biased agonism. Importantly, while the high level of signal amplification in the AequoScreen^® assay may have contributed to the overall lower EC₅₀ values found in this assay, the assay allowed identification of partial agonists in agreement with both employed NanoBiT^® assays. On the other hand, the observed ceiling effect in the AequoScreen^® assay limits the range of efficacies for agonists as compared to the NanoBiT^® assays. We show that, while absolute values can differ between assays with varying experimental set-ups, the rank order of potencies (and efficacies) remains largely the same. In addition, relative intrinsic activity values of the different compounds correlated well across the three assays. Finally, determination of MOR binding affinities may help understand the extent of MOR activation exerted by a combination of different opioids, and the obtained Ki values were found to correlate well with different functional parameters. Taken together, the data in this work show that, while in vitro research at MOR may offer a first estimation of the potential danger newly emerging opioids could evoke, one should be aware of the influence of distinct assay characteristics on the obtained results. Indeed, while deploying different assays offers the advantage of generating a more ‘global’ picture of a compound’s activity, the underlying intricacies should be carefully considered. Irrespective of the exact assay set-up used, derived potencies and efficacies have the most value when compared to (a set of) well-characterized reference agonists.