Background
Neuropathic pain is a widespread and highly debilitating condition commonly resulting from injury to peripheral nerves or from a variety of causes including trauma, cancer, HIV-AIDS or diabetes [
1]. Unlike inflammatory pain, for which many efficacious treatments exist, neuropathic pain is typically refractory to most current treatments and thus represents a major unmet medical need. Key symptoms of neuropathic pain are hyperalgesia, allodynia and spontaneous pain. Hyperalgesia involves enhanced pain perception to noxious stimuli; allodynia designates pain experienced in response to an innocuous stimulus and spontaneous pain refers to recurring pain, not necessarily related to an identifiable peripheral stimulus. Of these symptoms, tactile allodynia (e.g. pain induced by gentle mechanical stimulation of the skin) and spontaneous pain are the most prevalent and debilitating [
2]. Several cellular substrates for these symptoms have been proposed [
3], including suppression of the inhibition mediated by GABA
A- and glycine receptors in the dorsal horn of the spinal cord [
4‐
6]. We have discovered that following peripheral nerve injury such disinhibition of spinal dorsal horn neurons occurs by a collapse of the anion gradient in lamina I neurons [
7] via a novel microglia-neuron signalling pathway [
8], leading to weakening GABA
A- and glycine-mediated inhibitory synaptic transmission [
9,
10].
Lamina I is one of the main nociceptive output pathways from the spinal cord to the brain and, in contrast to lamina V neurons, lamina I neurons do not receive direct input from low-threshold Aδ and Aβ afferents [
11]; rather they normally encode and transmit essentially noxious or thermal information [
12‐
15]. However, tactile allodynia requires that innocuous inputs elicit a nociceptive percept or response. Because lamina I output neurons do not normally respond to innocuous mechanical input [
14], it is unclear how disinhibition of these neurons might cause innocuous input to trigger a noxious sensation at the supraspinal level. Similarly, spontaneous pain requires ongoing or episodic activity in nociceptive pathways, and lamina I nociceptive output neurons have little or no spontaneous activity [
14]. In the present study, we investigated the possibility that peripheral nerve injury and microglia-driven disinhibition in the dorsal horn transform lamina I output neurons from silent and strictly responsive to noxious stimuli to allow these neurons to display spontaneous activity and be driven additionally by innocuous peripheral stimulation.
Discussion
Our results show that in control conditions less than 25% of mechanosensitive lamina I projection neurons produce action potentials in response to low-threshold tactile stimuli, consistent with previous studies in rats [
12] and cats [
14,
24]. In contrast to the situation in naïve animals we show here for the first time that peripheral nerve injury causes a functional switch in the sensory specificity of this subpopulation of spinal output neurons whereby the majority of these neurons respond to low-threshold tactile stimuli. Finally, the same qualitative functional switch is triggered in naïve animals by acute local, spinal, application of ATP-stimulated microglia or disruption of chloride homeostasis.
Hyperalgesia can be explained by a quantitative change in response properties of nociceptive relay neurons whereby the same nociceptive input generates a greater action potential output. However, given that the majority of lamina I neurons respond to noxious input only, a quantitative change in nociceptive responsiveness appears insufficient to explain allodynia. Rather, allodynia implies a qualitative change, a miscoding of information such that innocuous inputs are converted into a nociceptive message. The switch in modality specificity we observed in lamina I output neurons is such a qualitative change in response properties and thus may be sufficient to explain allodynia. After peripheral nerve injury, administering ATP-stimulated microglia or disrupting chloride transport, innocuous inputs are transformed in the dorsal horn and become encoded by lamina I projection neurons. Consequently, action potential discharge is now generated by these neurons and sent to higher brain structures through output neurons that were previously nociceptive specific. It is thus logical to infer that such signals will be interpreted as noxious at the supraspinal level, providing a substrate to explain tactile allodynia. Similar logic can be applied to the finding of the appearance of spontaneous bursts of spikes in lamina I output neurons after nerve injury, treatment with ATP-stimulated microglia, or disruption of chloride homeostasis, providing a substrate of spontaneous pain that occur in neuropathic conditions.
A comparable loss of selectivity for noxious stimuli has been observed on unidentified nociceptive specific neurons in the superficial dorsal horn following application of mustard oil or capsaicin on the receptive field of the cells in the periphery [
25,
26]. Similarly, brushing of the skin after sciatic nerve crush leads to c-fos expression in the superficial dorsal horn [
27]. Under those conditions, however, as with nerve injury, the possibility remains that the central response to low-threshold stimuli is due to a peripheral change in selectivity (e.g. peripheral nociceptors responding to low-threshold mechanical stimuli). In contrast, the evidence presented here, using local spinal administration of agents, shows conclusively that responses to low threshold inputs by lamina I neurons can result from purely central mechanisms.
Based on previous studies, the proposed central mechanism that was affected in the current experimental conditions is a disruption of anion homeostasis [
7,
8], effectively weakening inhibition [
9]. The site at which the disinhibition occurs appears to be within the circuitry intrinsic to the dorsal horn, not within the network of afferent or descending terminals entering the dorsal horn. A loss of KCC2, which normally extrudes Cl
- from the cells, appears to be the underlying mechanism [
7,
8,
28] and KCC2 is not present in synaptic terminals nor on primary afferents [
7]. The other principal regulator of Cl
- in the dorsal horn is NKCC1, which normally causes Cl
- accumulation into cells. NKCC1 is very weakly expressed in adult dorsal horn neurons [
29,
30] but is the dominant cation-chloride co-transporter in primary afferents [
31], and thus NKCC1 leads to GABA exerting a depolarizing, albeit inhibitory, action on sensory terminals. Abnormal presynaptic excitation of sensory terminals has been proposed to occur via upregulation of NKCC1 in small diameter afferents after an inflammatory peripheral insult [
31]. This is thought to produce suprathreshold GABAergic depolarization in sensory terminals yielding cross excitation between low and high threshold afferents. The latter mechanism is unlikely to contribute to the effects observed in the present study for the following reasons. First, nerve injury is associated with a loss of KCC2 expression [
7] as mentioned above. Second, ATP-stimulated microglia has been shown to cause tactile allodynia via the release of BDNF [
8], and BDNF-trkB signalling is linked to downregulation of KCC2 [
28,
32]. Third, the blocker of cation-chloride co-transport used in the present study, DIOA, preferentially inhibits KCC2 and not NKCC1 [
33,
34]. Even if one doubts the specificity of DIOA and assumes that it also antagonized NKCC1 [
35], this site of action could not account for the effect of DIOA we observed, unmasking low-threshold input to lamina I neurons, because blocking NKCC1 would work against exaggerated depolarization in primary afferents and thus prevent rather than produce cross talk between them as proposed for inflammatory insult [
36]. In summary, our findings indicate that selective impairment of postsynaptic Cl
- homeostasis in the spinal dorsal horn is sufficient to unmask the relay of innocuous input through normally nociceptive specific pathways.
This aberrant relay of innocuous input may occur via unmasking polysynaptic connections in the superficial dorsal horn [
37‐
39] functionally linking low threshold afferents and nociceptive lamina I projection neurons [
6]. This can be mediated either via disinhibition of feedforward excitatory interneurons such that they can convey this input onto normally nociceptive lamina I neurons or by lowering the threshold of nociceptive lamina I neurons to normally subthreshold polysynaptic input [
40]. In addition, unmasking of low-threshold input to lamina I output neurons may occur via inversion of normally inhibitory post-synaptic events from inhibitory interneurons into excitatory ones in a subset of cells [
7].
The finding that altered chloride homeostasis compromises inhibitory control in dorsal horn neurons raises the question of the therapeutic avenues to compensate for this form of disinhibition. Indeed, a weakening of the hyperpolarizing action of GABA
A/glycine receptor activation suggests that drugs aimed at enhancing GABA
A/glycine receptor-mediated inhibition may be ineffective in reversing nerve injury-induced allodynia. However, three elements must be considered before making such inference. First, while raising intracellular [Cl
-] suppresses the component of inhibition caused by hyperpolarizing the neuron, it only minimally affects the component of inhibition caused by shunting the membrane (as discussed in detail in [
9]). Second, it is particularly important to note that a small depolarizing shift in reversal potential for GABA
A currents will cause a loss of inhibition without necessarily causing GABA
A-mediated net excitation. Thus, when intracellular Cl
- homeostasis is altered, activation of GABA
A and glycine receptors may continue to be inhibitory albeit less inhibitory [
9,
10], allowing for unmasking of latent excitatory inputs. Finally, it must be kept in mind that because presynaptic GABA
A receptor-mediated inhibition remains intact (see above), drugs activating or enhancing the function of GABA
A receptors may remain analgesic by inhibiting afferent input at its entry point into the spinal cord. This is consistent with reports of antiallodynic effect of intrathecally-applied GABA
A receptor agonists [
41,
42]. Thus, at least in some cases, sufficient residual inhibition may remain in neuropathic pain conditions permitting GABAergic drugs to be analgesic. Results from modelling studies show that while small reductions in anion gradient may be effectively compensated for by potentiating GABA
A/glycine receptor-mediated input, this can occur at the expense of stability of the system, and compensation will fail as the reduction of anion gradient exceeds a critical value [
9]. Depending on the extent of the pathology, compensation by enhancing GABAergic transmission may therefore be effective or not. Thus, measuring intracellular [Cl
-] may be important to guide treatments based on GABA-modulating agents, and restoring normal anion homeostasis or targeting excitatory transmission may represent more effective therapeutic strategies [
10].
Methods
All experimental procedures were performed in accordance with guidelines from the Canadian Council on Animal Care.
Animal Preparation
Male Sprague-Dawley rats (post-natal day 60+; 300–350 g, Charles River Laboratories, Wilmington, MA) were anaesthetized by intraperitoneal injection of pentobarbital (initial dose: 0.65 mg/kg; subsequent doses 0.35 mg/kg every hour). To maintain a stable level of anaesthesia, rats were given a supplemental dose of anaesthetic every hour post surgery. To test for possible influence of the anaesthesia on our recordings, Ketamine/Xylazine (0.1 ml/100 g) was used in some experiments. No difference was observed under either condition (data not shown).
Spinal cord segments L4-S1 were exposed by laminectomy. The jugular vein was canulated and a tracheotomy performed. The rat was then mounted on the stereotaxic frame and the vertebrae stabilised using two spinal clamps. A small recording chamber was made around the exposed spinal cord segment using agar (3%) to isolate the exposed spinal cord and prevent the diffusion of the liquid outside this area.
Animals were administered pancuronium bromide (Sigma, 5 mg/ml; 0.1 ml initially and then 0.05 ml hourly) and ventilated artificially. Expired CO2 was maintained at 4%. The body temperature was continuously controlled and maintained at 37.5°C. Immediately before recording, the meninges were carefully removed and the spinal cord covered with mineral oil at 34°C to avoid drying of the spinal cord.
Peripheral nerve injury
Peripheral nerve injury (PNI) was performed by surgically implanting a polyethylene cuff around the sciatic nerve of anaesthetized adult rats as described previously [
7,
43,
44]. Paw withdrawal threshold was measured using von Frey filaments to demonstrate tactile allodynia as described previously [
7,
45]. Within the nerve injured group, only animals that showed a gradual decrease in mechanical threshold (over 14–17 days) down to 2.0 g or less were used for recordings.
Recordings and stimulation
Extracellular single unit recordings were conducted using stainless steel metal electrodes (10 MΩ, FHC, USA; ER-1 extracellular amplifier, Cygnus technology). The recording electrode was mounted on a high precision manipulator (Burleigh 6000 controller) and the zero was set as the electrode touched spinal cord surface. The signal was filtered between 300 Hz and 1.5 kHz (Brownlee precision).
To identify lamina I projection neurons, the lateral parabrachial nucleus was stimulated as described previously [
12]. An array of 6 electrodes was specially designed to fit within the parabrachial nucleus (staggered by 200 μm; contact diameter 100 μm; contact length 150 μm; custom-made by Peter Rhodes, Rhodes Medical Instruments Inc). Stimulation of the parabrachial nucleus consisted of a train of 4 stimuli. Units were confirmed as projection neuron if they followed the train of 4 antidromic stimuli delivered at up to 500 Hz or if collision between antidromic and orthodromic action potentials was observed (Fig.
1b). Importantly, only one lamina I projection neuron per rat was recorded as each neuron was taken as its own control during treatments with either drug or microglia.
We used an established routine to characterize the response of lamina I projection neurons to innocuous and noxious natural stimulation of the receptive field: 20s of brush, 20s of repetitive touch (with the tip of finger, yielding 80 touch/20s) and 10s of noxious pinch using calibrated forceps (100 g; 1 mm diameter tip; Electrotechnology Selem Inc, Québec, Qc, Canada). Each set of stimuli was delivered at 30 min interval. Detailed mapping of the receptive fields as well as potential changes in receptive field size was not conducted to avoid applying repeating stimuli too often, which has been shown to produce sensitization of lamina I cells [
14].
Data acquisition and analysis
Extracellular recordings, parabrachial stimulation and expired CO2 were sampled at 20 kHz and stored on a computer using an analog-to-digital conversion system (Powerlab 8SP, AD instruments Inc.). Analysis was performed offline using Neuroexplorer (Nextechnology, USA) and locally designed software. Mean response to sensory stimulation (i.e. cumulative number of spikes detected during the stimulation period or mean frequency) was calculated for each neuron before and after drug treatment.
Microglia cultures
Rat primary cultures were prepared from neonatal cortex or spinal cord as previously described [
6,
7] and maintained for 10–14 days in DMEM medium with 10% fetal bovine serum. Microglia were separated from the primary culture by gentle shaking of the flask and were replated on plastic dishes. The cells were removed from the dish surface with a cell scraper and collected in 100 ml of PBS buffer. This method produces microglia cultures of >98% purity. The density of microglia was measured using a haemocytometer and the volume of PBS was adjusted to give a final density of 1,000 cells per 10 μl. Microglia were stimulated by incubation with 50 μM ATP for 1 hour.
Drugs
Drugs were applied directly on the top of the spinal dorsal horn using the chamber described above. R(+)DIOA (Sigma, 100 μM) was first prepared in ethanol at 100× concentration before being diluted in 0.9% NaCl. Bicuculline HCl (Tocris, 50 μM) was dissolved in saline 0.9%NaCl. Microglia were suspended in PBS. All solutions were applied at 34°C.
Statistics
All results are expressed as mean ± SEM. An arcsine transformation was performed to correct for binomial distributions when data were expressed as percentages or proportions [
46]. Chi squared test for contingency tables were used to compare proportion of lamina I projection neurons with and without responses to innocuous stimuli in control rats before and after drug application and in rats with nerve injury. Nonparametric Mann-Whitney post-hoc test was used to compare unpaired data whereas Wilcoxon post-hoc paired test was used to assess the effect of drug applications. Statistical significance was set at p < 0.05.
Competing interests
The author(s) declare that they have no competing interests.