Acute inhibition of signalling phenotype of spinal GABAergic neurons by tumour necrosis factor-α

Non-technical summary

Spinal application of tumour necrosis factor-α (TNFα) is shown to suppress inhibitory synaptic transmission and enhance excitatory synaptic transmission in spinal dorsal horn, but the underlying mechanisms are not fully known. We show that acute application of TNFα inhibits the excitability of a subset of spinal GABAergic neurons through TNF receptor 1 probably by suppressing hyperpolarization-activated cation currents (I_h) through the activation of p38 mitogen-activated protein kinase within these neurons. These results suggest a novel cellular mechanism of how TNFα modulates spinal synaptic transmission which may be involved in the development of pain.

Abstract

Abstract

Spinal application of TNFα induces both allodynia and hyperalgesia, and at least part of the pronociceptive effects of TNFα have been suggested as due to the impaired function of spinal inhibitory neurons (disinhibition). The present study explores the effects of TNFα on the signalling phenotype of spinal GABAergic neurons identified in transgenic mice expressing green fluorescent protein at the glutamic acid decarboxylase 67 (GAD67) promoter. Acute application of TNFα directly inhibits the excitability of a subset of GAD67⁺ spinal neurons. TNFα-induced inhibition was dependent on the activation of p38 mitogen-activated protein kinase (MAPK) within these GAD67⁺ neurons. TNFα receptor 1 (TNFR1) but not receptor 2 (TNFR2) was identified on spinal GAD67⁺ neurons, suggesting that TNFα signals through TNFR1. Voltage-clamp recordings of GAD67⁺ neurons indicated that the inhibitory effect of TNFα was through suppression of the hyperpolarization-activated cation current (I_h). This study defines a novel mechanism of spinal disinhibition mediated by a TNFα–TNFR1–p38 pathway within GABAergic inhibitory interneurons.

Introduction

Pro-inflammatory cytokines, such as tumour necroscis factor-α (TNFα), interleukin 1-β (IL-1β) and interleukin 6 (IL-6), play key roles in the pathogenesis of inflammatory and neuropathic pain (Rutkowski & DeLeo, 2002; Wieseler-Frank et al. 2005; Verri et al. 2006; Watkins et al. 2007). The pronociceptive effects and underlying mechanisms of these inflammatory mediators have been well documented in the peripheral nervous system (see Verri et al. 2006 for review), but are much less well studied in the central nervous system. The spinal dorsal horn consists of multiple types of neurons including projection neurons, excitatory and inhibitory interneurons, and has long been recognized to play a critical role in nociceptive transmission (Willis & Coggeshall, 2004). Spinal application of pro-inflammatory cytokines induces mechanical allodynia and thermal hyperalgesia (DeLeo et al. 1996; Arruda et al. 1998; Gao et al. 2009), whereas in contrast, spinal administration of neutralizing antibodies to these cytokines prevents the development of inflammatory and neuropathic pain (Arruda et al. 2000; Sweitzer et al. 2001; Schafers et al. 2001, 2003c; Schoeniger-Skinner et al. 2007; Choi et al. 2010).

Although it has recently been found that pro-inflammatory cytokines including TNFα, IL-1β and IL-6 modulate excitatory and inhibitory synaptic transmission in spinal dorsal horn (Kawasaki et al. 2008; Gao et al. 2009), the specific functional subtypes of neurons that are affected, the receptors and their locations, and the second messenger systems involved are not clear. Our previous study has shown that the enhancement of excitatory spinal transmission by TNFα is actually dependent upon suppression of on-going inhibitory synaptic transmission. Furthermore, acute application of TNFα inhibits the activities of spontaneously firing GABAergic neurons in spinal lamina II (Zhang et al. 2010). Since γ-aminobutyric acid (GABA) is a major inhibitory neurotransmitter in spinal dorsal horn and dysfunction of spinal GABAergic inhibitory tone has been shown to be involved in the development and maintenance of neuropathic pain (Yaksh, 1989; Castro-Lopes et al. 1993; Sivilotti & Woolf, 1994; Ibuki et al. 1997; Eaton et al. 1998, 1999; Moore et al. 2002; Baba et al. 2003; Coull et al. 2003; Torsney & MacDermott, 2006), we hypothesized that spinal GABAergic neurons might be cellular targets contributing to the spinal effects of pro-inflammatory cytokines. To this end we tested whether TNFα inhibits the excitability and evoked discharges of spinal GABAergic neurons identified by the transgenic expression of enhanced green fluorescent protein (EGFP) at the glutamic acid decarboxylase 67 (GAD67) promoter site in mice. The cellular mechanisms of the inhibition that was observed were also defined.

Methods

Ethical approval

All experiments were approved by the Institutional Animal Care and Use Committee for the University of Texas MD Anderson Cancer Centre and adhered to the guidelines set forth by the National Institutes of Health Guidelines for the Use and Care of Laboratory Animals and by the Committee for Research and Ethical Issues of the International Association for the Study of Pain (Zimmermann, 1983).

Spinal cord slice preparation

Thirty-four young (3- to 4-week-old) CB6-Tg (Gad1-EGFP) G42Zjh/J mice of either sex were used (Jackson Laboratory, ME, USA). Lumbar spinal cord slices (300 μm) were prepared as described previously (Zhang et al. 2009, 2010). Briefly, mice were deeply anaesthetized with inhaled isoflurane (3%). A laminectomy was performed, the lumbar spinal cord was quickly removed and placed into ice-cold oxygenated (95% O₂+ 5% CO₂) artificial cerebrospinal fluid solution consisting of (in mm): 117 sucrose, 3.6 KCl, 1.2 NaH₂PO₄, 1.2 MgCl₂, 2.5 CaCl₂, 25 NaHCO₃ and 12 glucose. The pia-arachnoid membrane was carefully peeled off and a block of the spinal cord from L3 to S1 was embedded in 4% agar. Transverse slices (300 μm thick) from lumbar segments L4 to L5 were cut on a vibratome (series 1000, Technical Products International Inc., St Louis, MO, USA). The slices were then returned to bubbled Krebs solution (in mm): 117 NaCl, 3.6 KCl, 1.2 NaH₂PO₄, 1.2 MgCl₂, 2.5 CaCl₂, 12 glucose and 25 NaHCO₃ at room temperature (∼22°C) and allowed to equilibrate at least for 1 h before recording. The mice were killed by anaesthetic overdose and exsanguination.

Electrophysiological recording

Whole-cell patch-clamp recordings from spinal dorsal horn neurons were obtained at room temperature as previously described (Zhang et al. 2009, 2010). Cells in substantia gelatinosa were first visualized using a 60× water-immersion objective with infrared and differential interference contrast (DIC) optics (Olympus, BX50WI, Japan). GAD67⁺ neurons were then identified by green fluorescence. Electrode resistances were 3–5 MΩ when filled with pipette solution containing (in mm): 145 potassium gluconate, 5 NaCl, 1 MgCl₂, 0.2 EGTA, 10 Hepes, 2 Mg-ATP and 0.1 Na₃-GTP (pH 7.2 adjusted with KOH). Standard whole-cell patch-clamp recording (Hamill et al. 1981) was made using a Multiclamp 700B amplifier and Clampex 10.0 software (Axon Instruments, Sunnyvale, CA, USA). Signals were filtered at 5–10 kHz and sampled at 10 kHz in digital forms using a Digidata 1322A digitizing board (Axon Instruments) interfaced with a computer system. Current-clamp recordings were made in bridge mode with liquid junction potential corrected. Access resistance, typically 15–30 MΩ, was measured after whole-cell configuration was established and monitored throughout the recording. The cells were abandoned if the access resistance changed more than 20%. The input resistance (R_in) of cells was calculated based on the steady-state current change during application of a 10 mV depolarizing and/or 10 mV hyperpolarizing pulse. The membrane capacitance (C_m) was calculated from the transient currents observed during the application of a 10 mV depolarizing or hyperpolarizing pulse, using a single spherical compartment model. The resting membrane potential (RMP) was recorded in current-clamp mode and only cells with RMP more negative than −45 mV were collected. The firing patterns of neurons were determined by their responses to a series of depolarizing currents injected to the recorded neurons through patch pipette (from 0 to 400 pA with 20 pA increment, 1 s duration). The current threshold was defined as the minimal current that evoked an action potential. The number of action potentials evoked by each 1 s depolarizing current was counted. The voltage dependence of firing patterns was tested with neurons initially held at their RMP and then again when held at −80 mV (Heinke et al. 2004). All chemicals were applied directly into the bath unless otherwise stated. DNQX (10 μm), d—AP5 (25 um), bicuculline (10 μm) and strychnine (5 μm) were routinely added to the bath to block synaptic transmission during the recording.

For voltage-clamp recordings, series resistance was compensated at 50–80%. To record voltage-gated potassium currents, 0.5 μm tetrodotoxin (TTX) and 2 mm CoCl₂ were added into the bath perfusion to block voltage-gated Na⁺ currents, Ca²⁺ currents and Ca²⁺-activated K⁺ currents. To further differentiate the A-type potassium current (K_a) and delayed rectifier potassium current (K_dr), voltage steps of 200 ms were applied at 5 s intervals in 10 mV increments from −80 mV to +70 mV with cells held at either −90 mV or −40 mV. As K_a is inactivated at −40 mV and the sustained K_dr was evoked from −40 mV, a subtraction of the current waveforms elicited at the two holding potentials yielded K_a. To record hyperpolarization-activated cation current (I_h), 0.5 μm TTX, 10 mm tetraethylammonium-Cl (TEA-Cl), 5 mm 4-aminopyridine (4-AP), 0.1 mm CdCl₂ and 1 mm BaCl₂ were added to the bath perfusion. I_h was elicited from a holding potential of −40 mV by voltage steps of 2 s from −120 mV to −40 mV (Maccaferri & McBain, 1996). The data were analysed off-line with leak correction using pClampfit 10.1 (Axon Instruments).

Immunohistochemistry

Mice were transcardially perfused with 4% paraformaldehyde (PFA) and lumber spinal cord segments were extracted and post-fixed in 4% PFA for 4 h. After cryoprotection in 30% sucrose solution, 20 μm transverse spinal cord sections were cut on a cryostat and processed while free-floating for immunohistochemical detection of several markers (Zhang et al. 2010). Slices were sequentially incubated with blocking solution (10% normal donkey serum + 0.2% Triton X-100 in PBS) and primary antibodies. The following primary antibodies were used: TNFR1 (rabbit, 1: 1000, Abcam), TNFR2 (rabbit, 1:500, Abcam), microtubule-associated protein 2 (MAP2, mouse, 1:1000, Millipore), NeuN (mouse, 1:1000, Millipore) and p-p38 (rabbit, 1:100, Cell Signalling Technology). After three washes with PBS, Cy3- or DyLight 649-conjugated secondary antibodies (1:500, Jackson ImmunoResearch) were added. To test the specificity of TNFR1 antibody, a pre-absorption experiment was conducted by incubating TNFR1 antibody (1 μg ml⁻¹) with TNFR1 peptide (5 μg ml⁻¹, Abcam) for 1 h at room temperature before incubating with spinal cord slices. For p—p38 staining, spinal cord slices (300 μm) were first prepared as described in the electrophysiological studies and then incubated with either TNFα (10 ng ml⁻¹) or vehicle for 5 min. Ten slices per mouse were collected for a total of 50 slices processed in this manner. Slices were then postfixed in 4% PFA, cryoprotected and incubated with 10% normal donkey serum and 1% Triton X-100 for 30 min followed by standard steps as detailed above.

Image acquisition and data analysis

All immunohistochemical images were viewed and captured under an Olympus FluoView 500 confocal microscope (Tokyo, Japan). For a given experiment, all images were taken using identical acquisition parameters and final representative figures are presented as the original images without further modification. Background fluorescence was subtracted in all quantitative intensity and cell count analyses. Using a 40× oil-immersion objective, three to five circular regions in the substantia galetinosa adjacent to immune-positive neurons were averaged to obtain a mean background value for each slice. The mean fluorescent intensity less background for the whole of spinal lamina II as well as the percentage of individual p-p38-positive neurons using comparative numbers of total GAD67⁺ neurons between groups was determined. All images were analysed using NIC Elements imaging software (Nikon, Japan).

Reagents

6,7-Dinitroquinoxaline-2, 3-dione (DNQX) and d-2-amino-5-phosphonopentanoic acid (d-AP5) were obtained from Tocris (St Louis, MO, USA). Recombinant mouse TNFα was purchased from R&D System (Minneapolis, MN, USA) and prepared as a stock solution at 10 μg ml⁻¹ in PBS with 0.1% BSA. SB202190 (4-(4-fluorophenyl)-2-(4-hydroxyphenyl)-5-(4-pyridyl)-1H-imidazole, C₂₀H₁₄FN₃O) was obtained from Calbiochem (La Jolla, CA, USA) and dissolved in DMSO (final concentration less than 0.1%). All other chemicals were obtained from Sigma-Aldrich (St Louis, MO, USA) unless otherwise noted.

Statistical analysis

All results are presented as means ± SEM. Differences between means were tested for significance using paired or non-paired t test, Wilcoxon rank-sum or Wilcoxon signed-rank test where appropriate with an alpha value of P < 0.05.

Results

Electrophysiological properties of spinal cord lamina II GAD67⁺ neurons

GAD67⁺ neurons in spinal lamina II were identified by visualization of green fluorescence (Fig. 1A). Whole-cell configuration was established and the passive membrane properties and spike-burst responses to intracellular current injection were measured. These data are summarized in Fig. 1 and Table 1. The cells were categorized by the patterns of action potential discharges that were evoked by intracellular depolarizing currents (Prescott & DeKoninck, 2002; Grudt & Perl, 2002; Ruscheweyh & Sandkuhler, 2002; Szûcs et al. 2003). Each cell was initially held at its resting membrane potential while tested with intracellular current pulses (n = 50). Four different patterns of discharges were observed. Representative cells for each firing pattern are shown in Fig. 1. Tonic firing cells (n = 32), those showing continuous, regular action potential discharges throughout the depolarization interval were the most commonly observed (Fig. 1B), consistent with previous studies of GAD67⁺ neurons in spinal cord (Hantman et al. 2004; Dougherty et al. 2005; Wilson et al. 2010). Initial bursting cells (n = 13), those showing a self-terminating burst of action potential discharges at the onset of depolarization, were the next most commonly observed (Fig. 1C). Most of the spike bursts in these cells showed a gradual reduction in action potential amplitude throughout the burst (11/13). Delayed firing cells (n = 3), those with a prolonged non-spiking interval prior to the first action potential discharge following membrane depolarization, were the third most commonly observed (Fig. 1D). Finally, single spike cells (n = 2), those showing only one action potential discharge at the onset of depolarization, were the least frequently observed (Fig. 1E). Gap firing cells, those showing a similar firing pattern as the delayed firing cell to suprathreshold current injections but with a long first interspike interval followed by tonic firing to stronger depolarizing currents, were not observed. All cells showed graded action potential responses with current magnitude except the single firing cells (Fig. 1B–E). The passive membrane properties were compared only between tonic firing and initial bursting neurons due to the limited numbers of delayed firing and single spike neurons that were sampled. No differences in membrane properties were observed between tonic firing and initial bursting neurons based on the samples obtained here (Table 1).

Figure 1. Firing patterns of GAD67⁺ GABAergic neurons in spinal lamina II.

A, image of GAD67⁺ GABAergic neurons in spinal lamina II under confocal microscope (a) and DIC view (b). Scale bar: 100 μm. Firing pattern of the recorded cell is defined by its response to a series of depolarizing currents injected from the soma (0–400 pA with 20 pA increment, 1 s duration). B–E, representative traces of typical firing patterns are shown as tonic firing (B), initial bursting (C), delayed firing (D) and single spike (E). Insets show traces of three injected currents to each representative cell. F, distributions of firing patterns when cells were held at either their RMP or −80 mV.

Table 1.

The passive membrane properties of GAD67⁺ GABAergic neurons in spinal lamina II

	Recorded with normal pipette solution				Recorded with SB202190 in pipette
	Tonic firing	Initial bursting	P	Delayed firing	Single spike	Tonic firing	Initial bursting	P
Cm (pF)	42.1 ± 2.5	42.0 ± 5.4	0.98*	37.3 ± 10.6	19.0 & 41.2	43.5 ± 3.0 (P = 0.77*)	42.7 ± 2.2 (P = 0.58†)	0.92*
Rin (MΩ)	458.7 ± 47.4	350.8 ± 62.0	0.16†	499.4 ± 58.2	337.9 & 218.7	449.0 ± 35.2 (P = 0.47†)	380.2 ± 43.6 (P = 0.34†)	0.32*
RMP (mV)	−50.7 ± 1.1	−52.8 ± 2.4	0.35*	−61.3 ± 1.9	−42.0	−52.7 ± 1.4 (P = 0.26*)	−53.6 ± 2.6 (P = 0.85*)	0.75*
CT (pA)	39.7 ± 4.8	48.5 ± 6.5	0.14†	76.7 ± 38.5	100.0 & 60.0	34.8 ± 3.1 (P = 0.66†)	42.9 ± 5.2 (P = 0.57*)	0.21†
N	32	13			3	2	23	7

Values are mean ± SEM. P value in the column: tonic vs. initial bursting neurons recorded with the same pipette solution. P value in parentheses: recorded with vs. without SB202190 in pipette solution for the same type of neurons.

^*Student's t test

^†Wilcoxon rank-sum test.

C_m, membrane capacitance; R_in, input resistance; RMP, resting membrane potential; CT, current threshold for action potential.

It has been suggested that both delayed firing and gap firing neurons are only observed when held more negative than their RMPs, presumably due to the partial inactivation of A-type potassium currents (Heinke et al. 2004). Since the RMPs of most neurons recorded were more positive than −55 mV (−51.6 ± 1.0 mV, n = 50), the effects of membrane hyperpolarization to −80 mV on the firing patterns were further tested in 26 GAD67⁺ neurons (Fig. 1F). Of the 15 cells that showed a tonic firing pattern at RMP, seven now changed signalling phenotype to an initial bursting firing pattern when held at –80 mV. All eight cells with an initial bursting discharge pattern at RMP maintained this pattern when hyperpolarized. A few single spike and delayed firing but not gap firing neurons were observed among the cells sampled in this study. These did not change signalling phenotype when hyperpolarized.

TNFα inhibits the excitability of tonic firing GAD67⁺ neurons

We have shown that bath application of TNFα (10 ng ml⁻¹, 2–3 min) induces a marked suppression of spinal inhibitory synaptic transmission (Zhang et al. 2010). The same concentration and application time of TNFα was again used in this study. The passive membrane properties and excitability of GAD67⁺ neurons were tested before and after the application of TNFα. Only one cell was tested in each slice. As shown in Table 2, TNFα significantly increased rheobase and decreased the input resistance in tonic firing neurons. The resting membrane potential of this group also became more negative following TNFα. In 12 tonic firing neurons tested, 11 neurons switched to initial bursting pattern after TNFα application. The recording of a representative neuron is shown in Fig. 2A. The number of action potentials was plotted against injected current as summarized in Fig. 2B. TNFα significantly decreased the number of current-evoked action potentials in the sampled GAD67⁺ tonic firing neurons.

Table 2.

The effect of TNFα on passive membrane properties of GAD67⁺ GABAergic neurons in spinal lamina II

	Neurons recorded with normal pipette solution					Neurons recorded with SB202190 in pipette
	Tonic firing			Initial bursting		Delayed firing		Tonic firing
	Before	After	P	Before	After	Before	After	Before	After	P
Cm (pF)	48.3 ± 2.9	44.7 ± 3.6	0.07	44.3 ± 13.3	44.8 ± 13.1	58.1	58.6	51.7 ± 3.5	49.5 ± 3.6	0.10
Rin (MΩ)	461.3 ± 40.9	272.3 ± 56.4	< 0.01	366.0 ± 143.7	441.4 ± 167.0	426.7	118.3	370.4 ± 36.1	332.4 ± 31.2	0.06
RMP (mV)	−51.6 ± 1.5	−61.8 ± 1.6	<0.01	−51.0 ± 3.8	−54.3 ± 3.5	−65.0	−75.0	−51.9 ± 1.8	−52.6 ± 1.9	0.39
CT (pA)	23.3 ± 1.9	65.0 ± 13.0	0.01	33.3 ± 13.3	26.7 ± 6.7	20.0	100.0	28.6 ± 3.5	35.7 ± 5.6	0.19
N	12	12		3	3	1	1	14	14

Values are mean ± SEM. P values: recorded before vs. after TNFα treatment in tonic firing neurons (paired t test). C_m, membrane capacitance; R_in, input resistance; RMP, resting membrane potential; CT, current threshold for action potential; AP, action potential.

Figure 2. TNFα decreases the excitability of tonic firing GAD67⁺ GABAergic neurons in spinal lamina II.

A, responses of a representative neuron to the same injected currents (insets) before (left panel) and after (right panel) the application of TNFα (10 ng ml⁻¹, 3–5 min). TNFα increases the current threshold (bottom traces), decreases the number of action potentials to current injections and switches the firing pattern of the recorded cell from tonic firing to initial bursting (middle and upper traces). B, summarized data of neuronal responses to injected currents before and after application of TNFα. TNFα significantly decreases the number of action potentials evoked by a series of depolarizing currents in tonic firing GAD67⁺ GABAergic neurons (n = 11). Values represent mean ± SEM. P < 0.001, Wilcoxon signed-rank test for slopes of each sample before and after the application of TNFα. C, there is no significant change detected on initial bursting GAD67⁺ GABAergic neurons by TNFα (n = 3). Values represent mean ± SEM.

The effect of TNFα was also tested on three initial bursting cells. No changes in passive membrane properties (Table 2) or evoked discharges (Fig. 2C) were observed in this small sample. Only one delayed firing cell was tested and it showed similar changes after TNFα to that observed for tonic neurons, including an increase in current threshold for action potentials, a decrease in input resistance (Table 2) and a decrease in the number of evoked action potentials (data not shown).

The inhibitory effect of TNFα on spinal GAD67⁺ GABAergic neurons depends on intracellular activation of p38 MAPK

We have shown that p38 MAPK is predominantly activated in spinal dorsal horn neurons and contributes to the altered spinal synaptic transmission induced by application of TNFα (Zhang et al. 2010). The role of p38 in the response of GAD67⁺ neurons to TNFα was explored using immunohistochemistry and whole-cell patch-clamp recordings. As shown in Fig. 3, a modest baseline expression of activated p38 (phosphorylated, p-p38) was confirmed to be present in superficial dorsal horn (Svensson et al. 2008; Zhang et al. 2010) (Fig. 3C). Incubation of spinal cord slices with TNFα (10 ng ml⁻¹ in Krebs solution for 5 min) induced a marked increase of p-p38 expression in spinal dorsal horn (Fig. 3D), with numerous GAD67⁺ neurons showing strong activation of p38 (Fig. 3F). Double immunohistochemical staining confirmed that the expression of p-p38 was predominantly in spinal dorsal horn neurons as opposed to glial cells (Fig. 3G). To compare the expression of p-p38, both the total level of p-p38 and the number of GAD67⁺ neurons expressing p-p38 in spinal lamina II were quantified. As shown in Fig. 3H, incubation of TNFα significantly increased both the mean intensity of p-p38 staining (from 5.7 ± 0.3 to 23.1 ± 2.2, n = 21 slices for each group, P < 0.001) and the mean percentage of GAD67⁺ neurons expressing p-p38 (from 13.6 ± 2.3% to 65.2 ± 1.8%, n = 21 slices for each group, P < 0.001) in spinal lamina II. A total of 367 GAD67⁺ neurons were counted in spinal lamina II for control animals and 434 were sampled in the TNFα-treated groups.

Figure 3. TNFα activates p38 MAPK (p-p38) in spinal dorsal horn.

A and B, GAD67⁺ GABAergic neurons in spinal lamina II. C, a baseline expression of p-p38 isdetected in spinal dorsal horn in naïve mice. D, a robust increase in p-p38 expression in spinal dorsal horn is detected after spinal slices are incubated with TNFα (10 ng ml⁻¹, 5 min). E (merged image of A and C) and F (merged image of B and D), shows TNFα (F) but not vehicle (E) induces a strong activation of p38 in some GAD67⁺ GABAergic neurons (arrows and inset). G, double immunohistochemical staining shows that activation of p38 is mostly in spinal dorsal neurons after TNFα stimulation. H, quantification of p-p38 expression in spinal dorsal horn. TNFα significantly increases both the total level of p-p38 and the percentage of GAD67⁺ GABAergic neurons expressing p-p38 in spinal lamina II. Values represent mean ± SEM. *P < 0.001, Student's t test. Scale bar: 20 μm in A–F; 10 μm in F inset.

Since the activation of p38 is not only in GAD67⁺ neurons but also in non-GAD67⁺ neurons (Fig. 3F and G) and some spinal astrocytes (Zhang et al. 2010), to further test the involvement of p38 within GAD67⁺ neurons in mediating the effects of TNFα a p38 MAPK blocker, SB202190, was delivered intracellularly into GAD67⁺ neurons through the patch pipette. Neuronal excitability and firing patterns were first tested when SB202190 (20 μm) was added to the patch pipette solution to intracellularly block the activation of p38. Comparison of the passive membrane properties of neurons recorded with SB202190 with neurons of the same firing pattern but recorded with the normal pipette solution did not reveal any clear differences (Table 1). The distribution in firing patterns of GAD67⁺ neurons was similar when recorded with SB202190 (Fig. 4A). However, when SB202190 was delivered intracellularly, further application of TNFα failed to inhibit the excitability or change the firing patterns in 8 of 12 tonic firing GAD67⁺ neurons (Fig. 4B and C), suggesting that the inhibitory effect of TNFα is largely dependent on the activation of a p38 MAPK-dependent pathway within the recorded GAD67⁺ GABAergic neurons. Cells with other firing patterns have not yet been tested given their general lack of sensitivity to TNFα as described above.

Figure 4. Intracellular delivery of SB202190 (20 μm), a p38 blocker, prevents the inhibitory effect of TNFα on tonic firing GAD67⁺ GABAergic neurons in spinal lamina II.

A, distributions of firing patterns when pipette solution contained SB202190 (20 μm). SB202190 did not change the distribution of firing patterns. B, responses of a representative neuron to the same injected currents (insets) before (left panel) and after (right) the application of TNFα (10 ng ml⁻¹, 3–5 min). TNFα shows no inhibitory effect on the excitability of the representative GAD67⁺ GABAergic neuron. C, summarized data show that intracellular delivery of SB202190 (20 μm) prevents the inhibitory effect of TNFα on tonic firing GAD67⁺ GABAergic neurons (n = 8). Values represent mean ± SEM.

TNFα receptor 1 mediates the inhibitory effect of TNFα on GAD67⁺ GABAergic neurons

TNFα receptors are generally expressed in spinal cord of naïve animals (Holmes et al. 2004; Ohtori et al. 2004) with TNFR1 being the major receptor mediating the downstream effect of TNFα (Sommer et al. 1998; Parada et al. 2003; Cunha et al. 2005; Jin & Gereau, 2006; Zhang et al. 2010). To examine the expression of TNFα receptors, immunohistochemical staining of TNFR1 and TNFR2 was performed in EGFP-GAD67 mice. Pre-incubation of TNFR1 peptide and TNFR1 antibody prevented the specific staining of TNFR1 in spinal cord (Fig. 5B), confirming the specificity of the antibody. TNFR1 was found throughout the spinal cord grey matter while TNFR2 was expressed only in ventral horn of naïve mice (Figs 5A and 6A, respectively). It was further observed that TNFR1 was expressed on a subset of GAD67⁺ neurons in spinal lamina II (Fig. 5D–G). In contrast, TNFR2 was only expressed in large ventral horn neurons but not found localized to any GAD67⁺ neurons in either dorsal or ventral horn (Fig. 6C–H). These data suggest that the activation of intracellular p38 MAPK and the following decreased excitability of GAD67⁺ GABAergic neurons by TNFα is medicated by TNFR1 not TNFR2.

Figure 5. Localization of TNFR1 in spinal cord of naïve mice.

A, expression of TNFR1 was throughout the whole spinal grey matter. No signal was detected when either the blocking peptide was pre-incubated with the primary antibody (B) or the primary antibody was omitted (C). D–G, co-localization of TNFR1 (red), MAP2 (pseudo-coloured blue) and one GAD67⁺ GABAergic neurons (green). Scale bar: 100 μm in A–C; 10 μm in D–G.

Figure 6. Localization of TNFR2 in spinal cord of naïve mice.

A, expression of TNFR2 is only detected in spinal ventral horn. B, no signal was detected when the primary antibody to TNFR2 was omitted as a negative control. Insets in A and B are DIC view of the slices. C–E, no expression of TNFR2 was found in spinal dorsal horn and none of GAD67⁺ GABAergic neurons in spinal dorsal horn expressed TNFR2. F–H, TNFR2 is expressed predominantly in large neurons of spinal cord ventral horn, but none of GAD67⁺ GABAergic neurons in ventral horn expressed TNFR2. Scale bar: 100 μm in A and B and insets: 60 μm in C–H.

TNFα inhibits the excitability of tonic firing GAD67⁺ GABAergic neurons through down-regulating I_h currents

The comparison of action potential spikes in GAD67⁺ neurons before and after the application of TNFα revealed a prolonged interspike interval, a reduced slope of the depolarization ramp leading to action potential threshold and only minimal effects on the properties of action potential waveforms, suggesting an effect of TNFα on intrinsic currents contributing to action potential pacemaker activity. Three different currents possibly contributing to rhythmic activities of spinal dorsal horn neurons, including A-type potassium current (K_a), delayed rectifier potassium current (K_dr) and hyperpolarization-activated cation current (I_h), were examined separately.

Voltage-dependent potassium currents were evoked by voltage steps when cells were held at two different potentials (Tan et al. 2006). Only 3 out of 16 spinal GAD67⁺ neurons showed clear K_a in voltage-clamp recordings (Fig. 7Aa). Acute application of TNFα did not evoke changes in K_a currents in these cells (Fig. 7Ac). K_dr currents were compared before and after application of TNFα by measuring the steady-state currents during depolarizing steps (Fig. 7Ab and e). Eleven out of 16 GAD67⁺ neurons showed a slight but not significant increase in K_dr following TNFα (Fig. 7Af). I_h current was tested in another set of GAD67⁺ neurons. Nine out of 11 cells showed I_h currents and TNFα significantly decreased this current in seven cells (Fig. 7B). These data suggest that TNFα inhibits the excitability of GAD67⁺ GABAergic neurons mainly through suppressing I_h currents.

Figure 7.

TNFα down-regulates I_h currents of GAD67⁺ neurons. A, K_a and K_dr were not significantly changed by TNFα. Total voltage-gated potassium currents were first elicited by a series of 200 ms voltage steps ranging from −80 mV to +70 mV in 10 mV increments when cells were held at −90 mV (Aa). K_dr currents were then elicited by the same voltage steps when cells were held at −30 mV (Ab). K_a currents were obtained by subtracting K_dr currents in Ab from total currents in Aa. The voltage-dependent activation of K_a currents was then obtained by plotting peak currents against voltage steps and is shown in Ac. The same voltage steps were tested on the same cell after the application of TNFα (Ad and Ae). The voltage-dependent activation of K_dr was obtained by plotting peak currents against voltage steps before (Ab) and after (Ae) the application of TNFα and is displayed in Af. B, I_h currents were elicited by a series of 2 s voltage steps ranging from −120 mV to −40 mV in 10 mV increments when cells were held at −40 mV before (Ba) and after (Bb) the application of TNFα. Comparison of voltage-dependent activation of I_h shows that TNFα significantly decreased I_h currents on GAD67⁺ neurons (Bc). P < 0.05, Wilcoxon signed-rank test for slopes of each sample before and after the application of TNFα.

Discussion

Membrane properties of GAD67⁺ GABAergic neurons in spinal lamina II

The passive membrane properties of GAD67⁺ neurons we recorded in spinal lamina II are similar to that of spinal lamina II islet cells that were considered as inhibitory interneurons (Grudt & Perl, 2002; Maxwell et al. 2007). In fact, about 70% GAD67⁺ neurons in spinal lamina II are islet cells (Heinke et al. 2004). The diversity in the firing patterns of lamina II GAD67⁺ neurons, including tonic, initial bursting, delayed and single spike firing, is consistent with previous reports of GAD67⁺ neurons in both superficial (Heinke et al. 2004; Schoffnegger et al. 2006) and deeper spinal lamina (Wilson et al. 2010) when cells are held at RMP. In the present study, the tonic spike burst response was the most common firing pattern observed, followed by the initial burst response when cells were depolarized from their resting membrane potentials. Only few neurons showed either a delayed or single spike firing pattern. This general distribution of firing patterns is in agreement with previous studies on spinal islet neurons (Grudt & Perl, 2002; Maxwell et al. 2007; Yasaka et al. 2010) and also on identified GAD67⁺ neurons (Hantman et al. 2004; Dougherty et al. 2005; Wilson et al. 2010). There were no differences observed here in passive membrane properties between tonic, initial bursting, delayed firing or single spike neurons, though this analysis may be underpowered due to the limited number of non-tonic firing neurons that were sampled.

Of note, when the possible effect of A-type potassium currents on the generation of action potentials in GAD67⁺ neurons was eliminated by hyperpolarization in a group of neurons (held at −80 mV) before current injections were initiated (Heinke et al. 2004; Schoffnegger et al. 2006), a switch of phenotype from tonic firing to initial bursting pattern was often observed in many cells and resulted in the initial bursting becoming the most common firing pattern. This phenotype change in spinal superficial neurons induced by hyperpolarization has also been reported in previous studies (Ruscheweyh et al. 2004). The small number of cells with either delayed or gap firing patterns in both recording conditions suggest that A-type potassium currents are not a common property of spinal GAD67⁺ neurons (Yasaka et al. 2010).

Plasticity of GAD67⁺ GABAergic neurons in spinal lamina II

In the present study, the excitability of tonic firing GAD67⁺ neurons and a delayed firing neuron, but not other subtypes, was inhibited by TNFα. This result is tempered by the relatively low numbers of non-tonic firing neurons that were sampled, yet the majority of GAD67⁺ neurons show the tonic firing signal phenotype (Hantman et al. 2004; Dougherty et al. 2005; Wilson et al. 2010). Hence, the inhibitory effect of TNFα on the dominant tonic subtype will result in decreased inhibitory tone in the spinal cord and therefore alter the spinal processing of innocuous and noxious afferent inputs from peripheral receptive fields. Indeed, acute application of TNFα has been shown to enhance excitatory synaptic transmission in spinal dorsal horn (Kawasaki et al. 2008) and this enhancement of spinal excitatory synaptic transmission is due to an inhibition of on-going inhibitory synaptic transmission (Zhang et al. 2010). Since TNFα, as a major pro-inflammatory cytokine, plays a crucial role in the development of both inflammatory and neuropathic pain (Cunha et al. 1992; Perkins & Kelly, 1994; Woolf et al. 1997; Schäfers et al. 2003a,b; Wacnik et al. 2005; Jin & Gereau, 2006; Gao et al. 2009), our data here provide a novel mechanism of spinal disinhibition in the context of elevated spinal pro-inflammatory cytokine levels. In agreement with the role of local spinal inhibition, blocking spinal inhibitory tone by topical application of bicuculline or strychnine causes hyperalgesia and allodynia (Yaksh, 1989; Sivilotti & Woolf, 1994; Miraucourt et al. 2009). Loss of spinal GABAergic inhibition has been found after peripheral nerve injury (Yaksh, 1989; Castro-Lopes et al. 1993; Sivilotti & Woolf, 1994; Ibuki et al. 1997; Eaton et al. 1998, 1999; Moore et al. 2002; Scholz et al. 2005) and spinal cord injury (Drew et al. 2004; Lu et al. 2008), but it has also been claimed that selective loss of spinal inhibitory neurons is not necessary for the development of thermal hyperalgesia after nerve injury (Polgar et al. 2003). A recent study suggests that a subset of spinal inhibitory neurons controls the formation of itch as well (Ross et al. 2010).

The present study confirms the activation of the TNFα–TNFR1–p38 pathway in TNFα-induced rapid inhibition of spinal lamina II GAD67⁺ neurons. Although it cannot be completely excluded that other non-GAD67⁺ neurons or spinal astrocytes may also play some role in mediating the altered excitability of GAD67⁺ neurons after TNFα application, the fact that intracellular delivery of p-p38 blocker largely blocked the effect of TNFα on GAD67⁺ neurons strongly suggest the direct involvement of p38 within GAD67⁺ neurons. Activation of intracellular p38 MAPK can modulate sodium channels and rapidly change neuronal excitability (Jin & Gereau, 2006). Activated p38 is also translocated to the nucleus and activates transcriptional factors such as activating transcription factor 2, c-Jun and cAMP response element-binding protein (Ji et al. 2009). In addition, the long-term effects of TNFα include the induction of apoptosis by sequentially activating caspase-8/-10 and caspase-3. It has been shown that blocking caspase activity could prevent the neuronal apoptosis, including loss of GABAergic neurons and loss of inhibition in spinal lamina II after peripheral nerve injury, suggesting the involvement of apoptosis in spinal disinhibition (Scholz et al. 2005).

Possible ionic mechanisms of TNFα-induced hypoexcitability in GAD67⁺ GABAergic neurons

I_h current is a mixed non-selective cation current that typically activates with hyperpolarization steps and slowly depolarizes the cell toward the equilibrium potential (Pape, 1996). One fundamental function of I_h current is contributing to pacemaker activity (spontaneous repetitive action potential firing or rhythmic-oscillatory activity) in CNS neurons (McCormick & Pape, 1990; Pape, 1996; Maccaferri & McBain, 1996; Gasparini & DiFrancesco, 1997; Luthi & McCormick, 1998). A gene family encoding the mammalian hyperpolarization-activated and cyclic-nucleotide-gated non-selective cation channels (HCN) has been cloned and shown to generate I_h current (Ludwig et al. 1998; Santoro et al. 1998). Four subunits of HCN (HCN 1–4) have been found and each subunit can form homomeric or heteromeric channels with different functional properties. It has been found that HCN2 but not HCN1 is widely expressed in spinal dorsal horn, revealed by in situ hybridization and immunohistochemical approaches (Santoro et al. 2000; Milligan et al. 2006; Papp et al. 2006). I_h current has been recorded in most of spinal inhibitory interneurons including GAD67⁺ neurons in several different studies (Grudt & Perl, 2002; Hantman et al. 2004; Wilson et al. 2010; Zhang et al. 2010). Our data showed that TNFα induced a downregulation of I_h currents without significantly changing K_a or K_dr currents in GAD67⁺ spinal neurons. This downregulation of I_h currents is consistent with other changes of characteristics in GAD67⁺ neurons, including the negative shift of RMP and increased interspike intervals during firing in the present study and the decrease of ‘sag’ recorded in bridge-mode in spontaneously firing GAD67⁺ neurons in our previous study (Zhang et al. 2010). Studies have shown that the biophysical properties including voltage dependence and activation kinetics of HCN channels can be modulated by cyclic adenosine monophosphate (cAMP), protons and chloride ions through direct combination or indirect interaction with these channels (for review see Wahl-Schott & Biel, 2009). How TNFα might inhibit the activity of I_h currents of spinal GAD67⁺ neurons through the activation of p38 MAPK remains to be further determined.

K_a current also plays an important role in modulating neuronal excitability, but the low expression of this current in GAD67⁺ neurons makes it an unlikely mediator of cytokine-induced alteration of excitability in these cells. The slight increase of K_dr current induced by TNFα could also contribute to the decrease of input resistance in GAD67⁺ neurons.

In summary, we have shown that TNFα inhibits the excitability of tonic firing GAD67⁺ neurons in spinal dorsal horn. This inhibition is through TNFR1 receptors and largely depends on the activation of p38 MAPK. We also provide evidence that the decreased excitability of GAD67⁺ neurons involves suppression of I_h currents by TNFα. These data provide a novel cellular and molecular mechanism of spinal disinhibition induced by TNFα.

Glossary

Abbreviations

BSA

bovine serum albumin

C_m

membrane capacitance

DIC

differential interference contrast

EGFP

enhanced green fluorescent protein

GAD67

glutamic acid decarboxylase 67

HCN

hyperpolarization-activated and cyclic-nucleotide-gated non-selective cation channels

I_h

hyperpolarization-activated cation current

IL-1β

interleukin 1-β

IL-6

interleukin 6

K_a

A-type potassium current

K_dr

delayed rectifier potassium current

MAP2

microtubule-associated protein 2

MAPK

mitogen-activated protein kinase

PFA

paraformaldehyde

R_in

input resistance

RMP

resting membrane potential

TNFα

tumour necroscis factor-α

TNFR1

TNFα receptor 1

Author contributions

H.Z. led the design and performed the experiments, analysed the data and drafted the manuscript. P.D. oversaw all aspects of the project and critically revised the manuscript. Both authors approved the final version.