Cannabinoid CB₂ Agonist GW405833 Suppresses Inflammatory and Neuropathic Pain through a CB₁ Mechanism that is Independent of CB₂ Receptors in Mice

Abstract

GW405833, widely accepted as a cannabinoid receptor 2 (CB₂) agonist, suppresses pathologic pain in preclinical models without the unwanted central side effects of cannabinoid receptor 1 (CB₁) agonists; however, recent in vitro studies have suggested that GW405833 may also behave as a noncompetitive CB₁ antagonist, suggesting that its pharmacology is more complex than initially appreciated. Here, we further investigated the pharmacologic specificity of in vivo antinociceptive actions of GW405833 in models of neuropathic (i.e., partial sciatic nerve ligation model) and inflammatory (i.e., complete Freund’s adjuvant model) pain using CB₂ and CB₁ knockout (KO) mice, their respective wild-type (WT) mice, and both CB₂ and CB₁ antagonists. GW405833 (3, 10, and 30 mg/kg i.p.) dose dependently reversed established mechanical allodynia in both pain models in WT mice; however, the antiallodynic effects of GW405833 were fully preserved in CB₂KO mice and absent in CB₁KO mice. Furthermore, the antiallodynic efficacy of GW405833 (30 mg/kg i.p.) was completely blocked by the CB₁ antagonist rimonabant (10 mg/kg i.p.) but not by the CB₂ antagonist SR144528 (10 mg/kg i.p.). Thus, the antinociceptive properties of GW405833 are dependent on CB₁ receptors. GW405833 (30 mg/kg i.p.) was also inactive in a tetrad of tests measuring cardinal signs of CB₁ activation. Additionally, unlike rimonabant (10 mg/kg i.p.), GW405833 (10 mg/kg, i.p.) did not act as a CB₁ antagonist in vivo to precipitate withdrawal in mice treated chronically with Δ⁹-tetrahydrocannabinol. The present results suggest that the antiallodynic efficacy of GW405833 is CB₁-dependent but does not seem to involve engagement of the CB₁ receptor’s orthosteric site.

Introduction

Cannabinoids exhibit antinociceptive effects in behavioral and electrophysiologic studies in rodents (Iversen and Chapman, 2002; Guindon and Hohmann, 2009). Two well characterized cannabinoid receptors are cannabinoid receptor 1 (CB₁) (Matsuda et al., 1990) and cannabinoid receptor 2 (CB₂) (Munro et al., 1993). Agonists that bind to CB₁ receptors have many desirable therapeutic properties, but they also induce undesirable psychoactive side effects that result from the abundant expression of CB₁ receptors in the central nervous system (CNS) (Herkenham et al., 1991; Galiègue et al., 1995). By contrast, CB₂ receptors are restricted mainly to the periphery and immune tissues (Galiègue et al., 1995) but may nonetheless be induced in the CNS by inflammation or injury (Zhang et al., 2003; Guindon and Hohmann, 2008; Viscomi et al., 2009; Atwood and Mackie, 2010). Therefore, much research interest has emerged in developing CB₂ agonists as suitable analgesics.

GW405833 (also known as L768242; see Fig. 1) binds with preferential affinity to CB₂ over CB₁ receptors (Gallant et al., 1996; Valenzano et al., 2005; Yao et al., 2008) (Table 1). Antihyperalgesic effects of GW405833 were first described in a carrageenan model of inflammatory nociception (Clayton et al., 2002). GW405833 was subsequently shown to reduce allodynia in other inflammatory pain (Valenzano et al., 2005; Whiteside et al., 2005; Beltramo et al., 2006), traumatic nerve injury (Valenzano et al., 2005; Whiteside et al., 2005; Beltramo et al., 2006; Brownjohn and Ashton, 2012; Hu et al., 2009; Leichsenring et al., 2009), and incisional injury (LaBuda et al., 2005; Valenzano et al., 2005) models. Despite substantial penetration to the CNS (Bouchard et al., 2012), GW405833 did not produce cannabimimetic deficits at doses lower than 100 mg/kg i.p. (Valenzano et al., 2005; Whiteside et al., 2005). Antiallodynic efficacy of GW405833 was opioid independent as systemic administration of naltrexone did not block the antihyperalgesic or antinociceptive effects of GW405833 (Whiteside et al., 2005). Thus, GW405833 was advanced as a more specific CB₂ agonist compared with AM1241, which exhibited naloxone sensitivity under some (Ibrahim et al., 2005), but not all, conditions (Rahn et al., 2010) and showed substantial protean agonism (Yao et al., 2006).

Fig. 1.

(A) Chemical structure of GW405833, also referred to as L768242 (Gallant et al., 1996; Valenzano et al., 2005). Cayman Chemical (cat. no. 20219) and Tocris Bioscience (cat. no. 2374) identify this ligand as a selective, high-affinity CB₂ receptor partial agonist. (B and C) Timeline for the dose response of GW405833 and blockade with antagonist in PSNL and CFA model. (B) In the PSNL model, the baseline was tested a day before the surgery. Animals were allowed (≥14 days) to fully develop neuropathic pain before pharmacologic manipulations. After the full establishment of neuropathic pain, different doses of GW405833 were tested in ascending order with variable intervals to ensure no carryover effect. Specifically, 0 mg/kg was tested, followed by 3 mg/kg on the same day with 3- to 4-hour intervals; 10 mg/kg was then tested the next day, followed by 30 mg/kg tested 2 days later. The CB₁ or CB₂ antagonist was tested 3 or 4 days after the last dose of 30 mg/kg. Groups: C57BL/6J, n = 6 (males); CB₂KO, n = 6 (males); CD1, n = 8 (mixed sex); CB₁KO, n = 8 (mixed sex). (C) In the CFA model, the baseline was tested on the same day before the CFA injection; the test of dose response started 48 hours after the injection. The timeline for the dose response of GW405833 was the same as the timeline in PSNL model. Groups: C57BL/6J, n = 8 (mixed sex); CB₂KO, n = 8 (mixed sex); CD1, n = 8 (males); CB₁KO: n = 6 (males).

TABLE 1.

In vitro binding profile of GW405833

Ki (nM)					rCB2/rCB1 Selectivity	Assay	Reference
hCB1	hCB2	hCB2/hCB1 Selectivity	rCB1	rCB2
4772	3.92	1217	273	3.6	76	[3H] CP55,940 binding	Valenzano et al., 2005
282	7.62	37	NA	11.1	NA	[3H] CP55,940 binding	Yao et al., 2008
1917	12	160	NA	NA	NA	[3H] WIN55,212-2 binding	Gallant et al., 1996

NA, not available; hCB, human cannabinoid receptor; rCB, rat cannabinoid receptor.

In in vitro studies, GW405833 was reported to behave as a partial agonist at human CB₂ receptors (Valenzano et al., 2005) and, alternately, a potent inverse agonist at both human and rat CB₂ receptors and weak agonist at rat CB₁ receptors (Yao et al., 2008). GW405833 was further proposed to be a protean agonist whose pharmacologic properties were dependent on the constitutive activity of the CB₂ receptor (Mancini et al., 2009). More recently, GW405833 was suggested to act as a noncompetitive CB₁ antagonist as GW405833 noncompetitively antagonized CP55,940-induced adenylyl cyclase activity, extracellular signal-regulated kinase 1/2 phosphorylation, phosphatidylinositol 2 signaling, and CB₁ internalization in vitro in human embryonic kidney cells transfected with CB₁ and showed a complex, time-dependent effect on arrestin recruitment in Chinese hamster ovary cells (Dhopeshwarkar et al., 2017).

Antiallodynic effects of GW405833 have been reported in several preclinical pain models, but pharmacologic specificity has been poorly characterized, and discrepancies among studies do exist. In a few studies, the antihyperalgesic effect of GW405833 was blocked by the CB₂ antagonist SR144528 (Clayton et al., 2002; Beltramo et al., 2006), whereas the effects of GW405833 did not differ from vehicle in CB₂ KO mice (Valenzano et al., 2005). The possible involvement of CB₁ was never investigated in these studies. By contrast, the antinociceptive effect of GW405833 in reducing lactic acid–stimulated stretching was partially blocked by the CB₁ antagonist rimonabant but not by the CB₂ antagonist SR144528, whereas GW405833-induced attenuation of acid-induced depression of intracranial self-stimulation was not blocked by either SR144528 or rimonabant (Kwilasz and Negus, 2013). Because of these discrepancies, we elected to further characterize the pharmacologic specificity of GW405833 in two mechanistically distinct animal pain models, neuropathic pain induced by partial sciatic nerve ligation (PSNL) and inflammatory pain induced by intraplantar injection of complete Freund’s adjuvant (CFA). To characterize pharmacologic specificity of the in vivo profile of GW405833, we used CB₂ KO, CB₁KO, and respective wild-type (WT) mice, as well as CB₁ and CB₂ antagonists. The highest effective dose of GW405833 (30 mg/kg i.p.) was also evaluated for cardinal signs of CNS CB₁ activation in the cannabinoid tetrad of tests. GW405833 also behaved as noncompetitive CB₁ antagonist in vitro (Dhopeshwarkar et al., 2017), but it remains unclear whether the reported in vitro effects could account for the pharmacologic effects of this compound observed in vivo. We therefore also evaluated whether GW405833 would behave similarly to the CB₁ antagonist rimonabant and precipitate a withdrawal syndrome in mice treated chronically with Δ⁹-tetrahydrocannabinol (Δ⁹-THC).

Materials and Methods

Subjects

Mice (5 ± 2 months old) of both sexes from different strains were used as specified in each study. CB₂KO mice and WT littermates on a C57BL/6J background (The Jackson Laboratory, Bar Harbor, ME) and CB₁KO mice and WT mice on a CD1 background (Charles River Laboratories, Wilmington, MA) were included. Animals were single-housed at relatively constant temperature (73 ± 2°F) and humidity (45%) under light-dark cycles of 12/12 hours. All the experimental procedures were approved by Bloomington Institutional Animal Care and Use Committee of Indiana University and followed the guidelines for the treatment of animals of the International Association for the Study of Pain (Zimmermann, 1983).

Drugs and Chemicals

GW405833 was purchased from two different commercial sources (Tocris Bioscience, Bristol, UK, and Cayman Chemical Company, Ann Arbor, MI). CFA and WIN55,212-2 were purchased from Sigma-Aldrich (St. Louis, MO). Rimonabant (SR141716A), SR144528, and Δ⁹-THC were obtained from the National Institutes of Health Drug Supply Program (Research Triangle Institute, Research Triangle Park, NC). All drugs except Δ⁹-THC were dissolved in a vehicle of dimethylsulfoxide (Sigma-Aldrich, St. Louis, MO), emulphor (Alkamuls EL 620L, Solvay, Princeton, NJ), ethanol (Sigma-Aldrich), and sterile saline (Aquilite System, Hospira Inc, Lake Forest, IL) at a ratio of 5:2:2:16, respectively. Δ⁹-THC was dissolved in a vehicle of 95% ethanol, Cremophor EL (Sigma-Aldrich), and sterile saline (Aquilite System) in a ratio of 1:1:18, respectively. All compounds were delivered via i.p. injection in a volume of 5 ml/kg.

PSNL

PSNL was performed to induce neuropathic pain. Mice were anesthetized with isoflurane (5% for induction followed by 2%–2.5% for maintenance) in 1.5 liter/min of oxygen. The right high thigh area was then shaved and disinfected, followed by a 1- to 1.5-cm longitudinal incision. The muscle layers were bluntly dissected to expose the underlying sciatic nerve. One third to half of the sciatic nerve at the location just above its trifurcation was tightly ligated using 8-0 silk suture (Sharpoint DA-2526N). The muscle layers were then closed with 5-0 silk thread (Ethicon 682 G; Ethicon Inc., Somerville, NJ), and the skin was closed by staples (Autoclips 9 mm no. 7631). Animals were allowed at least 2 weeks to recover and fully develop neuropathic pain.

CFA-Induced Inflammatory Pain

A single intraplantar injection was used to deliver CFA (20 µl; diluted 1:1 in sterile saline) unilaterally into the right hind paw. Animals were given 48 hours to fully develop inflammatory pain before pharmacologic manipulations.

Assessment of Mechanical Allodynia

Animals were habituated in individual transparent Plexiglass test chambers (10.5 × 9 × 7 cm) atop a metal mesh platform for at least 30 minutes before the testing. An electronic von Frey anesthesiometer with a 90-g probe (IITC Life Science, Woodland Hills, CA) was used to determine the paw withdrawal threshold to mechanical stimulation of the hind paw. A rigid tip with a diameter of 0.8 mm attached to the probe was applied vertically against the plantar surface of the paw with gradually increasing force. The force in grams necessary to produce paw withdrawal was recorded in duplicate for each paw.

Baseline mechanical paw withdrawal thresholds were measured in each paw for each animal before performing either PSNL or injecting CFA. Another predrug baseline was then taken after painful peripheral neuropathy or inflammatory pain was fully established. GW405833 was administered (i.p.) 30 minutes before evaluation of the impact of drug manipulations on mechanical paw withdrawal thresholds. Different doses of GW405833 were injected (i.p.) within subjects in the order of vehicle (0), 3, 10, and 30 mg/kg. Sufficient time was allowed to lapse between each dose to verify that mechanical paw withdrawal thresholds returned to the predrug levels before dose escalation. (For detailed timeline information, please see Fig. 1, B and C). In the groups where CB₁ or CB₂ antagonists were tested, rimonabant or SR144528 (10 mg/kg i.p.) was administered 20 minutes before GW405833 injection.

Tetrad Test

Rotarod.

Motor performance was measured using an accelerating Rotarod (IITC Life Science) (4–40 rpm, 300-second cutoff time). Male CD1 mice were trained over 2 consecutive days, and on day 3, the baseline latencies to descend from the rotating drum were measured. Mice that did not meet exclusion criteria before baseline measurements (i.e., ability to stay on rotating drum for at least 30 seconds on baseline day) were removed from the study and did not receive pharmacologic treatments.

Rectal Temperature.

Rectal temperature was measured to assess hypothermia by insertion of a mouse rectal probe (Braintree Laboratories Inc., Braintree, MA) 1.2 cm into the rectum using a thermometer (Physitemp Instruments Inc., Clifton, NJ).

Tail-Flick Test.

To assess the effects of drug treatments on acute antinociception, mice were gently restrained in a towel, and the tip of the tail (∼1 cm) was submerged in a hot water bath maintained at 54 to 55°C. The latency to withdraw the tail was measured.

Ring Test.

The ring test (Pertwee, 1972) was performed to assess possible cataleptic effects produced by drug administration. Briefly, each mouse was positioned on top of an elevated ring attached to a vertical stainless steel rod suspended at a height of 16 or 17 cm above a flat platform for a total duration of 5 minutes. The amount of time spent immobile was recorded as an index for catalepsy.

Behavioral Assessment of the Development of Tolerance to Δ⁹-THC

Male mice of the C57 background strain received once-daily injections of Δ⁹-THC (50 mg/kg i.p.) for 9 consecutive days. Mice were assessed in three of the behavioral assays from the tetrad test to quantify the development of tolerance to the behavioral effects of Δ⁹-THC. Motor coordination was measured using the rotarod test, body temperature was measured using a rectal temperature probe, and antinociception was measured using the tail-flick test. Responses were measured at baseline and 30 minutes after administration of Δ⁹-THC on days 1 and 8 of Δ⁹-THC administration. Mice received Δ⁹-THC in the same environment each day and remained for 4 hours post-Δ⁹-THC injection before being returned to the colony room.

Assessment of Antagonist-Precipitated Withdrawal Symptoms in Δ⁹-THC–Treated Mice

On day 9 of Δ⁹-THC administration, all mice were challenged with vehicle (i.p.), followed 30 minutes later by a second challenge (i.p.) with either rimonabant (10 mg/kg) or GW405833 (10 mg/kg). Mice were placed in clear Plexiglass chambers on an observation platform, and withdrawal behaviors were videotaped and scored by an investigator blinded to treatment conditions. On the day of withdrawal testing, mice were injected with Δ⁹-THC (50 mg/kg, i.p.) and allowed to acclimate to the chamber for 30 minutes. Next, all mice received a single injection of vehicle (i.p.), and behavior was videotaped for 30 minutes. Immediately after vehicle challenge, mice received a single i.p. injection of either rimonabant or GW405833, and behavior was videotaped for 30 minutes. The numbers of front paw tremors, headshakes, scratching, grooming, and rearing behaviors were counted. All mice received vehicle, followed by drug challenge so that within-subject comparisons could be made.

Statistical Analysis

Differences in mechanical paw withdrawal thresholds were evaluated using mixed analysis of variance (ANOVA) (with different pharmacologic treatments as the within-subject variable and genotype as between-subject variable), followed by Bonferroni post hoc tests. For tetrad tests, one-way ANOVA was performed separately on predrug (baseline) and postdrug responses, followed by Bonferroni post hoc tests. Paired or independent t tests were performed with the specific comparisons of interest as indicated. ANOVA for repeated measures was used to determine the time course of tolerance development to the effects of Δ⁹-THC, followed by paired-samples t tests to confirm the development of tolerance. Paired samples t tests were used to assess differences in withdrawal symptoms elicited by vehicle versus drug challenge. Statistical analyses were performed using IBM SPSS Statistics version 24 (IBM Corporation, Armonk, NY) or GraphPad Prism version 5.02 statistical software for windows (GraphPad Software, San Diego, CA). Figures were generated using GraphPad Prism version 5.02 statistical software. All data are presented as mean ± S.E.M.; P < 0.05 was considered significant.

Results

GW405833-Induced Suppressions of Neuropathic Pain in the PSNL Model Are CB₁ Receptor-Dependent and CB₂ Receptor-Independent

GW405833-Induced Antiallodynic Effects in PSNL Model Are Fully Preserved in CB₂ KO Mice and Are Not Blocked by a CB₂ Antagonist.

PSNL lowered mechanical paw withdrawal thresholds in CB₂KO and respective WT mice relative to baseline, consistent with the development of neuropathic pain (F_1,10 = 27.434, P < 0.001), but these effects did not differ between WT and CB₂ KO mice (F_1,10 = 0.003, P = 0.988) (Fig. 2A). Vehicle (0 mg/kg, i.p.) alone did not alter mechanical paw withdrawal thresholds in either group (P > 0.05) (Fig. 2A). ANOVA revealed a main effect of GW405833 treatment on mechanical allodynia (F_3,30 = 11.381, P < 0.001), and Bonferroni post hoc tests indicated that GW405833 elevated paw withdrawal thresholds in a dose-dependent manner (Fig. 2A). No main effects of genotype (F_1,10 = 3.620, P = 0.086) or significant interactions were observed (P = 0.598) (Fig. 2A). These results indicate that the antiallodynic effect of GW405833 was fully preserved in CB₂KO mice. GW405833 (10 and 30 mg/kg i.p.) fully restored mechanical paw withdrawal thresholds relative to baseline at doses as low as 10 mg/kg i.p. in both WT (t₅ = 0.270, P = 0.798) and CB₂KO (t₅ = 1.959, P = 0.107) mice (Fig. 2A). Paired t tests indicated that the CB₂ antagonist SR144528 (10 mg/kg i.p.) did not block the antiallodynic efficacy of GW405833 (30 mg/kg i.p.) in CB₂KO mice (t₅ = 1.790, P = 0.133), but it attenuated the efficacy of GW405833 in WT mice (t₅ = 2.833, P = 0.037). Although the antiallodynic efficacy of GW405833 (30 mg/kg i.p.) was attenuated in WT mice by SR144528, residual efficacy was nonetheless comparable to that observed in WT mice receiving GW405833 alone at a highly efficacious dose of 10 mg/kg i.p. (t₅ = 0.800, P = 0.460). Moreover, the antiallodynic efficacy of GW405833 in WT mice did not differ between doses of 10 mg/kg i.p. and 30 mg/kg i.p (P = 1.0). These observations suggest that, in WT mice, SR144528-induced blockade of the antiallodynic efficacy of GW405833 was minimal, if present at all.

Fig. 2.

GW405833 dose dependently reversed PSNL-induced mechanical allodynia in both WT and CB₂KO mice (A). The effect of GW405833 (30 mg/kg i.p.) was only slightly attenuated by pretreatment with the CB₂ antagonist SR144528 (10 mg/kg i.p.). Mechanical thresholds were not changed in the contralateral paw after PSNL or by drug treatments (B). WT (C57BL/6J): n = 6 males; CB₂KO: n = 6 males. ^#P < 0.05 vs. baseline; *P < 0.05 vs. vehicle (0 mg/kg i.p.); ^P < 0.05 vs. GW405833 (3 mg/kg i.p.); ^&P < 0.05 vs. GW405833 (30 mg/kg i.p.) in WT. BL, preinjury baseline. SR2, CB₂ antagonist SR144528; ipsilateral, injured side; contralateral, uninjured side.

GW405833-Induced Antiallodynic Effects in PSNL Model Are Absent in CB₁ KO Mice and Are Blocked by a CB₁ Antagonist.

PSNL lowered mechanical paw withdrawal thresholds in CB₁KO and respective WT mice relative to baseline, consistent with the development of robust mechanical allodynia (F_1,14 = 42.594, P < 0.001), but these effects did not differ between WT and CB₁ KO mice (F_{1, 14} = 1.079, P = 0.317) (Fig. 3A). Vehicle (0 mg/kg, i.p.) did not alter mechanical paw withdrawal thresholds after the neuropathic pain was fully established (P > 0.05). There was a main effect of GW405833 on mechanical allodynia (F_3,42 = 5.025, P = 0.005), a main effect of genotype (F_1,14 = 14.820, P = 0.002), and a trend was observed for the interaction that approached statistical significance (F_3,42 = 2.487, P = 0.074) (Fig. 3A). Bonferroni post hoc tests indicated that GW405833 dose dependently elevated mechanical paw withdrawal thresholds in WT mice, but these antiallodynic effects were absent in CB₁KO mice (Fig. 3A). The highest dose of GW405833 (30 mg/kg i.p.) did not fully restore mechanical paw withdrawal thresholds to baseline (preinjury) levels (t₇ = 3.33, P = 0.013). In WT mice, pretreatment with the CB₁ antagonist rimonabant (10 mg/kg i.p.) before GW405833 (30 mg/kg i.p.) completely blocked the antiallodynic effect of GW405833 (t₇ = 3.097, P = 0.017), and the resulting mechanical thresholds did not differ from those of CB₁KO mice receiving GW405833 alone (30 mg/kg i.p.) (t₁₄ = −0.582, P = 0.597) (Fig. 3A).

Fig. 3.

GW405833 dose dependently reversed PSNL-induced mechanical allodynia in WT but not in CB₁KO mice (A). The antiallodynic effect of GW405833 (30 mg/kg i.p.) was completely blocked by the CB₁ antagonist rimonabant (10 mg/kg i.p.) (A) Mechanical threshold was not changed in the contralateral paw after PSNL or by drug treatments (B). Mixed-sex groups were used; WT (CD1), n = 8; CB₁KO, n = 8. P < 0.05 vs. baseline; *P < 0.05 vs. vehicle (0 mg/kg i.p.); ^P < 0.05 vs. GW405833 (3 mg/kg i.p.); ⁺P < 0.05 vs. CB₁KO; ^&P < 0.05 vs. GW405833 (30 mg/kg i.p.) in WT. BL, preinjury baseline. Ipsilateral, injured side; contralateral, uninjured side.

Neither PSNL nor pharmacologic manipulations altered the mechanical thresholds in the contralateral (uninjured) paw in any genotype (P > 0.27) (Figs. 2B and 3B).

GW405833-Induced Antiallodynic Effects in the CFA Model of Inflammatory Pain are CB₁ Receptor–Dependent and CB₂ Receptor–Independent

GW405833-Induced Antiallodynic Effects in CFA Model Are Fully Preserved in CB₂ KO Mice and Are Not Blocked by a CB₂ Antagonist.

Intraplantar CFA injection reduced mechanical paw withdrawal thresholds in both WT and CB₂KO mice (F_1,14 = 222.594, P < 0.001) (Fig. 4A). There was a main effect of GW405833 treatment on mechanical allodynia (F_3,42 = 37.257, P < 0.001), and Bonferroni post hoc tests indicated that GW405833 suppressed CFA-induced allodynia in a dose-dependent manner (Fig. 4A). There was no main effect of genotype (F_{1, 14} = 0.146, P = 0.708) or interaction (F_3,42 = 1.156, P = 0.338). Efficacy of the 10 mg/kg and 30 mg/kg i.p. doses of GW405833 did not differ from each other (P = 0.236). Together, these data indicate that the antiallodynic efficacy of GW405833 was fully preserved in CB₂KO mice relative to WT mice. The highest dose of GW405833 (30 mg/kg i.p.) did not produce a full reversal of mechanical allodynia in WT relative to pre-CFA baseline (t₇ = 5.024, P = 0.002), but there was no statistically significant difference between pre-CFA baseline and GW405833 (30 mg/kg i.p.)-reversed mechanical withdrawal thresholds in CB₂KO (t₇ = 2.097, P = 0.074) mice. Moreover, there was no difference in responding between WT and CB₂KO mice (t₁₄ = 0.437, P = 0.165) (Fig. 4A). Paired-samples t tests showed that the CB₂ antagonist SR144528 did not block the antiallodynic efficacy of GW405833 in either WT (t₇ = 0.858, P = 0.419) or CB₂KO (t₇ = 0.760, P = 0.472) mice (Fig. 4A).

Fig. 4.

GW405833 dose dependently reversed CFA-induced mechanical allodynia in both WT and CB₂KO mice (A). The antiallodynic effect of GW405833 (30 mg/kg i.p.) was unaffected by the CB₂ antagonist SR144528 (10 mg/kg i.p.). Mechanical paw withdrawal thresholds in the contralateral paw were not changed by CFA injection or drug treatments (B). Mixed-sex groups were used: WT (C57BL/6J), n = 8; CB₂KO, n = 8. ^#P < 0.05 vs. baseline; *P < 0.05 vs. vehicle (0 mg/kg i.p.); ^P <0.05 vs. GW405833 (3 mg/kg i.p.) BL, preinjection baseline; SR2, CB₂ antagonist SR144528. Ipsilateral, injected paw; contralateral, uninjected paw.

GW405833-Induced Antiallodynic Effects in CFA Model Are Absent in CB₁ KO Mice and Blocked by a CB₁ Antagonist in WT Mice.

Intraplantar CFA injection lowered mechanical paw withdrawal thresholds in both WT and CB₁KO mice, suggesting that both genotypes developed inflammatory pain (F_{1, 12} = 145.506, P < 0.001) (Fig. 5A). Vehicle (i.p.) did not alter mechanical paw withdrawal thresholds after CFA-induced inflammation was established (P > 0.05). There was a main effect of GW405833 treatment on mechanical allodynia (F_{3, 36} = 26.618, P < 0.001), a main effect of genotype (F_1,12 = 8.348, P < 0.001), and the interaction was significant (F _{3, 36} = 8.212, P = 0.014) (Fig. 5A). Bonferroni post hoc tests revealed that GW405833 produced a dose-dependent elevation of mechanical thresholds in WT mice, but such antiallodynic effects were absent in CB₁KO mice (Fig. 5A). The highest dose of GW405833 (30 mg/kg i.p.) fully restored mechanical paw withdrawal thresholds relative to baseline (pre-CFA) levels (t₇ = 1.374, P = 0.212) in WT mice. Moreover, the CB₁ antagonist rimonabant completely blocked the antiallodynic efficacy of GW405833 in WT mice (t₇ = 5.58, P = 0.001) (Fig. 5A). Mechanical paw withdrawal thresholds observed in CB₁KO mice treated with GW405833 (30 mg/kg i.p.) did not differ from those observed in WT mice pretreated with rimonabant before GW405833 (30 mg/kg i.p.; t₁₂ = 1.150, P = 0.189) (Fig. 5A). Moreover, local administration of GW405833 (30 µg or 300 µg in 10 µl) in the CFA-injected paw of CB₁ KOs did not produce antiallodynic efficacy (data not shown).

Fig. 5.

GW405833 dose dependently reversed CFA-induced mechanical allodynia in WT but not in CB₁KO mice (A). The antiallodynic effect of GW405833 (30 mg/kg i.p.) in WT mice was completely blocked by the CB₁ antagonist rimonabant (10 mg/kg i.p.) (B) Mechanical withdrawal thresholds in the contralateral paw were not changed by CFA injection or treatments (B). WT (CD1), n = 8 males; CB₁KO, n = 6 males. ^#P < 0.05 vs. baseline; *P < 0.05 vs. vehicle (0 mg/kg i.p.); ^P < 0.05 vs. GW405833 (3 mg/kg i.p.); ^$P < 0.05 vs. GW405833 (10 mg/kg i.p.); ⁺P < 0.05 vs. CB₁KO; ^&P < 0.05 versus GW405833 (30 mg/kg i.p.) in WT. BL, preinjection baseline; ipsilateral, injected paw; contralateral, uninjected paw.

Neither CFA injection nor pharmacologic manipulations altered mechanical thresholds in the contralateral (noninflamed) paw in any genotype (P > 0.33) (Figs. 4B and 5B).

GW405833 Does Not Induce Typical Cannabimimetic Effects in the Tetrad Tests

Before pharmacologic manipulations, responding rates did not differ in the three groups in the ring test (F_{2, 15} = 0.344, P = 0.715), rotarod test (F_2,15 = 0.027, P = 0.973), rectal temperature assessment (F_2,15 = 0.443, P = 0.650), or tail-flick test (F_2,15 = 0.411, P = 0.670) (Fig. 6). Pharmacologic manipulations altered postdrug responding in the ring test (F_2,15 = 31.010, P < 0.001) (Fig. 6A). Bonferroni post hoc tests revealed that the positive control WIN55, 212-2 (5 mg/kg i.p.) increased immobility in animals compared with the vehicle or GW405833 (30 mg/kg i.p.) injection (P < 0.001) (Fig. 6A). Immobility time in the ring test in groups receiving GW405833 did not differ from vehicle (P = 1.0) or the preinjection responding (P = 0.156) (Fig. 6A). Therefore, GW405833 (30 mg/kg i.p.) failed to induce catalepsy.

Fig. 6.

GW405833 (30 mg/kg i.p.) did not induce catalepsy (A), motor ataxia (B), hypothermia (C), or acute tail-flick antinociception (D) in mice (n = 6 CD1 male mice per group). *P < 0.05; ***P < 0.001 vs. all the other postinjection groups (vehicle and WIN55, 212-2, 5 mg/kg i.p.).

Drug manipulations tended to alter rota-rod descend latencies (F_2,15 = 3.264, P = 0.067). Nonetheless, planned comparison independent t tests revealed that rotarod latencies were lower in groups receiving WIN55, 212-2 (5 mg/kg i.p.) compared with GW405833 (P = 0.046). By contrast, the effects of GW405833 (30 mg/kg i.p.) did not differ from vehicle (P = 0.302). Thus, GW405833 at a dose of 30 mg/kg i.p. did not produce motor ataxia in WT mice.

Pharmacologic manipulations altered body temperature (F_2,15 = 52.875, P < 0.001). Bonferroni post hoc tests revealed that the positive control WIN55 212-2 (5 mg/kg i.p.) reduced rectal temperature compared with body temperatures measured in mice receiving either vehicle or GW405833 (P < 0.001 for each comparison) (Fig. 6C). By contrast, postinjection body temperatures in mice receiving GW405833 did not differ from those observed in mice receiving vehicle (P = 0.555) and did not differ from preinjection (baseline) body temperatures (P = 0.119) (Fig. 6C). Therefore, GW405833 failed to induce hypothermia.

Pharmacologic manipulations altered postinjection tail-flick latencies (F_2,15 = 7.993, P = 0.004). Bonferroni post hoc tests indicated that the positive control WIN55, 212-2 (5 mg/kg i.p.) increased tail-flick latencies compared with either vehicle or GW405833 treatments (P < 0.05) (Fig. 6D). By contrast, postinjection tail-flick latencies in groups receiving GW405833 did not differ from vehicle (P = 0.772) or the preinjection latency (P = 0.691) (Fig. 6D). Therefore, GW405833 failed to induce acute tail-flick antinociception.

Mice Develop Tolerance to Motor Ataxic, Hypothermic, and Antinociceptive Properties of Δ⁹-THC

Before withdrawal manipulations, there were no differences between any of the groups in any of the dependent measures used to assess the development of tolerance (P > 0.1). Therefore, prewithdrawal responses of all mice were pooled for statistical analysis of tolerance development (Fig. 7).

Fig. 7.

Mice develop behavioral tolerance to Δ⁹-THC administration. Δ⁹-THC (50 mg/kg i.p.) produces motor impairment (A), hypothermia (B), and antinociception (C) on day 1 of repeated dosing relative to baseline and day 8 of Δ⁹-THC administration (A–C), indicating that mice had become tolerant to the pharmacologic effects of Δ⁹-THC. Data are expressed as mean ± S.E.M. (n = 12 C57BL/6J male mice). **P < 0.01; ***P < 0.0001 vs. baseline or day 8; ^#P < 0.05; ^###P < 0.0001 vs. baseline, paired-samples t test.

Mice treated chronically with Δ⁹-THC (50 mg/kg/d i.p. × 8 days) exhibited motor impairment (F_2,20= 108.8, P < 0.0001; Fig. 7A), and there was no main effect of group (P > 0.8) or interaction (P > 0.3). Δ⁹-THC produced motor impairment on day 1 of administration relative to baseline (t₁₁ = 8.899, P < 0.0001; Fig. 7A), but full tolerance developed by day 8 of drug administration (t₁₁ = 12.39, P < 0.0001; Fig. 7A), indicating that animals developed tolerance to the motor ataxic effects of Δ⁹-THC. Latencies to descend from the rotarod were also higher on day 8 relative to baseline (t₁₁ = 6.154, P < 0.0001; Fig. 7A), indicating that rotarod performance may have improved with additional training when acute motor impairment from Δ⁹-THC was no longer present.

Δ⁹-THC decreased rectal temperature (F_2,20 = 140.2, P < 0.0001; Fig. 7B), and there was no main effect of group (P > 0.3) or interaction (P > 0.3). Δ⁹-THC decreased rectal temperature on day 1 of administration relative to baseline (t₁₁ = 14.66, P < 0.0001; Fig. 7B) but reached ∼95% tolerance by day 8 of administration (t₁₁ = 10.62, P < 0.0001; Fig. 7B), indicating that the mice had become tolerant to the hypothermic effects of Δ⁹-THC. Δ⁹-THC also decreased rectal temperature on day 8 relative to baseline (t₁₁ = 2.367, P < 0.05; Fig. 7B).

In the hot water tail-flick test, Δ⁹-THC treatment (i.p.) produced antinociception in all mice (F_2,20 = 19.18, P < 0.0001; Fig. 7C), and there was no effect of group (P > 0.1) or interaction (P > 0.4). Δ⁹-THC increased the latency to withdraw the tail on day 1 relative to baseline (t₁₁ = 4.644, P < 0.001; Fig. 7C) and responding on day 8 (t₁₁ = 4.134, P < 0.01; Fig. 7C) of administration. Tail withdrawal latencies did not differ on day 8 of administration relative to baseline (P > 0.1), indicative of the development of tolerance to the antinociceptive effects of Δ⁹-THC.

Rimonabant Elicits Classic Signs of Cannabinoid Withdrawal, but GW405833 Does Not

In mice chronically treated with Δ⁹-THC (50 mg/kg/d × 9 days, i.p.), rimonabant challenge (10 mg, i.p.) increased paw tremors (t₅ = 3.95, P < 0.05; Fig. 8A), head shakes (t₅ = 2.716, P < 0.05; Fig. 8A), grooming behavior (t₅= 2.963, P < 0.05; Fig. 8A) and rearing behaviors (t₅= 3.943, P < 0.05, Fig. 8A) relative to vehicle (i.p.) challenge.

Fig. 8.

Rimonabant challenge elicits cannabinoid CB₁-dependent withdrawal behaviors, whereas GW405833 does not. Rimonabant (10 mg/kg i.p.) challenge increases the number of paw tremors, headshakes, grooming, and rearing behaviors in mice chronically treated with Δ⁹-THC (see tolerance development for these mice in Fig. 7). (A) GW405833 (10 mg/kg i.p.) decreases the levels of paw tremors in mice chronically treated with Δ⁹-THC. (B) Data are expressed as mean ± S.E.M. (n = 6 C57BL/6J male mice per group). *P < 0.05 vs. vehicle, paired-samples t test.

In mice chronically treated with Δ⁹-THC (50 mg/kg/day × 9 days, i.p.), GW405833 (10 mg/kg, i.p.) decreased paw tremors (t₅ = 3.955, P < 0.05; Fig. 8B) but failed to alter head shakes, scratching, grooming, or rearing behaviors (P > 0.05) relative to vehicle challenge (i.p.).

Discussion

GW405833 has been widely described in the literature as a cannabinoid CB₂ agonist, although the pharmacological specificity of its in vivo profile of actions has not been thoroughly evaluated. We characterized the pharmacologic specificity of the antiallodynic effects of GW405833 using both neuropathic (i.e., induced by PSNL) and inflammatory (i.e., induced by CFA) pain models in mice. Consistent with other studies (Valenzano et al., 2005; Whiteside et al., 2005), GW405833 dose dependently reversed mechanical hypersensitivity in both neuropathic and inflammatory pain models after systemic administration. GW405833 (3 mg/kg i.p.) did not produce antiallodynic efficacy, but doses of 10 and 30 mg/kg i.p. robustly reversed established allodynia. PSNL, CFA, and GW405833 treatment did not alter mechanical thresholds in the contralateral paw, suggesting that GW405833 suppressed established allodynia induced by neuropathic or inflammatory insults without altering basal nociceptive thresholds observed in the absence of injury.

The most striking observation of our study was that the antiallodynic efficacy of GW405833 was not meditated by CB₂ receptors. GW405833-induced antinociception was, in fact, fully preserved in CB₂ KO mice in models of established neuropathic and inflammatory pain. GW405833 reversed established allodynia with similar efficacy in WT and CB₂KO mice in both pain models. By contrast, antiallodynic efficacy of GW405833 was absent in CB₁ KO mice and blocked in WT mice by the CB₁ antagonist rimonabant. In neuropathic WT mice, the CB₂ antagonist SR144528 (10 mg/kg i.p.) produced only a modest attenuation of the antiallodynic efficacy of GW405833 (30 mg/kg i.p.); antiallodynic efficacy was suppressed to a level comparable to that of a fully efficacious lower dose (10 mg/kg i.p.). Moreover, the antiallodynic efficacy of GW405833 was not blocked at all by SR144528 in the CFA model. The antinociceptive effects of GW405833 are widely believed to be CB₂ receptor-mediated because of the high selectivity of GW405833 to CB₂ receptor over CB₁ receptor described in vitro (Valenzano et al., 2005), reports of blockade of GW405833 effects by a CB₂ antagonist (Clayton et al., 2002; Beltramo et al., 2006), and an absence of GW405833 antinociceptive efficacy in CB₂KO mice (Valenzano et al., 2005); however, our results suggest that GW405833 does not behave as a CB₂ agonist in mice after systemic administration.

Strikingly, the antiallodynic effects of GW405833 observed in the current study were CB₁ receptor-mediated. These observations are consistent with a previous report suggesting that the antinociceptive effects of GW405833 on lactic acid-induced nociceptive stretch responses were partially blocked by the CB₁ antagonist rimonabant but not by the CB₂ antagonist SR144528 (Kwilasz and Negus, 2013). In the few studies suggesting that the antinociceptive effects of GW405833 are CB₂-mediated (Clayton et al., 2002; Valenzano et al., 2005; Beltramo et al., 2006), neither CB₁ antagonists nor CB₁KOs were used to investigate the possible contribution of CB₁ receptors to the in vivo antinociceptive profile of GW405833. Only one prior study included CB₂KO mice (Valenzano et al., 2005). To our knowledge, our study is the first to include CB₁KO mice to study possible involvement of CB₁ receptors in the antiallodynic effects of GW405833. GW405833 has been reported to behave as a partial agonist of the orphan G protein–coupled receptor 55 (GPR55) in evaluating l-α-lysophosphatidylinostiol–induced mitogen-activated protein kinase activation in vitro (Anavi-Goffer et al., 2012). It is unclear, however, how GPR55 agonism would produce antinociception. The involvement of GPR55 in pain is controversial (Staton et al., 2008; Carey et al., 2017) and difficult to address owing to the lack of sufficiently selective agonists and antagonists for this receptor.

In the two studies in which the CB₂ antagonist SR144528 was used to block the antinociceptive effect of GW405833 (Clayton et al., 2002; Beltramo et al., 2006), different pain models (i.e., carrageenan and formalin inflammatory pain models), different methods (i.e., weight bearing, paw edema, paw licking, and flinching), and different species (i.e., rats were used by Clayton and colleagues) were used compared with our study; these differences may have contributed to the different pattern of results obtained. One possibly noteworthy difference in methods is that GW405833 was administered before the noxious insult in both prior studies, whereas in our study, GW405833 was administered after the establishment of both neuropathic pain and inflammatory pain. Thus, it is possible that GW405833 blocks pain through a CB₂-mediated mechanism during the development of pain but through a CB₁-mediated mechanism during the maintenance of pathologic pain. In the study by Valenzano et al. (2005), the antiallodynic effect of GW405833 (30 mg/kg i.p.) in the CFA model in CB₂KO mice is reported to not differ from vehicle; however, this latter study did not evaluate whether the efficacy of GW405833 differed statistically in WT and CB₂KO mice. Inadequate statistical power could account for failure to observe differences in responding between CB₂KO mice receiving vehicle and GW405833 at the single time point evaluated previously (Valenzano et al., 2005). By contrast, in our study, GW405833 exhibited similar antiallodynic effects in CB₂KO and WT mice at doses of 10 mg/kg i.p. and 30 mg/kg i.p. Both our study and that of Valenzano et al. (2005) used CB₂KO mice on the C57BL/6 background and the same concentration of CFA in the paw (20 µl, 50% CFA diluted in 0.9% saline). The effect of GW405833 (30 mg/kg i.p.) was studied 24 hours after CFA injection by Valenzano et al., whereas in our study, GW405833 was tested ≥48 hours after CFA injection. If GW405833 engages different antinociceptive mechanisms during different stages of the maintenance of inflammatory pain, then inclusion of different testing time points between studies could potentially contribute to the observed discrepancies.

Although GW405833 binds with higher affinity to the orthosteric site on CB₂ receptors relative to CB₁ receptors in vitro, the degree of selectivity for CB₂ over CB₁ is quite variable among studies, ranging from 37-fold up to ∼1200-fold selective to human CB₂ receptors over human CB₁ receptors (Gallant et al., 1996; Valenzano et al., 2005; Yao et al., 2008). Although it is possible that species differences exist between rat and mouse CB₂ receptors, they are much more highly conserved compared with human CB₂ receptors (Liu et al., 2009). Binding of GW405833 to native mouse CB₂ receptors, to our knowledge, has never been determined. Considering that GW405833 penetrates into the CNS at high levels (Valenzano et al., 2005; Evens et al., 2011), cannabimimetic CB₁-like effects could be expected in the tetrad tests if GW405833 binds directly to the orthosteric binding site of CB₁ receptors; however, the highest effective dose of GW405833 (30 mg/kg i.p.) identified in the present study was inactive in the tetrad tests. Specifically, GW405833 (30 mg/kg i.p.) did not produce catalepsy, motor ataxia, hypothermia, or acute antinociception in the tail-flick test in mice. These observations are in agreement with prior studies (Valenzano et al., 2005; Whiteside et al., 2005) suggesting that GW405833 did not produce cannabimimetic effects at doses lower than 100 mg/kg i.p. A recent study has shown that GW405833 can act as a noncompetitive CB₁ antagonist in vitro, because it noncompetitively antagonized CP55,940-induced adenylyl cyclase inhibition, extracellular signal-regulated kinase 1/2 phosphorylation, phosphatidylinositol 2 signaling, and CB₁ internalization in vitro (Dhopeshwarkar et al., 2017). In addition, GW405833 induced a complex, time-dependent activation of arrestin through CB₁ (Dhopeshwarkar et al., 2017).Therefore, to assess whether GW405833 behaves as an orthosteric CB₁ antagonist in vivo, we evaluated whether challenge with GW405833 would induce CB₁-dependent cannabinoid withdrawal in mice treated chronically with Δ⁹-THC (50 mg/kg, i.p. ×9 days). In contrast to challenge with the classic CB₁ antagonist/inverse agonist rimonabant (10 mg/kg, i.p.), challenge with GW405833 (10 mg/kg, i.p.) did not induce classic withdrawal responses in THC-tolerant mice. The lack of precipitated withdrawal by GW405833 observed in our withdrawal assay indicates that GW405833 does not behave as a CNS-penetrant competitive CB₁ antagonist in vivo. Thus, caution must be exerted before attributing the mechanism of action underlying in vivo effects of novel ligands to mechanisms identified in vitro.

In summary, the antiallodynic effect of GW405833 in the established neuropathic and inflammatory pain models examined here is CB₁-mediated and not CB₂-mediated. The antiallodynic effects of GW405833 were absent in CB₁KO mice and blocked by a CB₁ antagonist but were fully preserved in CB₂KO mice and largely unaffected by a CB₂ antagonist. Although GW405833 is brain penetrant (Valenzano et al., 2005) and reduced neuropathic and inflammatory pain through a CB₁-mediated mechanism in our study, it did not induce cannabimimetic CB₁-mediated side effects at the highest therapeutic dose evaluated (30 mg/kg i.p.). Moreover, GW405833 (10 mg/kg i.p.) did not function as a competitive CB₁ antagonist in our attempts to precipitate withdrawal with this agent in Δ⁹-THC tolerant mice. Together, our results indicate that GW405833 reversed established neuropathic and inflammatory pain through a CB₁-mediated mechanism without producing characteristic cannabimimetic effects associated with direct orthosteric binding to the CB₁ receptor. GW405833 could possibly act as a CB₁ allosteric modulator or interact with the endocannabinoid system. More research is needed to elucidate the mechanism of antinociceptive actions of GW405833.

Abbreviations

ANOVA

analysis of variance

CB₁ or CB₂

cannabinoid receptor 1 or 2

CFA

complete Freund’s adjuvant

CNS

central nervous system

KO

knockout

PSNL

partial sciatic nerve ligation

Δ⁹-THC

Δ⁹-tetrahydrocannabinol

WT

wild type

Authorship Contributions

Participated in research design: Hohmann.

Conducted experiments: Li, Carey.

Performed data analysis: Li, Carey.

Wrote or contributed to the writing of the manuscript: Li, Carey, Mackie, Hohmann.