Prophylactic cannabinoid administration blocks the development of paclitaxel-induced neuropathic nociception during analgesic treatment and following cessation of drug delivery

Background Chemotherapeutic treatment results in chronic pain in an estimated 30-40 percent of patients. Limited and often ineffective treatments make the need for new therapeutics an urgent one. We compared the effects of prophylactic cannabinoids as a preventative strategy for suppressing development of paclitaxel-induced nociception. The mixed CB1/CB2 agonist WIN55,212-2 was compared with the cannabilactone CB2-selective agonist AM1710, administered subcutaneously (s.c.), via osmotic mini pumps before, during, and after paclitaxel treatment. Pharmacological specificity was assessed using CB1 (AM251) and CB2 (AM630) antagonists. The impact of chronic drug infusion on transcriptional regulation of mRNA markers of astrocytes (GFAP), microglia (CD11b) and cannabinoid receptors (CB1, CB2) was assessed in lumbar spinal cords of paclitaxel and vehicle-treated rats. Results Both WIN55,212-2 and AM1710 blocked the development of paclitaxel-induced mechanical and cold allodynia; anti-allodynic efficacy persisted for approximately two to three weeks following cessation of drug delivery. WIN55,212-2 (0.1 and 0.5 mg/kg/day s.c.) suppressed the development of both paclitaxel-induced mechanical and cold allodynia. WIN55,212-2-mediated suppression of mechanical hypersensitivity was dominated by CB1 activation whereas suppression of cold allodynia was relatively insensitive to blockade by either CB1 (AM251; 3 mg/kg/day s.c.) or CB2 (AM630; 3 mg/kg/day s.c.) antagonists. AM1710 (0.032 and 3.2 mg/kg /day) suppressed development of mechanical allodynia whereas only the highest dose (3.2 mg/kg/day s.c.) suppressed cold allodynia. Anti-allodynic effects of AM1710 (3.2 mg/kg/day s.c.) were mediated by CB2. Anti-allodynic efficacy of AM1710 outlasted that produced by chronic WIN55,212-2 infusion. mRNA expression levels of the astrocytic marker GFAP was marginally increased by paclitaxel treatment whereas expression of the microglial marker CD11b was unchanged. Both WIN55,212-2 (0.5 mg/kg/day s.c.) and AM1710 (3.2 mg/kg/day s.c.) increased CB1 and CB2 mRNA expression in lumbar spinal cord of paclitaxel-treated rats in a manner blocked by AM630. Conclusions and implications Cannabinoids block development of paclitaxel-induced neuropathy and protect against neuropathic allodynia following cessation of drug delivery. Chronic treatment with both mixed CB1/CB2 and CB2 selective cannabinoids increased mRNA expression of cannabinoid receptors (CB1, CB2) in a CB2-dependent fashion. Our results support the therapeutic potential of cannabinoids for suppressing chemotherapy-induced neuropathy in humans.


Background
Cannabinoids attenuate or, in some cases, prevent pain associated with surgery [1], inflammation [2], internal organs [3], and neuropathies (for review see [4]). Neuropathic pain is associated with abnormal changes in the peripheral and/or central nervous system resulting in non-adaptive, chronic pain. Clinical manifestations of neuropathic pain are notoriously unresponsive to traditional analgesics. Chemotherapeutic treatment with antineoplastic agents, while effective at eliminating harmful malignancies, is also associated with severe side effects. Of these side effects, emesis, alopecia, and myelosuppression have received the spotlight; however, a new front runner has recently emerged. Neuropathic pain associated with chemotherapeutic treatment is dose-limiting and a major factor influencing discontinuation of treatment [5,6]. Chemotherapy-induced neuropathy is positively correlated with cumulative chemotherapeutic dose [7], and affected patients are more likely to experience other neuropathies [8]. An aging US population, coupled with diagnostic and medical advances in cancer treatment, means that more cancer survivors will be impacted by, and living longer with, chemotherapy-induced neuropathy. Thus, identification of prophylactic treatments that block development of chemotherapy-induced neuropathy represents an urgent medical need.
Chemotherapeutic agents are divided into three mechanistically distinct classes. These classes include the vinca alkaloids, platinum-derived agents, and taxanes. Taxanes (e.g., paclitaxel, docetaxel) produce antineoplastic effects by stabilizing microtubules through binding to β-tubulin, thereby disrupting normal cell mitosis and triggering the mitochondrial apoptosis pathway [9]. Paclitaxel is a preferred agent for treatment of ovarian, breast, and lung cancers; however, a high percentage of patients experience neuropathic paina type of pain poorly treated with available drugs [10]. Mechanisms underlying development of paclitaxel-induced neuropathy remain incompletely understood but may involve changes in glial activation [11,12].
Cannabinoid agonists suppress paclitaxel-induced neuropathic nociception in animal models through activation of both CB 1 [13] and CB 2 [14][15][16] cannabinoid receptor subtypes. Our laboratory first demonstrated CB 2 receptormediated suppression of neuropathic allodynia induced by chemotherapeutic treatment with vincristine [17], paclitaxel [14,18], and cisplatin [15]. Previous prophylactic treatment strategies with cannabinoids in a traumatic nerve injury model demonstrated that pre-emptive cannabinoids produced greater antinociception relative to post-injury treatment [19]. Here we investigate the therapeutic efficacy of prophylactically administered WIN55,212-2, a mixed cannabinoid (CB 1 /CB 2 ) agonist, and AM1710, a CB 2 -preferring agonist, on the development of chemotherapy-induced neuropathy in the paclitaxel model. Osmotic mini pumps were used to continuously infuse cannabinoids before, during, and after paclitaxel treatment, to emulate a prophylactic analgesic strategy achievable in clinical oncology settings. We compared development of mechanical and cold allodynia, both common clinical manifestations of paclitaxel-induced neuropathy [10,20]. We hypothesized that chronic prophylactic cannabinoid infusion would produce sustained suppression of paclitaxel-induced behavioral sensitization to mechanical and cold stimulation. Furthermore, we evaluated whether long-term transcriptional changes in mRNA markers of astrocytes (GFAP), microglia (CD11b), and cannabinoid receptors (CB 1 , CB 2 ) would accompany long lasting anti-allodynic efficacy of cannabinoids.

General results
Paclitaxel-treated animals showed reduced sensitivity to heat on day 6 (F 1,10 = 20.745, P < 0.01; Figure 1a), but not at subsequent time points (P > 0. 16), while the same animals developed hypersensitivity to mechanical stimulation (i.e., mechanical allodynia) (F 1,10 = 6.191, P < 0.05; Figure 1b). Based upon these results, animals implanted with osmotic pumps were evaluated for responsiveness to mechanical and cold stimulation only.
Osmotic mini pump dispersion volume was calculated by subtracting the fill volume from the residual volume in the pump reservoir following pump removal (day 22). The pump dispersion volume differed between groups in which drugs were dissolved in the DMSO:PEG400 vehicle (F 19,180 = 2.213, P < 0.01). Post-hoc analysis revealed that pump dispersion volume for the Taxol-WIN55,212-2 (1 mg/kg/day s.c.) group was less than half (< 43%) of other groups dissolved in the same vehicle. No other differences were found. Mechanical withdrawal thresholds did not differ between either the right or left paw on any given day for animals tested up to 20 (P > 0.98) or 50 (P > 0.71) days post-chemotherapy treatment; therefore, withdrawal thresholds are presented as the mean of duplicate measurements, averaged across paws. Two dependent measures for cold allodynia were evaluated: percentage of paw withdrawals and duration of paw withdrawal. Duration of paw withdrawal in response to topical acetone application is a reported measure of cold allodynia [21][22][23]. However, we found this measure highly variable in rat subjects (data not shown) and consequently only the percentage of paw withdrawals is reported here. Percentage of paw withdrawals to cold stimulation did not differ between either paw on any given day for animals tested up to 21 (P > 0.33) or 51 (P > 0.82) days post-paclitaxel; therefore, the percentage of paw withdrawals is presented as the mean of duplicate measurements averaged across paws.
To control for any possible effects associated with the vehicle used to dissolve cannabinoids (DMSO:PEG 400 in a 1:1 ratio), a subset of animals treated with either paclitaxel or cremophor received saline in their osmotic mini pumps. No differences were detected between paclitaxel-treated animals that received vehicle (DMSO: PEG 400; n = 14) or saline (n = 4) in any behavioral parameter assessed (i.e., mechanical threshold, cold withdrawal frequency, and locomotor activity). Similarly, no differences were noted between cremophor-treated animals receiving chronic infusions of vehicle (DMSO:PEG 400; n = 8) or saline (n = 4). Therefore, vehicle and saline groups were combined for each condition and are referred to as the Taxol-vehicle group and cremophor-vehicle group, respectively.

Body weight
Body weight did not differ between paclitaxel-or cremophor-treated animals receiving infusions of vehicle (P = 0.69; Figure 2a). Moreover, no differences in body weight were observed between paclitaxel-treated animals receiving either vehicle or saline (data not shown). However, cremophor-treated animals receiving saline infusions exhibited greater weight gain on days 14-21 (F 12,204 = 8.455, P < 0.001, P < 0.05 for each day) relative to those receiving vehicle.

Discussion
Prophylactic administration of cannabinoid analgesics protected against the development of paclitaxel-induced hypersensitivities to mechanical and cold stimulation in a preventative fashion. Both the mixed cannabinoid CB 1 /CB 2 agonist WIN55,212-2 and the CB 2 agonist AM1710 blocked development of paclitaxel-induced mechanical and cold allodynia. Strikingly, the protective prophylactic effects of both WIN55,212-2 and AM1710 were preserved following drug removal, with the CB 2specific agonist providing a longer duration of protection against allodynia development for both mechanical and cold modalities. In our study, paclitaxel produced marked mechanical and cold allodynia but not heat hyperalgesia, as observed in a different dosing paradigm (cumulative dose: 4 mg/kg i.p.) [28]. In vehicle (cremophor) treated  controls, the most efficacious doses of these cannabinoids also failed to produce antinociception, suggesting that cannabinoids were anti-allodynic rather than analgesic under these conditions.
Prophylactic WIN55,212-2 suppresses paclitaxel-induced mechanical and cold allodynia WIN55,212-2 (0.5 and 0.1 mg/kg/day s.c.) suppressed the development of paclitaxel-induced mechanical and cold allodynia both during drug delivery and following drug removal. Our study is the first to evaluate duration of efficacy, dose response, and pharmacological specificity of prophylactic WIN55,212-2. Anti-allodynic effects of both doses were present 11 (mechanical) and 12 (cold) days following cessation of drug delivery. WIN55,212-2 (0.5 mg/ kg/day s.c.)-induced suppression of mechanical hypersensitivity was dominated by CB 1 receptor activation because anti-allodynic efficacy was blocked by AM251. The CB 2 antagonist AM630 (3 mg/kg/day s.c.) prevented anti-allodynic efficacy of AM1710 but failed to eliminate WIN55,212-2-mediated anti-allodynia. Interestingly, blockade of WIN55,212-2-mediated anti-allodynic effects to cold was not achieved with either antagonist. However, the same antagonist infusion conditions blocked either WIN55,212-2 mediated suppression of mechanical allodynia (AM251) or AM1710-mediated suppression of both mechanical and cold allodynia (AM630), documenting efficacy of antagonist infusion conditions employed here.
We could find only one report of WIN55,212-2-induced suppression of cold allodynia in a neuropathic pain model (spinal nerve ligation) where pharmacological specificity was assessed; anti-allodynic effects were blocked by a CB 1 (SR141716a) but not a CB 2 (SR144528) antagonist [29]. Blockade of both CB 1 and CB 2 receptors may be required to fully prevent anti-allodynic effects of WIN55,212-2. However, limitations in compound solubility prohibited co-administration of both antagonists in one pump.
Few studies have examined cannabinoid-mediated modulation of cold allodynia and/or its development in neuropathic pain models and more work is necessary to determine functional contributions of each receptor. WIN55,212-2 (0.5 mg/kg/day s.c.) treatment increased both CB 1 and CB 2 receptor mRNA expression within lumbar spinal cord of paclitaxel-treated animals on day 22, an effect blocked by concurrent AM630 (3 mg/kg/ day s.c.) treatment. WIN55,212-2 also ameliorates established paclitaxel-induced nociception [13] and repeated administration (1 mg/kg i.p. × 14 days) prevents nociception development during drug delivery [11].

Prophylactic AM1710 suppresses paclitaxel-induced mechanical and cold allodynia
Strikingly, doses of AM1710 as low as 0.032 mg/kg/day blocked the development of paclitaxel-induced allodynia in our study and these effects were preserved for approximately three weeks following cessation of drug delivery. Prophylactic AM1710 treatment suppressed development of both paclitaxel-induced mechanical and cold allodynia, with high (3.2 mg/kg/day s.c.) and low (0.032 mg/ kg/day s.c.) doses exhibiting the greatest efficacy. A similar U-shaped dose response curve was obtained for thermal antinociception (plantar test) in naive animals without observable CB 1 -mediated side effects [30].

Lack of side effects after either WIN55,212-2 or AM1710 treatment
In our study, animals remained in good health and showed either normal or enhanced weight gain. WIN55,212-2 (0.1 mg/kg/day s.c.) increased weight gain in paclitaxeltreated rats. CB 1 activation can produce both orexigenic and metabolic effects to promote weight gain [33]. Interestingly, higher doses of WIN55,212-2 (1-2 mg/ kg/day i.p.) failed to attenuate anorexia or weight loss in animals treated with cisplatin [34].
Activity meter assessments conducted during prophylactic treatment (day 19) and following drug removal (day 31), failed to reveal major differences between groups. Thus, chronic infusion of either the mixed CB 1 /CB 2 agonist or the CB 2 agonist was unlikely to nonselectively activate CB 1 receptors; no evidence for hypoactivity [35], a cardinal sign of CB 1 activation, was observed. These findings are consistent with the results documenting absence of cardinal signs of CB 1 receptor activation following acute administration of AM1710 [30].

Potential mechanisms of action for cannabinoid-mediated suppression of paclitaxel-induced neuropathic nociception
Glial activation mediates alterations in synaptic transmission for a number of excitatory and inhibitory mediators known to be important for the maintenance of neuropathic pain states [36]. Because of the prolonged suppression of paclitaxel-induced neuropathy after removal of cannabinoid agonists, we chose to analyze transcriptional changes in markers of glial activation. GFAP mRNA expression in lumbar spinal cord on day 22 (i.e., approximately 24 h after the pump ceased to release drug) showed a trend toward increased expression in paclitaxel-relative to cremophor-treated controls, while no change in CD11b expression was observed. Increases in astrocytic activation (GFAP) with no corresponding changes in microglial activation (OX42, Iba1, and phosphorylated p38) were also recently observed with the same paclitaxel-induced neuropathy dosing protocol used here [12]. In another study, paclitaxel failed to produce microglial activation (% of cremophor-control staining) on day 27 posttreatment [37]. By contrast, MDA7 and WIN55,212-2 suppressed paclitaxel-induced glial activation (on days 28 and 29 post-treatment, respectively) when immunohistochemical staining for astrocytes (GFAP) and microglia (CD11b) was compared with naive animals [11], and it remains unclear whether vehicle or cremophor administration alters glial activation [16]. Cremophor can produce side effects in both clinical use and animal models [38], and assumptions that it is inert are not appropriate.
Prophylactic treatment with either a mixed CB 1 /CB 2 agonist or a CB 2 agonist, while failing to produce robust alterations in lumbar spinal cord glial expression, increased CB 1 and CB 2 mRNA expression. This effect was blocked by CB 2 receptor blockade. Upregulation of endocannabinoids and cannabinoid receptors is associated with several neuropathic pain models (for review see [39]). However, to our knowledge, very few, if any, studies have evaluated alterations following prophylactic cannabinoid treatment. Increased receptor densities could increase the potency or efficacy of prophylactic cannabinoids in this model. More work is necessary to determine whether changes in CB 1 and CB 2 mRNA levels observed here are also associated with changes in receptor protein. Alternatively, increased CB 1 and CB 2 mRNA expression could reflect compensatory changes in transcription following chronic agonist-induced downregulation of receptors. More work is necessary to fully characterize the duration of these effects and their therapeutic implications.

Translation to the clinic
Our preclinical studies [14,15,17,18] motivated completion of a pilot clinical trial utilizing Sativex, an oromucosal extract containing Δ 9 -tetrahydrocannabinol and cannabidiol, for treatment of chemotherapy-induced neuropathy. Sativex suppressed established chemotherapyinduced neuropathic pain in a subset of responders (5 of 18) in this double-blind placebo-controlled crossover pilot [40], supporting a further evaluation of the clinical viability of cannabinoid-based pharmacotherapy.

Conclusion
Prophylactic treatment has been tested as a preventive strategy for paclitaxel-induced neuropathic nociception with multiple drugs (for review see [41]). Here, we demonstrate that cannabinoid agonists with different mechanisms of action prevent development of paclitaxel-induced neuropathic nociception during treatment and approximately two to three weeks following cessation of drug delivery. Paclitaxel treatment marginally altered long-term GFAP mRNA expression in lumbar spinal cord and this expression was unaffected by prophylactic cannabinoids, whereas CD11b mRNA expression was unchanged. Prophylactic treatment with either WIN55,212-2 (0.5 mg/ kg/day s.c.) or AM1710 (3.2 mg/kg/day s.c.) in paclitaxel animals did, however, increase both CB 1 and CB 2 receptor mRNA expression, an effect blocked by concurrent AM630 (3 mg/kg/day s.c.) administration. Some inroads have been made toward discovering mechanisms for cannabinoid-mediated suppression of paclitaxelinduced neuropathy, but more work is necessary to determine the scope and time course of this complex interaction. Our study suggests that further clinical cannabinoid trials [40] for chemotherapy-induced peripheral neuropathy are warranted.

Subjects Rats
One hundred seventy-six adult male Sprague-Dawley rats (beginning weight: 300-400 g; Harlan, Indianapolis, IN) were used in these experiments. All procedures were approved by the University of Georgia Animal Care and Use Committee and followed the guidelines for the treatment of animals of the International Association for the Study of Pain. Animal experiments were conducted in full compliance with local, national, ethical and regulatory principles, and local licensing regulations of the Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) International's expectations for animal care and use/ethics committees.
Animals were allowed a minimum of one week habituation prior to beginning the study. Animals were single housed and maintained in a temperature (70-72°F ± 4°F) and humidity (30-70%) controlled facility on a 12 hour light cycle (lights on: 07:00 and lights off: 19:00). Food and water was available ad libitum. Following the initial pilot study (n = 17), all animals with osmotic mini pumps were allowed nyla bones (BioServe; Frenchtown, NJ) due to the study duration. Corn cob bedding containing metabolized paclitaxel was treated as chemical hazard waste and disposed of according to appropriate institutional guidelines. (4-methoxyphenyl)methanone (Iodopravadoline) were syn thesized in the Center for Drug Discovery by one of the authors (by GT, VKV, and AZ respectively). Rat subjects received paclitaxel dissolved as previously described [42], administered in a volume of 1 ml/kg. Briefly, paclitaxel was dissolved in a 1:2 ratio of working stock (1:1 ratio of cremophor EL and 95% ethanol) to saline. AM1710, WIN55,212-2, AM251, and AM630 were dissolved in a vehicle of DMSO:PEG 400 in a 1:1 ratio. The selected vehicle was the most compatible for dissolving cannabinoids to be used in Alzet osmotic mini pumps with no reported adverse side effects [43][44][45].

General experimental methods
In an initial study, animals were evaluated for development of paclitaxel-induced behavioral sensitization to mechanical and heat stimulation. Responsiveness to different modalities of cutaneous stimulation was assessed on alternate days to avoid sensitization. All subsequent studies used animals surgically implanted with osmotic mini pumps. Baseline assessments of withdrawal thresholds to mechanical and cold (acetone drops) stimulation of the hind paw occurred 48 h (day −8) and 24 h (day −7) prior to surgery, respectively. Osmotic mini pumps (Alzet model 2ML4, Cupertino, CA) were implanted subcutaneously through an incision between the scapulae. Responsiveness to mechanical and cold stimulation was reassessed post-surgery (i.e., after pump implantation but within 48 h prior to initiation (on day 0) of paclitaxel dosing). Animals were weighed on all testing and surgical/sacrifice dates. A subset of animals was sacrificed via live decapitation (day 22) to extract lumbar spinal cords. Certain groups (e.g., antagonist alone conditions, submaximal doses of agonists, cremophor-agonist groups) were only tested through day 22. Osmotic mini pumps were removed in all remaining animals (day 22), and following a short recovery period, responses to mechanical and cold stimulation were reassessed until day 51 post-paclitaxel.
Drug doses were estimated based on the peak osmotic mini pump performance reported by the manufacturer (2.5 μl/hr) and an average rat weight of 375 grams. A small percentage of animals (4.2%) presented with edema around the pump site (seromas). Alzet reports this side effect in < 5% of animals. Treatment for these animals was supervised by the attending veterinarian and consisted of draining fluid every 3 days, or as needed. Six animals (3.6%) were re-sutured following surgery. One of the six animals developed an infection and was treated (from days [16][17][18][19][20][21][22] with daily injections of an antibiotic (Enrofloxacin 4.5 mg/ml, 0.4 cc s.c., 2× daily) and sterile water (1 cc s.c., 1× daily) as prescribed by the staff veterinarian. One animal died during the first paclitaxel injection and this animal was excluded from all analyses.
Behavioral measurements, surgeries, chemotherapeutic treatment, and tissue removal were performed by a single experimenter (EJR). Coded testing sheets were used to preserve blinding. Behavioral testing was performed in the presence of a white noise generator to mask extraneous noise.

Surgical implantation and removal of osmotic mini pumps
Osmotic mini pumps were implanted under isoflurane anesthesia (Isoflo®, Abbott Laboratories, Chicago, IL). The osmotic mini pump was inserted through a surgical incision made between the scapulae; incisions were sutured closed. In the instances where two pumps were implanted (i.e., agonist and antagonist co-administration conditions), pumps were placed in the same pocket. The Alzet model 2ML4 pump has an approximate 2 ml reservoir that releases a preloaded drug or vehicle at a rate of 2.5 ul/hr for approximately 28 days. The pump begins to release the preloaded drug approximately 4-6 hours after implantation; the flow rate is not subject to variations in body temperature. Osmotic mini pumps were weighed before and after being filled with drug or vehicle. The difference of these two values provided an approximate pump fill volume. The animals were given three days (days −5 through −3) to recover from surgery before testing resumed. Animals were either sacrificed or underwent surgery on day 22 to remove pumps; this time point corresponds to the 29th day following pump implantation, at which point the pump should have released its contents. Following pump removal, the residual pump volume was estimated by withdrawing the remaining fluid within the pump reservoir. Animals that underwent surgical removal of osmotic mini pumps were allowed three days of recovery (days [23][24][25] prior to resumption of behavioral testing.

Assessment of paw withdrawal latencies to heat stimulation
Paw withdrawal latencies to radiant heat were measured in duplicate for each paw using the Hargreaves test [46] and a commercially available plantar stimulation unit (IITC model 336; Woodland Hills, CA). Rats were placed underneath inverted plastic cages positioned on an elevated glass platform and allowed a minimum of 20 min to habituate prior to testing. Radiant heat was presented to the hind paw midplantar region through the floor of the glass platform. The intensity of the heat source was adjusted such that an average baseline latency of approximately 20 sec was achieved [47]. Stimulation was terminated upon paw withdrawal or after 40 s to prevent tissue damage. Approximately 4 minute interstimulation intervals were allowed between tests. Thermal withdrawal latencies were evaluated before (day 0) and on days 2, 6, 10, 14 and 18 following initiation of paclitaxel dosing. The same animals were tested for the development of mechanical allodynia. Baseline responses to mechanical stimulation (methodology below) were measured (on day 0) before baseline responses to thermal stimulation. A minimum of 1 hour was allowed to elapse between baseline measurements.

Assessment of mechanical withdrawal thresholds
Mechanical withdrawal thresholds were assessed using a digital Electrovonfrey Anesthesiometer (IITC model Alemo 2390-5; Woodland Hills, CA) equipped with a rigid tip. Rats were placed underneath inverted plastic cages positioned on an elevated mesh platform and allowed a 20 min habituation period prior to testing. Stimulation was applied to the hind paw midplantar region through the floor of a mesh platform. Mechanical stimulation was terminated upon paw withdrawal; consequently, no upper threshold limit was set for termination of a trial. Two thresholds were taken for each paw. Approximately 2 minute interstimulation intervals were allowed between tests. Mechanical withdrawal thresholds were measured on days 0, 4, 8, 12, 16 and 20 for animals that did not receive osmotic mini pumps (Figure 1). Mechanical withdrawal thresholds were measured every 2-6 days (i.e., days −8, −2, 0, 2, 4, 6, 8, 10, 12, 14, 16, 18, and 20) for all animals that received osmotic mini pumps. A subset of osmotic mini pump animals were tested until day 50 (testing continued with the following schedule: days 26, 28, 30, 32, 34, 36, 38, 40, 42, 44, 46, 48, and 50).

Assessment of cold allodynia
Cold allodynia was assessed using acetone drops applied to the hind paw midplantar surface as previously described [15,48]. Rats were placed underneath inverted plastic cages positioned on an elevated mesh platform and allowed a 20 min habituation period prior to testing. Acetone was loaded into a one cc syringe barrel with no needle tip. One drop of acetone (approximately 20 μl) was applied through the mesh platform onto the hind paw midplantar surface. Care was taken to gently apply the bubble of acetone to the skin without inducing mechanical stimulation by syringe barrel contact with the paw.
Paw withdrawal was recorded as a binary response (presence or absence) and was frequently accompanied by nocifensive behaviors (e.g., rapid flicking of the paw, chattering, biting, and/or licking of the paw). These nocifensive behaviors were recorded as duration of acetone response. Five measurements were taken for each paw. Testing order alternated between paws (i.e., right, left). No cut-off latency was enforced. Approximately 2 min interstimulation intervals were allowed between testing of right and left paws. A minimum interstimulation interval of 5 min was allowed between testing each pair of paws (right and left). Cold allodynia testing took place on days −7, −1, 5, 11, 17 and 21 for all animals with osmotic mini pumps. Five days were allowed between assessments of cold allodynia to avoid hypersensitivity with one exception. Animals were tested on day 21 because osmotic mini pumps would purportedly still be releasing drug (i.e., 28 days following pump implantation). A subset of animals was tested to day 51 (i.e., testing for these animals continued with the following schedule: days 27, 33, 39, 45, and 51).

Locomotor activity
Total distance traveled (cm) was assessed using an activity monitor chamber (Coulbourn Instruments, Whitehall, PA) measuring 40.64 cm 3 . The apparatus was housed in a darkened room and red light was used to provide illumination. Tracking beams were positioned 2.54 cm apart giving 1.27 cm in spatial resolution. Activity was automatically measured by computerized analysis of photobeam interrupts (TruScan 2.0; Coulbourn Instruments, Whitehall, PA). Animals were allowed a minimum of 15 minutes to habituate to the room prior to being placed undisturbed in the activity meter for 15 min. Chlorhexidine was used to clean the activity meter after each animal. Activity meter assessment took place both during (day 19) and following termination (day 31) of drug delivery in a subset of animals that received chronic infusions.

Statistical analyses
Percentage of paw withdrawals from acetone application to the hind paws was calculated using the following formula: ((Total number of paw withdrawals) * 100)/10. Data were analyzed using analysis of variance (ANOVA) for repeated measures, one-way ANOVA, or planned comparison t-test as appropriate. SPSS 19.0 (SPSS Incorporated, Chicago, IL, USA) statistical software was employed. The Greenhouse-Geisser correction was applied to all repeated factors where the epsilon value from Mauchly's Test of Sphericity was < 0.75 and significance level was P < 0.05. Degrees of freedom reported for interaction terms of repeated factors are uncorrected values in cases where the Greenhouse-Geisser correction factor was applied. Post-hoc comparisons between the primary control group (paclitaxel-vehicle) and other experimental groups were performed using the Dunnett test (2-sided). Post-hoc comparisons between different experimental groups were also performed to assess dose-response relationships and pharmacological specificity using the Tukey test. Levene's test for homoscedasticity was applied to all planned comparison t-tests. P < 0.05 was considered statistically significant.

Competing interests
Dr. Alexandros Makriyannis serves as a consultant for MAK Scientific. No other authors declare competing interests.
Authors' contributions EJR contributed to experimental design, completed all surgeries, behavioral studies, and tissue extractions, analyzed data and drafted the manuscript. LD isolated RNA, carried out the RT-PCR studies and analyzed data. GAT synthesized AM1710. VKV synthesized AM251 and AM630. AMZ synthesized AM630. YYL assisted with RT-PCR and contributed to manuscript preparation and data interpretation. AM provided cannabinoid compounds and contributed to data interpretation. AGH designed the study, participated in its coordination and implementation, and wrote the manuscript with EJR. All authors read and approved the final manuscript.