Pharmacological characterization of levorphanol, a G-protein biased opioid analgesic

Abstract

Background

Levorphanol is a potent analgesic that has been used for decades. Most commonly used for acute and cancer pain, it also is effective against neuropathic pain. The recent appreciation of the importance of functional bias and the uncovering of multiple mu opioid receptor splice variants may help explain the variability of patient responses to different opioid drugs.

Methods

Here, we evaluate levorphanol in a variety of traditional in vitro receptor binding and functional assays. In vivo analgesia studies using the radiant heat tail flick assay explored the receptor selectivity of the responses through the use of knockout mice, selective antagonists and viral rescue approaches.

Results

Receptor binding studies revealed high levorphanol affinity for all the mu, delta and kappa opioid receptors. In ³⁵S-GTPγS binding assays, it was a full agonist at most mu receptor subtypes, with the exception of MOR-1O, but displayed little activity in β-arrestin2 recruitment assays, indicating a preference for G-protein transduction mechanisms. A knockout mouse and selective antagonists confirmed that levorphanol analgesia was mediated through classical mu receptors, but there was a contribution from 6TM targets, as illustrated by a lower response in an exon 11 knockout mouse and its rescue with a virally transfected 6TM receptor splice variant. Compared to morphine, levorphanol had less respiratory depression at equianalgesic doses.

Conclusions

While levorphanol shares many of the same properties as the classic opioid morphine, it displays subtle differences that may prove helpful in its clinical use. Its G-protein signaling bias is consistent with its diminished respiratory depression while its incomplete cross-tolerance with morphine suggests it may prove valuable clinically with Opioid Rotation.

Introduction

Levorphanol was first introduced into clinical practice in 1953. Clinically, it is an effective analgesic (1,2), including activity against neuropathic pain (3), with an advantageous longer half-live than most opioids. Furthermore, studies have identified subtle, yet significant, differences between levorphanol and morphine, suggesting potential differences in its mechanism of action (2,4,5). Prommer termed it “the forgotten opioid’ (6) due to its decline in its use, a result of extended release formulations of other opioids and its limited availability (2).

The mu opioid receptor gene, Oprm1, undergoes extensive 5′ and 3′ splicing to generate a series splice variants (7) (Supplemental Fig. 1). Most correspond to classical 7 transmembrane (7TM) domain G-protein coupled receptors differing only at the intracellular tip of their C-terminus where alternative splicing replaces exon 4 with its 12 amino acids with combinations of alternative exons, yielding unique amino acid sequences. The ligand binding pockets of the 7TM variants, encoded by exons 1, 2 and 3, are defined by the conserved transmembrane domains, consistent with the similar binding affinities of various opioids (7–9). In contrast to their similar binding affinities, the differing C-terminal amino acid sequences significantly impact efficacy, potency and bias of the various variants to different opioids (8,10)

A second set of variants results from 5′ splicing with a second promoter to generate truncated forms with only 6TM. While atypical in structure, these 6TM variants provide novel targets for compounds producing analgesia without many side-effects (7,11).

Levorphanol was developed and characterized before the classification of the standard opioid receptor families (12,13), the identification of their genes (14–18) and the recognition of the importance of biased signaling in GPCR drug action (19). While it is generally assumed that levorphanol acts through the traditional mu opioid receptor MOR-1, its subtle pharmacological differences from morphine suggest a more complex mechanism of action. It reportedly can interact with both NMDA receptors and monoamine uptake function (20,21), but at concentrations far greater than those needed for opioid receptor binding. The current study re-explores levorphanol in the context of our current understanding of mu opioid action.

Methods

Drugs

Morphine sulfate, levorphanol tartrate and naloxone HCl were obtained from the Research Technology Branch of the National Institute on Drug Abuse (Rockville, MD). Naloxonazine (22) and IBNtxA (3-iodobenzoyl naltrexamine) (11) were synthesized in our laboratory as previously described and structures validated.

Animals

All animal studies were approved by the Institutional Animal Care and Use Committee of the Memorial Sloan Kettering Cancer Center and were conducted in strict accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals in facilities accredited by the American Association for the Accreditation of Laboratory Animal Care (AAALAC). Animals were maintained on a 12-h light/dark cycle with Purina rodent chow and water available ad libitum, and were housed in groups of 5 until testing.

CD-1 mice were obtained from Charles River Laboratories. Exon 11 MOR-1 KO animals were generated by our laboratory as previously described (23), and were backcrossed to a 129SvEv^j background through speed congenic breeding (The Jackson Laboratory). Double exon 1/exon 11 knockout mice (E1/E11 KO) on a mixed 129SvEv-C57BL/6 background were generated as previously reported (24). All in vivo testing of KO models used the corresponding background strain for comparison. Since the E1/E11 KO were not congenic, we used mixed background wildtype controls.

Lentivirus production and injection

Lentiviral constructs and lentivirus production were produced as described previously (24). Briefly, the 6TM mMOR-1G cDNA was subcloned into a modified pWPI vector that independently expresses an enhanced green fluorescence protein (EGFP) (a gift from Dr. Didier Trono, Ecole Polytechnique Federale de Lausanne, Switzerland) to construct the lenti-mMOR-1G. Two lentiviruses, one expressing mMOR-1G and EGFP and the other, which expresses only EGFP (as a vector control), were generated in human embryonic kidney (HEK) 293T cells by co-transfecting the lenti-mMOR-1G construct or pWPI vector with PAX2 (a packaging vector) and pMD2 (an envelope vector) using FuGENE HD transfection reagent (Promega). The viral titer of the concentrated lentiviral particles was determined by quantifying EGFP-expressing cells in infected HEK293T cells with different dilutions using fluorescent microscope. Two μl of the lentiviral particles expressing mMOR-1G or vector alone without insertion (1.5 × 10⁹ transducing units/ml) were administrated intrathecally or intracerebroventricularly in E11 KO or E1/E11 KO mice on days 1, 3, and 5 under general halothane anesthesia. Under these conditions, protein expression from the virus progressively increases over a month and then remains stable for at least 14 weeks (24). All drug testing was carried out between 5 and 14 weeks post viral injection.

Tail Flick Analgesia

Analgesia was assessed with the radiant heat tail flick assay using a Ugo Basile radiant heat tail flick machine (Varese, Italy) with baseline values between 2–3 sec and a maximum latency of 10 sec. ED₅₀ values were determined using a cumulative dose-response approach to measure tail flick latency following a radiant heat stimulus (11). Following baseline latency determinations, each animal was injected subcutaneously with escalating doses of drug and tested 30 minutes after the previous injection at peak effect. Tailflick latencies were converted to %MPE by the formula %MPE = (Observed latency − Baseline latency)/(10 − Baseline Latency) * 100%. Dose-response curves were fit by nonlinear regression by GraphPad Prism (Carlsberg, CA). Dose-response curves were compared using an extra sum-of-squares F test.

Gastrointestinal motility assay

Gastrointestinal transit was measured as previously described (23). Briefly, animals (n=6–7 per group) were injected with either saline or opioid and 10 minutes later received a charcoal meal (10% charcoal and 2.5% gum tragacanth in distilled water) by gavage. Thirty minutes after administration of the charcoal meal, animals were sacrificed by cervical dislocation and the charcoal meal transit distance was measured and expressed as the total distance from the pyloric sphincter to the cecum.

Opioid receptor binding assays

¹²⁵I-IBNtxA (0.1 nM) competition binding assays were performed in membranes prepared from Chinese Hamster Ovary (CHO) cells stably expressing the indicated mouse opioid receptor clones or from CD-1 mouse brain as previously described (25,26). Nonspecific binding was defined in the presence of levallorphan (8 μM) and was subtracted from total binding to yield specific binding. All points were determined in triplicate and individual experiments independently replicated the indicated number of times. Inhibition curves for levorphanol were generated and the IC₅₀ value determined by nonlinear regression analysis and Ki values were calculated based upon the Cheng/Prusoff conversion (27) using Prism (GraphPad, San Diego, CA, USA). Protein concentrations were determined using the Lowry method.

Stimulation of ³⁵S-GTPγS binding

³⁵S-GTPγS Assays were performed using published methods (8). Membrane homogenates from CHO cells stably transfected with mMOR, mDOR, or mKOR (50 μg protein) were incubated for 1 hr at 30°C with the indicated drug, ³⁵S-GTPγS (0.05 nM) and GDP (60 μM in cell lines and 40 μM in brain) in a final volume of 1mL assay buffer containing Tris HCl (50 mM; pH 7.4 at 37°C), MgCl₂ (3 mM), EGTA (0.2 mM), NaCl (100 mM), and a protease inhibitor cocktail (leupeptin, bestatin, aprotinin, and pepstatin). GDP concentrations were optimized for each receptor assay: DOR-1 and KOR-1, 10μM; MOR-1, 30μM; brain, 60μM). Nonspecific binding was assessed by the addition of 10 μM cold GTPγS. Binding was terminated by vacuum filtration through Whatman GF/C glass fiber filters which were rinsed 3×2mL with cold Tris HCl. Filters were cut out and 3mL of scintillation fluor (Liquiscint, National Diagonistics, Atlanta, GA) was added to each tube and incubated at room temperature for at least 2 hours before being counted on a Packard Tri-Carb TR-2900 liquid scintillation counter.

β-Arrestin-2 recruitment assay

β-arrestin-2 recruitment was determined using the PathHunter^© enzyme complementation assay (DiscoveRx, Fremont, CA) using engineered CHO cells expressing the indicated engineered Oprm1 splice variant (DiscoveRx, Fremont, CA) (10). Cells were plated at a density of 2500 cells/well in a 384-well plate as described in the manufacturer’s protocol. The following day, cells were treated with the indicated compound for 90 minutes at 37°C followed by incubation with PathHunter^© detection reagents for 60 minutes. Chemiluminescence was measured with an Infinite M1000 Pro plate reader (Tecan, Männedorf, Switzerland).

Respiratory Depression

Respiratory rate was assessed in freely moving adult mice with the MouseOx pulse oximeter system (Starr Life Sciences) (11). Mice were shaved around the neck 24 hours prior to testing. Mice were habituated to the device for at least 1 hour prior to testing. A 5 second average breath rate was assessed at 5 minute intervals. A baseline was obtained over a 25 minute period before drug injection. Testing began 15 minutes post injection. Data are reported as % of baseline readings.

Statistical analysis

Data analysis was carried out using Prism (GraphPad, Carlsbad, CA). Behavioral dose-response curves were evaluated using nonlinear regression analysis to determine ED₅₀ values with 95% confidence limits. The model constrained the maximal response to 100% and the minimum response to 0% with a variable slope. Cumulative dose-response curves involved administering escalating doses of drug to each animal and testing the animal after each dose. The data was pooled and analyzed. In vitro studies examining ³⁵S-GTPγS binding and β-arrestin2 recruitment were evaluated using nonlinear regression analysis of dose-response curves without constraints and a variable slope to determine EC₅₀ values with 95% confidence limits. In vitro studies utilized pooled data from three independent determinations. Receptor binding studies yielded IC₅₀ values based upon the inhibition of control binding that were fit using nonlinear regression analysis with a model that constrained the maximal response to 100% and the minimum response to 0% with a variable slope. K_i values were obtained based upon the Cheng-Prusoff conversion (27). Values are the means ± s.e.m. of independent replications. Group comparisons utilized analysis of variance.

Results

Levorphanol effects on opioid receptors in vitro

Levorphanol potently competed binding to the classical mu, delta and kappa receptors expressed in CHO (Chinese Hamster Ovary) cells (Table 1; Supplemental Fig. 2). Its affinity was greatest for the full length (7TM) mu receptors, followed by delta and then kappa (Table 1). There was little difference in affinity for levorphanol among a series of full length Oprm1 splice variants (Table 1). Similar binding affinities were anticipated since they all share identical binding pockets comprised of the conserved transmembrane domains (TM) (7).

Table 1.

Receptor affinities of levorphanol

Receptor	Ki (nM)
Oprm1 (Mu)
7TM
MOR-1	2.4 ± 0.9
MOR 1-A	1.5 ± 0.3
MOR 1-B	1.6 ± 0.17
MOR 1-C	2.4 ± 0.9
MOR 1-D	0.8 ± 0.11
MOR 1-E	1.7 ± 0.32
MOR 1-F	1.4 ± 0.2
6TM
125I-IBNtxA target	54.8 ± 11.9
Oprd1 (Delta)	12.6 ± 0.7
Oprk1 (Kappa)	23.6 ± 0.3

K_i values were determined from IC₅₀ values obtained in binding studies with ¹²⁵I-IBNtxA against the indicated variants stably expressed in CHO cells, with the exception of the E11 site, which was determined in brain membranes in the presence of mu (DAMGO, 250 nM), delta (DPDPE, 250 nM) and kappa₁ (U50,488H, 250 nM) blockers. Values represent the means ± s.e.m. of at least 3 independent replications.

A second set of Oprm1 variants resulting from 5′ splicing are truncated, with only 6 transmembrane domains (6TM) (7,9). ¹²⁵I-IBNtxA binding in brain membranes in the presence of blockers of traditional mu (DAMGO), delta (DPDPE) and kappa (U50,488H) receptors labels a 6TM-dependent binding site (11). Although classical opioids like morphine and methadone have no appreciable affinity for this target (K_i >1000 nM), levorphanol competed it with a K_i of 55 nM (Table 1), similar to our prior observations (11).

Levorphanol was an agonist in ³⁵S-GTPγS binding assays against the classical opioid receptors (Table 2; Supplemental Fig. 3). Again, its potency was greatest against the mu receptor MOR-1, with lower potencies against DOR-1 and KOR-1. Against both MOR-1 and DOR-1, levorphanol elicited a full response when compared to a fixed concentration of DAMGO (1 μM) or DPDPE (1 μM), respectively. Against the kappa receptor KOR-1, levorphanol was a partial agonist, stimulating binding only to a level approximately half that of U50,488H.

Table 2.

Stimulation of ³⁵S-GTPγS binding to classical opioid receptors by levorphanol

	Stimulation of 35S-GTPγS Binding
	EC50 (nM)	Maximal Response
		% basal	% Control
MOR-1
Levorphanol	2.5 (1.2, 5.3)	339% (317, 360)	110%
DAMGO (1 μM)		307% ± 11
DOR-1
Levorphanol	41 (19, 88)	182% (172, 192)	89%
DPDPE (1 μM)		205% ± 3.3
KOR-1
Levorphanol	64 (33, 123)	171% (162, 180)	57%
U50,488H (1 μM)		302% ± 13

Stimulation of ³⁵S-GTPγS binding was determined in CHO cells stably expressing native MOR-1, DOR-1 or KOR-1 receptors and reported as stimulation over basal levels, as described in Methods. Full dose-response curves were carried out for levorphanol. Results are from three pooled independent experiments. Levorphanol EC₅₀ values were determined using nonlinear regression analysis (Prism) and are the EC₅₀ and 95% confidence limits. Values for the internal standards (DAMGO, DPDPE and U50,488H) are the means ± s.e.m. for stimulation of these compounds at 1 μM in three independent experiments. For comparison, in similar experiments, morphine stimulated ³⁵S-GTPγS binding with and EC₅₀ of 42 ± 8 nM and a Bmax of 72 ± 9% DAMGO (10).

Recent work has revealed the importance of biased signaling in the actions of many drugs at G-protein coupled receptors (19,28,29). In the opioid field, morphine tolerance and respiratory depression was attenuated in a β-arrestin2 knockout mouse (30). We therefore examined levorphanol in parallel ³⁵S-GTPγS binding and β-arrestin2 recruitment assays (10) against several full length Oprm1 variants (Table 3). To assess β-arrestin2 recruitment, we utilized the PathHunter complementation assay (DiscoverX) in which two inactive fragments of β-galactosidase are fused to either the GPCR (ProLink tag, PK) or β-arrestin2 (enzyme acceptor, EA). When β-arrestin2-EA attaches to the GPCR-PK β-galactosidase, activity from the fragments is restored, leading to cleavage of a reagent that induces chemiluminesence. The required fusion at the C-terminus of the GPCR with the ProLink tag did not significantly affect the binding affinity of the variants.

Table 3.

Levorphanol stimulation of ³⁵S-GTPγS binding and β-arrestin2 recruitment in mu opioid receptor splice variants

Variant	35S-GTPγS Binding		β-Arrestin2 Recruitment		Bias Factor
	EC50 (nM)	Max (% DAMGO)	EC50 (nM)	Max (% DAMGO)
MOR-1	28 (15.8, 49.6)	84.5 (74, 95)	103 (49.8. 215)	30.1 (25.4, 35)	−2.6
MOR-1A	2.3 (1.3, 3.9)	112 (106, 118)	37.8 (1.98, 72)	10.5 (8.8, 12.3)	−1.2
MOR-1B1	6.9 (3.1, 15.6)	90.6 (82, 99)	30.6 (10.5, 89)	12.9 (10, 15.9)	−4.4
MOR-1E	0.96 (0.57, 1.6)	109 (103, 115)	1,550 (0.07, 33,000)	20.7 (9.2, 32.1)	1.2
MOR-1O	1.15 (0.62, 2.1)	69 (64.3, 73.9)	54 (15.6, 188)	16.4 (11.8, 20.9)	9.4

Stable cell lines expressing the indicated splice variant, engineered with the ProTag to enable testing in the β-arrestin2 PathHunter assay, were examined for levorphanol-induced stimulated ³⁵S-GTPγS binding as previously described or β-arrestin2 recruitment using the PathHunter assay from DiscoverX (10). Full dose-response curves were carried out. Both EC₅₀ and B_max values determined from three independent replications were pooled and analyzed by nonlinear regression analysis with Prism. Max effects are given relative to a fixed DAMGO concentration (1 μM). Results are presented along with 95% confidence limits. Full dose-response curves are illustrated in Fig. 3. Bias factors were calculated as previously described (10). Bias factors greater than zero imply G-protein bias while those less than zero suggest β-arrestin2 bias. For comparison, morphine has a bias factor at of −1.5 at MOR-1 and of −5 at MOR-1O (10).

C-terminal splicing impacted levorphanol function in both ³⁵S-GTPγS and β-arrestin binding assays (Table 3; Supplemental Fig. 4). Levorphanol potently stimulated ³⁵S-GTPγS binding with all the splice variants, but with differing efficacies. Furthermore, the relative variations in efficacy and potency did not co vary. Levorphanol was most efficacious against MOR-1A and MOR-1E, where stimulation was slightly greater than DAMGO control. Maximal levorphanol stimulation of ³⁵S-GTPγS binding was only slightly less against MOR-1 and MOR-1B1, with values that were quite similar despite 4-fold differences in their potency (ED₅₀). The largest differences were observed with MOR-1O. Here, levorphanol was a partial agonist, with maximal stimulation of only 70% that of DAMGO despite its very high potency.

Levorphanol stimulated β-arrestin2 recruitment far less effectively than ³⁵S-GTPγS binding. Compared to DAMGO, levorphanol was a partial agonist in the β-arrestin recruitment assays against all the variants, with EC₅₀ values higher than those for stimulating ³⁵S-GTPγS. Levorphanol was most efficacious against MOR-1, but even here it only had a maximal recruitment equivalent to 30% that of DAMGO. It was even less efficacious against the other mu receptor splice variants, displaying a maximal response of less than 15% against MOR-1A and MOR-1B1. The differences between the two functional assays are readily seen with their dose-response curves (Supplemental Fig. 4). As with ³⁵S-GTPγS binding, potency did not correspond to efficacy.

Levorphanol analgesia

Levorphanol was equally potent in CD-1 (ED₅₀ 0.38 mg/kg) and 129 mice strains (ED₅₀ 0.38 mg/kg; Fig. 1a) in the radiant heat tail flick assay and it was approximately 5-fold more potent than morphine (ED₅₀ 2.3 mg/kg, s.c.) and oxycodone (ED₅₀ 2.0 mg/kg, s.c.) in CD-1 mice. Levorphanol analgesia was dependent upon mu opioid receptors, as illustrated by its inactivity in a full Oprm1 knockout mouse (Fig. 1b). Increasing the dose to 100 mg/kg still failed to significantly increase tail flick latencies. Antagonist studies confirmed the importance of mu receptors (Fig. 1c). Levorphanol analgesia was reversed by naloxone and levallorphan and the mu-selective antagonist β-funaltrexamine. The delta antagonist naltrindole (31) and the kappa antagonist norBinaltorphimine (32) were ineffective at doses active against DPDPE and U50,488H, respectively. Naloxonazine is a mu-selective antagonist active against a subpopulation of mu opioid receptors (22,33,34). As previously observed, naloxonazine completely blocked this dose of morphine. However, naloxonazine only partially inhibited levorphanol analgesia, implying both sensitive and insensitive mechanisms and illustrating its differences from morphine.

Fig. 1. Levorphanol analgesia.

a) Analgesic activity in mouse strains: Analgesic dose-response curves were determined in groups of CD-1 mice receiving levorphanol (n=10–30), morphine (n= 20) or oxycodone (n= 10) at the indicated dose s.c. and were assessed for analgesia in the radiant heat tailflick assay 30 min later. ED₅₀ values with 95% confidence limits were determined by nonlinear regression analysis using GraphPad Prism: levorphanol, 0.38 mg/kg (0.28, 0.51); morphine, 2.2 mg/kg (1.4, 3.6); oxycodone, 2.0 mg/kg (1.3, 3.1). A group of 129/Sv mice (n=10) received levorphanol at the indicated dose and were assessed for analgesia in the radiant heat tailflick assay 30 min later, yielding an ED₅₀ 0.38 (0.23, 0.64).

b) Analgesic activity in a mu knockout mouse: Groups of mixed background wildtype (n=9) or exon 1/exon 11 knockout (E1/E11 KO) mice (n=10) generated by heterozygous mating were administered levorphanol s.c. and assessed for analgesia in the radiant heat tailflick assay 30 min later. The dose-response curves were generated through a cumulative dosing paradigm. The ED₅₀ in the wildtype mice was 0.19 mg/kg (0.11, 0.31).

c) Sensitivity of levorphanol analgesia to selective antagonists: Groups of mice (n=20) received saline, naloxone (1 mg/kg, s.c.), levallorphan (1 mg/kg, s.c.), or naltrindole (10 mg/kg, s.c.) 5 min before levorphanol (1 mg/kg, s.c.). Mice treated with norBNI (20 mg/kg, s.c.), β-FNA (40 mg/kg, s.c.) or naloxonazine (35 mg/kg, s.c.) were given the antagonist 18–24 hr prior to levorphanol. Groups of mice (n=10) received saline or naloxonazine (35 mg/kg, s.c.) 18–24 hr prior to morphine (3.5 mg/kg, s.c.). Analysis was carried using ANOVA followed by Tukey’s multiple comparisons test.

d) Rescue of levorphanol analgesia in an exon 11 MOR-1 knockout: Groups of wildtype C57/BL6 mice (n=9), exon 11 knockout mice on a C57/BL6 background (E11KO; n=11) or exon 11 knockout mice on a C57/BL6 background that had been administered a lentivirus containing mMOR-1G (E11KO/MOR-1G; n=7) to reconstitute a 6TM variant were administered levorphanol (0.8 mg/kg, s.c.) and assessed for analgesia in the radiant heat tailflick assay or for Straub tail. ANOVA revealed significant differences among the groups for analgesia (p<0.0001; F_3,27 = 15.3). Bonferroni’s multiple comparisons test revealed significant differences between wildtype and exon 11 KO mice (p<0.0001) and between exon 11 KO mice with and without treatment with the lentivirus (p<0.05). Wildtype mice and MOR-1G virus treated exon 11 mice were not significantly different. The groups did not differ for the Straub tail, as determined by the Fisher Exact test (WT vs E11KO, p=1.0; WT vs E11KO/MOR-1G, p=0.203; E11KO vs E11KO/MOR-1G, p=0.240).

Earlier work observed that levorphanol analgesia was partially sensitive to elimination of a set of truncated Oprm1 variants containing only 6TM in an E11 knockout model (E11 KO) (11). In the current study, levorphanol analgesia (0.8 mg/kg, s.c.) was lower in the E11 KO mice (Fig. 1d). Reconstituting expression of the E11-dependent variant MOR-G through a lentiviral vector in the knockout mouse restored analgesia, establishing a role for truncated 6TM variants in its analgesic actions. The Straub tail seen with levorphanol in wildtype mice remained intact in the E11 KO mice, consistent with a traditional 7TM receptor mechanism.

Cross tolerance is a measure of whether drugs share a common mechanism of action. Prior work looking at morphine and levorphanol intravenous infusions in rats suggested that levorphanol displays incomplete cross tolerance to morphine (4). In the current studies, we examined groups of mice treated with levorphanol, morphine or oxycodone for 4 days and assessed tolerance on the 5^th day (Table 4). Levorphanol was approximately 6-fold less potent after chronic administration while chronic morphine displayed a similar shift of 5.5-fold of its dose-response curve. However, cross tolerance between the drugs was incomplete. Levorphanol analgesia was shifted only 3-fold in chronic morphine animals, compared to 5.5-fold for morphine. Similarly, both morphine and oxycodone showed only a 2-fold shift in the levorphanol-tolerant mice while levorphanol was shifted 6-fold.

Table 4.

Effects of chronic dosing on tolerance among opioids

	Naïve	Levorphanol tolerant		Morphine tolerant
	ED50 mg/kg, s.c.	ED50 mg/kg, s.c.	Shift	ED50 mg/kg, s.c.	Shift
Levorphanol	0.40 (0.3, 0.5)	2.3 (1.4, 3.6)	6	1.3 (0.8, 2.2)	3
Morphine	2.3 (1.5, 3.6)	5.3 (3.2, 8.7)	2	12.1 (7.6, 19.4)	5.5
Oxycodone	2.0 (1.3, 3.1)	4.6 (2.8, 7.6)	2

Groups of mice (n=10) were administered the indicated dose of drug for 4 days and ED₅₀ values determined for the indicated drug on the 5^th day using a cumulative dose-response paradigm. Chronic dosing was carried out with levorphanol (0.8 mg/kg, s.c.; 2× ED₅₀) or morphine (5 mg/kg, s.c.; 2.5× ED₅₀). ED₅₀ values were determined nonlinear regression analysis (Prism) and are presented with 95% confidence limits. All drugs were fully efficacious.

Other levorphanol actions

Respiratory depression is a major safety issue with opioid use. We assessed respiratory function by measuring respiratory rate and administered both morphine and levorphanol at doses approximately 5-fold greater than their respective analgesic ED₅₀ values. As anticipated, morphine produced a profound decrease in respiratory rate of approximately 50% that persisted over the full length of observation (Fig. 2a). Although we observed a decrease in respiratory rate following levorphanol, it was not as great as morphine. Analyzing the areas under the curve for the three groups showed that morphine was significantly different from both saline (p<0.001) and from levorphanol (p<0.05) while the difference between saline and levorphanol did not reach statistical significance (Fig. 2b).

Fig. 2. Non-analgesic levorphanol actions.

a) Respiratory depression: Groups of mice received saline (n=4), morphine (n=4) or levorphanol (n=5) after establishing a baseline and were monitored for 50 minutes.

b) The areas under the curves were determined for each treatment using Prism and compared using ANOVA followed by Tukey’s multiple comparison test. The three groups were significantly different (F_2,10=16.64, p<0.0007). Morphine was significantly different from both saline (p<0.001) and levorphanol (p<0.05). Levorphanol was not significantly different from saline.

c) Inhibition of gastrointestinal transit: Groups of mice (n=9–10) received either saline, morphine (3 mg/kg, s.c.) or levorphanol (0.4 or 4 mg/kg, s.c.). They then were given a charcoal meal and the distance traveled determined after 30 min, as described in methods. ANOVA showed that the groups were significantly different (F_3,35=34.8; p<0.0001). Tukey’s multiple comparison test showed that saline was significantly different from all the others (p<0.001), but the morphine and levorphanol groups were not significantly different from each other.

Morphine and levorphanol both decreased gastrointestinal transit of a charcoal meal (Fig. 2c). Both opioid treatments significantly depressed transit when compared to saline. The decrease from morphine and levorphanol at their analgesic ED₅₀ doses were comparable. Higher morphine doses typically totally inhibit transit in this assay. However, increasing the levorphanol dose 10-fold did not further reduce transit, implying a ceiling effect in its actions.

Discussion

Not all mu opioids are the same and patients do not respond similarly to them (35). Several mechanisms may be responsible for these subtle clinical differences. With biased signaling, two drugs acting through a common receptor may have different pharmacological profiles due to their differential activation of transduction pathways (19). For example, one drug may preferentially activate G-protein transduction pathways while another may preferentially recruit β-arrestin2. With opioids, this can be important since β-arrestin2 has been implicated in many side-effects, such as respiratory depression (19,30).

The identification of a multitude of splice variants of the mu opioid receptor gene Oprm1 provides additional complexity since they provide multiple mu opioid receptor targets. Oprm1 generates dozens of distinct proteins through alternative splicing, with similar patterns in humans, mice and rats (for review, see (7)). These variants can be divided into three classes based upon the protein structure. The majority are classical 7TM GPCRs with identical binding pockets defined by exons 1, 2 and 3 that differ only at the tip of the intracellular C-terminal. While drugs have similar receptor binding affinities of each of these variants, their efficacies and transduction bias varies from variant to variant and from drug to drug (10). The overall pharmacological profile of a drug reflects the simultaneous activation of many splice variants with varying efficacies and bias, contributing to the variability in response among mu drugs.

In addition to the classical 7TM receptors, Oprm1 also generates a set of truncated proteins with 6TM that are an essential component of a novel binding site in brain and that also are involved with analgesia (11,24,26,36). The role of 6TM variants in mu opioid analgesia varies markedly among different drugs. For example, morphine analgesia is totally independent of these truncated forms (23), but buprenorphine analgesia is lost in knockout mice lacking them (26).

Levorphanol is an effective analgesic, with its relatively long half-life and activity against neuropathic pain being major advantages (3). Although it acts through mu receptors, its actions subtly differ from morphine, as illustrated by their unidirectional cross tolerance in rats (4,5). Like other mu opioids, levorphanol labels the full length mu opioid receptors with similar low nanomolar affinities, while also showing affinity for delta and kappa receptors and the 6TM-dependent ¹²⁵I-IBNtxA binding site in brain. It was a full agonist in ³⁵S-GTPγS binding studies at the classical mu (MOR-1) and delta receptors and a partial agonist at kappa ones. However, its ³⁵S-GFTPγS activity profile varied among the mu receptor spice variants. Its ED₅₀ for MOR-1 was about 30-fold higher than MOR-1E and MOR-1O. It also failed to fully stimulate binding in the MOR-1O (69% DAMGO) compared to approximately 110% for MOR-1 and MOR-1E. This partial agonism of levorphanol at MOR-1O was similar to morphine (64% DAMGO), but quite different from fentanyl (110% DAMGO) (10).

Among the full length mu opioid receptor variants, levorphanol transduction showed a pronounced G-protein preference with markedly lower potency and maximal stimulation of β-arrestin compared to DAMGO. Its greatest β-arrestin response was only 30% that of DAMGO with MOR-1 and 10–20% for the others. This preference for G-protein stimulation was consistent with its effects on respiratory rate. Many mu side-effects have been associated with β-arrestin activity (30,37), including respiratory depression (38). Although levorphanol depressed the respiratory rate, its effects were significantly lower than those of an equianalgesic morphine dose.

Levorphanol analgesia is dependent upon mu opioid receptors. Elimination of all Oprm1 gene products in the E1/E11 knockout animals blocked levorphanol analgesia. The small residual effect in the E1/E11 KO mice remained less than 15% MPE at doses as high as 250-fold greater than its analgesic ED₅₀ and is unlikely to be clinically relevant. However, the almost total dependence of levorphanol on Oprm1 gene products does not rule out a potential contribution from delta or kappa receptors, although the antagonist studies make a delta or kappa contribution less likely. The mu-selective antagonist β-funaltrexamine fully blocked levorphanol analgesia while neither the kappa antagonist norbinaltorphimine nor the delta antagonist naltrindole were effective.

Despite its mu classification, levorphanol actions could be distinguished from morphine. The possibility of mechanistic differences between the two drugs was suggested a number of years ago by their unidirectional cross tolerance in the rat (4). In these studies, rats tolerant to levorphanol showed cross tolerance to morphine while morphine tolerant rats still responded to levorphanol. This suggested that, in addition to a shared target with morphine, levorphanol may also be acting through a second target not utilized by morphine. The incomplete cross tolerance between levorphanol and both morphine and oxycodone in the current study in mice supports this concept. Animals tolerant to one drug were only partially cross tolerant to the other. While the cross tolerance studies between levorphanol and morphine in both mice and rats were consistent with subtly different analgesic mechanisms, the mice showed bidirectional incomplete cross tolerance while rats displayed unidirectional cross tolerance. This difference could be due to a variety of reasons. In addition to potential species differences, the methodology also differed. Whereas the current study was carried out over 5 days with intermittent subcutaneous dosing, the earlier rat report used continues intravenous infusions over 24 hr, resulting in a smaller degree of tolerance.

Unlike β-funaltrexamine which antagonizes all mu receptor actions, naloxonazine antagonizes a restricted to a subset of mu receptors and morphine actions (22,39). Naloxonazine fully blocked morphine analgesia but was only partially effective against an equianalgesic levorphanol dose. Levorphanol and morphine actions also could be differentiated genetically. Eliminating 6TM variants in the E11 KO mice modestly lowered levorphanol responses, a response that was rescued by re-expressing the 6TM variant MOR-1G using a lentiviral vector. In contrast, morphine analgesia was fully retained in the E11 KO mice (23). However, the identity of this secondary target is still uncertain. ¹²⁵I-IBNtxA labels a 6TM-dependent binding site in brain unrelated to traditional full length mu, delta or kappa receptors that can mediate analgesia that is insensitive to norbinaltorphimine or to naltrindole (11,24) that may contribute to levorphanol actions.

The current studies addressed the potential role of opioid receptors, but levorphanol also reportedly will interact with non-opioid systems (6,20,21). It binds to NMDA receptors with a K_i 600 nM (21) and inhibits both norepinephrine (1.2 μM) and serotonin uptake (90 nM) (20). The importance of these interactions, however, is unclear since they occur at concentrations 50–100-fold greater than those needed to occupy opioid receptors. In naïve patients, where the dose of drug is titrated for activity on opioid receptors, the responses are less likely to involve these extra-opioid receptor effects, but they may become more prominent at the higher doses used in tolerant subjects. For example, the brain concentrations of levorphanol in tolerant mice were noted to be 4-fold greater than naïve mice (40).

In conclusion, levorphanol is a potent analgesic valuable in the pain management. While its early classification as a mu opioid is appropriate, new insights into the Oprm1 gene and its products has revealed a previously unrecognized complexity that may be contributing to the subtle, but potentially important, differences between levorphanol and other classical mu opioids such as morphine. These preclinical studies may help explain the variable clinical responses observed among different opioids and the utility of Opioid Rotation, and illustrate the advantages of having a range of opioid choices in the management of pain.

Supplementary Material

3

Supplemental Figure 1: Schematic of Oprm1 gene products

Supplemental Figure 2: Levorphanol competition of mu, delta and kappa receptor binding in cloned cell lines.

Supplemental Figure 3: Levorphanol stimulation of ³⁵S-GTPγS binding in mu, delta and kappa receptors in cloned cell lines

Supplemental Figure 4: Levorphanol stimulation of ³⁵S-GTPγS binding and β-arrestin2 recruitment in mu opioid receptor splice variant expressing cell lines

Key points.

• Question: To determine if levorphanol is pharmacologically different from morphine, we explored its actions on a range of different opioid receptors.

• Findings: Levorphanol was a G-protein biased agonist at a number of mu opioid receptor splice variants, consistent with diminished respiratory depressant activity, and displayed incomplete cross tolerance to morphine and oxycodone.

• Meaning: Despite acting through mu opioid receptors, levorphanol actions can be differentiated from morphine and may provide advantages clinically in pain mananagement, particularly with regards to Opioid Rotation.