The authors sought to characterize the pharmacologic characteristic and site of action of gabapentin (Neurontin) in a model of thermal hyperalgesia induced by intrathecal substance P administration.
Rats were prepared with long-term lumbar intrathecal catheters. Hind paw withdrawal latency was determined using a radiant heat stimulus focused through a glass surface onto the plantar surface of the paw.
Within 5 min after intrathecal injection of substance P (30 nmol), hind paw withdrawal latency fell from 11 to 8 s. Gabapentin given intrathecally or intraperitoneally produced dose-dependent reversal of the thermal hyperalgesia, with complete reversal (ED100) occurring at 163 microg for intrathecal and 185 mg/kg for intraperitoneal administration. S(+)-3-isobutyl-gamma aminobutyric acid, but not R(-)-3-isobutyl-gamma aminobutyric acid, also produced dose-dependent reversal of the intrathecal substance P-induced thermal hyperalgesia (intrathecal ED100, 65 microg and intraperitonal ED100, 31 mg/kg). The effects of intraperitoneally administered gabapentin and 3-isobutyl-gamma aminobutyric acid were reversed by intrathecal pretreatment with D-serine (100 microg) but not by L-serine. All effects were observed at doses that had little effect on motor function or spontaneous activity. Intrathecal N-methyl-D-aspartate (2 nmol) induced thermal hyperalgesia, which was blocked by gabapentin (100 mg/kg intraperitoneally) and S(+)-3-isobutyl-gamma aminobutyric acid (30 mg/kg intraperitoneally).
The structure-activity relationship and the stereospecificity noted after intrathecal delivery suggest that gabapentin and S(+)-3-isobutyl-gamma aminobutyric acid act at a common spinal locus to modulate selectively a facilitated state of nociceptive processing.
Gabapentin, first synthesized as a gamma-aminobutyric acid (GABA) analogue with improved biologic stability and distribution, has been developed for clinical use as an anticonvulsant agent. Several recent clinical case reports have suggested that gabapentin also is efficacious in treating human neuropathic pain. [2,3]Subsequent preclinical studies demonstrated that this agent, when given intrathecally, is able to reverse tactile allodyniadose in a dose-dependent manner in the Chung model of neuropathy and to reverse thermal hyperalgesia in the Bennett model of neuropathy after systemic and intrathecal administration, the second phase of the formalin test after systemic delivery (Singh, personal communication, 1996), and the thermal hyperalgesia induced by thermal injury to the paw (Jun and Yaksh, unpublished data, 1997). Importantly, these effects have been observed at doses that appear to have little effect on behavior or motor function.
Functionally, the actions of gabapentin appear limited to those models of nociception that involve a facilitated state of processing. No changes in escape latencies have been seen with this agent in animal models using only short-term noxious stimuli, such as the hot plate or tail flick tests; effects have been manifested principally in the context of injury or inflammation. 
Many such models of facilitated processing are known to depend on an increase in the spinal release of glutamate and the subsequent activation of spinal N-methyl-D-aspartate (NMDA) receptors. [6,7]Gabapentin has been shown to have moderate inhibitory effects in vitro on branched-chain amino acid transferase (an enzyme that metabolizes several cytosolic amino acids to form glutamate). Although some data suggest that gabapentin may reduce NMDA-evoked currents,* other studies have not shown effects at drug concentrations achieved in vivo. Moreover, binding studies have reportedly shown no affinity of gabapentin for NMDA, alpha-amino-3-hydroxy-5-methylisoxazole-4-propionic acid hydrobromide (AMPA) or nonstrychnine-sensitive glycine sites associated with the NMDA receptor. It is noteworthy, however, that despite an apparent lack of affinity for the NMDA receptor complex, it has been reported that D-serine, an agent known to reverse the effects of drugs acting as antagonists at the nonstrychnine-sensitive glycine site, antagonizes the anticonvulsant effects of gabapentin. 
Although gabapentin is structurally similar to GABA, binding studies fail to show any affinity for either GABA-A or -B sites. The compound has been shown to increase the rate of synthesis of GABA in several brain regions, however. In addition, although gabapentin itself has no effect on the GABA transporter, it may enhance extracellular levels of GABA released from astrocytes by reversal of the GABA transporter. In contrast, in studies on the Chung model of neuropathy, the spinal delivery of either GABA-A or -B receptor antagonists, at doses capable of blocking the intrathecal effects of the respective agonists, did not reduce the antiallodynic effects of intrathecal gabapentin. These observations jointly appear to exclude an anticipated action at a conventional GABA site. 
Despite a lack of interaction at many sites, there appears to be a high-affinity binding site in the brain for gabapentin. Importantly, this gabapentin binding is displaced in a stereospecific fashion by S(+)-3-isobutyl-GABA but not by R(-)-3-isobutyl-GABA, and these agents show a similar activity profile as anticonvulsant agents. [15,16]
Although a supraspinal effect is not excluded, it is clear that the effects of gabapentin can be mediated by a spinal action. The dose required after spinal delivery is considerably lower than that required after systemic delivery; therefore, peripheral redistribution after spinal delivery appears unlikely to account for the antihyperalgesia conferred by this route of administration.
In the current studies, we sought to further characterize the actions of this family of agents by examining the dose-dependent spinal and systemic effects of (1) gabapentin, (2) the isomers of 3-isobutyl-GABA, and (3) D- and L-serine on the thermal hyperalgesia observed after the spinal delivery of substance P. This model was chosen because it reflects the activation of a spinal facilitatory cascade by a neurokinin (NK)-1 receptor leading to a behaviorally defined hyperalgesia that does not depend on a previous peripheral injury. We further examined in a limited number of studies whether the hyperalgesia induced by intrathecal NMDA was similarly sensitive to the effects of gabapentin.
Materials and Methods
All animal surgery and testing procedures were approved by the Institutional Animal Care Committee of the University of California, San Diego (San Diego, CA).
Male Holtzmann Sprague-Dawley rats, weighing 300–400 g each, were housed separately after catheter implantation. They were allowed access to standard rat chow and water ad libitum and were maintained on a 12-h light/dark cycle. Rats were tested up to four times at 4–6-day intervals. Approximately 55 rats were used in this study.
The surgical implantation of intrathecal catheters was performed under halothane anesthesia using a modification of the previously described technique. Briefly, each rat was anesthetized under a mixture of 1–4% halothane in 50% O2/air and shaved along the occiput and neck. The rat was then placed in a stereotaxic head holder, and the shaved skin was prepared with alcohol and Povidone iodine solution. To maintain anesthesia, 2% halothane was delivered through a face mask. An incision was made overlying the atlantooccipital junction, and the dura mater was exposed by blunt dissection. An incision was made in the dura, and a polyethylene catheter (PE-10), with a loose knot cemented with dental acrylic 8.4 cm from the end, was threaded caudally through the incision. The external end of the catheter was tunneled subcutaneously to exit atop the head. The catheter was then flushed with 10 micro liter of saline and plugged. The animals were allowed to recover for approximately 5 days after surgery before testing was begun. Motor function, body weight, and sensory threshold were assessed before testing, and those animals that showed impairment were killed.
Drugs administered intrathecally in this study were substance P (molecular weight = 1,347; Peninsula Laboratories, Belmont, CA); NMDA (molecular weight = 147; Sigma Chemical, St. Louis, MO), gabapentin (1-[aminomethyl]cyclohexanacetic acid [Neurontin [registered sign]]; molecular weight = 171; Parke-Davis, Ann Arbor, MI), S(+)-3-isobutyl-GABA (molecular weight = 157; Parke-Davis), and D-serine and L-serine (molecular weight = 105; Sigma Chemical). Stock solutions of substance P or NMDA were prepared in deionized water, lyophilized, stored at -70 [degree sign] Celsius, and used immediately after reconstitution. Gabapentin was stored at 5 [degree sign] Celsius in an opaque container. All drugs were dissolved in 0.9% sterile saline. Control experiments were performed with normal saline at volumes comparable to those of the drug solutions and injected intrathecally, intraperitoneally, or both, as required.
For intrathecal injection, the catheter was unplugged and connected to a gear-driven microinjection syringe via a length of calibrated PE-90 tubing. Drugs were injected in a 10-micro liter solution, followed by an additional 10 micro liter of saline flush.
For intraperitoneal injection, drugs were injected in a volume of 3 ml/kg. The same volume of normal saline was administered as a control. All drug doses were randomized.
Thermal Escape Latencies. The latency to hind paw withdrawal from a thermal stimulus was determined by exposing the plantar hind paw to radiant heat using a modified Hargreaves-type thermal testing device. The apparatus used (University Anesthesiology Research and Development Group, University of California, San Diego) has been described elsewhere. Briefly, rats were placed in individual enclosures on a glass plate maintained at 30 [degree sign] Celsius. The radiant thermal stimulus, underneath the glass plate, was positioned directly under the hind paw. Activation of the bulb simultaneously activated a timer; both bulb and timer were immediately turned off by paw withdrawal or after 20 s (cutoff time). After acclimation, a measurement was taken for each hind paw to establish an average baseline latency (counted as time = 0). Measurements were then made at intervals after injection of intrathecal substance P. The mean of the response latencies from each paw was taken as the latency for that testing time. Data are presented as the escape latency in seconds. For analysis, the area under the time versus change-in-escape latency curve (AUC) was calculated between the intervals of 10 and 60 min after intrathecal substance P. The AUC was calculated using a program that computed the sum of the areas of a series of trapezoids created by the time-effect curves. The dimensions of the AUC are thus Delta s [center dot] h. As the response latencies were reduced during this interval, the AUC yielded a negative area. This value was used to define the magnitude of the thermal hyperalgesia.
General Behavior. Pinna twitching in response to touching a small probe to the meatus, blink in response to lightly touching the cornea, and the righting reflex (coordinated response made when the animal is placed on its back) were tested periodically to evaluate possible sedative effects.
Raw hyperalgesia data are presented as the response latency in seconds (mean +/- SEM). For statistical analysis of the magnitude of the hyperalgesia, the AUC was determined for the interval between 10 and 60 min after the injection of substance P; the 10-min delay was used to avoid including the initial increase in response latency that was typically observed immediately after injection of substance P. The AUC was used to construct dose-response curves with the computer programs of Tallarida and Murray. These programs also were used to verify the parallelism of the dose-response curves. In the calculation of the complete reversal (ED100), a complete response was defined as the dose of drug that returned the post-substance P response latency to the same value as the pre-substance P response latency. Comparisons of potency were made using ED100values. Comparisons between groups were accomplished with repeated-measures or one-way analysis of variance. When significance was achieved, post hoc comparisons were performed using the Bonferroni-Dunn method. These statistical analyses were performed with Statview 4.5 for the Macintosh (Abacus Concepts, Inc., Berkeley, CA).
Intraperitoneal Gabapentin and Intraperitoneal S(+)-3-Isobutyl GABA
Intraperitoneal doses of gabapentin and S(+)-3-isobutyl GABA up to 100 mg/kg and 30 mg/kg, respectively, did not change thermal response latencies and were without significant side effects (Figure 1). The intraperitoneal injection of 300 mg/kg of gabapentin caused some mild sedation and motor weakness. At this dose, there was a small increase in baseline thermal escape latency. Mortality was not observed, however, even at doses of gabapentin up to 500 mg/kg.
Effects of Intrathecal Substance P
In the absence of treatment, the mean latency to hind paw withdrawal was 11.3 +/- 0.4 s (n = 50). The intrathecal injection of substance P (30 nmol) resulted in a brief period of agitation, with scratching of the flank and licking of the paws. Concurrent with this behavior, there was a brief increase in the thermal escape threshold that lasted approximately 10 min. By 30 min, the thermal escape latency had decreased relative to the pre-substance P latency to 7.7 +/- 0.3 s (Delta s = 3.6 +/- 0.3 s). This reduced latency persisted for [nearly =] 40–50 min (Figure 2).
Intraperitoneal Gabapentin and 3-Isobutyl-GABA with Substance P
The intraperitoneal injection of either gabapentin (10–100 mg/kg;Figure 2) or S(+)-3-isobutyl-GABA (1–30 mg/kg;Figure 3) resulted in a dose-dependent reversal of the intrathecal substance P-induced lowering of the thermal escape latencies (Figure 2). Slopes of the respective dose-effect curves were statistically different from 0 (P < 0.05) and were not significantly different from parallel. Calculated ED100doses (and 95% confidence intervals) for intraperitoneal gabapentin and S(+)-3-isobutyl GABA were 184 (91–363) and 31 (19–49) mg/kg, respectively.
Duration of Action of Intraperitoneal Gabapentin
To determine the time course of drug action, the just maximally effective dose of gabapentin (100 mg/kg intraperitoneally) was injected at intervals before and after intrathecal substance P. As indicated, injections at intervals at short as 2 min and up to 60 min but not 180 min before intrathecal substance P produced a significant reduction in the thermal hyperalgesia (Figure 4). Similarly, gabapentin given 10 min after treatment also resulted in a complete reversal of the developed thermal hyperalgesia.
Intrathecal Gabapentin and 3-Isobutyl-GABA
The intrathecal delivery of gabapentin (30–300 micro gram) and S(+)-3-isobutyl-GABA (1–30 micro gram) but not R(-)-3-isobutyl-GABA (30 micro gram)(Figure 5and Figure 6) produced a significant dose-dependent reversal of the thermal hyperalgesia. By themselves, however, these compounds caused no significant change from baseline (Figure 1). Slopes of the respective dose-effect curves were significantly different from 0 (P < 0.05) but not significantly different from parallel. Calculated ED100doses (and 95% confidence intervals) for intrathecal gabapentin and 3-isobutyl-GABA were 165 (101–262) and 65 (45–93) micro gram, respectively.
Intrathecal D-Serine or L-Serine and Intraperitoneal Gabapentin/S(+)-Isobutyl-GABA
The intrathecal injections of D- or L-serine (100 micro gram) had no effect on thermal escape threshold alone or after the injection of intrathecal substance P (Figure 1). In contrast, attenuation of the thermal hyperalgesia otherwise produced by gabapentin (100 mg/kg intraperitoneally) or S(+)-3-isobutyl-GABA (30 mg/kg intraperitoneally) was significantly reversed by D-serine but not L-serine (Figure 7).
Effects of Intrathecal NMDA and Intraperitoneal Gabapentin
Intrathecal NMDA at 2 nmol evoked a short-lasting thermal hyperalgesia, the maximum effect of which was similar to that produced by intrathecal substance P, although the overall effect was shorter lasting. Pretreatment (at -60 min) with gabapentin (100 mg/kg intraperitoneally) or S(+)-3-isobutyl-GABA (30 mg/kg intraperitoneally) prevented the appearance of hyperalgesia (Figure 8).
Gabapentin or S(+)-3-isobutyl-GABA had no effect on the righting response or pinna or blink reflexes at doses up to the maximum used after either intrathecal (Figure 2) or intraperitoneal (Figure 5) injection. Examination of ambulatory behavior revealed that there were no effects at intrathecal doses up to 30 or 100 micro gram or intraperitoneal doses of 30 or 100 mg/kg of S(+)-3-isobutyl-GABA and gabapentin, respectively. At the next higher dose by either route, animals receiving gabapentin or S(+)-3-isobutyl-GABA were clearly able to ambulate but typically displayed signs of weakness characterized by splaying of the hind paws. Even at these doses, the rats displayed a placing and stepping response, although it was less brisk than that observed at lower doses or in animals in the control group. It is important to emphasize that at doses required to produce a “normal” thermal escape latency in the injured paw, rats displayed no motor deficit detectable to the observer.
The current work demonstrates the potent antihyperalgesic effect of systemic and intrathecal delivery of gabapentin and its analogue, S(+)-3-isobutyl-GABA. Several points should be emphasized.
Site of Action
Although a supraspinal action is not excluded, the potent effect of the spinally delivered agent compared with the systemic activity demonstrated in the current study and in other works [4,5]suggests that the spinal cord is a certain site of drug action. As is discussed further later, the effects of systemic gabapentin and 3-isobutyl-GABA were readily antagonized by intrathecally delivered D-serine. This antagonism by intrathecal injection provides additional evidence that the primary site of action of gabapentin and 3-isobutyl-GABA is mediated at a spinal level.
Time Course of Action
The current studies indicate that by the intraperitoneal and intrathecal routes, the onset of action for gabapentin and 3-isobutyl-GABA is rapid. Significant changes from baseline values could be detected at intervals as short as 2–5 min. This contrasts somewhat with the anticonvulsant effects, which take as long as 20 min to appear. The rapid onset after systemic delivery was unexpected given the low lipid solubility (log P [octanol/water]=-1.2). Gabapentin is actively transported into the bloodstream by an L-amino acid transporter in the gut. L-Amino acid transporters also have been demonstrated to be present at the blood-brain barrier and likely account in part for the immediate appearance of gabapentin in the brain [26,27](as measured by intracranial microdialysis) after intraperitoneal delivery and thus for the rapid onset of action observed here. It is also interesting to note that gabapentin was effective even 10 min after treatment with substance P. This demonstrates that gabapentin, unlike many drugs, is effective before and after sensitization.
Intrathecal Substance P/NMDA-evoked Hyperalgesia
Spinal delivery of substance P and NMDA produces a thermal hyperalgesia by an action at NK-1 and NMDA sites, respectively. Substance P and NMDA effects are attenuated by intrathecal delivery of cyclooxygenase and nitric oxide synthase inhibitors. These results suggest a positive feedback loop in which spinal prostanoids and nitric oxide enhance the excitability of the presynaptic terminal (see ). These findings are consistent with our work showing that that intrathecal substance P increases the gain of the function that relates response latency to stimulus intensity. Although intrathecal NMDA receptor antagonism has not been reported to reverse the thermal hyperalgesia of intrathecal substance P, recent studies have shown that exogenous substance P increases spinal glutamate release [31,32](Hua, Marsala, and Yaksh, unpublished data). This raises the possibility that at least a component of the thermal hyperalgesia might be mediated by an indirect action at one or another spinal NMDA or non-NMDA receptors.
Mechanism of Action
GABAergic Receptor. Gapapentin and 3-isobutyl-GABA are GABA analogues. They have no reported affinity for the GABA-A or GABA-B binding sites. It has been shown, however, that gabapentin can increase the extracellular levels of GABA in striatal slice preparations and evoke a nonsynaptic release of GABA from glial cells. Given that gabapentin may increase spinal GABAergic tone, its actions after spinal delivery are consistent with the ability of GABA-A and GABA-B receptor agonists to reduce the second phase of the formalin test, the allodynia in the Chung model of neuropathy, and the thermal hyperalgesia in the Bennett model of neuropathy. In the work with the Chung model of neuropathy, however, the effects of gabapentin were not antagonized by doses of bicuculline and CGP 35348, which were able to reverse selectively the effects of muscimol and (GABA-A) and baclofen (GABA-B), respectively. [4,35]
Glutamatergic Receptor. The several animal models noted previously (formalin test, second phase; allodynia in the Chung model of neuropathy; thermal hyperalgesia in the Bennett model of neuropathy), in which gabapentin has been shown to display a significant effect, reflect diverse underlying mechanisms. Common to all, however, is the role played by spinal NMDA receptors (see ). The interaction of gabapentin with the NMDA receptor appears to be indirect. Binding studies failed to indicate an affinity of gabapentin for NMDA, nonstrychnine-sensitive glycine sites, or non-NMDA sites. Gabapentin has been reported not to alter NMDA channel responses. Moreover, as with the GABAergic system, the lack of prominent motor effects renders it unlikely that the primary effects reflect a simple antagonism of the NMDA site.
Gabapentin has been shown to bind to the alpha2/delta subunit of voltage-sensitive calcium channels. Whether this mechanism is relevant to the spinal sites of gabapentin action is not known. Importantly, intrathecal N-type calcium channel blockers have been shown in rats to have a similar antihyperalgesic action, although gabapentin's side-effect profile is clearly less serious. The doses of gabapentin required to induce hyperalgesia were similar to those at which anticonvulsant effects are seen. It is possible, then, that these effects operate through similar mechanisms. In recent studies, Singh et al. have shown that the anticonvulsant effect of gabapentin can be reversed by D-serine. D-Serine is said to be an agonist at the nonstrychnine-sensitive glycine site on the NMDA receptor. Our studies show a similar reversal of the antihyperalgesic activity of gabapentin and 3-isobutyl-GABA. These results confirm the observation that D-serine but not L-serine was able to reduce flinching in the second phase of the formalin test (Lakhbir Singh, personal communication, 1996) and block the thermal hyperalgesia induced by thermal injury to the paw (Jun and Yaksh, unpublished data, 1997). The results also suggest that gabapentin and 3-isobutyl-GABA may be acting at the nonstrychnine-sensitive glycine site on the NMDA receptor; however, binding studies have failed to show any effect of gabapentin at this site. After intrathecal delivery, D-serine alone was reported to facilitate the thermally evoked tail flick. In the current study, however, neither D-serine nor L-serine altered the thermal escape response in healthy rats nor in rats rendered hyperalgesic by intrathecal substance P. The reason for this difference is not known.
The efficacy of gabapentin in several models of hyperalgesia, including hyperalgesia induced by intrathecal substance P and NMDA, suggests a common mechanisms associated with the generation of a facilitated state of processing. It also implies that gabapentin is not acting as a direct antagonist at the NK-1 or NMDA receptors.
Site specificity of gabapentin may be in doubt, but there are accumulating data to suggest that there is a specific mode of action. Gabapentin has been shown to bind to cellular plasma membranes with high nanomolar affinity. Its binding is displaced in a stereospecific fashion by S(+)-3-isobutyl-GABA but not by the stereoisomer R(-)-3-isobutyl-GABA. This stereoselectivity in binding is also reflected in the biologic activity of this family of agents. Therefore, as an anticonvulsant agent, S(+)-3-isobutyl-GABA is approximately 4 to 10 times more potent than gabapentin, whereas the stereoisomer R(-)-3-isobutyl-GABA is inactive. [19,20]The current studies reveal a similar structure-activity relationship for the spinal antihyperalgesic effects, with S(+)-3-isobutyl-GABA being approximately three times more potent than gabapentin as an antihyperalgesic agent, whereas R(-)-3-isobutyl-GABA is without effect at the highest dose examined. Moreover, the effects of both active agents are reversed by intrathecal D-serine. These data jointly suggest that the gabapentinoids, gabapentin and S(+)-3-isobutyl-GABA, may share a common, if as-yet undefined, mechanism of action.
The current study reveals a potent, dose-dependent, and stereospecific antihyperalgesic effect for gabapentin and 3-isobutyl-GABA. Comparison of the doses required for spinal versus systemic delivery, and the ability of intrathecal D-serine to reverse the effects of these agents given systemically, suggest the likely importance of a common spinal site of action. These results are in accord with a growing body of preclinical and clinical literature suggesting that gabapentin may exert potent effects on a variety of anomalous pain states in which facilitated spinal processing is induced by tissue or nerve injury.
The authors thank Dr. Michael Rafferty for his suggestions and critical reading of this manuscript.
*Oles RJ, Singh L, Hughes J, Woodruff GN: The anticonvulsant action of gabapentin involves the glycine/NMDA receptor. Soc Neurosci Abstr 1995; 6:783