Spinal Endocannabinoids and CB1 Receptors Mediate C-Fiber–Induced Heterosynaptic Pain Sensitization

See allHide authors and affiliations

Science  07 Aug 2009:
Vol. 325, Issue 5941, pp. 760-764
DOI: 10.1126/science.1171870

Plastic Pain Perception

Drugs and endocannabinoids acting on cannabinoid (CB) receptors have potential in the treatment of certain types of pain. In the spinal cord they are believed to suppress nociception, the perception of pain and noxious stimuli. Pernia-Andrade et al. (p. 760) now find that endocannabinoids, which are released in spinal cord by noxious stimulation, may promote rather than inhibit nociception by acting on CB1 receptors. Endocannabinoids not only depress transmission at excitatory synapses in the spinal cord, but also block the release of inhibitory neurotransmitters, thereby facilitating nociception.


Diminished synaptic inhibition in the spinal dorsal horn is a major contributor to chronic pain. Pathways that reduce synaptic inhibition in inflammatory and neuropathic pain states have been identified, but central hyperalgesia and diminished dorsal horn synaptic inhibition also occur in the absence of inflammation or neuropathy, solely triggered by intense nociceptive (C-fiber) input to the spinal dorsal horn. We found that endocannabinoids, produced upon strong nociceptive stimulation, activated type 1 cannabinoid (CB1) receptors on inhibitory dorsal horn neurons to reduce the synaptic release of γ-aminobutyric acid and glycine and thus rendered nociceptive neurons excitable by nonpainful stimuli. Our results suggest that spinal endocannabinoids and CB1 receptors on inhibitory dorsal horn interneurons act as mediators of heterosynaptic pain sensitization and play an unexpected role in dorsal horn pain-controlling circuits.

Activity-dependent central hyperalgesia can be induced in the absence of any inflammation or nerve damage by selective activation of glutamatergic C-fiber nociceptors; for example, with the specific transient receptor potential channel (TRP) V1 agonist capsaicin. Local subcutaneous injection of capsaicin induces primary hyperalgesia at the site of injection and a purely mechanical secondary hyperalgesia in the surrounding healthy skin (1). This secondary hyperalgesia originates from changes in the central processing of input from mechanosensitive A fibers and is characterized by an exaggerated sensitivity to painful stimuli and by pain evoked by light tactile stimulation (allodynia or touch-evoked pain). These symptoms are mimicked by the blockade of inhibitory γ-aminobutyric acid–mediated (GABAergic) and glycinergic neurotransmission in the spinal dorsal horn (2, 3), suggesting that a loss of synaptic inhibition also accounts for C-fiber–induced secondary hyperalgesia. Activity-dependent hyperalgesia can thus be regarded as a correlate of heterosynaptic depression of inhibition (4). In many neuronal circuits of the central nervous system, endocannabinoids [2-arachidonoyl glycerol (2-AG) and anandamide (AEA)] are released upon intense activation of metabotropic glutamate receptors and serve as retrograde messengers mediating either homosynaptic feedback inhibition or heterosynaptic depression of (GABAergic) inhibition (5, 6). Type 1 cannabinoid (CB1) receptors are densely expressed in the superficial dorsal horn of the spinal cord (7), where they exert anti-hyperalgesia in different inflammatory or neuropathic disease states (8, 9).

To define the role of CB1 receptors in dorsal horn neuronal circuits, we first characterized the effects of CB1 receptor activation on neurotransmission in mouse transverse spinal cord slices (Fig. 1). Excitatory and inhibitory postsynaptic currents (EPSCs and IPSCs) were evoked by extracellular electrical field stimulation at a frequency of four per minute and recorded from visually identified neurons in the superficial spinal dorsal horn (laminae I and II) (10). The mixed CB1/CB2 receptor agonist WIN 55,212-2 (3 μM) reversibly reduced the amplitudes of glycine receptor IPSCs to 64.3 ± 3.5% of control amplitudes (mean ± SEM, n = 13 neurons, P < 0.001, paired Student’s t test) (Fig. 1A). Similarly, GABAA receptor IPSCs were reduced to 64.7 ± 3.0% (P < 0.001, n = 8 neurons, paired Student’s t test) (Fig. 1B). The inhibition of IPSCs by WIN 55,212-2 was confined to the superficial dorsal horn, reversed by the CB1 receptor antagonist/inverse agonist AM 251 (5 μM) (Fig. 1, A and B), and absent in global CB1 receptor–deficient mice (CB1−/− mice) (11) and in mice lacking CB1 receptors specifically in dorsal horn inhibitory interneurons (ptf1a-CB1−/− mice) (12) (fig. S1). WIN 55,212-2 had virtually no effect on EPSCs mediated by glutamate receptors of the α-amino-3-hydroxy-5-methylisoxazole (AMPA) subtype (AMPA-EPSCs) [Fig. 1C, see also (13)]. It did, however, reduce the amplitudes of monosynaptic AMPA-EPSCs evoked by stimulation of dorsal root primary afferent nerve fibers at C-fiber intensity by 34.5 ± 3.3% (n = 9 neurons). This inhibition was not reversed by AM 251 (fig. S2). We next tested whether stimulation of endogenous endocannabinoid production through the activation of group I metabotropic glutamate receptors (mGluR1/5) would have a similar effect on inhibitory synaptic transmission. These experiments were carried out in GlyT2–enhanced green fluorescent protein (GlyT2-EGFP) transgenic mice, which allowed targeted recordings from glycinergic (EGFP-positive) and nonglycinergic (EGFP-negative), presumed excitatory, interneurons (14). (S)-3,5-dihydroxyphenylglycine (DHPG, 10 μM), an agonist at mGluR1/5, reduced IPSC amplitudes in nonglycinergic presumed excitatory superficial dorsal horn neurons by 40.6 ± 4.5% (n = 8 neurons) (Fig. 1D). This inhibition was reversed by AM 251 (5 μM) (Fig. 1D) and partially prevented by mGluR1 and mGluR5 antagonists (LY 367,385, 100 μM, remaining inhibition 21.0 ± 3.9%, n = 5 neurons; and MPEP, 10 μM, 25.0 ± 3.4%, n = 5 neurons) (fig. S3). Glycinergic input to EGFP-positive (glycinergic) neurons was less sensitive to DHPG, with an average reduction of only 10.3 ± 3.6% (n = 8 neurons) (Fig. 1D). Depolarization-induced suppression of inhibition (DSI) could be induced in six out of eight nonglycinergic neurons but was not seen in glycinergic neurons (n = 5 neurons) (Fig. 1E).

Fig. 1

Synaptic effects of CB1 receptor activation in dorsal horn neuronal circuits. (A to C) Effects of the mixed CB1/CB2 receptor agonist WIN 55,212-2 (3 μM) on glycinergic IPSCs (A), GABAergic IPSCs (B), and AMPA-EPSCs (C). Left panels: Current traces averaged from 10 consecutive stimulations under control conditions, after addition of WIN 55,212-2 and after the additional application of AM 251 (5 μM). Right panels: Time course. Mean ± SEM, n = 7 to 13 neurons. (D) Inhibition of glycinergic IPSCs in nonglycinergic (EGFP-negative) neurons (n = 8 neurons) by the mGluR1/5 agonist DHPG (10 μM) and its reversal by AM 251 (5 μM). Only a minor inhibition was observed in glycinergic (EGFP-positive) neurons (n = 8 neurons). (E) DSI (1-s depolarization of the postsynaptic neuron to 0 mV) in nonglycinergic neurons (six out of eight neurons) and its prevention by AM 251 (5 μM). No DSI occurred in glycinergic neurons (n = 5 neurons). Glycinergic IPSCs were evoked at a frequency of 0.2 Hz.

The reduction of inhibitory synaptic transmission by endocannabinoids was due to diminished release of GABA and glycine from inhibitory nerve terminals. In paired pulse experiments, WIN 55,212-2 (3 μM) increased the amplitude ratio of two consecutive IPSCs, 70 ms apart, from 1.14 ± 0.07 to 1.61 ± 0.15 (n = 5 neurons, P < 0.05, paired Student’s t test) (Fig. 2A). Accordingly, the coefficient of variation [CV = (SD2/mean2)1/2] of IPSC amplitudes (15) increased from 0.190 ± 0.012 under control conditions to 0.306 ± 0.031 in the presence of WIN 55,212-2, again indicative of a presynaptic action (n = 13 neurons, P < 0.01, paired Student’s t test) (Fig. 2B). We directly demonstrated the presence of CB1 receptors on the presynaptic terminals of inhibitory mouse superficial dorsal horn neurons by electron microscopy (EM) (Fig. 2, C to F). Peroxidase-based and immunogold labeling of CB1 receptors and high-resolution EM unequivocally showed the presence of CB1 receptors on presynaptic terminals of symmetrical (inhibitory) synapses (Fig. 2, C and D) and the colocalization of CB1 with the vesicular inhibitory amino acid transporter (VIAAT) (Fig. 2, E and F), a marker of inhibitory axon terminals (16).

Fig. 2

Inhibition of glycinergic and GABAergic synaptic transmission via presynaptic CB1 receptors. (A) Paired pulse experiments. Current traces of two consecutive glycinergic IPSCs (P1 and P2) under control conditions (black) and in the presence of 3 μM WIN 55,212-2 (red) are shown. (B) Variation analysis. Top panel: Individual traces of glycinergic IPSCs recorded under control conditions and in the presence of WIN 55,212-2 (3 μM). Bottom panel: Changes in the CV in 13 cells are plotted versus changes in the mean amplitude induced by WIN 55,212-2. (C to F) Electron microscopic analysis (a to c and a and b are serial sections) of CB1 receptor localization in the superficial spinal dorsal horn. Arrowheads indicate symmetric synapses; arrows indicate immunogold labeling. (Ca to Cc) CB1 immunostaining coupled to immunoperoxidase reaction [3,3′-diaminobenzidine (DAB)]. CB1 receptors are present in an axon terminal (t) forming a symmetric (inhibitory) synapse on an immunonegative dendritic shaft (d) in lamina II. The asterisk labels a CB1-negative bouton of another symmetric synapse on the same dendrite. (Da and Db) High-resolution pre-embedding immunogold staining for CB1. The CB1 receptor is located presynaptically on the plasma membrane of an inhibitory axon terminal (t). (Ea and Eb) DAB staining for VIAAT and pre-embedding immunogold labeling for CB1. CB1 cannabinoid receptors (indicated by arrows) are on an inhibitory (VIAAT-positive) axon terminal (t). In this reaction, silver intensification results in weaker electron density of the DAB precipitate. (Fa and Fb) Immunoperoxidase staining for CB1 combined with pre-embedding immunogold labeling for VIAAT demonstrates colocalization of the two proteins. Similar results were obtained in four animals. Scale bar, 0.1 μm.

We next studied the role of endocannabinoids in secondary hyperalgesia in intact rats and performed in vivo extracellular single-unit recordings (10) from neurons with a wide dynamic range (that is, neurons responding to both noxious and innocuous stimulation) with receptive fields in the hindpaw and located in the deep lumbar dorsal horn (Fig. 3). Intracutaneous injection of capsaicin (200 μg) into the receptive field of the recorded neuron led to a robust increase in action potential firing in response to mechanical stimulation in an area surrounding the capsaicin injection site, akin to secondary hyperalgesia and allodynia. This increase was reversed by local spinal application not only of the mGluR1 antagonist LY 367,385 (10 μM, n = 5 neurons) but also of the CB1 receptor blocker AM 251 (5 and 50 μM, n = 5 or 6 neurons).

Fig. 3

Extracellular single-unit recordings from deep dorsal horn neurons in intact rats. Frequency histograms of action potentials evoked by mechanical stimulation within the receptive field on one hindlimb with brush (br), pressure (pr), and pinch (pi) but outside the capsaicin-injected area are shown. (A) Responses (spikes/s) of three representative neurons. (B) Statistical analysis of background-corrected action potential activity (mean ± SEM). (C) Dose-response analyses by repeated-measures analysis of variance (ANOVA), followed by Newman-Keuls multiple comparison post-hoc tests; n = 5 or 6 neurons per group. *P ≤ 0.05, **P < 0.01, ***P < 0.001, against control (pre-capsaicin). +P ≤ 0.05, ++P < 0.01, +++P < 0.001, against capsaicin.

In mice, we tested the effects of pharmacological and genetic manipulation of the endocannabinoid system on capsaicin-induced secondary hyperalgesia (Fig. 4). Subcutaneous injection of capsaicin (30 μg) into one hindpaw of wild-type mice led to a reduction in paw withdrawal thresholds in response to mechanical stimulation with dynamic von Frey filaments from 2.85 ± 0.04 g under control conditions to 0.53 ± 0.10 g (mean ± SEM, n = 6 mice) at 2 hours after capsaicin injection (10). Intrathecal injection (injection into the lumbar spinal canal) of the mGluR1 antagonist LY 367,385 (1.0 nmol per mouse) 2 hours after capsaicin reduced mechanical sensitization by 64.9 ± 2.9% (n = 6 mice) (17). Consistent with the role of CB1 receptors in synaptic disinhibition, intrathecal AM 251 (0.5 nmol) reversed mechanical sensitization by 71.2 ± 9.0% (n = 6 mice). Accordingly, inhibition of endocannabinoid degradation with URB 597 or of endocannabinoid reuptake with UCM 707 (each 1.0 nmol) (18) prolonged secondary hyperalgesia (Fig. 4A). In naïve mice, all five compounds exerted only minor effects on mechanical sensitivity (fig. S4).

Fig. 4

Effects of pharmacological and genetic manipulations of the endocannabinoid system on capsaicin-induced mechanical hyperalgesia in mice. (A) Mechanical paw withdrawal thresholds (mean ± SEM) were determined with dynamic von Frey filaments at 20-min intervals for 2 hours after capsaicin injection into the left hindpaw and for another 2 hours after intrathecal injections of vehicle (10% dimethyl sulfoxide), AM 251 (0.5 nmol per mouse), URB 597 (1.0 nmol), UCM 707 (1.0 nmol), LY 367,385 (1.0 nmol), or MPEP (150 nmol). Left panel: Time course (mean ± SEM). Right panel: Treatment-induced changes in hyperalgesia. Areas under the curve (AUC) were integrated over time from 2 to 4 hours after capsaicin injection. The time course of sensitization in wild-type mice treated with intrathecal vehicle is the same as in wild-type mice that did not receive intrathecal injections (B). n = 5 or 6 mice per group; for statistical analyses, three groups of vehicle-injected mice were pooled. Analyses were by one-way ANOVA followed by Dunnett’s post-hoc test F (11,74) = 21.18; *P ≤ 0.05, **P < 0.01, ***P < 0.001. (B) Capsaicin-induced secondary hyperalgesia in wild-type mice versus CB1−/− mice (n = 9 mice per group) and in ptf1a-CB1−/− mice (n = 7 and 11 mice per group) and sns-CB1−/− mice versus mice carrying a CB1 receptor gene flanked by two loxP sites (CB1fl/fl mice) (n = 5 mice per group). Left: Time course. Right: AUC (0 to 4 hours after capsaicin injection). ***P < 0.001, unpaired Student’s t test.

Global CB1−/− mice and ptf1a-CB1−/− mice were protected from capsaicin-induced mechanical sensitization. In contrast, mice devoid of CB1 receptors only in primary afferent nociceptors (sns-CB1−/− mice) (19) developed normal secondary hyperalgesia (Fig. 4B), indicating that the CB1 receptors on inhibitory dorsal horn neurons, and not those on primary nociceptors, mediated capsaicin-induced secondary hyperalgesia. The unchanged responses of sns-CB1−/− mice also indicate that possible direct interactions of CB1 receptors with TRPV1 channels (20, 21) expressed on the spinal terminals of primary nociceptors were not involved.

Mechanical sensitization could also be evoked by intrathecal injection of the CB1/CB2 agonist CP 55,940 (fig. S5). Intrathecal CP 55,940 (10 nmol) significantly decreased the thresholds of mechanical stimulation with von Frey filaments in wild-type (CB1fl/fl) and sns-CB1−/− mice and rendered both types of mice extremely sensitive to touch. In both tests, mechanical sensitization by CP 55,940 was absent in global CB1−/− mice. The pronociceptive effects of endocannabinoids suggested here are specific for C-fiber–mediated, activity-dependent hyperalgesia. In models of mild inflammatory pain (produced by subcutaneous injection of zymosan A) (10) and neuropathic pain (produced by chronic constriction injury) (10), CB1−/− mice behaved normally (fig. S6, A and B). AM 251 had only negligible effects (fig. S6, C and D), whereas CP 55,940 exerted anti-hyperalgesic actions in these models (fig. S6, E and F). Both of these models also involve spinal disinhibitory processes, but the underlying mechanisms are most likely different and involve the spinal release of pronociceptive prostaglandin E2 (22) and changes in the transmembrane chloride gradient (23).

Finally, we tested the effect of CB1 receptor blockade on C-fiber–induced secondary hyperalgesia and allodynia in human volunteers (fig. S7). Secondary hyperalgesia was induced by intracutaneous electrical stimulation at C-fiber strength (2 Hz, 15 to 100 mA) of a small skin area of the left forearm (10). In the first session, the intensity of electrical stimulation was adjusted to yield a value of 6 on a numeric rating scale ranging from 0, no pain, to 10, maximum imaginable pain, and pain ratings and the sizes of hyperalgesic skin areas surrounding the site of electrical stimulation were determined for 100 min at regular intervals. In a second session, 28 days later, the volunteers were tested again after a 10-day treatment with either placebo or rimonabant (20 mg/day, aken rally), a CB1 receptor antagonist/inverse agonist closely related to AM 251. Rimonabant treatment had no effect on acute pain ratings induced by electrical stimulation (–2.0 ± 5.7%, n = 8 volunteers per group) but decreased the sizes of hyperalgesic and allodynic skin areas to 53.7 ± 5.2 and 57.4 ± 5.0%, respectively.

The contribution of endocannabinoids to activity-dependent pain sensitization, which we propose here, builds on a model of secondary hyperalgesia and allodynia (fig. S8), in which normally pain-specific dorsal horn neurons receive not only monosynaptic input from C-fiber nociceptors but also polysynaptic input from non-nociceptive fibers (24). The suprathreshold activation of these neurons by such non-nociceptive input is normally prevented by the activity of dorsal horn inhibitory interneurons. The present study shows that intense glutamatergic input from C-fiber nociceptors diminishes this inhibitory control through endocannabinoids acting at CB1 receptors located on dorsal horn inhibitory interneurons. Our findings thus attribute to endocannabinoids an unexpected role in dorsal horn neuronal circuits as mediators of spinal activity–dependent pain sensitization. They are also an example of a distinctive phenotype of mice lacking CB1 receptors specifically in inhibitory interneurons, whereas most previously reported phenotypes of global CB1 receptor–deficient mice could be ascribed to the lack of CB1 receptors on glutamatergic neurons (25).

Supporting Online Material

Materials and Methods

Figs. S1 to S8

Table S1


  • * These authors contributed equally to this work.

  • Present address: Institute of Physiology, University of Freiburg, Engesserstrasse 4, D-79108, Freiburg, Germany.

  • Present address: Department of Anesthesiology, Medical School Hannover, D-30625 Hannover, Germany.

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. This research was supported in part by grants from the Schweizerischer National Fonds to H.U.Z. (3100A0-116064/1), from the Deutsche Forschungsgemeinschaft to W.K. (KO 1878/2-2) and H.U.Z. (ZE 377/8-2), from NIH (NS38261 and NS11255) to V.N., from the European Union (LSHM-CT-2004-005166) to T.F.F., and from the OTKA (F046407) and ETT (561/2006) to I.K. A.J.P.A. was supported partly by scholarships from the German DAAD and the Venezuelan FONACIT for graduate study at IVIC. I.K. was supported by a János Bolyai scholarship. The authors thank R. Kuner and C. V. Wright for providing sns-cre and ptf1a-cre mice, respectively; J.-M. Fritschy, H. Handwerker, H. Möhler, and M. Schmelz for critical reading of the manuscript; T. Müller for very valuable suggestions; and I. Camenisch and L. Scheurer for genotyping of the mice.
View Abstract

Navigate This Article