The monoacylglycerol lipase inhibitor KML29 with gabapentin synergistically produces analgesia in mice

Abstract

Background and Purpose

Gabapentin is commonly prescribed for nerve pain but may also cause dizziness, sedation and gait disturbances. Similarly, inhibition of the endogenous cannabinoid enzyme monoacylglycerol lipase (MAGL) has antinociceptive and anti‐inflammatory properties but also induces sedation in mice at high doses. To limit these side effects, the present study investigated the analgesic effects of coadministering a MAGL inhibitor with gabapentin.

Experimental Approach

Mice subjected to the chronic constriction injury model of neuropathic pain were administered the MAGL inhibitor KML29 (1–40 mg·kg⁻¹, i.p.), gabapentin (1–50 mg·kg⁻¹, i.p.) or both compounds. Mice were tested for mechanical and cold allodynia. The function and expression of cannabinoid CB₁ receptors in whole brain homogenates and lipid profile of spinal cords were assessed after repeated drug administration.

Key Results

The combination of low‐dose KML29:gabapentin additively attenuated mechanical allodynia and synergistically reduced cold allodynia. The CB₁ antagonist, rimonabant, partially reversed the anti‐allodynic effects of KML29:gabapentin in mechanical allodynia but not cold allodynia. The anti‐allodynic effects of KML29:gabapentin did not undergo tolerance in mechanical allodynia after repeated administration but produced mild tolerance in cold allodynia. High dose KML29 alone reduced CB₁ receptor expression and function, but KML29:gabapentin reduced the density of CB₁ receptors but did not alter their function. KML29:gabapentin influenced additional signalling pathways (including fatty acids) other than the pathways activated by a higher dose of either drug alone.

Conclusion and Implications

These data support the strategy of combining MAGL inhibition with a commonly prescribed analgesic as a therapeutic approach for attenuating neuropathic pain.

Abbreviations

2‐AG

2‐arachidonylglycerol

2‐LG

2‐linoleoyl glycerol

anandamide

arachidonoylethanolamine

CB receptor

cannabinoid receptor

CCI

chronic constriction injury

CI

confidence interval

FAAH

fatty acid amide hydrolase

MAGL

monoacylglycerol lipase

NAEs

N‐acylethanolamines

NAGly

N‐arachidonoyl glycine

PEA

N‐palmitoyl ethanolamine

TRPV

transient receptor potential cation channel subfamily V

Z_add

predicted additive ED₅₀ values

Z_mix

experimentally derived ED₅₀ values

Introduction

Neuropathic pain is characterized by altered nerve function stemming from peripheral nerve injury, autoimmune and other disease states, or toxic insult. Treating neuropathic pain is difficult, because it is often refractory to traditional analgesics (Attal et al., 2006; Rahn and Hohmann, 2009). As compared with the well‐documented side effects and abuse potential of traditional analgesics, such as opioids, anticonvulsants have a relatively strong safety profile with minimal adverse drug interactions and are the recommended first‐line treatment for neuropathic pain (Kukkar et al., 2013). However, anticonvulsants are associated with negative side effects, including drowsiness, dizziness and ataxia (Beal et al., 2012; Stahl et al., 2013). Gabapentin is the most commonly prescribed anticonvulsant for neuropathy and, unlike opioids, repeated administration of gabapentin does not undergo tolerance (Hao et al., 2000; Gottrup et al., 2004), which reduces abuse potential. Gabapentin is a structural analogue of GABA but binds to α2δ subtype 1 of voltage‐gated calcium channels (Sills, 2006; Kukkar et al., 2013). The specific mechanisms through which gabapentin produces analgesia are not completely clear. However, gabapentin is thought to suppress central sensitization by blocking α2δ channels, which are densely expressed on presynaptic dorsal horn neurons (Tuchman et al., 2010; Stahl et al., 2013), thus reducing hyperexcitability of spinal nociceptive neurons.

Despite being routinely prescribed, gabapentin is effective in only half of chronic pain patients (Moore et al., 2014). Using an alternative approach targeting multiple systems with combination therapies could increase the efficacy of treating neuropathic pain (Chaparro et al., 1996; Gilron et al., 2005; Grim et al., 2014; Crowe et al., 2015). Combination therapy combines two or more drugs with different mechanisms of action, thereby increasing potential efficacy as compared with administration of either drug alone (Raffa, 2001; Perez et al., 2013). Using this approach offers multiple advantages, such as allowing for lower doses of each drug, thereby potentially reducing negative side effects of each drug, while maintaining pain relief (Raffa, 2001). For example, gabapentin has been characterized alone (Moore et al., 2014) and in combination with other analgesics, including opioids and antidepressants (Chaparro et al., 1996; Gilron et al., 2005), in neuropathic pain patients.

Preclinical studies have evaluated the effects of endogenous cannabinoid ligands (i.e. endocannabinoids) anandamide (arachidonoylethanolamine) and 2‐arachidonylglycerol (2‐AG) by inhibiting their respective catabolic enzymes, fatty acid amide hydrolase (FAAH) or monoacylglycerol lipase (MAGL), which have analgesic and anti‐inflammatory properties (Klein, 2005; Schlosburg et al., 2009; Kinsey et al., 2011; Crowe et al., 2015). Further, inhibiting FAAH or MAGL increases brain levels of anandamide or 2‐AG, respectively (Lichtman et al., 2004; Kinsey et al., 2009; Long et al., 2009a), thus increasing the bioavailability of the endocannabinoids to bind to cannabinoid CB₁ and CB₂ receptors. Specifically, inhibiting FAAH reduces neuropathic (Russo et al., 2007; Kinsey et al., 2009), inflammatory (Schlosburg et al., 2009; Booker et al., 2012) and visceral pain (Naidu et al., 2009). However, the FAAH inhibitor PF‐04457845 did not reduce pain in a clinical trial of osteoarthritis (Huggins et al., 2012), raising questions about the translation of preclinical to clinical application of this enzyme. Similarly, the MAGL inhibitors, JZL184 and KML29, attenuate pain in models of neuropathic pain (Kinsey et al., 2010; 2013; Ignatowska‐Jankowska et al., 2014) and inflammatory pain (Guindon et al., 2011; Ghosh et al., 2013). MAGL inhibitors attenuate allodynia (i.e. the painful perception of non‐noxious stimuli) and hyperalgesia (i.e. increased sensitivity to noxious stimuli) after nerve injury through a CB₁ receptor‐mediated mechanism of action (Kinsey et al., 2009; 2010; Ignatowska‐Jankowska et al., 2014).

Despite the benefits of MAGL inhibition, chronic administration of high‐dose JZL184 or KML29 induces physical dependence and cross‐tolerance to cannabinoid receptor agonists and down‐regulates and desensitizes CB₁ receptors (Ignatowska‐Jankowska et al., 2014; Schlosburg et al., 2014). To circumvent the effects of high doses and maximize analgesia, research has focused on using lower doses of endocannabinoid modulators in combination with other analgesics (Gunduz et al., 2011; Grim et al., 2014; Kazantzis et al., 2016). For example, the combined pharmacological inhibition of MAGL and COX1 and 2 synergistically attenuated mechanical allodynia and additively attenuated cold allodynia (Crowe et al., 2015), indicating that MAGL might be a good candidate for combination therapy. However, repeated administration of MAGL combined therapy has not been studied. In addition, commonly prescribed anticonvulsants have not been evaluated in conjunction with drugs affecting the endocannabinoid system. Therefore, the objective of the present study was to determine the anti‐allodynic effects of acute and chronic administration of the preclinical MAGL inhibitor, KML29, in combination with the clinically available anticonvulsant, gabapentin.

Methods

Animals

All animal care and experimental procedures were in accordance with ARRIVE guidelines (Kilkenny et al., 2010; McGrath and Lilley, 2015) and were approved by the Institutional Animal Care and Use Committee at West Virginia University prior to the start of any experiments. A total of 95 male C57BL/6 J mice (Jackson Laboratory, Bar Harbor, ME, USA) approximately 20 weeks old at the start of the experiment were used. Every effort was made to reduce animal suffering and to minimize the number of mice used whenever possible. Mice were housed in polysofone plastic NextGen cages (Allentown, Inc.) with corncob bedding. Mice were housed three to five per cage in a temperature (20–22°C) and humidity‐controlled environment with ad libitum access to food and water in an AAALAC‐accredited facility at West Virginia University. C57BL/6 J mice are widely studied and have been evaluated extensively in pain and behavioural assays (Crowe et al., 2015; Ignatowska‐Jankowska et al., 2015; Deng et al., 2015b). Mice were randomly assigned, and all experiments were carried out by trained technicians who were blinded to treatment conditions.

Chronic constriction injury (CCI)

Surgery was performed as described previously (Russo et al., 2007; Kinsey et al., 2009). Mice were anaesthetized with inhaled isoflurane (Phoenix Pharmaceuticals, Burlingame, CA, USA) with oxygen. Anesthesia was confirmed by toe pinch. The right hind leg was shaved and cleaned with three alternating wipes of Betadine solution, followed by 70% ethanol. An incision was made on the skin lateral to the femur. After separating the muscle, the sciatic nerve was isolated and partially ligated with a single, double‐knotted 5–0 silk suture. The muscle and skin were then closed with 6–0 nylon suture. Mice recovered in clean, heated cages and were observed for ataxia before being returned to the vivarium. Mice were administered ketoprofen (5 mg·kg⁻¹, i.p.) for 3 days, as a postoperative analgesic. The mice were tested repeatedly, starting 7 days post‐surgery, to confirm that allodynia developed. Testing with experimental drug treatments started 4 weeks post‐surgery.

Behaviour assessments

The mice were brought into the testing room, weighed, injected and subjected to behavioural testing. Mice were injected with gabapentin or vehicle 60 min before testing. Pilot data in mice subjected to CCI were administered gabapentin (50 mg·kg⁻¹, i.p.) either 60 or 120 min prior to allodynia testing, and no differences were found between pretreatment times in mechanical (P = 0.70; data not show) and cold allodynia (P = 0.12; data not shown). For experiments using KML29 alone and in combination with gabapentin, mice were injected with drug or vehicle 120 min before testing (Ignatowska‐Jankowska et al., 2014). The mice subjected to CCI were tested for allodynia, starting 7 days after CCI surgery, and placed inside ventilated polycarbonate chambers on an aluminium mesh table and allowed to acclimatize for 60 min before testing. In the CCI antagonist studies, rimonabant (3 mg·kg⁻¹, i.p.), SR144528 (3 mg·kg⁻¹, i.p.) or vehicle was given 15 min before administering the KML29:gabapentin combination (Crowe et al., 2015). A separate group of mice without CCI were subjected to the ‘Billy Martin tetrad battery’ to evaluate potential negative side effects of the drugs used in the present study, including non‐selective motor impairment.

A statistical power analysis was performed for sample size estimation of n = 14 (80% power, α 0.05, effect size 0.40) for acute allodynia testing, n = 8.5 (80% power, α 0.05, effect size 0.40) for repeated allodynia testing and n = 9 (80% power, α 0.05, effect size 0.60) for the tetrad battery. Final sample sizes are reported in the figure legends.

Mechanical allodynia test

Mechanical allodynia was tested using von Frey filaments (North Coast Medical, Morgan Hill, CA, USA) using the ‘up‐down’ method (Chaplan et al., 1994; Crowe et al., 2015). The plantar surface of either hind paw was stimulated with each filament, ranging from 0.16 to 6.0 g, starting with the 0.6 g filament, five times at a frequency of ~2 Hz (Kinsey et al., 2010; 2011; Grim et al., 2014; Crowe et al., 2015). The filaments were tested in ascending order until the mouse lifted its paw after three out of the five stimulations (i.e. a positive response). Once a positive response occurred, the filaments were tested in descending order until a positive response was no longer recorded, thus establishing a sensory threshold.

Acetone‐induced cold allodynia test

Immediately following the von Frey test (i.e. 30 min after starting von Frey test), 10 μL of acetone (99% HPLC grade; Thermo Fisher Scientific, Waltham, MA, USA) was applied via a 100 μL pipette (USA Scientific, Ocala, FL, USA) onto the plantar surface of each hind paw to test cold allodynia (Choi et al., 1994; Decosterd and Woolf, 2000). Acetone was applied from below the testing table via air burst by ‘expressing’ the pipette, thereby avoiding mechanical stimulation of the paw with the pipette tip. Cold allodynia was operationally defined as total time lifting or clutching the hind paw, which included paw lifting when walking or grooming. A maximum cut‐off time of 20 s was used (Decosterd and Woolf, 2000).

In order to assess the effects of repeated administration of the low dose combination of KML29:gabapentin, mice were administered KML29 (40 mg·kg⁻¹, i.p.; Ignatowska‐Jankowska et al., 2014), gabapentin (50 mg·kg⁻¹, i.p.; Kinsey et al., 2010), KML29:gabapentin combination (13.33:4 mg·kg⁻¹, i.p.) or 1:1:18 vehicle for 7 days, with each administration separated by approximately 24 h. The low combination dose was chosen due to its similar effect on both mechanical and cold allodynia. On the first and sixth day, mice were tested for mechanical and cold allodynia, as described above, 2 h after drug administration. Two hours after the final injection on the seventh day, mice were killed via CO₂ asphyxia, and brains and lumbar spinal cords were dissected, snap frozen in liquid nitrogen and stored at −80°C until assayed.

Tetrad

Testing was conducted 120 min following administration of KML29 (40 mg·kg⁻¹, i.p.), gabapentin (50 mg·kg⁻¹, i.p.), KML29:gabapentin (13.33:4 mg·kg⁻¹, i.p.) or vehicle in the following order: locomotor activity, bar test (catalepsy), tail immersion test and rectal temperature. Testing was performed according to previously described procedures (Martin et al., 1991; Long et al., 2009b; Schlosburg et al., 2010). Briefly, locomotor activity was video recorded for a 5 min period in a Plexiglas chamber placed within a lighted, sound‐attenuating chamber. ANYmaze (Stoelting, Wood Dale, IL, USA) software was used to determine time spent immobile. Catalepsy was assessed using a horizontal bar (0.75 cm diameter) placed 4.5 cm off the benchtop. The mouse was placed with its front paws on the bar and a timer was started. Time spent immobile on the bar was recorded, for up to 60 s maximum. If the mouse moved off the bar, it was placed back on in the original position up to three times. Nociception was then assessed in the tail immersion assay. The mouse was placed head first into a custom‐fabricated restraint from absorbent under pads (VWR Scientific Products, Radnor, PA, USA) with the tail protruding out. The distal 1 cm of the tail was submerged into a 56°C water bath. The latency for the mouse to withdraw its tail (within a 10 s cut‐off time) was scored. Rectal temperature was assessed by inserting a mouse thermocouple probe 2 cm into the rectum, and temperature was determined by telethermometer (Physitemp Bat‐12).

Homologous competition receptor binding

Increasing concentrations of the non‐radioactive CB₁/CB₂ full agonist CP‐55,940 (10⁻¹⁰ to 10⁻⁷ M) were incubated in triplicate with 0.5 nM of [³H]‐CP‐55,940 in a final volume of 1 mL of binding buffer (50 mM Tris, 0.05% BSA, 5 mM MgCl₂, pH 7.4) as described previously (Brents et al., 2011). Each binding assay contained 50 μg of crude whole mouse brain homogenates. To achieve equilibrium, reactions were incubated for 90 min at room temperature and terminated by rapid vacuum filtration through Whatman GF/B glass fibre filters, followed by three washes with ice‐cold binding buffer. Filters were immediately placed into scintillation vials with 4 mL of scintiverseTM BD cocktail scintillation fluid (Fisher Scientific, Fair Lawn, NJ, USA). Samples were incubated overnight in scintillation fluid, vortexed and bound reactivity determined by employing a liquid scintillation spectrophotometer (Tri Carb 2100 TR Liquid Scintillation Analyser, Packard Instrument Company, Meriden, CT, USA).

[³⁵S]‐GTPγS binding

[³⁵S]‐GTPγS assays were conducted as described previously (Brents et al., 2012) in buffer containing 20 mM HEPES, 100 mM NaCl, 10 mM MgCl₂, 0.05% BSA, 10 μM GDP and 20 U L⁻¹ of adenosine deaminase. Assays were performed in triplicate in a final volume of 1 mL, with all reactions containing 0.1 nM [³⁵S]‐GTPγS and increasing concentrations of the CB₁/CB₂ full agonist CP‐55,940 (10⁻⁹ to 3 × 10⁻⁷), and 10 μg of crude whole mouse brain homogenates. To achieve maximal G‐protein activation, reactions were incubated at 30°C for 30 min and terminated by rapid vacuum filtration through Whatman GF/B glass fibre filters and followed by four 1 mL washes with ice‐cold filtration buffer (20 mM HEPES, pH 7.4, 0.05% BSA). Filters were immediately placed into scintillation vials with 4 mL of scintiverseTM BD cocktail scintillation fluid (Fisher Scientific, Fair Lawn, NJ, USA). Samples were incubated overnight in scintillation fluid, vortexed and bound reactivity determined by employing a liquid scintillation spectrophotometer (Tri Carb 2100 TR Liquid Scintillation Analyser, Packard Instrument Company, Meriden, CT, USA).

Lipid extraction, HPLC/MS/MS analysis, lipid quantification

Lipid extraction was performed on frozen spinal cord tissue as previously described (Leishman et al., 2016a). In brief, frozen tissue underwent mechanical, methanolic extraction in the presence of deuterated standards, then partial purification on C18 columns (Agilent Technologies, Santa Clara, CA, USA). Eluent fractions were analysed using HPLC/MS/MS (Shimadzu autosampler and pumps, Columbia, MD, USA; API 3000 triple quadrupole, Applied Biosystems/MDS Sciex; Foster City, CA, USA). Quantification of lipids is through a series of standard curve analyses using Analyst software (Applied Biosystems/MDS Sciex; Foster City, CA, USA) and was previously described in greater detail (Leishman et al., 2016b).

To determine the magnitude change in one treatment group relative to another and, therefore, the number of arrows to assign each significant difference, the mean level of a particular lipid in a selected treatment group was divided by that same lipid's mean level in the comparison group (Table 3). For example, the average level of 2‐AG in the spinal cord of KML29‐treated mice was 6.39 × 10⁻⁹ mol·g⁻¹, and the average level of 2‐AG in the spinal cord of the vehicle‐treated mice was 1.90 × 10⁻⁹ mol·g⁻¹; 6.39 × 10⁻⁹ divided by 1.90 × 10⁻⁹ equals 3.36, meaning that 2‐AG levels are over three times as high in the spinal cord of the KML29‐treated mice and assigning it four up arrows in the 2‐AG cell for the ‘Change with KML relative to vehicle’ column because the magnitude of change was between 3 and 10 times higher. For decreases, the process was very similar: the mean level in spinal cord of one group was divided by the same lipid's mean level in the spinal cord of the comparison group; however, the reciprocal of the decimal was taken to express a fold decrease (if the level in the one group is half of the level of another then that is a twofold decrease). As an example, the mean level of PGE₂ was 4.90 × 10⁻¹¹ mol·g⁻¹ in the spinal cord of the KML29‐treated mice and 8.18 × 10⁻¹¹ mol·g⁻¹ in the corresponding gabapentin‐treated spinal cord. 4.90 × 10⁻¹¹ divided by 8.18 × 10⁻¹¹ is 0.599 and the reciprocal of 0.599 is 1.67, meaning that the decrease is between 1.5 and 2 fold and giving it two down arrows on our scale in PGE₂ cell of the ‘Change with KML relative to gabapentin’ column.

Data analysis

The data and statistical analysis comply with the recommendations on experimental design and analysis in pharmacology (Curtis et al., 2015). All results are presented as means ± SEM. Results were considered significant at P < 0.05. Ipsilateral paw sensitivity data were analysed using a t‐test compared to the contralateral paw. Dose–response data were analysed using one‐way ANOVA, followed by Dunnett's post hoc test. The antagonist studies were analysed by two‐way (combo vs. antagonist) between‐subjects ANOVA followed by Bonferroni post hoc tests. The repeated administration study was analysed by a two‐way (acute vs. chronic) ANOVA followed by paired t‐tests. No animals were excluded from analyses.

For the mechanical allodynia assay, raw paw threshold was expressed as % maximum possible effect (%MPE) using the equation %MPE = (test threshold / max threshold)*100, where ‘max threshold’ was the assay's maximum filament (i.e. 6 g), and ‘test threshold’ was the paw's established filament threshold. For the cold allodynia assay, raw seconds the paw was lifted was expressed as %MPE using the equation %MPE = [(max cut‐off − test time)/(Max Cut‐off)]*100, where ‘max cut‐off’ was the assay's maximum cut‐off point (i.e. 20 s), ‘test time’ was the time (s) the paw was lifted off the testing table. The ED₅₀ values were calculated by interpolation when only two data points were available (one below and one above 50% MPE) or by standard linear regression analysis when at least three data points were available on the linear portion of the dose‐effect curve (Crowe et al., 2015). To determine synergistic, additive or subadditive interactions, the theoretical additive ED₅₀ value of the combined drugs was calculated from the individual dose–response curves. The combination is assumed to equal the sum of the effects of each drug.

Dose‐addition analysis was carried out as previously published (Tallarida, 2006; Naidu et al., 2009; Crowe et al., 2015). For graphical display, the ED₅₀ of gabapentin was plotted on the abscissa (x axis), and the isoeffective dose of KML29 was plotted on the ordinate (y axis) to generate the isobologram. A line connecting the two points represents the theoretical additive effect of KML29 and gabapentin dose combinations. The drug mixture ED₅₀ value was determined by linear regression as the overall mixture dose (KML29 and gabapentin doses were summed). The experimentally derived ED₅₀ values (Z_mix) from the dose–response curves of the ratios were compared to the predicted additive ED₅₀ values (Z_add). Z_add values were calculated using the equation: Zadd = fA + (1 − f)B, where Α was the KML29 alone ED₅₀ value, Β was the gabapentin alone ED₅₀ value and ƒ was a fractional multiplier of Α in the computation of the additive total dose. The experiments described in this manuscript tested mixtures that yielded values of ƒ=0.25, ƒ=0.5 and ƒ=0.75, where ƒ is related to the proportion of KML29 in a mixture per the equation ρA=ƒ/Zadd. If the ED₅₀ values of the Z_mix are below those of Z_add and the confidence limit (CI) does not overlap, then the interaction is considered synergistic. The statistical difference between the theoretical additivity ED₅₀ value (Z_add) and the experimental ED₅₀ value (Z_mix) was analysed using a Fisher's exact test.

For receptor binding assays, curve‐fitting and statistical analyses were conducted utilizing nonlinear regression; the one site homologous competition binding equation was used to determine the affinity (K_D) of [³H]‐CP‐55,940 and CB₁ receptor density (B_MAX) expression in crude whole mouse brain homogenates. Curve fitting of concentration–effect curves via nonlinear regression was also employed to determine the EC₅₀ (a measure of potency, in nM) and E_MAX (a measure of efficacy, in pmol·mg⁻¹) for the [³⁵S]‐GTPγS binding experiments. Measures of affinity (K_D) and potency (EC₅₀) were converted to pK_D or pEC₅₀ values by taking the negative log of each value so that parametric tests could be used for statistical comparisons. Data were analysed for statistical differences by a one‐way ANOVA, followed by Tukey's post hoc comparisons. For the HPLC/MS/MS data, concentrations of each detected lipid in mol·g⁻¹ adjusted for % recovery were analysed using one‐way ANOVA followed by post hoc Fisher's Least Significant Differences Test to determine significant differences between treatment groups. All statistical tests for HPLC/MS/MS data were carried out using SPSS Statistics (IBM, Armonk, NY, USA).

Materials

Gabapentin was purchased from Sigma‐Aldrich (St. Louis, MO, USA). KML29, rimonabant (SR141716; CB₁ receptor antagonist) and SR144528 (CB₂ receptor antagonist) were purchased from Cayman Chemical (Ann Arbor, MI, USA). All compounds were dissolved in a vehicle consisting of ethanol, Cremophor (Sigma‐Aldrich, St. Louis, MO, USA) and normal saline in a ratio of 1:1:18 parts (Pinto et al., 2010). All solutions were warmed to room temperature and injected i.p. at a volume of 10 μL·g⁻¹ body mass. While both KML29 and gabapentin are orally bioavailable, i.p. and p.o. routes of administration of KML29 have comparable levels of MAGL inhibition (Chang et al., 2012) and gabapentin pharmacokinetics are not influenced by route of administration (Gambelunghe, Mariucci, Tantucci, & Ambrosini, 2005). CP‐55,940 was purchased from Tocris Bioscience (Bristol, UK) and prepared as a stock solution in 100% DMSO at a concentration of 10 mM, divided into aliquots and maintained at −4°C until use. [³H]‐CP‐55950 (168 Ci mmol⁻¹) and [³⁵S]‐GTPγS (1250 Ci mmol⁻¹) were purchased from Perkin Elmer (Boston, MA, USA). All other reagents for binding assays were purchased from Fisher Scientific Inc (Pittsburgh, PA, USA).

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY (Southan et al., 2016), and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 (Alexander et al., 2015a,b,c).

Results

CCI induces mechanical and cold allodynia

Following the CCI surgery, mice were tested repeatedly, in the absence of any drug administration, to establish allodynia development before treatment testing. The mice developed mechanical [t(29) = 9.47, P < 0.05] and cold [t(29) = −9.50, P < 0.05] allodynia in the ipsilateral paw [mean(SEM) = 1.23(0.22)g and 13.27(0.79)s], as compared to the contralateral paw [mean(SEM) = 4.74(0.28)g and 3.98(0.53)s].

Coadministration of KML29 and gabapentin additively attenuated mechanical allodynia and synergistically reduced cold allodynia

Administration of KML29 has been previously shown to significantly attenuate mechanical and cold allodynia at ≥30 mg·kg⁻¹ (Crowe et al., 2015). Gabapentin significantly reduced mechanical and cold allodynia at ≥10 mg·kg⁻¹ (data not shown). To determine the overall ED₅₀ for KML29, the ED₅₀s for mechanical and cold allodynia, 16.62 and 27.26 mg·kg⁻¹, respectively, were averaged, resulting in an overall ED₅₀ of 22 mg·kg⁻¹ (Crowe et al., 2015). For the overall ED₅₀ of gabapentin, the ED₅₀s for mechanical and cold allodynia, 6.68 and 7.65 mg·kg⁻¹, respectively, were averaged, resulting in an overall ED₅₀ of 7 mg·kg⁻¹. Using the overall ED₅₀s for KML29 and gabapentin, the 1:1 ratio reflects 1 part KML29 to 0.3 parts gabapentin, the 3:1 ratio reflects 1 part KML29 to 0.1 parts gabapentin and the 1:3 ratio reflects 1 part KML29 and 0.9 parts gabapentin (Figure 1).

Figure 1.

Coadministration of the selective MAGL inhibitor KML29 and gabapentin additively attenuated mechanical allodynia and synergistically reduced cold allodynia in mice subjected to CCI. A separate group of mice was subjected to CCI and then administered KML29, gabapentin or vehicle and tested for mechanical and cold allodynia. Then, KML29 was coadministered with gabapentin and mechanical (A) and cold allodynia (E) were assessed. The isobole of additivity is shown as a solid line connecting the ED₅₀ values of KML29 and gabapentin and depicting the theoretical line of additivity. The experimental points of the collective mechanical allodynia tests (B–D) do not differ significantly from line of additivity, indicating an additive interaction. The experimental points of the collective cold allodynia tests (F–H) significantly lie below the theoretical line of additivity, indicating a synergistic interaction. Data are expressed as mean (±) SEM. *P < 0.05 versus vehicle (n = 15).

The Z_mix in the 1:1, 3:1 and 1:3 ratios in the mechanical allodynia test was lower than the Z_add with some CI overlap, indicating that the interaction was additive (Table 1). The Z_mix in the 1:1 ratio in the cold allodynia test was less than the Z_add; however, there was some CI overlap, and thus, the interaction was considered additive (Table 1). The Z_mix in the 3:1 and 1:3 ratios in the cold allodynia test was less than the Z_add and did not have any CI overlap, indicating the interactions were synergistic (Table 1).

Table 1.

ED₅₀ values of KML29 and gabapentin (GBP) in combination

Combination (doses in mg·kg−1)	Combination Ratio		ED50 mg·kg−1 (95% confidence interval) KML29:GBP
			Zadd (theoretical)	Zmix (experimental)
KML29:GBP	1:1	Mechanical Allodynia	11.80 (8.68–14.93)	11.94 (7.35–19.36)
0.5:0.15		Cold Allodynia	16.62 (12.97–20.27)	13.65 (9.02–20.65)
1:0.3
4.44:1.3
13.33:4
30:9
40:12
KML29:GBP	3:1	Mechanical Allodynia	14.63 (10.36–18.90)	7.38 (4.83–11.27)
0.5:0.05		Cold Allodynia	21.13 (16.04–26.23)	10.06 (6.52–15.52)*
1:0.1
4.44:0.4
13.33:1.3
30:3
40:4
KML29:GBP	1:3	Mechanical Allodynia	8.98 (6.49–11.46)	4.27 (2.76–6.60)
0.5:0.45		Cold Allodynia	12.11 (9.40–14.82)	5.28 (3.43–8.11)*
1:0.9
4.44:4
13.33:12
30:27
40:36

The Z_add and Z_mix values reflect the total amount of both drugs combined, where KML29 and gabapentin were summed for each combination.

Experimentally determined Z_mix values and predicted Z_add values (95% CI) for mixtures of KML29 and gabapentin in assays of acetone and von Frey. Asterisks indicate a synergistic interaction as evidence of non‐overlapping 95% CI between Z_mix and Z_add values.

^* P < 0.05 compared with Z_add using the Fisher test.

Because the experimental points of the collective mechanical allodynia tests do not differ significantly from the theoretical line of additivity, the interaction was additive. The experimental points of the collective cold allodynia of the 1:1 ratio did not differ significantly from the theoretical line of additivity, and thus, the interaction was additive (Figure 1A–D). Because the experimental points of the cold allodynia in the 3:1 and 1:3 ratios were significantly different from the line of additivity, the interaction was synergistic (Figure 1E–H).

Anti‐allodynia effects were partially blocked by a CB₁ receptor antagonist

To determine the relative contribution of either cannabinoid receptor to the observed anti‐allodynia, the CB₁ antagonist, rimonabant (3 mg·kg⁻¹; Crowe et al., 2015), or the CB₂ antagonist, SR144528 (3 mg·kg⁻¹; Crowe et al., 2015), was administered prior to the combination of KML29 (13.33 mg·kg⁻¹) and gabapentin (4 mg·kg⁻¹) to assess receptor mechanism. There was an overall main effect of treatment in mechanical allodynia [F(3,52) = 5.15, P < 0.05; Figure 2A] and cold allodynia [F(3,52) = 9.49, P < 0.05; Figure 2B]. In mechanical allodynia, post hoc analyses did not indicate a difference between KML29:gabapentin and rimonabant or SR144528. Similarly, in cold allodynia, post hoc analyses did not indicate a difference between KML29:gabapentin and rimonabant or SR144528.

Figure 2.

Anti‐allodynic effects of KML29:gabapentin combination treatment are partially mediated by CB₁ receptors. Mice subjected to CCI were administered a combination of KML29:gabapentin (13.33:4 mg·kg⁻¹). Pretreatment with rimonabant (SR1; 3 mg·kg⁻¹), but not SR144528 (SR2, 3 mg·kg⁻¹), partially reduces anti‐allodynia in the von Frey mechanical allodynia test (A), but neither had any effect in the acetone‐induced cold allodynia test (B), (n = 14). KML29 (40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹) or KML29:gabapentin (13.33:4 mg·kg⁻¹) did not elicit any behavioural changes in spontaneous locomotor activity (C), catalepsy (D), tail immersion (E) or body temperature (F), (n = 8). Data are expressed as mean (±) SEM. *P < 0.05 versus vehicle.

Neither drug nor the combination elicited classic cannabimimetic effects

KML29 (40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹) and KML29:gabapentin (13.33:4 mg·kg⁻¹) were assessed for classic cannabinoid effects using the ‘Billy Martin tetrad battery’. Neither drug, alone or in combination altered spontaneous locomotor activity [F(3,28) = 0.13, P = 0.94; Figure 2C], bar test catalepsy [F(3,28) = 1.11, P = 0.36; Figure 2D], tail immersion analgesia [P = 0.94; Figure 2E] or core body temperature [F(3,28) = 1.04, P = 0.39 Figure 2F].

No tolerance to repeated KML29 and gabapentin combination in mechanical allodynia

Mice were treated repeatedly with KML29 (40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹), KML29:gabapentin (13.33:4 mg·kg⁻¹) or vehicle for 7 days and were assessed for mechanical and cold allodynia on the first (i.e. acute effect) and sixth day (i.e. chronic effect). On Day 1 of the experiment, KML29, gabapentin and KML29:gabapentin attenuated mechanical allodynia [F(3,27) = 21.88, P < 0.05; Figure 3A], which replicated the acute results reported above. On Day 6, gabapentin differed from vehicle (P < 0.05), but KML29 (P = 0.29) and the KML29:gabapentin combination (P = 0.02) did not differ from chronic vehicle. There was an interaction between the time of treatment (Day 1 or Day 6) and drug group [F(3,54) = 8.35, P < 0.05; Figure 3A]. Post hoc analyses revealed that acute and chronic administration of KML29 (40 mg·kg⁻¹) differed (P < 0.05), indicating that repeated administration of high dose KML29 leads to tolerance. Acute and repeated gabapentin differed from each other (P < 0.05), whereas the acute and chronic combination of KML29:gabapentin group did not differ (P = 0.28), indicating that repeated administration of the low dose combination does not lead to tolerance.

Figure 3.

The anti‐allodynic effects of KML29:gabapentin do not undergo tolerance in mechanical allodynia (A) after chronic administration but show partial tolerance in cold allodynia (B). Mice were administered KML29 (40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹) or KML29:gabapentin (combo; 13.33:4 mg·kg⁻¹) daily for 6 days. Data are expressed as mean (±) SEM. *P < 0.05 versus vehicle, #P < 0.05 versus acute treatment. For vehicle, KML29 and KML29:gabapentin groups n = 8 and for the gabapentin group n = 7. The difference in experimental numbers reflects an odd number of animals evenly distributed in the experimental design.

Similarly, KML29, gabapentin and KML29:gabapentin attenuated cold allodynia [F(3,27) = 13.16, P < 0.05; Figure 3B] on Day 1. On Day 6, gabapentin differed from vehicle (P < 0.05), but KML29 (P = 0.60) and the combination (P = 0.04) were not statistically significantly different from chronic vehicle. There was an interaction between the time of treatment (Day 1 or Day 6) and drug group [F(3,54) = 7.60, P < 0.05; Figure 3B]. Acute and chronic administration of KML29 (40 mg·kg⁻¹) differed (P < 0.05), indicating that repeated administration of high dose KML29 leads to tolerance. Acute and chronic gabapentin did not differ, whereas the acute and chronic combination of KML29:gabapentin differed (P < 0.05), indicating partial tolerance developed in the low combination group.

To assess whether there was a change in efficacy of the mixtures to produce anti‐allodynia that went along with the potency shifts allodynia, the maximum effective doses of each drug alone and each of the three mixtures was compared by between‐subjects ANOVA. There was no significant effect of treatment condition on the maximal effect on either mechanical allodynia (P = 0.45) or cold allodynia (P = 0.20).

Chronic KML29:gabapentin reduces CB₁ receptor density but not function in whole brain homogenates

The CB₁/CB₂ agonist CP‐55,940 produced concentration‐dependent and complete displacement of radiolabeled [³H]‐CP‐55,940 (i.e. homologous competition) from CB₁ receptors expressed in brain homogenates and was best fit by a one‐site model. In mice injected daily with vehicle, [³H]‐CP‐55,940 bound to CB₁ receptors with an affinity (K_D) of 1.69 nM and a receptor density (B_MAX) of 5.294 pmol·mg⁻¹ protein. Repeated gabapentin administration failed to alter the B_MAX of CB₁ receptors, while repeated treatment with the MAGL inhibitor KML29 significantly decreased the density of CB₁ receptors in brain homogenates by 32% [F(3, 27) = 5.69, P < 0.05; Figure 4; Table 2a]. Finally, in mice receiving KML29:gabapentin, the B_MAX of CB₁ receptors was reduced to levels comparable to that produced by KML29 alone (29%) when compared to vehicle‐treated controls. Chronic drug treatment produced slight but significant effects on the affinity (K_D) of CP‐55,940 for CB₁ receptors [F(3, 27) = 16.85, P < 0.05; Figure 4; Table 2a]. For example, KML29 treatment alone and in combination with gabapentin significantly reduced CP‐55,940 affinity from 1.69 nM in control animals, to 2.82 and 2.85 nM in KML29 and KML29:gabapentin‐treated mice respectively.

Figure 4.

Homologous [³H]‐CP‐55,940 receptor binding and CP‐55,940 stimulated G‐protein activation in mouse whole brain homogenates. (A) Specific binding was determined as described in the Methods by incubating 0.5 nM [³H]‐CP‐55950 with increasing concentrations (10^‐10to 10⁻⁷M) of CP‐55,940 and 50 μg of whole mouse brain membranes prepared from mice chronically treated with vehicle (black circles), gabapentin (blue circles), KML29 (red triangles) or a combination of gabapentin and KML29 (green squares). Receptor affinity (K_D in nM) and density (B_MAX in pmol·mg⁻¹ protein) was determined by nonlinear regression analysis of specific [³H]‐CP‐55950 binding, and values are presented in Table 1. (B) Whole brain homogenates (10 μg) prepared from mice chronically treated as labelled were incubated in the presence of 0.1 nM [³⁵S]‐GTPγS with increasing concentrations (10⁻⁹ to 3 × 10⁻⁷) of the full CB₁ receptor agonist CP‐55,940 as described in the Methods. The potency (EC₅₀ in nM) and efficacy (E_MAX in pmol·mg⁻¹ protein) of G‐protein activation was determined by nonlinear regression analysis, and values are presented in Table 2a. (n = 7–8).

Table 2a.

Homologous [³H]‐CP‐55,940 receptor binding in mouse whole brain homogenates

Group	[3H]‐CP‐55,940 binding
	KD (nM)	pKD (−Log[KD])	BMAX (pmol·mg−1)	n
Vehicle	1.69	8.88 ± 0.04a,b	5.29 ± 0.42a	8
GBP	1.75	8.76 ± 0.03a,b	4.81 ± 0.23a,b	7
KML29:GBP	2.85	8.55 ± 0.03a,b	3.76 ± 0.37a,b	8
KML29	2.82	8.56 ± 0.03a,b	3.60 ± 0.31b	8

^a,b pK_D and B_MAX values not sharing a letter are significantly different from values within the same column; That is, differences were observed with B_MAX, but no differences were observed with pKD. P < 0.05, one‐way ANOVA, Tukey's post hoc test.

Table 2b.

CP‐55,940 stimulated [³⁵S]‐GTPγS binding in mouse whole brain homogenates

Treatment	[35S]‐GTPγS binding
Group	EC50 (nM)	pEC50 (−Log[EC50])	EMAX (pmol·mg−1)	n
Vehicle	9.6	8.03 ± 0.04a,b	0.61 ± 0.06a	8
GBP	9.7	8.03 ± 0.05a,b	0.58 ± 0.04a,b	7
KML29:GBP	10.4	8.01 ± 0.06a,b	0.49 ± 0.05a,b	8
KML29	17.6	7.77 ± 0.05a,b	0.41 ± 0.03b	8

^a,b pEC₅₀ and E_MAX values not sharing a letter are significantly different from values within the same column; That is, differences were observed with E_MAX, but no differences were observed with pEC50. P < 0.05, one‐way ANOVA, Tukey's post hoc test.

The CB₁ cannabinoid receptor is a GPCR that produces intracellular effects via interaction with the G_i/G_o‐subtype of G‐proteins (Dalton et al., 2009). Upon binding to CB₁ receptors, agonists produce activation of G‐proteins that can be quantified in membrane preparations by measuring increases in agonist‐induced binding of [³⁵S]‐GTPγS, a nonhydrolyzable GTP analogue (Harrison and Traynor, 2003). Therefore, to measure the function of CB₁ receptors expressed in brain membranes of treated mice, the ability of increasing concentrations of CP‐55,940 to increase [³⁵S]‐GTPγS binding was examined. In vehicle‐treated mice, CP‐55,940 produced a concentration‐dependent increase in [³⁵S]‐GTPγS binding in brain membranes with a potency (EC₅₀) of 9.6 nM and efficacy (E_MAX) of 0.610 pmol·mg⁻¹ protein. Although repeated treatment with gabapentin alone did not change the efficacy of CP‐55,940‐mediated G‐protein activation, daily injections of KML29 significantly decreased the E_MAX of CP‐55,940 in mouse brain homogenates by 34% [F(3, 27) = 3.83, P < 0.05; Figure 4B]. Importantly, unlike that observed for effects on CB₁ receptor density, the combination of gabapentin and KML29 produced a slight (19%), but non‐significant, reduction in the efficacy of CP‐55,940 to activate G‐proteins in brain homogenates.

Similar to effects on CP‐55,940 affinity (K_D) repeated KML29 also had a slight but significant [F(3, 27) = 6.25, P < 0.05; Table 3] effect on the potency (EC₅₀) of CP‐55,940 to activate G‐proteins. Specifically, KML29 treatment alone significantly reduced the potency of CP‐55,940 from 9.6 nM in controls to 17.6 nM in treated mice.

Table 3.

Eicosanoids in lumbar spinal cords

	Change with GBP relative to vehicle	Change with KML29 relative to vehicle	Change with combo relative to vehicle	Change with KML29 relative to GBP	Change with combo relative to GBP	Change with combo relative to KML
N‐acyl ethanolamine
N‐palmitoyl ethanolamine			↑↑
N‐stearoyl ethanolamine
N‐oleoyl ethanolamine
N‐linoleoyl ethanolamine			↑↑↑
N‐arachidonoyl ethanolamine			↑↑			↑
N‐docosahexaenoyl ethanolamine			↑
N‐acyl glycine
N‐palmitoyl glycine
N‐stearoyl glycine
N‐oleoyl glycine
N‐arachidonoyl glycine	↑↑↑↑
N‐acyl taurine
N‐arachidonoyl taurine			↑
Free Fatty Acids
Oleic acid
Linoleic acid			↑↑		↑↑	↑
Arachidonic acid
2‐acyl‐sn‐glycerol
2‐palmitoyl‐sn‐glycerol			↑↑
2‐oleoyl‐sn‐glycerol			↑↑		↑
2‐linoleoyl‐sn‐glycerol		↑↑↑	↑↑↑	↑↑	↑↑↑
2‐arachidonoyl‐sn‐glycerol		↑↑↑↑	↑↑↑↑	↑↑↑↑	↑↑↑↑	↑
PGs
PGE2	↑	↓		↓↓	↓	↑

Summary of lumbar spinal cord levels of eicosanoids from mice treated repeatedly with KML29 (KML; 40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹) or KML29:gabapentin (combo; 13.33:4 mg·kg⁻¹). Tissue was collected from mice subjected to CCI 2 h after drug administration. Arrows indicate fold increase or decrease ↑↑↑↑ (≥10), ↑↑↑↑ (3–9.99), ↑↑↑ (2–2.99), ↑↑ (1.5–1.99), ↑ (1–1.49), ↓ (1–1.49), ↓↓ (1.5–1.99).

The increase in [³⁵S]‐GTPγS binding produced by CP‐55,940 in whole mouse brain homogenates was due to activation of CB₁ receptors, because G‐protein activation is blocked by co‐incubation with the CB₁ antagonist rimonabant (data not shown).

Chronic KML29:gabapentin increases 2‐AG, anandamide and NAEs in the spinal cord

Nineteen of the 26 lipids screened were present in all samples. The remaining seven were detectable in only some of the samples; therefore, those were not statistically analysed here. A series of comparisons of each of the 19 lipids in each of the four treatment groups was performed so that each of the three drug treatments was compared directly to the vehicle. We then compared each of the individual drug treatments to the combination treatments, and then, we compared the individual drug treatments. A full list of all lipids measured, the mean values for each and statistical comparisons for all interactions are available in supplemental data (Supporting Information Table S1). A composite of these series of comparisons is provided in Table 3.

Compared to vehicle controls, 40 mg·kg⁻¹ of KML29 for 7 days caused a predictable increase in 2‐AG [F(3,27) = 70.58; P < 0.05] and 2‐linoleoyl glycerol (2‐LG) [F(3,27) = 8.39; P < 0.05]. In comparison to KML29 treatment, 50 mg·kg⁻¹ gabapentin for 7 days showed an identical profile to high dose KML29, except gabapentin decreased PGE₂ [F(3,27) = 8.19; P < 0.05;Figure 5D].

Figure 5.

Lumbar spinal cord levels of eicosanoids expressed as mean (±SEM) in pmol·g⁻¹. Tissue was collected from mice subjected to CCI and administered KML29 (40 mg·kg⁻¹), gabapentin (50 mg·kg⁻¹) or KML29:gabapentin (combo; 13.33:4 mg·kg⁻¹) daily for 7 days. Samples were collected 2 h after the final drug administration. *P < 0.05 versus vehicle (n = 7–8).

The greatest changes in lipid levels occurred after the combination dose of KML29:gabapentin (13.33:4 mg·kg⁻¹) was administered for 7 days. Unlike the treatment of KML29 or gabapentin alone, the combination therapy significantly increased levels of anandamide [F(3,27) = 3.07; P < 0.05; Figure 5A] and its endogenous structural analogues, the N‐acylethanolamines (NAEs) and N‐arachidonoyl taurine. Specifically, the individual NAEs increased by the combination therapy were N‐palmitoyl ethanolamine (PEA), N‐linoleoyl ethanolamine and N‐docosahexaenoyl ethanolamine (Table 3). Linoleic acid was also significantly increased [F(3,27) = 4.41; P < 0.05; Figure 5F]; however, no change was observed in levels of arachidonic acid. Levels of N‐arachidonoyl glycine (NAGly), which were significantly increased with gabapentin alone, were not significantly different from vehicle with the combination therapy. Conversely, levels of 2‐AG (Figure 5C) and 2‐LG (Figure 5E) that were significantly increased with KML29 treatment remained elevated. Interestingly, levels of 2‐oleoyl glycerol and 2‐palmitoyl glycerol that were unchanged with KML29 treatment alone were also significantly increased with KML29:gabapentin. Levels of PGE₂, which were increased with gabapentin, were unchanged with the combination therapy.

Discussion

The main goal of this study was to evaluate the interaction between MAGL inhibition and gabapentin in a model of neuropathic pain in mice. The selective MAGL inhibitor, KML29, dose‐dependently attenuated mechanical and cold allodynia, as previously reported (Crowe et al., 2015). Gabapentin also dose‐dependently reduced mechanical and cold allodynia. Fixed‐dose proportions of KML29 and gabapentin additively reduced mechanical allodynia and synergistically attenuated cold allodynia. The reductions in allodynia were achieved by combining subthreshold doses of both drugs. Cannabinoid receptor involvement in the observed anti‐allodynia was also probed. The CB₁ antagonist rimonabant, but not the CB₂ antagonist SR144528, partially attenuated the anti‐allodynic effect of KML29:gabapentin in mechanical allodynia, indicating that the anti‐allodynic effects were at least partly mediated by CB₁ receptors. However, neither rimonabant nor SR144528 affected cold allodynia, indicating a non‐cannabinoid mechanism, and possibly also a stronger contribution of gabapentin in reducing cold allodynia.

Using lower doses of candidate analgesics reduces drug exposure and may subsequently reduce the risk of drug tolerance. For example, chronic administration of high‐dose, but not low‐dose, MAGL inhibitors causes functional antagonism of CB₁ receptors. Specifically, the MAGL inhibitor, JZL184 (≥16 mg·kg⁻¹), causes down‐regulation and desensitization of CB₁ receptors (Schlosburg et al., 2010; Kinsey et al., 2013), thereby reversing 2‐AG‐mediated antinociception. The tolerance to JZL184 is spared at lower doses (Kinsey et al., 2013). Similarly, the anti‐oedematous and anti‐allodynic effects of KML29 following carrageenan injection were reversed after repeated administration (Ignatowska‐Jankowska et al., 2014), indicating that the chronic blockade of MAGL with high‐dose KML29 produces tolerance. Although acute administration of a high‐dose MAGL inhibitor is beneficial for acute preclinical models of pain, chronic pain conditions, such as neuropathic pain, require long‐term treatments.

As expected, in the present study, prolonged MAGL inhibition by high‐dose KML29 produced analgesic tolerance in mechanical and cold allodynia, which was confirmed by a reduction in CB₁ function (desensitization) and density (down‐regulation) in whole brain homogenates. In contrast to KML29, gabapentin produced anti‐allodynia that persisted following chronic administration in both tests. The combination treatment of low‐dose KML29:gabapentin did not show tolerance in mechanical allodynia, but tolerance may have developed in the cold allodynia test. Given the marginal effect, CB₁ binding and GTPγS were assayed in whole brain homogenates. The combination reduced CB₁ density but did not alter CB₁ function. This reduced CB₁ density may account for the incomplete anti‐allodynic tolerance of the drug combination evident in cold allodynia not observed in mechanical allodynia.

To further elucidate the mechanisms through which the KML29:gabapentin combination synergistically interact to attenuate allodynia, we quantified spinal cord lipid levels following repeated administration of each drug alone and in combination. High dose KML29 increased 2‐AG by threefold as compared to vehicle, while gabapentin increased NAGly and PGE₂. Surprisingly, the combination of KML29:gabapentin not only increased 2‐AG but also increased PEA, N‐acyl glycines, 2‐acyl glycerols and N‐arachidonoyl taurine, which may indicate that dual administration of KML29 and gabapentin influences additional pathways other than those pathways activated by the two drugs alone. For example, KML29:gabapentin drive increases in all 2‐acyl glycerols tested. Levels of PGE₂ are at baseline with combination treatment but were affected in opposite ways with individual treatment, suggesting that the suppression of PGE₂ may be one of the ways the combination therapy is functioning. NAEs and N‐arachidonyl taurine are potent transient receptor potential cation channel subfamily V (TRPV) receptor ligands; however, the increases shown with the combination therapy was a low magnitude, which likely means that their effects on TRPVs would be minimal.

Mechanistically, we suggest that it is the dramatic and sustained increases in 2‐acyl glycerols, which appear to be involved in stabilizing PGE₂ levels that is part of the lipid signalling driving the decrease in pain‐related behaviours observed with the combination therapy. Lipid‐based signalling molecules were quantified in the lumbar enlargement of the spinal cord. CCI surgery induces cFos expression in the lumbar spinal cord and activating CB receptors decreases cFos in the lumbar spinal cord (Rodella et al., 2005). Similarly, it is plausible that gabapentin‐induced reductions in pain pathway signalling indirectly alter endocannabinoid levels in discrete regions that were not captured in the present study, which examined homogenates of the lumber spinal cord. In rats, CCI surgery site specifically increases anandamide and 2‐AG in the lumbar spinal cord, as compared with sham operated rats (Petrosino et al., 2007). Because pain perception is mediated by afferent signalling in the dorsal spinal cord, quantifying endocannabinoid levels in the dorsal, lumbar spinal cord is a priority for future studies.

The efficacy of low dose gabapentin can be potentiated when combined with other non‐cannabinoid drugs. For example, patients with diabetic neuropathy or postherpetic neuralgia reported lower pain scores after receiving gabapentin and morphine than when administered morphine or gabapentin alone (Gilron et al., 2005). However, combining gabapentin and morphine was also associated with a higher incidence of constipation and dry mouth, compared to gabapentin or morphine alone respectively (Gilron et al., 2005). Similarly, patients reported an increase in pain relief after receiving a combined treatment of gabapentin and oxycodone but experienced an increase in opioid‐associated side effects (Hanna et al., 2008). In other words, the interaction of the combined drugs non‐selectively enhanced both analgesia and negative side effects of either drug. To evaluate potential changes in cannabimimetic side effects in the current study, KML29, gabapentin, KML29:gabapentin and vehicle were assessed for classic cannabinoid effects in the tetrad battery. None of the treatment groups elicited catalepsy, locomotor activity, tail immersion or body temperature differences, indicating that none of the treatments produced overt cannabimimetic effects. Moreover, the drug mixture results in the tetrad support that the synergistic antinociceptive effects were not due to a generalized suppression of all behaviour. However, an important limitation of the present study is that the treatment side effects were evaluated in non‐operated mice, and thus, it is possible that mice subjected to CCI may respond differently.

Patients with neuropathic pain tend to be over‐represented in pain clinics due to higher reported pain intensity, longer duration of pain and less effective pain relief, as compared with other forms of chronic pain (Torrance et al., 2006; 2007), underscoring the need to develop new, effective therapeutic treatments. Examining the effectiveness and safety of a MAGL inhibitor in clinical trials is an important step towards the development of new analgesics that target MAGL.

Clinically, synthetic cannabinoids are prescribed to cancer patients to stop emesis (Sticht et al., 2015); however, less is known about the efficacy of cannabinoids to control pain associated with chemotherapy‐induced neuropathy in patients. While there are relatively few clinical studies evaluating cannabinoids for chemotherapy‐induced neuropathy in humans, one study evaluated Sativex (an inhaled combination of Δ⁹‐THC and cannabidiol) in combination with patients' existing opioid treatment to increase the efficacy of reducing cancer pain (Johnson et al., 2010). Sativex, in combination with the existing opioid regimen, reduced pain severity, as compared to placebo, although it did not decrease the dose of opioids taken (Johnson et al., 2010). Murine models of chemotherapy‐induced neuropathic pain indicate that cannabinoids attenuate mechanical and cold allodynia (Hohmann, 2005; Guindon et al., 2013; Deng et al., 2015a,b) and prevent neuropathy when given in conjunction with chemotherapy (Ward et al., 2011; 2014; Rahn et al., 2014). Taken together, the results from clinical and preclinical research indicate that drugs targeting the endocannabinoid system alone and in conjunction with other non‐cannabinoid targets alleviate multiple types of neuropathic pain.

In summary, coadministration of KML29 and gabapentin additively reduced mechanical allodynia and synergistically reduced cold allodynia. Repeated administration of low dose KML29:gabapentin did not undergo tolerance in mechanical allodynia and did not alter CB₁ receptor function in the brain. The synergistic interaction of the drugs was also evident in lipidomic analyses of the spinal cord, which indicate that the combination treatment activates pathways in addition to those altered by either drug alone. The combination of these drugs may be beneficial for increasing analgesia, while administered at relatively low doses. These data provide support to the strategy of targeting the endocannabinoid system in conjunction with non‐cannabinoid systems, to develop new candidate analgesics with limited negative side effects.

Author contributions

M.S.C., P.L.P., H.B.B., M.L.B. and S.G.K. participated in research design. M.S.C., C.D.W. and E.L. conducted the experiments. M.S.C., C.D.W., E.L., P.L.P., M.L.B., H.B.B. and S.G.K. performed the data analyses. M.S.C., E.L., P.L.P., H.B.B., M.L.B. and S.G.K. wrote or contributed to the writing of the manuscript.

Conflict of interest

The authors declare no conflicts of interest.

Declaration of transparency and scientific rigour

This Declaration acknowledges that this paper adheres to the principles for transparent reporting and scientific rigour of preclinical research recommended by funding agencies, publishers and other organisations engaged with supporting research.

Supporting information

Table S1 Lumbar spinal cord levels of eicosanoids from mice treated repeatedly with KML29 (KML; 40 mg·kg⁻¹), gabapentin (GBP; 50 mg·kg⁻¹), or KML29: GBP (combo; 13.33:4 mg·kg⁻¹). Tissue was collected from mice subjected to chronic constriction injury 2 h after drug administration Spinal cord levels are expressed as mean (SEM) in pmol·g⁻¹. Tissue was collected from mice with CCI 2 h after drug administration. *P < 0.05 versus vehicle, Dunnett's post hoc test.

Crowe, M. S. , Wilson, C. D. , Leishman, E. , Prather, P. L. , Bradshaw, H. B. , Banks, M. L. , and Kinsey, S. G. (2017) The monoacylglycerol lipase inhibitor KML29 with gabapentin synergistically produces analgesia in mice. British Journal of Pharmacology, 174: 4523–4539. doi: 10.1111/bph.14055.