Experimental Neurology 185 (2004) 160 – 168 www.elsevier.com/locate/yexnr Cyclooxygenase inhibition in nerve-injury- and TNF-induced hyperalgesia in the rat Maria Schäfers, a,b,* Martin Marziniak, a Linda S. Sorkin, b Tony L. Yaksh, b and Claudia Sommer a a b Department of Neurology, University of Würzburg, 97080 Würzburg, Germany Anesthesiology Research Laboratory, University of California San Diego, La Jolla, CA 92093-0818, USA Received 28 January 2003; revised 12 September 2003; accepted 25 September 2003 Abstract After nerve injury, cyclooxygenase-2 (COX-2) is upregulated in spinal cord and peripheral nerve, the latter being dependent on tumor necrosis factor-alpha (TNF). Here we asked whether COX inhibitors attenuate pain behavior induced by chronic constrictive sciatic nerve injury (CCI) or intraneural injection of TNF (2.5 pg/ml). Rats received either 0.9% saline, the nonselective COX inhibitor ibuprofen (40 mg/ kg) or the selective COX-2 inhibitor celecoxib (10 or 30 mg/kg) twice daily by gavage started 2 days before, 12 h or 7 days after surgery. Mechanical allodynia and thermal hyperalgesia induced by CCI was moderately, but consistently attenuated by early (day 2 or 12 h after CCI), but not late (7 days after CCI) ibuprofen and celecoxib treatment. Mechanical allodynia, but not thermal hyperalgesia induced by intraneural TNF, was reduced by ibuprofen, but not by celecoxib treatment 5 and 7 days after injection. Sciatic nerves, lumbar dorsal root ganglia (DRG) and spinal cords from rats with treatment started 12 h after surgery were analyzed for prostaglandin E2 (PGE2) levels 10 days after CCI. In injured nerves and ipsilateral DRG, PGE2 levels were increased. Ibuprofen treatment reversed PGE2 levels in injured nerves and DRG, whereas celecoxib blocked increased PGE2 levels only in nerves. In spinal cord, no change in PGE2 levels was observed. In contrast to the marked inhibition of nerve-injury-induced upregulation of PGE2 by COX inhibitors, the effect on pain behavior was modest. Nerveinjury- and TNF-induced pain-related behavior seem to be only partly dependent on peripheral prostaglandins. D 2003 Elsevier Inc. All rights reserved. Keywords: Chronic constriction injury; Cyclooxygenase inhibition; Prostaglandin E2; Neuropathic pain; Tumor necrosis factor-alpha Introduction Prostaglandins are critically involved in the processing of pain at the level of peripheral nerve terminals and in the spinal cord (Malmberg and Yaksh, 1992; Taiwo and Levine, 1988). Cyclooxygenase-2 (COX-2), the inducible prostaglandin-synthesizing enzyme, is constitutively present in normal rat lumbar spinal cord (Beiche et al., 1998) and increased by inflammation or administration of cytokines (Cao et al., 1998; Lacroix and Rivest, 1998; Samad et al., 2001). In contrast to their established role in inflammatory hyperalgesia, the function of prostaglandins in neuropathic pain is less clear. Nonsteroidal anti-inflam- * Corresponding author. Department of Neurology, University of Würzburg, Josef-Schneider Straße 11, 97080 Würzburg, Germany. Fax: +49-931-201-23697. E-mail address: schaefers_m@klinik.uni-wuerzburg.de (M. Schäfers). 0014-4886/$ - see front matter D 2003 Elsevier Inc. All rights reserved. doi:10.1016/j.expneurol.2003.09.015 matory drugs (NSAIDs) are generally considered to be ineffective in the treatment of neuropathic pain and the few published clinical trials support this notion (Max et al., 1988; Weber et al., 1993). However, one open clinical study reports an analgesic effect of high dose ibuprofen on pain in diabetic neuropathy (Cohen and Harris, 1987). Although some animal studies implicate prostaglandins in the pathogenesis of neuropathic pain, reports are conflicting. Local (intraneural and intraplantar), but not systemic, administration of the nonselective COX inhibitor indomethacin reduces mechanical allodynia and thermal hyperalgesia after nerve injury (Syriatowicz et al., 1999). However, the nonselective, but COX-1-preferring COX inhibitor ketorolac attenuates pain behavior if applied locally (perineural or intraplantar), as well as systemically after nerve injury (Ma and Eisenach, 2002). In contrast, preemptive local (intraplantar and perineural) injection of ketorolac given at the time of nerve lesion failed to prevent the development of tactile allodynia, but had a delayed antiallodynic effect 3 M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 weeks after nerve lesion (Ma and Eisenach, 2002). Intrathecal administration of indomethacin reduces tactile allodynia if treatment is initiated early, but not if it starts 2 weeks after nerve injury (Zhao et al., 2000). In one study, a single intrathecal injection of ketorolac applied 4 weeks after nerve injury almost completely reversed tactile allodynia (Ma et al., 2002), while in others only a minimal antiallodynic effect was noted (Lashbrook et al., 1999). Conflicting results may derive from different administration routes of COX inhibitors, different degrees of CNS penetration of the various agents and different roles for the two COX isozymes during development and maintenance of neuropathic pain. The mechanisms by which COX inhibitors exert their analgesic action is not certain. Prostaglandin E2 (PGE2) is a key mediator in the processes of peripheral and spinal sensitization (Schaible and Schmidt, 1988; Yaksh et al., 2001). It evokes ectopic activity when applied to a neuroma after axotomy (Michaelis et al., 1998) and increases neuronal activity (Ahmadi et al., 2001) and hyperalgesia (Minami et al., 1994) when administered spinally. Following injury to peripheral nerve, Wallerian degeneration occurs and results in an upregulation of several inflammatory mediators in the endoneurium, among them COX-2 (Ma and Eisenach, 2002; Marziniak and Sommer, 2000) and cytokines like tumor necrosis factor-alpha (TNF) (George et al., 1999). Endoneurial upregulation of COX-2 after nerve injury is TNF-dependent (Marziniak and Sommer, 2000). TNF itself is increased in nerve and dorsal root ganglia (DRG) after peripheral nerve injury (Schäfers et al., 2002), and its intraneural injection yields hyperalgesia (Wagner and Myers, 1996). In light of these observations, we investigated whether systemic administration of a nonselective (ibuprofen) or selective COX-2 inhibitor (celecoxib) altered pain-related behavior generated either by partial nerve injury or by intraneural injection of TNF in rats. PGE2 concentrations were measured in sciatic nerve, DRG and spinal cord to look for covariance between PGE2 increases at different sites and hyperalgesia and to determine if COX inhibition reduced injury-associated increases in PGE2 and pain behavior in parallel. Materials and methods Animals All experiments were approved by the Bavarian State and by the Institutional Animal Care and Use Committee of the University of California, San Diego and adhered to the guidelines for pain experiments with awake animals (Zimmermann, 1983). All behavioral studies were performed on rats (200 – 250 g Harlan Sprague – Dawley rats, Charles River, Sulzfeld, Germany) at the Department of Neurology, 161 University of Würzburg, in Germany. Tissue analysis was performed in separate sets of animals (200 –250 g Harlan Sprague – Dawley rats, Indianapolis, IN) at the Anesthesiology Research Laboratory, UCSD, San Diego. Rats were kept on a standard 14 h/10 h light/dark cycle with standard rat chow and water ad libitum in groups of three in plastic cages with soft bedding. Behavioral analysis and tissue harvest were carried out exclusively during the light phase (09:00 –14:00 h) of the cycle. Experimental paradigms Treatment for all rats consisted of twice daily oral gavage of ibuprofen, celecoxib or vehicle starting at different time points. Rats were regularly assessed for thermal hyperalgesia and mechanical allodynia. The studies were carried out in four phases. Phase 1: Rats were prepared with a chronic constrictive injury (CCI) of the sciatic nerve and received treatment beginning 2 days before surgery and lasting until 9 days after surgery. Phase 2: Rats were prepared with a CCI and treatment was initiated 12 h after surgery. Phase 3: Rats were prepared with CCI, those with ipsilateral mechanical withdrawal thresholds < 10 g on day 3 after CCI were included in this phase and received treatment from day 7 through day 16. Phase 4: Rats received intraneural injection of TNF, which evoked thermal hyperalgesia and mechanical allodynia. The effects of treatment beginning 2 days before surgery were assessed. Surgical procedures A CCI of the sciatic nerve modified after Bennett and Xie (1988) was performed on one sciatic nerve under deep pentobarbital anesthesia (50 mg/kg) as previously described (Sommer et al., 1995). Briefly, the sciatic nerve was exposed unilaterally at the mid-thigh level. Three ligatures (chromic gut 4.0) were placed around the nerve with 1-mm spacing. The ligatures were tied until they just constricted the diameter of the nerve and a transient twitch was seen in the hindlimb. A sham operation that just exposed the nerve was performed contralaterally. Intraneural injection Under deep pentobarbital anesthesia (50 mg/kg), the right sciatic nerve was exposed at the mid-thigh level. Recombinant rat TNF (R&D Systems) was dissolved in 0.1% bovine serum albumin/0.1 M phosphate buffer to a final concentration of 2.5 pg/ml. A total volume of 5 Al was injected into the mid-thigh sciatic nerve using a 30G needle and a Hamilton syringe. In preliminary experiments, this TNF concentration was found to reliably induce mechanical allodynia (data not shown). The diluent (0.1% bovine serum albumin/0.1 M phosphate buffer, 5 Al) was used as the vehicle control solution. All rats were carefully monitored upon awakening and no animal dis- 162 M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 Fig. 1. Treatment with the nonselective COX inhibitor ibuprofen and selective COX-2 inhibitor celecoxib prevents, but does not reverse, ipsilateral mechanical allodynia and thermal hyperalgesia in rats with chronic constrictive injury (CCI). (A, B) For preemptive treatment, continuous drug treatment started 2 days before CCI (bars). After CCI, withdrawal thresholds to von Frey hairs were reduced in all vehicle-treated animals indicating mechanical allodynia (A). A slight but significant attenuation of mechanical allodynia was observed in rats treated with ibuprofen (n = 8) on days 6 and 9 and in rats treated with celecoxib (n = 5) on day 9, but not in rats treated with vehicle (n = 9). Withdrawal latencies to heat were reduced in all rats after CCI, indicating thermal hyperalgesia (B). Thermal hyperalgesia was modestly attenuated in rats treated with celecoxib on days 6 and 9 and in rats treated with ibuprofen on days 3, 6 and 9 after CCI. (C, D) For early posttreatment, drugs were administered starting 12 h after placement of the CCI (bars). Most of the rats developed mechanical allodynia (C) and thermal hyperalgesia (D) by day 3 after CCI. In rats treated with either dose of celecoxib (n = 6 per dose) and ibuprofen (n = 11), but not vehicle (n = 11), mechanical allodynia and thermal hyperalgesia were significantly attenuated on days 6 and 9. (E, F) For late posttreatment, drugs were administered starting 7 days after CCI (bars). Treatment with either vehicle (n = 5), celecoxib (n = 5) or ibuprofen (n = 5) did not affect mechanical allodynia (E) or thermal hyperalgesia (F) at any time point tested. (*P < 0.05, **P < 0.005 vehicle- vs. drug-treated rats). M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 played overt signs of neurological damage (e.g. inverted or curled foot, limping). 163 adding 50 Al stop solution. Optical density (Dynex MRX Revelation plate reader, Chantilly, VA, USA) was read immediately at 405 nm with a correction at 570 nm. Behavioral tests Thermal hyperalgesia was assessed with an algesimeter (Ugo Basile, Comerio, Italy) as described previously (George et al., 2000). After accommodation to the testing apparatus, three consecutive thermal stimuli were applied to the rats’ hindpaws with at least a 1-min interval between paws. Means of latencies were calculated and used as a measure of the withdrawal threshold to heat. Mechanical sensitivity was assessed using von Frey hairs with hair values ranging from 1.65 to 5.46 and the up –down method (Chaplan et al., 1994; Dixon, 1965) was applied. The 50% probability withdrawal threshold (force of the von Frey hair to which an animal reacts to 50% of the presentations) was recorded. All behavioral testing was done by observers unaware of animal treatment. Drug treatments All drugs were delivered by gavage in a volume of 0.2 ml. Rats received either vehicle (0.9% saline), ibuprofen (IbuhexalR, Hexal, Holzkirchen, Germany, 40 mg/kg), or celecoxib (CelebrexR, Pharmacia, Erlangen, Germany, 10 or 30 mg/kg) every 12 h. Determination of PGE2 Sciatic nerves, L4 and L5 DRG, lumbar and cervical spinal cords were harvested from naı̈ve rats or rats with CCI on postoperative day 10 that had received drug treatment (vehicle, 10 mg/kg celecoxib, 40 mg/kg ibuprofen) starting 12 h after surgery. Since only ibuprofen, and not celecoxib, reduced CCI-induced increases in endoneurial PGE2 levels, we repeated the experiments and added another group, with celecoxib treatment at a higher dose (30 mg/kg). The L4 and L5 DRG of each side were pooled. Ipsilateral and contralateral spinal cords were analyzed separately. The tissue was snap frozen and stored at 70 jC until it was assayed for PGE2 using an enzyme-immunoassay (EIA, Assay Design, Ann Arbor, MI, USA) according to the manufacturer’s instructions. Briefly, tissues were placed in 0.2 ml cold (4jC) diluted buffer (supplied in the EIA kit), homogenized using a Polytron device and sonicated for 5 min. The homogenates were subsequently centrifuged (4jC, 1,000  g for 15 min), the supernates collected, probed with PGE2 EIA alkaline phosphatase conjugate and a monoclonal PGE2 EIA antibody and incubated at room temperature for 2 h at 500 rpm. After three washes with 200 Al wash buffer, 5 Al alkaline phosphatase conjugate was added with 200 Al p-Npp substrate solution and the mixture was incubated for 45 min without shaking until the reaction was stopped by Fig. 2. Mechanical allodynia induced by intraneural injection of TNF is only partially reduced by treatment with COX inhibitors: Rats were made hyperalgesic by a single intraneural injection of recombinant rat TNF (2.5 pg/ml); 0.1% BSA was used as the control solution (n = 3). Systemic treatment with vehicle (n = 3) or COX inhibitors celecoxib (n = 4) and ibuprofen (n = 3) was started 2 days before TNF injection (bars). (A) All rats with intraneural injection of TNF developed mechanical allodynia on day 1 after injection, which persisted until the end of the experiment in the vehicletreated group. Mechanical allodynia was attenuated in rats treated with ibuprofen on days 5 and 7 after intraneural TNF injection, but not in rats treated with celecoxib. (B) All rats with intraneural TNF injection developed thermal hyperalgesia, which was maximal on day 1 after injection. Thermal hyperalgesia induced by TNF injection was not influenced by either of the treatments. (*P < 0.05 vehicle- vs. drug-treated rats). 164 M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 Concentration of PGE2 was calculated from an eight-point standard curve. All experimental samples and PGE2 standards were run in duplicate. Sensitivity of the assay was 15.9 pg/ml. The assay cross-reacts 100% with PGE2, 70% with PGE1, 16.3% with PGE3 and 1.4% with PGF1a, but does not recognize PGF2a, 6-keto-PGF1a, PGA2, PGB1, thromboxane B2, 2-arachidonoylglycerol, anandamide or PGD2 ( < 1% at 39 –5000 pg/ml). post hoc tests was computed. A difference was accepted as significant if P < 0.05. Statistical analysis After CCI, withdrawal thresholds to von Frey hairs were reduced in all vehicle-treated animals indicating mechanical allodynia (Figs. 1A,C,E). Similarly, withdrawal latencies to heat were reduced after CCI in all rats treated with vehicle, indicating thermal hyperalgesia (Figs. 1B,D,F). Preemptive treatment with ibuprofen and celecoxib started 2 days before CCI resulted in a small, but significant, Data are given as mean F standard error. For comparison of change of hyperalgesia, a two-way ANOVA for repeated measures with Fisher PLSD post hoc tests was used. For comparison of pain-related behavior and EIA data between treatment groups, a one-way ANOVA with Fisher PLSD Results Systemic COX inhibition prevents, but does not reverse, CCI-evoked mechanical allodynia and thermal hyperalgesia Fig. 3. PGE2 levels are increased in the injured nerve and ipsilateral lumbar DRG. Systemic treatment with ibuprofen reduces PGE2 levels in injured nerves and lumbar DRG, whereas celecoxib reduces only endoneurial PGE2 levels: Tissue (sciatic nerve, L4 and L5 DRG, lumbar and cervical spinal cord) from naı̈ve rats and rats with drug treatment starting 12 h after CCI was harvested on day 10 after surgery and analyzed by EIA for PGE2 levels. (A) In the injured nerve of vehicle (n = 6) and low-dose celecoxib (n = 6) treated rats, PGE2 levels were increased compared to the naı̈ve rats (n = 6) and the sham-operated contralateral side. Ibuprofen (n = 6) and high-dose celecoxib (n = 6) treatment significantly reduced ipsilateral PGE2 levels compared to vehicle treatment. (B) In ipsilateral DRG, PGE2 levels were increased compared to DRG from naı̈ve rats. In ibuprofen-, but not in celecoxib-treated rats, PGE2 levels were reduced in ipsi- and contralateral DRG. (*P < 0.05, **P < 0.005, ***P < 0.001, ****P < 0.0005 vehicle- vs. drug-treated rats, +P < 0.05, + +P < 0.005 vs. naı̈ve controls). M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 attenuation of mechanical allodynia (on days 6– 9 and day 9, respectively, Fig. 1A). Thermal hyperalgesia was modestly attenuated in rats treated with celecoxib (here 10 mg/ kg) on day 6 and 9 and ibuprofen on days 3, 6 and 9 (Fig. 1B). Thus, preemptive treatment with either a nonselective COX inhibitor or a selective COX-2 inhibitor modestly, but consistently, attenuated thermal hyperalgesia and mechanical allodynia after CCI. Early postoperative treatment with ibuprofen and celecoxib (10 or 30 mg/kg) starting 12 h after CCI attenuated mechanical allodynia (Fig. 1C) and thermal hyperalgesia (Fig. 1D) on days 6 and 9. There was no difference between rats treated with either dose of celecoxib or ibuprofen. In a separate group of CCI rats, the development of painrelated behavior was monitored and rats received drug treatment starting on day 7. There was no difference between saline-treated rats and rats treated with a COX inhibitor at any time point tested (Figs. 1E,F). Thus, injuryinduced pain behavior, once established, was not attenuated by treatment with COX inhibitors. Treatment with COX inhibitors partially reduces thermal hyperalgesia and mechanical allodynia induced by intraneural injection of TNF Hyperalgesia was induced by intraneural injection of 5 Al of a 2.5 pg/ml solution of recombinant rat TNF. Rats were continuously treated starting 2 days before injection with either saline, celecoxib (10 mg/kg) or ibuprofen (40 mg/kg) twice daily. Mechanical allodynia was attenuated by ibuprofen, but not celecoxib treatment (Fig. 2A). Thermal hyperalgesia induced by the TNF injection was not influenced by any treatment (Fig. 2B). PGE2 levels are increased in the injured nerve and reduced by ibuprofen and celecoxib Tissue from rats with treatment starting 12 h after CCI was harvested on day 10 after surgery and analyzed for PGE2. In the injured nerve of vehicle and low-dose celecoxib-treated rats, PGE2 levels were increased compared to both naı̈ve and sham-operated controls (Fig. 3A). PGE2 levels were not significantly different between naı̈ve controls and the sham-operated contralateral nerves. Ibuprofen and high-dose celecoxib treatment significantly reduced PGE2 levels in injured nerves compared to vehicle treatment. Ibuprofen treatment also reduced PGE2 levels in contralateral nerves compared to contralateral nerves of vehicle-treated rats. In ipsilateral L4 and L5 DRG of vehicle-treated CCI rats, PGE2 levels were increased compared to DRGs from naı̈ve control rats. In ibuprofen-treated rats, PGE2 levels were reduced in DRG of both sides compared to vehicle treatment (Fig. 3B). Although celecoxib treatment did not significantly reduce CCI-induced increases in PGE2 levels compared to vehicle, the DRG content of PGE2 was not significantly 165 elevated compared to naı̈ve controls. In lumbar and cervical spinal cord, PGE2 levels were not different among treatment groups (data not shown). Discussion COX inhibition partially prevents hyperalgesia induced by CCI or intraneural TNF injection In the present study, loose ligation of the sciatic nerve or intraneural injection of TNF resulted in tactile and thermal hyperalgesia. Treatment with a nonselective COX or a selective COX-2 inhibitor beginning before injury or at 12 h after injury modestly, but consistently, attenuated thermal hyperalgesia and mechanical allodynia after CCI. If treatment was initiated on day 7 after CCI, when altered pain thresholds were established, there was no effect of the drugs on thermal and mechanical withdrawal thresholds. There are several possible explanations for these timedependent effects. First, pretreatment and early posttreatment after CCI may inhibit the initial local COX activity in the endoneurium and thus reduce later inflammation and plasma extravasation induced by the ligatures. Therefore, injury and consequently hyperalgesia could be attenuated due to reduced swelling and self-strangulation of the nerve. Secondly, endogenous COX-2 levels are increased after nerve injury not only in the endoneurium (Ma and Eisenach, 2002; Marziniak and Sommer, 2000), but also in the spinal cord dorsal horn (Zhao et al., 2000). Interestingly, this spinal increase is only transiently observed 1 day after nerve injury and not at later time points. Thus, if blockade of this transient spinal COX-2 peak is the mechanism by which COX inhibition 12 h after CCI reduces hyperalgesia, it implies that early spinal COX upregulation (Zhao et al., 2000) is a significant initiator of later pain behavior. Indeed, in the present study, COX inhibition beginning at later time points when spinal PGE2 levels are no longer increased were ineffective. Our results are consistent with previous reports that suggest that systemic, intrathecal or subcutaneous applications of COX inhibitors can moderately attenuate injuryinduced thermal hyperalgesia and tactile allodynia (Syriatowicz et al., 1999; Zhao et al., 2000). The twice-daily treatment regimen used in the present study should result in sufficient plasma levels of celexocib to inhibit tissue PGs, since celecoxib has a mean half-life of f 11– 14 h in rats (Paulson et al., 2000). Considering the relatively high celecoxib doses used in the present study, it is unlikely that there has been a ‘breakthrough’ COX-2 dependent PGE2 production. For ibuprofen, a breakthrough PGE2 production between doses cannot be fully excluded since the mean halflife in rats is f3 h (Shah and Jung, 1987). However, in the present study, PGE2 levels were markedly reduced after ibuprofen. In addition, other drugs such as ketorolac with comparable short half-lives ( f5 h) also show prolonged 166 M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 (>10 days) antiallodynic effects even after a single application (Ma and Eisenach, 2002). The only moderate antihyperalgesic effect is in accord with the typical lack of efficacy of NSAIDs in neuropathic pain in humans. Others have reported no effect of systemic administration of COX inhibitors in animal models of neuropathic pain, but pronounced effects have been associated with their local administration into the injured nerve itself or into the hyperalgesic hind paw (Syriatowicz et al., 1999). The effects of intrathecal COX inhibitors are controversial. The nonselective COX inhibitor indomethacin reduced tactile allodynia when given early (day 1) but not late (day 14) after nerve injury (Zhao et al., 2000). Intrathecal ketorolac, a nonselective, but COX-1-preferring inhibitor, reversed tactile allodynia when given 3 weeks after nerve lesion but did not prevent it when given at the time of surgery (Ma and Eisenach, 2002). COX inhibitors seem to be efficacious in models of radiculopathy. In a model of nerve root injury by local nucleus pulposus application, mechanical but not thermal hyperalgesia was reduced by a COX-2 inhibitor given by epidural injection (Kawakami et al., 2002). In another radiculopathy model, intrathecal COX-2 inhibition was much more effective than systemic inhibitors in reducing mechanical allodynia (Deleo et al., 2000). Thus, the route of administration and degree of CNS penetration may be an important factor for the antiallodynic effect of COX inhibitors. Increased PGE2 concentrations in the injured nerve after CCI Here we show for the first time that PGE2 is elevated in the injured nerve. Our findings correspond to the increase of COX-2 immunoreactive cell profiles seen in the ipsilateral nerve between 2 and 4 weeks after nerve injury (Ma and Eisenach, 2002). Interestingly, both selective and nonselective COX inhibition reduced PGE2 levels in injured nerves, nonselective COX inhibition additionally reduced PGE2 levels in sham-operated nerves compared to naı̈ve controls. PGE2 levels were also increased in ipsilateral DRG of CCI rats. Only nonselective COX inhibition, but not selective COX-2 inhibition, significantly reduced PGE2 levels compared to vehicletreated CCI rats, which may suggest a predominant role of COX-1 in regulating PGE2 levels in DRG. In contrast to sciatic nerve, we did not find a PGE2 increase in spinal cord 10 days after CCI. This is in accordance with previous studies reporting unchanged spinal COX-2 mRNA or protein expression at comparable time points several days after nerve injury (Ichitani et al., 1997; Zhao et al., 2000). It remains unclear if spinal PGE2 might be increased at earlier time points, in parallel to the early and transient spinal upregulation of COX-2 previously reported (Zhao et al., 2000). Furthermore, other prosta- noids were not measured and they may also play a role. Prostacyclin (PGI2), for example, has been shown to induce ectopic activity in DRG neurons (Omana-Zapata and Bley, 2001). The locally administered drugs may reduce pain through additional mechanisms independent of COX inhibition. Such effects have occasionally been described before, for example meloxicam, but not indomethacin, inhibits baseline reflexes and wind-up in spinal cord preparations, such that the spinal antinociceptive actions of meloxicam could not be explained by COX inhibition (Lopez-Garcia and Laird, 1998). COX and TNF Like prostaglandins, proinflammatory cytokines are mediators involved in the generation of pain (Sommer, 2001). Specifically, endogenous TNF mediates pain-related behavior in nerve injury models (Sommer et al., 1998, 2001), topical application of exogenous TNF on DRG cells excites DRG neurons in vitro (Liu et al., 2002; Schäfers et al., 2003) and induces pain-related behavior in vivo (Homma et al., 2002; Schäfers et al., 2003). Intraneural injection of TNF produces hyperalgesia (Wagner and Myers, 1996). Since the expression of COX-2 is dependent on TNF (Marziniak and Sommer, 2000), we speculated that hyperalgesia and allodynia induced by intraneural TNF injection might be mediated by prostaglandins via the induction of COX-2 and might resolve upon COX inhibition. However, neither a COX-2 selective nor a nonselective COX inhibitor had an effect on TNF-induced thermal hyperalgesia. Only the nonselective COX inhibitor ibuprofen modestly decreased mechanical allodynia. This is in accord with the recent observation that indomethacin co-injected with TNF into the plantar skin only incompletely reduces pain behavior (Sachs et al., 2002) and indomethacin only partially inhibits TNFevoked responses in DRG neurons in vitro (Liu et al., 2002). It is thus unlikely that TNF produces neuropathic pain behavior exclusively via prostaglandins. Many possible mechanisms by which TNF may enhance nerve injury-induced pain have been postulated (for review, see Sommer, 2001). Among these mechanisms are direct sensitization of neurons (Liu et al., 2002; Schäfers et al., 2003; Zhang et al., 2002), increased small afferent transmitter (CGRP) release (Opree and Kress, 2000), and/or anterograde transport to innervated areas (Schäfers et al., 2002). The present data indicate, however, that COX-2 induction is not a major factor in TNF-induced neuropathic pain. In conclusion, we show a consistent but modest effect of systemically administered COX inhibitors on pain-related behavior in CCI. In contrast to the marked inhibition of nerve-injury-induced endoneurial upregulation of PGE2 by the drugs, the effect on hyperalgesia and allodynia was limited. Monotherapy with COX inhibitors in the treatment of neuropathic pain seems to be insufficient. However, M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 combined treatment regimes with other analgesic drugs such as morphine as used in clinical studies (Matoba, 2001) or gabapentin might be useful. Acknowledgments We are grateful to Lydia Biko, Barbara Dekant and Alan Moore for expert technical assistance. We thank Klaus V. Toyka for his contributions in the planning process and for helpful suggestions during the preparation of the manuscript. This study was supported by Deutsche Forschungsgemeinschaft SCHA 924/1-1 (M.S.), Volkswagenstiftung I/ 75 833 (C.S.) and NIH NS35630 (L.S.S.). References Ahmadi, S., Lippross, S., Neuhuber, W.L., Zeilhofer, H.U., 2001. PGE2 selectively blocks inhibitory glycinergic neurotransmission onto rat superficial dorsal horn neurons. Nat. Neurosci. 5, 34 – 40. Beiche, F., Klein, T., Nusing, R., Neuhuber, W., Goppelt-Struebe, M., 1998. Localization of cyclooxygenase-2 and prostaglandin E2 receptor EP3 in the rat lumbar spinal cord. J. Neuroimmunol. 89, 26 – 34. Bennett, G.J., Xie, Y.K., 1988. A peripheral mononeuropathy in rat that produces disorders of pain sensation like those seen in man. Pain 33, 87 – 107. Cao, C., Matsumura, K., Yamagata, K., Watanabe, Y., 1998. Cyclooxygenase-2 is induced in brain blood vessels during fever evoked by peripheral or central administration of tumor necrosis factor. Brain Res. Mol. Brain Res. 56, 45 – 56. Chaplan, S.R., Bach, F.W., Pogrel, J.W., Chung, J.M., Yaksh, T.L., 1994. Quantitative assessment of tactile allodynia in the rat paw. J. Neurosci. Methods 53, 55 – 63. Cohen, K.L., Harris, S., 1987. Efficacy and safety of nonsteroidal antiinflammatory drugs in the therapy of diabetic neuropathy. Arch. Intern. Med. 147, 1442 – 1444. Deleo, T.A., Hashizume, H., Rutkowski, M.D., Weinstein, T.N., 2000. Cyclooxygenase-2 inhibitor SC-236 attenuates mechanical allodynia following nerve root injury in rats. J. Orthop. Res. 18, 977 – 982. Dixon, W., 1965. The up-and-down method for small samples. J. Am. Stat. Assoc. 60, 967 – 978. George, A., Schmidt, C., Weishaupt, A., Toyka, K.V., Sommer, C., 1999. Serial determination of tumor necrosis factor-alpha content in rat sciatic nerve after chronic constriction injury. Exp. Neurol. 160, 124 – 132. George, A., Marziniak, M., Schäfers, M., Toyka, K.V., Sommer, C., 2000. Thalidomide treatment in chronic constrictive neuropathy decreases endoneurial tumor necrosis factor-alpha, increases interleukin-10 and has long-term effects on spinal cord dorsal horn met-enkephalin. Pain 88, 267 – 275. Homma, Y., Brull, S.J., Zhang, J.M., 2002. A comparison of chronic pain behavior following local application of tumor necrosis factor-alpha to the normal, mechanically compressed lumbar ganglia in the rat. Pain 95, 239 – 246. Ichitani, Y., Shi, T., Haeggstrom, J.Z., Samuelsson, B., Hokfelt, T., 1997. Increased levels of cyclooxygenase-2 mRNA in the rat spinal cord after peripheral inflammation: an in situ hybridization study. NeuroReport 8, 2949 – 2952. Kawakami, M., Matsumoto, T., Hashizume, H., Kuribayashi, K., Tamaki, T., 2002. Epidural injection of cyclooxygenase-2-inhibitor attenuates pain-related behavior following application of nucleus pulposus to the nerve root in the rat. J. Orthop. Res. 20, 376 – 381. Lacroix, S., Rivest, S., 1998. Effect of acute systemic inflammatory re- 167 sponse and cytokines on the transcription of the genes encoding cyclooxygenase enzymes (COX-1 and COX-2) in the rat brain. J. Neurochem. 70, 452 – 466. Lashbrook, J.M., Ossipov, M.H., Hunter, J.C., Raffa, R.B., Tallarida, R.J., Porreca, F., 1999. Synergistic antiallodynic effects of spinal morphine with ketorolac and selective COX-1- and COX-2-inhibitors in nerveinjured rats. Pain 82, 65 – 72. Liu, B., Li, H., Brull, S.J., Zhang, J.M., 2002. Increased sensitivity of sensory neurons to tumor necrosis factor-alpha in rats with chronic compression of the lumbar ganglia. J. Neurophysiol. 88, 1393 – 1399. Lopez-Garcia, J.A., Laird, J.M., 1998. Central antinociceptive effects of meloxicam on rat spinal cord in vitro. NeuroReport 9, 647 – 651. Ma, W., Eisenach, J.C., 2002. Morphological and pharmacological evidence for the role of peripheral prostaglandins in the pathogenesis of neuropathic pain. Eur. J. Neurosci. 15, 1037 – 1047. Ma, W., Du, W., Eisenach, J.C., 2002. Role for both spinal cord COX-1 and COX-2 in maintenance of mechanical hypersensitivity following peripheral nerve injury. Brain Res. 937, 94 – 99. Malmberg, A.B., Yaksh, T.L., 1992. Hyperalgesia mediated by spinal glutamate or substance P receptor blocked by spinal cyclooxygenase-inhibition. Science 257, 1276 – 1279. Marziniak, M., Sommer, C., 2000. Upregulation of cyclooxygenase-2 is dependent on tumor necrosis factor receptor 1 and 2 after chronic constriction injury. J. Neurol. 247 (III), 125. Matoba, M., 2001. Use of morphine and adjuvant drugs according to the condition. Eur. J. Pain 5, 59 – 62. Max, M.B., Schafer, S.C., Culnane, M., Dubner, R., Gracely, R.H., 1988. Association of pain relief with drug side effects in postherpetic neuralgia: a single-dose study of clonidine, codeine, ibuprofen, and placebo. Clin. Pharmacol. Ther. 43, 363 – 371. Michaelis, M., Vogel, C., Blenk, K.H., Arnarson, A., Jänig, W., 1998. Inflammatory mediators sensitize acutely axotomized nerve fibers to mechanical stimulation in the rat. J. Neurosci. 18, 7581 – 7587. Minami, T., Uda, R., Horiguchi, S., Ito, S., Hyodo, M., Hayaishi, O., 1994. Allodynia evoked by intrathecal administration of prostaglandin E2 to conscious mice. Pain 57, 217 – 223. Omana-Zapata, I., Bley, K.R., 2001. A stable prostacyclin analog enhances ectopic activity in rat sensory neurons following neuropathic injury. Brain Res. 904, 85 – 92. Opree, A., Kress, M., 2000. Involvement of the proinflammatory cytokines tumor necrosis factor-alpha, IL-1beta, and IL-6 but not IL-8 in the development of heat hyperalgesia: effects on heat-evoked calcitonin gene-related peptide release from rat skin. J. Neurosci. 20, 6289 – 6293. Paulson, S.K., Zhang, J.Y., Breau, A.P., Hribar, J.D., Liu, N.W., Jessen, S.M., Lawal, Y.M., Cogburn, J.N., Gresk, C.J., Markos, C.S., Maziasz, T.J., Schoenhard, G.L., Burton, E.G., 2000. Pharmacokinetics, tissue distribution, metabolism, and excretion of celecoxib in rats. Drug Metab. Dispos. 28, 514 – 521. Sachs, D., Cunha, F.C., Poole, S., Ferreira, S.H., 2002. Tumor necrosis factor-alpha, interleukin-1beta and interleukin-8 induce persistent mechanical nociceptor hypersensitivity. Pain 96, 89 – 97. Samad, T.A., Moore, K.A., Sapirstein, A., Billet, S., Allchorne, A., Poole, S., Bonventre, J.V., Woolf, C.J., 2001. Interleukin-1beta-mediated induction of COX-2 in the CNS contributes to inflammatory pain hypersensitivity. Nature 410, 471 – 475. Schäfers, M., Geis, C., Brors, D., Yaksh, T., Sommer, C., 2002. Anterograde transport of tumor necrosis factor-alpha in experimental mononeuropathy. J. Neurosci. 22, 236 – 245. Schäfers, M., Lee, D.H., Brors, D., Yaksh, T.L., Sorkin, L.S.S., 2003. Increased sensitivity of injured and adjacent uninjured rat primary sensory neurons to exogenous tumor necrosis factor-alpha after spinal nerve ligation. J. Neurosci. 23, 3028 – 3038. Schaible, H.G., Schmidt, R.F., 1988. Excitation and sensitization of fine articular afferents from cat’s knee joint by prostaglandin E2. J. Physiol. 403, 91 – 104. Shah, A., Jung, D., 1987. Dose-dependent pharmacokinetics of ibuprofen in the rat. Drug Metab. Dispos. 15, 151 – 154. 168 M. Schäfers et al. / Experimental Neurology 185 (2004) 160–168 Sommer, C., 2001. Cytokines and neuropathic pain. In: Hansson, P., Fields, H., Hill, R., Marchettini, P. (Eds.), Neuropathic Pain: Pathophysiology and Treatment. IASP Press, Seattle, WA, pp. 37 – 62. Sommer, C., Lalonde, A., Heckman, H.M., Rodriguez, M., Myers, R.R., 1995. Quantitative neuropathology of a focal nerve injury causing hyperalgesia. J. Neuropathol. Exp. Neurol. 54, 635 – 643. Sommer, C., Schmidt, C., George, A., 1998. Hyperalgesia in experimental neuropathy is dependent on the TNF receptor 1. Exp. Neurol. 151, 138 – 142. Sommer, C., Lindenlaub, T., Teuteberg, P., Schäfers, M., Hartung, T., Toyka, K.V., 2001. Anti-TNF-neutralizing antibodies reduce pain-related behavior in two different mouse models of painful mononeuropathy. Brain Res. 913, 86 – 89. Syriatowicz, J.P., Hu, D., Walker, J.S., Tracey, D.J., 1999. Hyperalgesia due to nerve injury: role of prostaglandins. Neuroscience 94, 587 – 594. Taiwo, Y.O., Levine, J.D., 1988. Prostaglandins inhibit endogenous pain control mechanisms by blocking transmission at spinal noradrenergic synapses. J. Neurosci. 8, 1346 – 1349. Wagner, R., Myers, R.R., 1996. Endoneurial injection of TNF-alpha produces neuropathic pain behaviors. NeuroReport 7, 2897 – 2901. Weber, H., Holme, I., Amlie, E., 1993. The natural course of acute sciatica with nerve root symptoms in a double-blind placebo-controlled trial evaluating the effect of piroxicam. Spine 18, 1433 – 1438. Yaksh, T.L., Dirig, D.M., Conway, C.M., Svensson, C., Luo, D.Z., Isakson, P.C., 2001. The acute antihyperalgesic action of nonsteriodal, anti-inflammatory drugs and release of spinal prostaglandin E2 is mediated by the inhibition of constitutive spinal cyclooxygenase-2 (COX-2), but not COX-1. J. Neurosci. 21, 5847 – 5853. Zhang, J.M., Li, H., Liu, B., Brull, S.J., 2002. Acute topical application of tumor necrosis factor-alpha evokes protein kinase A-dependent responses in rat sensory neurons. J. Neurophysiol. 88, 1387 – 1392. Zhao, Z., Chen, S.R., Eisenach, J.C., Busija, D.W., Pan, H.L., 2000. Spinal cyclooxygenase-2 is involved in development of allodynia after nerve injury in rats. Neuroscience 97, 743 – 748. Zimmermann, M., 1983. Ethical guidelines for investigations of experimental pain in conscious animals. Pain 16, 109 – 110.