Acute inflammation reveals GABAA receptor‐mediated nociception in mouse dorsal root ganglion neurons via PGE 2 receptor 4 signaling

Abstract Gamma‐aminobutyric acid (GABA) depolarizes dorsal root ganglia (DRG) primary afferent neurons through activation of Cl− permeable GABAA receptors but the physiologic role of GABAA receptors in the peripheral terminals of DRG neurons remains unclear. In this study, we investigated the role of peripheral GABAA receptors in nociception using a mouse model of acute inflammation. In vivo, peripheral administration of the selective GABAA receptor agonist muscimol evoked spontaneous licking behavior, as well as spinal wide dynamic range (WDR) neuron firing, after pre‐conditioning with formalin but had no effect in saline‐treated mice. GABAA receptor‐mediated pain behavior after acute formalin treatment was abolished by the GABAA receptor blocker picrotoxin and cyclooxygenase inhibitor indomethacin. In addition, treatment with prostaglandin E2 (PGE 2) was sufficient to reveal muscimol‐induced licking behavior. In vitro, GABA induced sub‐threshold depolarization in DRG neurons through GABAA receptor activation. Both formalin and PGE 2 potentiated GABA‐induced Ca2+ transients and membrane depolarization in capsaicin‐sensitive nociceptive DRG neurons; these effects were blocked by the prostaglandin E2 receptor 4 (EP4) antagonist AH23848 (10 μmol/L). Furthermore, potentiation of GABA responses by PGE 2 was prevented by the selective Nav1.8 antagonist A887826 (100 nmol/L). Although the function of the Na+‐K+‐2Cl‐ co‐transporter NKCC1 was required to maintain the Cl‐ ion gradient in isolated DRG neurons, NKCC1 was not required for GABAA receptor‐mediated nociceptive behavior after acute inflammation. Taken together, these results demonstrate that GABAA receptors may contribute to the excitation of peripheral sensory neurons in inflammation through a combined effect involving PGE 2‐EP4 signaling and Na+ channel sensitization.


Introduction
Doral root ganglion (DRG) neurons are primary afferent neurons which conduct sensory information from the environment to the spinal cord. DRG neurons express the Clpermeable gamma-aminobutyric acid (GABA) A receptor (Morris et al. 1983;Farrant and Nusser 2005;Zeilhofer et al. 2012) which display membrane depolarization in response to GABA stimulation. It is suggested that GABA A receptor-mediated tonic primary afferent depolarization (PAD) may serve to inhibit primary afferent signaling through the inactivation of voltage-gated channels such as Na v and thus reduce neurotransmitter release from the afferent terminals to second-order neurons (Willis 1999;Kullmann et al. 2005;Guo and Hu 2014). In addition, GABA may itself exert an inhibitory effect via activation of metabotropic GABA B receptors expressed on DRG neurons (Hanack et al. 2015).
It has previously been demonstrated that co-application of a low dose (2 lmol/L, 30 lL) of the GABA A receptor agonist muscimol with formalin into the mouse hind paw reduced formalin-induced biphasic nocifensive behavior (Carlton et al. 1999), a result consistent with the inhibition associated with PAD. However, a high dose of muscimol (1 mmol/L) was in fact found to increase formalin-induced biphasic nocifensive behavior (Carlton et al. 1999;Bravo-Hernandez et al. 2014), a phenomenon that was blocked by pre-treatment with the GABA A receptor antagonist bicuculline (Bravo-Hernandez et al. 2014). These findings suggest that near-maximally activated GABA A receptors may participate in nociceptive sensory transduction in pathological conditions. However, the molecular mechanism(s) underlying the contribution of peripheral GABA A receptors to inflammatory pain remain unclear.
In the present study, we sought to investigate the mechanism behind this apparent shift in the role of peripheral GABA A receptors in acute inflammation. We found that peripheral GABA A receptor activation induces de novo pain behavior after formalin and prostaglandin E2 (PGE 2 ) pre-conditioning through a signaling pathway involving EP4 receptor activation. We also found that functional upregulation of tetrodotoxin-resistant voltagegated sodium channels in the presence of PGE 2 allows GABA signaling to evoke robust neuronal activity under inflammatory conditions. These findings may open new avenues of research into the contribution of GABA A receptors to sensory physiology.

Ethical approval
All surgical and experimental procedures were reviewed and approved by the Institutional Animal Care and Use Committees of Seoul National University (SNU-120710-5-1) and the National Institute of Physiological Sciences (NIPS), Japan. Experiments were carried out and are reported here in accordance with the ARRIVE guidelines (Kilkenny et al. 2010) and the principles and best practice of The Journal of Physiology (Grundy 2015).

Animals
Adult C57BL/6J (wild type) male mice (6-8 weeks) were purchased from Daehan Biolink (Korea) and Japan SLC (Hamamatsu, Japan). Adult male and female Na + -K + -2Cl À co-transporter 1 knockout (NKCC1 À/À ) mice were from a stock originally created by Dr. Gary Shull of the University of Cincinnati (Flagella et al. 1999) and maintained at the Department of Physiology, Korea University, Seoul. Animals were housed four to five mice per cage in a conventional facility with a 12 h light cycle (lights on 8.00 AM) and ad libitum access to water and chow. Mice were acclimatized for at least 1 week before experiments. All behavioral experiments were carried out between 12.00 PM and 6.00 PM. Drug treatments were assigned randomly to mice of the same litter by an independent observer. Mice were euthanized at the end of behavior experiments by rising concentration of CO 2 , or by overdose of urethane followed by cervical dislocation at the end of in vivo recording, in accordance with the Schedule 1 of the UK Home Office Animals (Scientific Procedures) Act 1986.
Formalin pre-conditioning and spontaneous pain behavior Mice were placed in an observation chamber (60 9 100 9 60 mm) and allowed to habituate for at least 30 min before drug administration. A mirror was positioned behind the observation chamber to provide an unobstructed view. After habituation, one experimenter restrained the mouse while another experimenter performed a 'pre-conditioning' injection of formalin (0.8%) or saline vehicle (20 lL) subcutaneously into the dorsum of the right hind paw using a 0.3 mL insulin syringe fitted with a 31 gauge needle. After subsidence of formalin-induced nocifensive behaviors (70 min after first injection) a second 'stimulating' injection of muscimol (1 mmol/L) or saline vehicle (20 lL) was given subcutaneously to the same dorsum area of the hind paw. For inhibition of GABA A receptors in vivo, picrotoxin was added at the indicated concentration to both the initial formalin solution and muscimol stimulation solution before paw injection. Care was taken with both injections to avoid leakage of the solutions from the paw. Spontaneous pain behaviors were assessed by measuring the time each animal spent licking the affected hind paw. The cumulative time spent licking was recorded during the 5 min immediately before drug administration and up to 110 min after the first drug administration. All behavior tests were video recorded and analyzed offline by an investigator who was blind to the treatment and genetic background of the mice.

In vivo spinal cord extracellular recording
The methods used for the present study were modifications of those used in preceding studies (Sugiyama et al. 2012;Funai et al. 2014). Adult C57BL/6J male mice were anesthetized with urethane (1.5 g/kg, i.p.) and monitored for loss of hind paw pinch reflex with supplemental injections of urethane (0.2 g/kg, i.p.) given if necessary during the experiment. A laminectomy was performed to expose the lumbar enlargement of the spinal cord. The mouse was placed in a stereotaxic apparatus (Model STS-A, Narishige, Tokyo, Japan). After the dura mater was opened, the pia-arachnoid membrane was cut to make a window to allow a tungsten electrode to enter into the spinal cord.
The surface of the spinal cord was irrigated with 95% O 2 and 5% CO 2 equilibrated Krebs solution (in mmol/L: 117 NaCl, 3.6 KCl, 2.5 CaCl 2 , 1.2 MgCl 2 , 1.2 NaH 2 PO 4 , 11 glucose, and 25 NaHCO 3 ) at a flow rate of 10-15 mL/ min at 38°C AE 1°C. The tungsten electrode (impedance, 1 MX, A-M systems, Sequim, WA) was advanced into the spinal cord using a micromanipulator (Model SM-11, Narishige). The tungsten electrode was placed into the spinal cord dorsal horn and action potentials in spinal cord neurons were extracellularly recorded with an AC differential amplifier (DAM 80,World Precision Instruments,Sarasota,FL). Wide dynamic range (WDR) neurons with a receptive field covering the ipsilateral hind paw were identified during electrode advancement by consistent spike responses to touch, brush and pinch stimuli applied to the hind paw. The firing rate of spinal cord neurons was analyzed with Offline Sorter software (version 3, Plexon, Dallas, TX). Injections were made into the dorsal surface of the ipsilateral hind paw in the same manner as for the behavioral experiments.

DRG preparation
DRG neurons were isolated from 6 to 8-week-old mice. Animals were killed in accordance with the Schedule 1 of the UK Home Office Animals (Scientific Procedures) Act 1986 by inhalation of a rising, lethal concentration of isoflurane (Hana Pharm. Co. Ltd., Korea) followed by decapitation. Bilateral DRG were rapidly removed under aseptic conditions and placed in ice-cold HBSS (Gibco) containing 20 mM HEPES. DRGs were digested in 1 mg/ ml collagenase A (Roche) and 2.4 U/ml dispase II (Roche) in HBSS for 60 min, respectively, followed by 5 min in 0.25% trypsin (Sigma), all at 37°C. The DRGs were then washed in DMEM (Gibco) and resuspended in DMEM medium supplemented with 10% FBS (Invitrogen) and 1% penicillin/streptomycin (Sigma). DRGs were then mechanically dissociated using fire-polished glass pipettes, centrifuged (200 g, 5 min, resuspended in Neurobasal media (Gibco) with B27 supplement (Invitrogen), L-glutamine and 1% penicillin/streptomycin (Invitrogen), and plated on 0.5 mg/ml poly-D-lysine (Sigma)-coated glass coverslips. Cells were maintained at 37°C in a 5% CO 2 incubator. All experiments using DRG neurons were performed 12-36 h after plating.

Patch-clamp electrophysiology
Electrophysiologic responses were recorded using the patch-clamp recording technique with EPC-10 amplifier and Pulse 8.30 software (both from HEKA). For perforated-patch recordings in DRG neurons, we used an external bath solution of the following composition (in mmol/L): 140 NaCl, 5 KCl, 2 CaCl 2 , 1 MgCl 2 , 10 glucose, and 10 HEPES, adjusted to pH 7.4 with NaOH. Patch pipettes with resistances of 3-5 MO were made from borosilicate glass capillaries. Pipette solution contained (in mM): 140 K-gluconate, 1 CaCl 2 , 2 MgCl 2 , 10 EGTA, 5 K 2 ATP, 10 HEPES, adjusted to pH 7.4 with KOH. A 50 mg/mL stock solution of gramicidin (Calbiochem, La Jolla, CA) was prepared in dimethylsulfoxide (DMSO; Sigma). Gramicidin was diluted into the pipette solution to a final concentration of 100 lg/mL and vortexed thoroughly before use. All drug solutions were applied to cells by local perfusion through a capillary tube (1.1 mm inner diameter) positioned near the cell of interest. After the formation of a tight seal, the progress of gramicidin perforation was evaluated by monitoring the capacitive current transient produced by a 10 msec hyperpolarizing voltage step (À5 mV) from a holding potential of À60 mV. Cells were accepted for recording if the access resistance dropped to 20 MO within 20 min after seal formation. The solution flow was driven by gravity (flow rate, 3-5 mL/min) and controlled by miniature solenoid valves (The Lee Company). Perforated-patch recordings of membrane potential (V m ) were corrected offline using the following formula: V m = V p +V pf -(LJP), where Vp is the recorded potential, V pf is the perforated-patch potential and LJP is the liquid junction potential between intracellular pipette and extracellular bath solutions. V pf (+3.6 mV) was measured directly as the difference in membrane potential between perforated and whole-cell patch configuration; LJP (+16.1 mV) was calculated using JPCalc for Windows (Barry 1994) (Molecular Devices). For whole-cell patch-clamp recordings of sodium currents the pipette solution contained (in mmol/L): 130 CsCl, 9 NaCl, 1 MgCl 2 , 10 EGTA, 10 HEPES, adjusted to pH 7.4 with CsOH. The external solution was composed of (in mmol/L): 131 NaCl, 10 TEACl, 10 CsCl, 1 CaCl 2 , 2 MgCl 2 , 0.1 CdCl 2 , 3 4-aminopyridine, 10 HEPES, 10 glucose adjusted to pH 7.4 with NaOH. Tetrodotoxin (TTX)-resistant currents were recorded from DRG neurons in the presence of TTX (300 nmol/L). For quantification of current amplitude in the presence of PGE 2 currents were evoked with a voltage pulse to 0 mV from a holding potential of À70 mV. To determine the voltage of activation of Na v 1.8-mediated sodium currents, a 500msec pre-pulse step at À50 mV, designed to inactivate TTX-resistant Na v 1.9 channels (Berta et al. 2008), was followed by a series of voltage steps increments of +5 mV at a frequency of 1 Hz. A liquid junction potential of +6.0 mV was corrected offline for calculation of the conductance-voltage relationship. Data were fit with a Boltzmann function. The shift in voltage dependence of Na v 1.8 activation was calculated from the difference in V 1/2 activation.

Ca 2+ imaging
We performed fura-2 AM-based (Molecular Probes) Ca 2+ imaging experiments. Briefly, DRG neurons prepared as above were loaded with fura-2 AM (2 lmol/L) in DMEM for 50 min at 37°C in a 5% CO 2 incubator. The cells were then rinsed with DMEM and incubated for an additional 20 min to de-esterify the dye. Cells on slides were placed onto an inverted microscope and illuminated with a 175 W xenon arc lamp; excitation wavelengths (340/ 380 nm) were selected by a monochromatic wavelength changer. Intracellular calcium concentrations ([Ca 2+ ] i ) were measured by digital video microfluorometry with an intensified charge-coupled-device camera (CasCade, Roper Scientific) coupled to the microscope and a computer with Metafluor software (Universal Imaging). All drugs were applied via bath perfusion at a flow rate of 3-5 mL/min. For Clgradient depletion experiments, NaCl was substituted for equimolar Na-gluconate in the bath solution. The left axis of all scale bars for Ca 2+ imaging traces represents the F 340/380 ratio.

GABA induces Ca 2+ transients by GABA A receptor activation in DRG neurons
We first employed ratiometric Ca 2+ imaging to study the physiologic role of GABA A receptors in acutely dissociated adult mouse DRG neurons without disrupting the intracellular ionic milieu. GABA (300 lmol/L, 10 sec) applied via the bath perfusion induced Ca 2+ transients ( Fig. 1A) in a subpopulation of DRG neurons (52.8%; n = 507/ 960) (Fig. 4D). These transients were reproducible and stable across sequential applications of GABA (Fig. 1B).
The GABA A receptor selectively conducts anions through its pore; therefore, we next sought the origin of Ca 2+ transients induced by GABA. The GABA-induced Ca 2+ transients were blocked by CdCl 2 (100 lmol/L), a non-selective blocker of calcium channels (data not shown) suggesting that GABA A receptor activation leads to voltage-gated Ca 2+ channel (VGCC) activation. We verified that GABA-induced Ca 2+ transients were abolished by the removal of extracellular Ca 2+ (Fig. 1E, F) but not by the depletion of intracellular Ca 2+ stores ( Fig. 1G, H). We confirmed that Ca 2+ transients induced by supramaximal GABA (300 lmol/L) were blocked by the non-competitive GABA A receptor antagonist picrotoxin (Fig. 1I) in a concentration-dependent manner (Fig. 1J). Ca 2+ responses could also be induced by the GABA A receptor selective agonist muscimol (10 lmol/L), which were again blocked by picrotoxin (100 lmol/L) (data not shown).
Gramacidin is an antibiotic agent which diffuses into the cell membrane forming small Climpermeable perforations, thereby giving electrical access without disruption of the intracellular Clconcentration (Ebihara et al. 1995;Kyrozis and Reichling 1995). Using gramicidinperforated patch-clamp technique we observed that GABA (300 lmol/L) evoked a fast, rapidly decaying depolarization of the membrane potential in small-sized (12-20 lm diameter) DRG neurons (Fig. 1K). The average membrane potential was recorded as À53.7 AE 0.8 mV at rest and a peak of À35.3 AE 0.6 mV during GABA (300 lmol/L) application (Fig. 1L); at this concentration of GABA action potentials were observed in 1/13 neurons (data not shown). Thus, GABA elicits membrane depolarization via activation of GABA A receptors, resulting in VGCC-mediated Ca 2+ influx in DRG neurons.

Peripheral GABA A receptors are nociceptive in acute inflammation
We investigated the modulatory effects of GABA A receptors on inflammatory pain by injecting the GABA A agonist muscimol (1 mmol/L, 20 lL) into the hind paw after cessation of 0.8% formalin-induced pain behavior in adult mice. This formalin pre-conditioning protocol revealed a novel peripheral GABA A receptor-mediated nocifensive behavior ( Fig. 2A, B). In contrast, mice injected with muscimol after saline pre-injection showed no pain-like behavior in the 30 min after muscimol injection ( Fig. 2A, B). These results suggest that GABA A receptors elicit nociceptive behavior in the presence of acute inflammation, but not in otherwise na€ ıve animals.
We next investigated whether muscimol-induced nociceptive signaling can be transmitted to the spinal cord after formalin inflammation by performing extracellular recording of WDR neurons in adult mice in vivo. After a period (20 min) of baseline recording, muscimol (1 mmol/L, 20 lL) was injected subcutaneously into the hind paw. The spike frequency of all units did not differ compared with baseline in the 20 min after the first muscimol injection (Fig. 2C). Units representing the action potential discharge of 12 individual WDR neurons were identified based on spike profile ( Fig 2D). Subsequent formalin (0.8%) injection augmented spike frequency in 8 out of 12 WDR neurons (Representative units 'a' and 'b',Fig. 2C). After the activity of formalin-responsive neurons returned to baseline (approximately 60 min) a second muscimol injection facilitated action potential firing in 10 out of 12 WDR neurons. Units identified at a low frequency during the first muscimol injection were observed firing at higher frequency during the second muscimol injection after formalin conditioning (Fig. 2D, E). The overall number of action potentials recorded during the 20 min after the second muscimol injection was also significantly increased (Fig. 2F). Representative unit 'c', which did not respond to the formalin injection, was not affected by the second muscimol injection (Fig. 2C). These results indicate that activation of GABA A receptors in peripheral nociceptive neurons after acute inflammation evokes nociceptive signal transmission to secondorder neurons in the spinal cord. In addition, we confirmed that muscimol-induced licking behavior after formalin pre-conditioning was dose-dependently inhibited by the non-competitive GABA A receptor antagonist picrotoxin (Fig. 2G, H). Interestingly, co-injection of picrotoxin with formalin also dose-dependently inhibited the second phase of the formalin-induced behavior. Frequency histogram of muscimol-induced AP discharge (1 min bins). (F) Mean frequency of total muscimol-induced AP discharge per min for 20 min, n = 12 units recorded from 6 animals, **P < 0.01, Wilcoxon signed-rank test. (G) Effect of GABA A antagonist picrotoxin (PTx) on time course of hind paw licking behaviors during pre-conditioning with formalin (0.8%, 20 lL) followed by injection of muscimol (1 mmol/L, 20 lL) into the same dorsum hind paw area. Picrotoxin was given at 1 mmol/L (gray triangle, n = 5 mice), 5 mmol/L (black squares, n = 6 mice) or vehicle (0.9% saline, open circles, n = 12 mice). (H) Picrotoxin significantly inhibited the second phase of the formalin response (*P < 0.05, *** P < 0.001, one-way ANOVA with Bonferroni post-test) as well as muscimol-induced pain behavior (**P < 0.01, P < 0.001, one-way ANOVA with Bonferroni post-test). WDR, wide dynamic range. after pre-treatment of formalin (0.001%, 2 min) (Fig. 3A, B). Formalin is known to directly activate nociceptive neurons through TRPA1 (McNamara et al. 2007); however, formalin potentiation of GABA-induced Ca 2+ transients occurred in neurons which had no response to formalin (0.001%) alone (Fig. 3A) and furthermore was not affected by the TRPA1 selective antagonist, HC030031 (30 lmol/L) (Fig. 3C, D).

Prostaglandin E2 reveals GABA A receptormediated nociception
We next sought to determine the mechanism underlying peripheral GABA A receptor-induced nociception in inflammation. Formalin-induced pain behavior can be attenuated by pre-treatment with the cyclooxygenase (COX) inhibitor indomethacin (Hunskaar et al. 1986), which reduces inflammation by blocking prostanoid synthesis. We found that systemic pre-treatment with indomethacin (40 mg/kg, i.p.) 30 min before formalin injection completely abolished the subsequent muscimolinduced licking behavior in adult mice (Fig. 4A, B). The prostaglandin E2 (PGE 2 ) derivative is a potent inflammatory mediator produced at the site of inflammation (Fulton et al. 2006). We observed that PGE 2 (500 lmol/L, 20 lL)-induced licking behavior was also enhanced by co-injection of muscimol (Fig. 4C), confirming that PGE 2 is sufficient for GABA A receptor-mediated nocifensive behavior. Next, we examined the effect of PGE 2 on GABAinduced Ca 2+ transients in cultured DRG neurons. Nociceptive DRG neurons were identified by their response to a 10 sec application of capsaicin (1 lmol/L) at the end of each experiment and made up 42.3% (406/960 neurons) of the total population, of which 17.6% (169/960 neurons) were also responsive to GABA (Fig. 4D). We found that PGE 2 (10 lmol/L) potentiated GABA-induced Ca 2+ transients (Fig. 4E, F) in a large subpopulation of capsaicin-responsive nociceptive DRG neurons (40.1%; n = 69/169 neurons) ( Fig. 4G) but only a small number of capsaicin-insensitive DRG (1.8%; n = 6/338 neurons) (Fig. 4H). In addition, PGE 2 (10 lmol/L) facilitated GABA-induced membrane depolarization resulting in action potential firing in a subpopulation of small-sized DRG neurons as recorded by gramicidin perforated patch (Fig. 4I-K). Collectively, our results suggest that PGE 2 is one of the possible pro-inflammatory mediators that may contribute to GABA A receptor-mediated nocifensive behavior during acute inflammation via action in a subpopulation of nociceptive DRG neurons.

Activation of EP4 receptors increases GABAinduced nociceptive neuron activity
PGE 2 signals through a family of EP receptors (EP1-4) of which subtypes EP1, EP2, and EP4 are excitatory and promote inflammation (Fulton et al. 2006). PGE 2 -potentiated GABA responses were inhibited by EP4 receptor antagonist, AH23848 (10 lmol/L) (Lin et al. 2006) (Fig 5A, B) but not by EP1-2 receptor antagonist, AH6809 (50 lmol/L) (St-Jacques and Ma 2011) (Fig. 5C,  D). GABA-induced action potential firing in the presence of PGE 2 was also abolished by AH23848 (Fig. 5E, F). Furthermore, potentiation of GABA-induced Ca 2+ transients by formalin was also blocked by AH23848 (10 lmol/L) (Fig. 5G, H). Together, these data suggest that formalin and PGE 2 may potentiate GABA-induced responses through an EP4 receptor signaling mechanism in nociceptive DRG neurons.  Voltage-gated Na + channels may contribute to GABA-induced neuronal excitability during PGE 2 sensitization PGE 2 is an important mediator in the development of peripheral inflammation and peripheral sensitization (Vane 1971;Julius and Basbaum 2001). For example, PGE 2 can directly potentiate the TTX-resistant Na + channels Na v 1.8 (England et al. 1996;Gold et al. 1998) and Na v 1.9 (Rush and Waxman 2004) in peripheral nociceptive neurons. We therefore conducted experiments to examine whether modulation of Na v channels contributes to GABA-induced responses during PGE 2 application. The archetypal Na v channel blocker lidocaine (Sheets et al. 2008) was applied to cultured DRG neurons during PGE 2 -induced potentiation of the GABA response (Fig. 6A). Lidocaine (300 lmol/L) inhibited the potentiation of GABA-induced Ca 2+ responses by PGE 2 (Fig. 6A, B) but had no effect on GABA-induced Ca 2+ transients in control conditions (Fig. 6C, D). We confirmed that PGE 2 acutely potentiates TTX-resistant Na v channel current (Fig 6E, F), as well as produce a leftward shift (À6.08 AE 1.8 mV, n = 6 cells) in the voltage dependence (G) Normalized conductance-voltage relationship of Na v 1.8 currents in the presence and absence of PGE 2 (10 lmol/L) (n = 6 cells) fit with a Boltzmann function. Test pulses were preceded by a 500-msec step to À50 mV to inactivate TTX-resistant Na v 1.9. (H) A887826 (100 nmol/L) inhibits the membrane depolarization caused by PGE 2 (10 lmol/L) application (n = 12-20 cells; *P < 0.05, unpaired Student's t-test). (I) Potentiation of GABA (300 lmol/L)-induced Ca 2+ transients by PGE 2 (10 lmol/L) is abolished by Na v 1.8 channel blocker A887826 (100 nmol/L) in DRG neurons (J) Quantification of GABA-induced Ca 2+ transients in the presence and absence of A887826 relative to peak amplitude of 1st GABA response (n = 10 cells, 4 coverslips from 3 mice; **P < 0.01, repeatmeasures one-way ANOVA with Bonferroni post-test). (K, L) The potentiating effect of PGE 2 on GABA-induced changes in membrane potentials (K) and generating action potentials (L) in small-sized DRG neurons is blocked by Na v 1.8 channel blocker A887826 (100 nmol/L) (n = 9 cells). DRG, dorsal root ganglia. of Na v 1.8 activation (Fig 6G). PGE 2 additionally induced a small depolarization from the resting membrane potential (Fig. 6H). Pre-incubation with the Na v 1.8 channel antagonist A887826 (100 nmol/L) (Zhang et al. 2010) not only blocked PGE 2 -induced depolarization (Fig. 6H) but also the PGE 2 potentiation of GABA-induced Ca 2+ responses (Fig. 6I, J) as well as GABA-induced action potential firing in the presence of PGE 2 (Fig. 6K, L). Together, these results suggest that the activation of voltage-sensitive TTX-resistant Na + channels may contribute to the nociceptive role of GABA A receptors during PGE 2mediated inflammation.

NKCC1 is not required for potentiated GABA responses by pro-inflammatory mediators
The Na + -K + -Cl --co-transporter (NKCC1) is thought to be responsible for maintaining intracellular Cllevels in DRG neurons (Sung et al. 2000). Furthermore, NKCC1 activity is dynamically regulated by cAMP-dependent protein kinase (PKA) and protein kinase C (PKC) phosphorylation (Smith et al. 2008;Flemmer et al. 2010). We therefore hypothesized that NKCC1 may be a downstream target of EP4 receptor signaling (via PKA) during formalin-induced inflammation leading to potentiation of GABA A receptor-mediated responses. To investigate whether NKCC1 activity is required for GABA-induced Ca 2+ transients, we applied the NKCC1 inhibitor bumetanide. Continuous application of bumetanide (10 lmol/L) led to a progressive reduction in amplitude of GABAinduced Ca 2+ transients (Fig. 7A, B). To confirm the Cl À dependency of GABA-induced Ca 2+ transients extracellular Cl À was substituted with gluconate; this has the effect of gradually depleting intracellular Cl À and effectively reducing the Cl À concentration gradient (Rocha-Gonzalez et al. 2008). Under such conditions, we saw a consistent reduction in GABA-induced Ca 2+ transient amplitude (Fig. 7C, D). Reduction of extracellular Cl À in the bath solution to 0 mmol/L over just 15 min reduced GABAinduced Ca 2+ transient amplitude by more than 60% (Fig. 7D). However, potentiated GABA-induced Ca 2+ transients by PGE 2 was still observed in DRG neurons isolated from NKCC1-deficient mice (Fig. 7E, F). In addition, application of PGE 2 for the duration necessary to potentiate GABA-induced Ca 2+ transients did not significantly alter the amplitude of GABA-induced currents (Fig. 7G, H). Furthermore, peripheral injection of muscimol still restored licking behavior after formalin pre-conditioning in mice lacking NKCC1 (Fig. 7I, J). These results suggest that NKCC1 activity is not required for peripheral GABA A -mediated nociception during acute inflammation.

Discussion
Peripheral GABA A receptor nociception in vivo GABA has long been observed to depolarize peripheral sensory nerves in rodents (Feltz and Rasminsky 1974;Deschenes et al. 1976;Sung et al. 2000) and humans (Carr et al. 2010); however, activation of peripheral GABA A receptors elicits neither spontaneous pain behavior nor mechanical sensitivity in na€ ıve animals. One proposal to this apparent paradox is the phenomenon of primary afferent depolarization (Rudomin and Schmidt 1999;Willis 1999) whereby sub-threshold depolarization results in inactivation of voltage-gated channels. Nevertheless, previous reports have suggested that the activation of peripheral GABA A receptors can facilitate inflammation-induced pain behavior (Carlton et al. 1999;Bravo-Hernandez et al. 2014). The formalin test is a well-characterized test of acute-inflammatory pain (Hunskaar and Hole 1987) which produces biphasic pain-like behavior that disappears 60 min after the original treatment. By a simple modification of the protocol -injection of muscimol after the return of formalin behavior to baseline -we were able to reveal a novel GABA A receptor-mediated nocifensive behavior that could be used for further mechanistic study. The phenomenon of GABA A -mediated nociception after acute inflammation appears to contrast with observations that either application of GABA A agonists directly to the DRG (Naik et al. 2008), or systemic administration of GABA A allosteric modulators (Munro et al. 2008), can have an alleviating effect in certain models of chronic inflammatory or neuropathic pain, although the contribution of spinal GABA A receptors cannot be completely excluded in these studies.

Ca 2+ imaging as an indirect measure of GABA A receptor function in vitro
Unlike whole-cell techniques, the fura2-AM Ca 2+ imaging technique allows the quantitative study of cell activity without disrupting the cell membrane. Thus, cells may be studied in a relatively undisturbed state with endogenous concentrations of intracellular ions and signaling factors. Application of GABA to DRG neurons induced consistent and concentration-dependent Ca 2+ transients that were mediated by GABA A receptors specifically and blocked by CdCl 2 , a non-selective voltage-gated Ca 2+ channel blocker. Moreover, manipulation of the Clconcentration gradient had a proportional effect on GABA-induced Ca 2+ transient amplitude. These results suggested that GABA elicited Ca 2+ influx through VGCC activation in DRG neurons of an amplitude proportional to the degree of membrane depolarization.
The high concentration of muscimol (>1 mmol/L) required to potentiate the formalin nociceptive response in vivo in this study and reported by others (Carlton et al. 1999) suggests that near-maximal activation of GABA A receptor is required for conversion to a nociceptive role. Correspondingly, we used a relatively high concentration of GABA (300 lmol/L) for the majority of our in vitro studies, which represented a supramaximal response in our Ca 2+ assay. The physiologic concentration of GABA has been reported at >500 lmol/L at GABAergic synapses in the brain (Maconochie et al. 1994) with a peak concentration of 1.5-3 mmol/L GABA (Mozrzymas et al. 2003). Further work will be required to establish whether such concentrations are reached at peripheral GABA A receptors in vivo.

Formalin and PGE 2 potentiation of GABA responses are mediated by EP4 receptor
The effect of formalin on DRG neurons in our in vitro culture system appeared to mimic the in vivo observation by potentiating the GABA-induced Ca 2+ response. Formalin-induced pain behavior is characterized by two phases: an acute first phase mediated by peripheral nociceptive transmission and a second phase thought to be driven by a combination of central sensitization (McNamara et al. 2007) and biphasic nociceptor activity (McCall et al. 1996). In addition, formalin is known to directly activate TRPA1 in nociceptive sensory neurons. However, a supramaximal concentration of the TRPA1 selective antagonist, HC030031, had no effect on the potentiation of the GABA-induced Ca 2+ responses by acute low concentration (0.001%) formalin suggesting a mechanism independent of TRPA1. PGE 2 is a potent inflammatory mediator produced during formalin-induced inflammation (Malmberg and Yaksh 1995). PGE 2 sensitizes peripheral nociceptive neurons through EP receptors present on the peripheral terminals of high-threshold sensory neurons (Omote et al. 2002). In our experiments, PGE 2 potentiated GABA-induced Ca 2+ transients almost exclusively in capsaicin-sensitive DRG neurons suggesting that the effect is restricted to a population of nociceptive sensory neurons. PGE 2 also revealed additional muscimol-induced pain-like licking behaviors in vivo.
There are four subtypes of PGE 2 receptor known as EP receptors (EP1-4), all of which are G-protein coupled receptors: EP1 coupled to G q /G11, EP2 and EP4 coupled to G s , and EP3 coupled to G s and G i . In particular, the EP4 receptor is highly expressed in primary sensory neurons and EP4 levels are known to increase in the DRG after peripheral inflammation (Lin et al. 2006). We found that GABA-induced Ca 2+ transients potentiated by PGE 2 were blocked by the EP4 receptor antagonist, AH23848, but not by EP1-2 receptor antagonist, AH6809. Furthermore, formalin-induced facilitation of GABAinduced Ca 2+ transients was also abolished by AH23848. These observations suggest that potentiation of the GABA response by formalin and PGE 2 may share the same downstream pathway through the EP4 receptor, although in the case of formalin we cannot exclude the possible contribution from other inflammatory mediators. Formalin application to DRG neurons may act indirectly on EP4 receptors via PGE 2 release. The source of PGE 2 released in our DRG cultures by formalin application in vitro is currently unknown but could include nonneuronal cells carried over during DRG dissociation and plating. consistent with a previous study in which systemic picrotoxin treatment reduced formalin behavior (Heidari et al. 1996) and suggests that endogenous activators of the GABA A receptor may indeed play a role in the behavioral response to acute formalin inflammation. One important remaining question is the endogenous source of GABA A receptor activation in DRG neurons. Expression of functional GABA A receptors at the central terminals of DRG (Labrakakis et al. 2003) creates the potential to receive spinal GABAergic signaling. Expression of GABA, or the GABA-synthesizing enzyme glutamine decarboxylase (GAD), has also been reported in various peripheral tissues, including keratinocytes (Ito et al. 2007) and macrophage (Tannahill et al. 2013). Whether these tissues are capable of synthesizing and more importantly releasing GABA remains to be explored.
In conclusion, our report suggests a nociceptive role of peripheral GABA A receptors in acute-inflammatory pain and presents a working model of the mechanism of GABA A receptor-mediated nociception.