Analgesic dorsal root ganglionic field stimulation blocks conduction of afferent impulse trains selectively in nociceptive sensory afferents

Abstract

Increased excitability of primary sensory neurons after peripheral nerve injury may cause hyperalgesia and allodynia. Dorsal root ganglion field stimulation (GFS) is effective in relieving clinical pain associated with nerve injury and neuropathic pain in animal models. However, its mechanism has not been determined. We examined effects of GFS on transmission of action potentials (APs) from the peripheral to central processes by in vivo single unit recording from lumbar dorsal roots in sham injured rats and rats with tibial nerve injury (TNI) in fiber types defined by conduction velocity. Transmission of APs directly generated by GFS (20Hz) in C-type units progressively abated over 20s, whereas GFS-induced Aβ activity persisted unabated, while Aδ showed an intermediate pattern. Activity generated peripherally by electrical stimulation of the sciatic nerve and punctate mechanical stimulation of the receptive field (glabrous skin) was likewise fully blocked by GFS within 20s in C-type units, whereas Aβ units were minimally affected and a subpopulation of Aδ units were blocked. After tibial nerve injury, the threshold to induce AP firing by punctate mechanical stimulation (von Frey) was reduced, which was reversed to normal during GFS. These results also suggest that C-type fibers, not Aβ, mainly contribute to mechanical and thermal hypersensitivity (von Frey, bush, acetone) after injury. GFS produces use-dependent blocking of afferent AP trains, consistent with enhanced filtering of APs at the sensory neuron T-junction, particularly in nociceptive units.

Electrical stimulation of the dorsal root ganglion provides effective analgesia by selectively blocking nociceptive sensory afferents from somata and peripheral endings

Introduction

Dorsal root ganglion (DRG) stimulation is an established clinical treatment for controlling chronic pain [8]. This analgesic approach avoids the risks of addiction and overdose that accompany opioid use, while lacking the irreversibility of neurodestructive techniques, and produces analgesia without anesthesia. DRG field stimulation (GFS), which is achieved with electrodes placed adjacent to the DRG in the intervertebral foramen, has proved effective for treating chronic neuropathic and non-neuropathic pain [2,7,13,39]. We have developed a rat model in which GFS relieved neuropathic pain from tibial nerve injury (TNI) [46,62]. GFS activates low threshold mechanoreceptor (LTMR) myelinated Aβ afferent units, which is perceived by patients as paresthesias, and it is expected that this activity may contribute to triggering analgesia by activating analgesic processes at brain and segmental levels as has been demonstrated to occur when SCS activates these same units at the level of the dorsal columns of the spinal cord [24]. However, GFS provides relief in subjects for whom SCS has failed [6,38], indicating the presence of an additional analgesic mechanism. Direct actions at the level of the DRG are likely, since the electrical fields are applied locally to the DRG and analgesia is limited to the stimulated segment [6,62]. It would be assumed that GFS should activate the full population of sensory neurons within the DRG exposed to the stimulation, including nociceptive units, which therefore should generate afferent activity leading to pain. However, this is clearly not the case in clinical experience, for which a hypothetical mechanism must account.

Sensory neurons are morphologically pseudo-unipolar, with the DRG containing the somata that give rise to a stem axon that bifurcates into peripheral and central processes at the T-junction. APs trains are generated at the peripheral terminals in response to the appropriate external stimuli and propagate along the peripheral process, then pass through the T-junction and hence into the central process (dorsal root) to continue via the dorsal roots into the dorsal horn and dorsal columns of the spinal cord. We have previously shown that the T-junction within the DRG is a site of natural low-pass filtering of AP trains in a use-dependent fashion, and that C-type units are particularly prone to this form of blockade, although peripheral nerve injury diminishes this filtering [9,20]. We therefore reasoned that GFS may produce analgesia in part by enhancing this filtering capability of the T-junction to limit the propagation of nociceptive impulse into the dorsal root, including APs triggered by GFS within the neuronal somata. This view has been supported by a computational model [28] and in vitro observations [30]. To directly test whether the T-junction is a site of action of GFS, we designed experiments using our established rat model [62] to determine if GFS modulates T-junction filtering in a fashion that can produce analgesia without anesthesia.

Materials and methods

Animals

Sprague Dawley rats weighing 200–250g were obtained from the Taconic Farms Biosciences (Rensselaer, NY), and were maintained and used according to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. As GFS was found to have comparable analgesic effects on male and female rats with neuropathic pain in our previous report [62], only male rats were used in this study. All animal experiments were performed according to protocols approved by the Institutional Animal Care and Use Committee of the Medical College of Wisconsin (animal protocol AUA00000454). Animals were housed in a pathogen-free facility, 2 animals per ventilated cage, in a room maintained at 25±1 °C at 35 to 45% humidity, with a 12/12-hour day/night cycle. Animals had ad libitum access to food and water. At the termination of the study, euthanasia was performed by decapitation during deep isoflurane anesthesia.

Neuropathic pain model

Tibia nerve injury (TNI) was performed as described in our previous report [46]. Briefly, animals were anesthetized with 2% isoflurane/O₂ mixture, a 2cm incision was made on the lateral mid-thigh of right leg, and the sciatic nerve was exposed at the point at which it divides into its distal branches. At a distance 5mm distal to this branch point, the tibial nerve was ligated with 5.0 silk sutures and 2–3mm of the nerve was removed distal to the ligation. Contact with the preserved sural and common peroneal nerves was avoided during surgery. Fascia was closed with synthetic absorbable sutures, and skin was closed with staples. Sham TNI control rats had only exposure of the nerves without further handling.

During surgery, rectal temperature of rats was maintained at 37°C with a thermostatically controlled heating pad. Animals were given a prophylactic dose of enrofloxacin (25mg/ml, 0.1ml) after surgery and in the following three days to prevent infection. Starting 10–17 days after the surgery, animals were used for behavioral testing and electrophysiological recording.

GFS electrode implantation

GFS bipolar electrodes were prepared from 2 insulated platinum-iridium wires (0.010 inch and 0.005 inch) connected to a threaded plastic pedestal (P1 Technologies, Roanoke, VA), as described in our previous report [46], where their implantation was also described. Briefly, rats were anesthetized with 2% isoflurane/O₂ while maintaining body temperature at 36.5°C. A dorsal paramedian incision was made to expose the external aspect of the intervertebral foramen at the required levels of GFS, which included fourth lumbar (L4). This requires removal of the overhanging accessory process at the L4 level. A probe with 0.4-mm diameter was inserted into the intervertebral foramen dorsolateral to the DRG, to create a passage into which the GFS electrode was inserted in juxtaposition to the dorsolateral aspect of the DRG at that level. A stainless-steel wire was used to stabilize the electrode to a screw inserted into the L5 transverse process caudal to the foramen. The leads, which were protected within a flexible plastic tubing from excess flexion of rats, were tunneled to the head, where the pedestal was secured to the skull with screws and dental cement and connected to stimulator to deliver stimulation.

GFS

Animals received GFS while awake and freely moving. In the clinic setting, the intensity used is typically that which can produce paresthesias [32,38,39,42,58], indicating that low-threshold mechanoreceptors are activated, but at current levels that do not produce motor activity, even though GFS with intensity that doesn’t produce paresthesias is also effective in some reports [42,59]. In studies on rats, it is impossible to know at what level rats do or do not perceive paresthesias. We previously tested effects of GFS with intensities at 40%, 60%, 80%, 98% motor threshold on pain behaviors of rats with TNI and found that GFS with 80% motor threshold produces maximum effectiveness both in reflex behavioral testing and in operant testing of spontaneous pain relief (conditioned place preference), while GFS with 60% MT fails to induce conditioned place preference [62]. Therefore, in studies on rats, we used current at 80% of the motor threshold. The motor threshold was determined as the current at which any further increase resulted in perceptible hind limb movement for GFS during stimulation at 2Hz with a pulse width of 0.2ms. Unless otherwise indicated, a frequency of 20Hz for was used for GFS, comparable to clinical settings used to obtain optimal analgesia [13].

Behavioral tests

Sensory testing of the plantar skin included eliciting reflexive behaviors induced at threshold intensity by punctate mechanical stimulation (von Frey test), by noxious mechanical stimulation (pin), dynamic non-noxious mechanical stimulation (brush), and cold stimulation (acetone). For GFS treatment, the behavioral tester was blinded to treatment patterns with different frequencies.

Threshold punctate mechanical stimulation (von Frey)

The von Frey test was performed using a set of calibrated monofilaments (Patterson Medical, Bolingbrook, IL). Briefly, beginning with the 2.8-gram filament, the filament was applied perpendicularly to the glabrous skin on the lateral third of the plantar aspect of the hind paw innervated by sural nerve for 1s, with just enough force to bend the fiber. If paw withdrawal was observed, then the next weaker filament was applied, and if no response was observed, then the next stiffer filament was applied, until a response occurred (termed a reversal event). After the reversal event, 4 more stimulations were performed following the same pattern. There were at least 10s interval between applications. The forces of the filaments before and after the reversal, and the 4 filaments applied after the reversal, were used to calculate the 50% withdrawal threshold [3]. Rats not responding to any filament were assigned a score of 25g. Since the magnitude of mechanical sense is perceived in a logarithmic scale, as described by Weber’s law, von Frey paw withdrawal threshold data were log transformed [44]. In the absence of a hypersensitivity state, animals often default to the 25g score.

Noxious punctate mechanical stimulation (pin test)

Pin test was performed using a standard 22-gauge spinal anesthesia needle that was applied to the lateral third of the hind paw with enough force to indent the skin but not puncture it. This was repeated for 5 applications, with at least 10s intervals between applications, and this set of applications was repeated after 1min, making a total of 10 touches. Each application induced a behavior that was categorized as either the type typical of uninjured rats, consisting of a very brief (<1s) withdrawal and immediate return of the foot to the cage floor, or an alternate behavior that we term a hyperalgesic response, consisting of a complex event with sustained elevation of the foot for at least 1s, variably combined with grooming that included licking and chewing of the paw, and with shaking of the limb [26]. This hyperalgesic behavior is associated with place avoidance [61], indicating an aversive experience. Hyperalgesia was quantified by tabulating the number of hyperalgesic responses as a percentage of the total touches. Naïve animals rarely show hyperalgesic behavior to pin stimulation.

Dynamic mechanical stimulation (brush test)

A camel hair brush (4 mm wide) was applied to the lateral plantar skin of the hind paw by light stroking in the direction from heel to toe over the span of 2s [26]. The response was scored as either positive if the paw was removed or none in the absence of movement. The test was applied three times to each paw, separated by intervals of at least 10s. Hypersensitivity was quantified by tabulating the number of responses as a percentage of the total applications. Naïve animals rarely respond to brush stimulation.

Cold stimulation (acetone test)

Sensitivity to cold was assessed using application of acetone, which was expelled through tubing perpendicularly to form a convex meniscus on the end of the tubing that was touched to the lateral plantar skin without contact of the tubing with the skin [4]. The response was scored as positive if the paw moved, and 3 repetitions were spaced at least 1 min apart. Hypersensitivity to acetone was quantified by tabulating the number of responses as a percentage of the total applications. Naïve animals rarely respond to acetone application.

Neuron isolation and plating

L4&L5 DRGs from rats were rapidly harvested following isoflurane anesthesia and decapitation. Ganglia were placed in a 35 mm dish containing Ca²⁺/Mg²⁺-free, cold HBBS (Life Technologies) and cut into four to six pieces that were incubated in in 0.01% blendzyme 2 (Roche Diagnostics, Indianapolis, IN) for 26 min followed by incubation in 0.25% trypsin (Sigma Aldrich, St. Louis, MO) and 0.125% DNase (Sigma) for 30 min, both dissolved in Dulbecco’s modified Eagle’s medium (DMEM)/F12 with glutaMAX (Invitrogen, Carlsbad, CA). After exposure to 0.1% trypsin inhibitor and centrifugation, the pellet was gently triturated in culture medium containing Neural Basal Media A with B27 supplement (Invitrogen), 0.5mM glutamine, 10ng/ml nerve growth factor 7S (Alomone Labs, Jerusalem, Israel) and 0.02 mg/ml gentamicin (Invitrogen). Dissociated neurons were plated onto poly-L-lysine coated glass cover slips and maintained at 37°C in humidified 95% air and 5% CO₂ for 2 hours, and were studied no later than 6 hours after harvest.

Calcium Microfluorometry

Calcium microfluorometry was performed following our previously published protocols [12,18,21]. Neurons plated on cover slips were exposed to the ratiometric Ca²⁺ indicator Fura-2-AM (5μM, Invitrogen) at room temperature in a solution that contained 2% bovine albumin to aid dispersion of the fluorophore. After 30 min, they were washed 3 times with regular Tyrode’s solution (in mM: NaCl 140, KCl 4, CaCl₂ 2, MgCl₂ 2, glucose 10, HEPES 10 with osmolarity of 297–300 mOsm and pH 7.40), given 30 minutes for de-esterification, and then mounted in the recording chamber. Cover slips were then mounted on a RC-21BRFS imaging chamber with two platinum wires on each side for field stimulation (Warner Instruments, Hamden, CT). The fluorophore was excited alternately with 340 nm and 380 nm wavelength illumination (Lambda DG-4, Sutter, Novato, CA), and images were acquired at 510 nm with a digital camera (Hamamatsu, Japan). After background subtraction, the fluorescence ratio R for individual neurons was determined as the intensity of emission during 340 nm excitation (I₃₄₀) divided by I₃₈₀. Only neurons with stable baseline R traces for over 2min were further evaluated. Field stimulation-induced Ca²⁺ transients were generated by step field stimulations (2V steps) from low to high with a stimulator (Master 9, AMPI, Israel).

Imaging of GFS-triggered Ca²⁺ influx with genetically encoded calcium indicator GCaMP

AAV vector solution was microinjected into the right L4 DRG using previously described techniques [16]. Briefly, the intervertebral foramen was exposed surgically and then the laminar bone was removed to expose DRG. Injection was performed with a microinjector (Nanoliter 2000, World Precision Instruments, Sarasota, FL, USA). A glass micropipette filled with vector was advanced ~100μm into the ganglion, followed by injection of 2μl of AAV1-CAG-GCaMP6s-WPRE-SV40 (Penn Vector Core, PA, USA) with titers containing a total of 2.0×10¹⁰ genome viral particles. Injection was performed over a 5-min period and removal of the pipette was delayed for an additional 5min to minimize the extrusion of the injectate. Overlying muscle and skin were closed, and the animals were used 2 weeks later for imaging sensory somata expressing GCaMP6s during GFS. Specifically, the animal was anesthetized with 2% isoflurane, a tracheostomy was inserted, and PE-10 cannula was placed into the carotid artery to monitor blood pressure and blood gases. The rat was mechanically ventilated. The L4 DRG was exposed by removing L4 transverse process and surrounding tissues and bones to fit the 40X water immersion objective in a Nikon infrared-differential interference contrast microscopy. Bone wax was used to stop bleeding from bone. The skin edges of the incision were sutured to a round metal ring to form a basin that was perfused with warm (36°C) artificial cerebrospinal fluid (aCSF) saturated with 95% O₂ and 5% CO₂ at 2ml/minute. This aCSF contained 119 mM NaCl, 2.5 mM KCl, 2.5 mM CaCl₂, 1 mM MgCl₂, 1.25 mM NaH2PO₄, 26 mM NaHCO₃, and 10 mM glucose. Two tungsten electrodes were inserted into nerves on each side of the DRG to provide field stimulation and motor threshold was determined. Neurons were identified by infrared LED illumination [48]. The fluorophore was excited alternately with 485 nm wavelength illumination (Lambda DG-4, Sutter, Novato, CA), and images were acquired at 510 nm using a digital camera (Hamamatsu, Japan) and analyzed with MetaFluor Imaging System (Molecular Devices, San Jose, CA). Regions of interest were chosen without overlap. Pancuronium (0.1mg/kg) was administrated to keep the recording stable. After stable baseline fluorescence intensity was recorded, GFS with 80% MT was applied. Changes from this baseline fluorescence were recorded and calculated as ΔF/F, calculated as (F_t−F₀)/F₀, where F_t is the fluorescence intensity at time t and F₀ is the baseline fluorescence intensity. Cells with ΔF/F > 0.2 were considered positive to GFS.

In vivo electrophysiological recording

Electrophysiological signals from dorsal root teased fiber recordings were collected with an Axoclamp 900A microelectrode amplifier (Molecular Devices, San Jose, CA), filtered at 2 kHz, and sampled at 10 kHz using a digitizer (DigiData 1440A, Molecular Devices). Individual action potentials were isolated and amplified by means of a window discriminator (Clampfit, Molecular Devices) or by template matching using Spike2 (Cambridge Electronic Design Limited, Cambridge, UK).

Animal preparation:

Anesthesia was induced with 2% inhalant isoflurane. Then, the rats were given urethane (100mg/kg, Subcutaneous injection, S.C). Thirty minutes after urethane injection, isoflurane was reduced to 1.25–1.5%. During electrophysiological recording, the isoflurane will be discontinued or run at levels of ≤0.2% such that the rats will remain immobile but metabolically and hemodynamically stable, which are confirmed by arterial blood analysis with a blood gas analyzer (ABL800FLEX, Radiometer, Copenhagen, Denmark), and continuous blood pressure was monitored via a carotid artery PE-10 cannula. A heating pad was used to maintain body temperature at 37 °C.

For sciatic nerve stimulation, the sciatic nerve was exposed by incision at the midthigh. The sciatic nerve was wrapped and insulated from surrounding tissues with a nerve cuff electrode (World Precision Instruments, Sarasota, FL). The wound was closed after the implantation. For acute DRG stimulation, the electrode (described above) was implanted beside the DRG and secured by suturing it to nearby tissues.

Dorsal root teased fiber recording:

The spinal cord and dorsal roots were exposed by a laminectomy at the level of vertebrae T13 to L3 and covered with warm mineral oil (36°C) to prevent it from drying. Bone wax was used to stop bleeding from the bone, and the dura was removed. Rats were then mounted on a spinal frame and the vertebral column was stabilized by clamps applied to the spinal process rostral and caudal to the exposure. The skin edges of the midline incision were sutured to a metal ring (3×4 cm) to form a basin that was filled with warm mineral oil. A fine glass probe with a rounded tip was used to gently release the L4 or L5 dorsal root from connective tissues, after which they were transected at the point where they divide into rootlets. The dorsal root was then placed on a glass platform, upon which the distal portion of the dorsal root was repeatedly teased to create fine neuronal bundles. Examination by electron microscope (EM) of 3 such bundles shows overall diameter of 21±0.6 µm and a composition including the full range of axon types (Supple. Fig. 1). Teased axon bundles were placed on a platinum/iridium recording electrode and single-unit activity was recorded. A reference electrode was attached to the adjacent muscle tissue. Initially, spontaneous activity was observed for 3 min. If the spontaneous activity was observed, 3 min of firing was recorded. The motor threshold for DRG stimulation was then determined by identifying the lowest current that could trigger a leg muscle twitch. Neuronal single units were identified either by electrical stimulation of the DRG or the sciatic nerve every 5s.

From a single axon bundle, one or several units with different CVs and AP morphologies can be recorded. Each unit’s receptive field was identified by gentle stimulation of hind paw with a small glass probe (with 1mm round tip), and a von Frey monofilament (29g). The skin with the receptive field was gently pinched with forceps to confirm that the receptive field was cutaneous. Muscle spindles and joint receptors that responded to slow pulling and bending of the leg were excluded. To characterize a unit’s firing properties, the mechanical threshold was defined as the minimal force of von Frey fiber that evoked firing. We then determined the threshold force to induce firing and the firing rate in response to noxious stimulation by both a von Frey filament (29g force, 1.1mm diameter) and a modified von Frey made by attaching a tungsten tip of 100 µm diameter with precisely squared ends to a von Frey filament with 16g force [26,53], applied to the receptive field for 10s.

Conduction velocity (CV) measurement:

CV was calculated as the distance from the point of stimulation at the DRG or sciatic nerve to the dorsal root teased fiber recording electrode (measured after the final recording in each rat), divided by the time latency between stimulation and initiation of the AP (measured directly from the recorded firing in pClamp). The length from the middle of DRG to the recording site was 17±0.2mm (n=41 rats) and the length of stimulated sciatic nerves to DRG was 56±0.3mm (n=41 rats). As CV is generally faster in peripheral axon than that in central branch [56], the CV in our recording from dorsal root with sciatic nerve stimulation should be faster than the CV with DRG stimulation that only central branch was involved. Recordings from teased dorsal root fibers were categorized to Aβ, Aẟ, and C-type fibers according to their CV. From observation of 23 units that could be activated by both DRG stimulation and sciatic nerve stimulation, CVs when stimulated at the sciatic nerve are around 3 times faster than when stimulated by GFS (Supple. Fig. 2). We used the following CV standards to distinguish different fiber types. For recordings from dorsal root with sciatic nerve stimulation, units with CV>20m/s are Aβ, 3<CV<20m/s for Aẟ, CV<3m/s for C-type. This was similar to other reports with recordings from either sciatic nerve or somata of DRG neurons by stimulating sciatic nerve [37,60]. For recordings from the dorsal root with DRG stimulation, units with CV>6m/s are Aβ, 1<CV<6m/s for Aẟ, CV<1m/s for C-type. These definitions are consistent with other reports performing similar dorsal root recordings [15,34,60].

Electron microscopic imaging (EM)

The procedure for preparing tissues for EM was similar to our previous report [23]. Teased fibers harvested after recording were fixed for 1 hr in cold fresh fixative containing 2.5% glutaraldehyde in 0.1 M cacodylate buffer and post-fixed with 1% OsO4in buffer for 1hr. Specimens were then washed with distilled water (2×5min); dehydrated in methanol (50%, 70%, 85%, 95% and 3×100%) followed by 2×10min acetonitrile; infiltrated with Epon 812 resin (Shell Chemical, Houston, TX) in acetonitrile 1:1 for 1hr followed by 100% Epon for 3hr; moved to fresh 100% Epon and polymerized at 70°C overnight. Sections of 70 nm thick were cut and stained with saturated uranyl acetate in 50% ethanol and Reynold’s lead citrate. Images were obtained by using a transmission electron microscope (JEM2100, Japanese Electron Optics Limited, Tokyo, Japan) with a digital camera (Ultrascan 1000, Gatan Inc., Pleasanton, CA). The number of Aβ and Aδ myelinated fibers were counted in low-magnification images and C-type fibers were counted in high-magnification images (Supple. Fig. 1). Aβ and Aδ fibers were distinguished by diameters of axons, which were calculated from their measured areas and assuming a round shape. Fibers with diameter>6.5μm were designated as Aβ, and diameter<6.5μm as Aẟ [35].

Protocol design

In TNI experiments, implantation of GFS electrodes and TNI surgeries were performed after baseline sensory testing on day 0. On 14–20d after electrode implantation surgery, effects of GFS with different frequencies were evaluated on the same group of rats with TNI. On each day, GFS was provided for 30min using one of the frequencies, which was randomly selected. The ensemble of pain behavior tests (von Frey, brush, cold, pin test, all done within 5min) was performed 15min before GFS, and again 15min after the initiation of GFS (i.e. during stimulation), and again immediately, 15min and 30min after the end of GFS stimulation. In teased fiber recordings from rats with TNI, recordings were performed 2 weeks after the TNI surgery. In recordings with trains of GFS from the same rat, there was a 45min rest interval between prior recordings to avoid the residual effects of the prior stimulation.

Statistical analysis

Statistical analyses were performed with Prism 8 (GraphPad Software, La Jolla, CA). To compare changes from baseline prior to simulation, responses to pin, brush and cold were evaluated nonparametrically using Friedman’s one-way repeated measures analysis of variance (ANOVA) with post hoc Dunn’s test. Responses to von Frey were evaluated using repeated measures ANOVA, with post hoc comparisons using Dunnett’s test. Responses to GFS were compared between groups by first calculating the area under the curve (AUC) for each subject by summing the product of the behavior value, normalized to the baseline just before the initiation of GFS, multiplied by the time interval between each pair of determinations, including timespans both during and after delivery of GFS. AUC values could be of either polarity. Differences between groups were then tested using one-way ANOVA of the AUC values, with post hoc comparison using Tukey test. Chi-Squared test was used to compare the incidence of blockade effects of GFS upon different groups. To control for familywise error for multiple planned comparison if there are 3 or more groups, post hoc correction for the P-value was multiplied by the total number of comparisons that were made. The effects of GFS on blocking propagation of AP trains from the sciatic nerve into the dorsal root were evaluated nonparametrically using two-way repeated measures ANOVA with post hoc Dunnett’s test. The effect of GFS on von Frey threshold for initiating firing and on firing rates induced by punctuate stimuli were evaluated using two-way repeated measures ANOVA with post hoc Dunnett’s test to compare values before and after GFS and between with and without nerve injury. Significance testing of P<0.05 was considered significant. Data are reported as mean ± SEM.

Results

GFS decreases neuropathy-induced hypersensitivity with a frequency dependent manner

In clinic for pain treatment, optimal analgesic effect is reached with 20Hz GFS [7], and in our preclinical model, we confirmed analgesic effectiveness with 20Hz GFS [47,62]. Here, we examined the range of frequencies at which GFS is effective. Two weeks after TNI and electrode implantation, hypersensitivity was evident by elevated rates for hyperalgesic responses to noxious mechanical stimulation (pin, Fig. 1A), reduced threshold for withdrawal from mechanical stimulation (von Frey, Fig 1B and Supple. Fig. 3), which used unmodified von Frey fibers as animals became sensitive to these following nerve injury increased frequency of withdrawal from dynamic soft touch (brush, Fig. 1C), and increased frequency of withdrawal from cold (acetone, Fig. 1D). Following this baseline testing session, testing was again performed at 15min and 30min timepoints during stimulation. GFS with frequencies of 20Hz and above diminished hypersensitivity to all tested modalities at both timepoints during GFS (Fig. 1A–D). Residual analgesia was evident 15min after terminating stimulation, although this was variable. Comparison between groups by AUC analysis (Fig. 1E–H) confirmed analgesic effect of treatments with frequencies of 20Hz and above, but no differences between GFS at frequencies of 20 through100Hz GFS. Motor threshold levels here were 0.24±0.05 mA, n=9 rats.

Figure 1.

Effects of GFS with different frequencies on nociceptive sensation on rats with TNI. Time course for effects on the testing day are shown in the left panels (A, B, C, D), and AUC analysis (E, F, G, H) for group comparisons are shown in the right panels, for sensitivity to noxious mechanical stimuli (pin, A, E), threshold mechanical stimuli (von Frey, B, F), brush (C, G), and acetone (F, H). The sham treatment (0Hz stimulation) group consisted of animals with GFS electrodes that were inserted but not activated. Red bars represent Means. Results are means ± SEM. The variance at the baseline timepoint is small such that the SEM is smaller than the radius of the data symbol. “1” P<0.05, “2” P<0.01, “3” P<0.001 compared to data immediately before GFS by the Dunnett’s test following two-way repeated-measures ANOVA. * P<0.05, *** P<0.001 by the Tukey test following one-way ANOVA. n represents animal numbers.

GFS with 80% motor threshold activates most DRG

Initiation of APs in C-type units at the level of their axons typically needs more than 10-fold more stimulation intensity than is required to activate Aβ axons [36,49]. Since the difference in somatic diameters of Aβ and C-type units (1.5–3 fold) is much smaller that the difference in their axonal diameters (>10 fold) [20,55], we speculated that the somata may not share this large difference in required stimulating intensities. We therefore employed single unit dorsal root teased fiber recording to identify what intensity is needed to activate somata of different sensory neuron types during GFS, which exposes the somata to excitatory electrical fields. Comparing stimulation levels needed to induce activity in dorsal root Aβ, Aẟ, and C-type units (Fig. 2A, B, C) showed that C-type units need on average a 1.3-fold higher intensity (0.14±0.00mA, n=84 units from 9 rats) than Aβ units (0.09±0.00mA, n=35 units from 9 rats, P<0.001) for initiating an AP (Fig. 2D, E). To determine if 80% motor threshold is sufficient to activate sensory neurons during GFS, we compared the averaged threshold to activate C-type units to the 80% motor threshold for that animal and found that 68 of the 72 recorded C-type units (94%, n=9 rats) were activated by current levels below 80%MT (Fig. 2F). To characterize the comparative excitability of different sensory neuron types when stimulating their axons alone, we performed a similar experiment but instead of stimulating at the DRG, we applied stimulation selectively to the sensory neuron peripheral axonal process at the level of the sciatic nerve using a cuff electrode for optimal contact with the nerve. We found that C-type units need a 60-fold higher intensity (3.02±0.17mA, n=23 units) than Aβ units (0.05±0.00mA, n=21 units, P<0.001 vs. C-type) to be activated (n=5 rats, Fig. 2G, H). We also examined the CVs while both recording and stimulating from saphenous nerve, which lacks motor neuron fibers and avoids T-junction filtering, and found a 29-fold higher stimulation threshold for C-type units (1.46±0.07mA, n=55 units from 5 rats) compared to Aβ units (0.05±0.00mA, n=17 units, P<0.001 vs. C-type) (Fig. 2I, J). Axon bundles in which both C-type and Aβ unit activity could be evoked were used for comparison of thresholds to needed activate C-type and Aβ units, thereby minimizing variance due to stimulating and recording conditions. With recordings from sciatic nerve stimulation, C-type units required 73±14 fold greater current than that for Aβ units (n=11 bundles). With recordings from saphenous nerve stimulation, C-type units required 33±4 fold greater current than Aβ units (n=12 bundles). These ratios contrast substantially with those derived from recordings during DRG stimulation, for which C-type units required only 1.5±0.1 fold greater current than Aβ units (n=28 bundles, P<0.001 vs. sciatic nerve stimulation; Fig. 2K). Taken together, these data indicate that exposing the axons to field stimulation highly favors Aβ activation rather than slower conducting units, whereas the somata of C-type units are readily activated by field stimulation at intensities only slightly higher than that needed to activate Aβ neurons.

Figure 2.

Recordings from dorsal root teased fibers. (A) The GFS lead was implanted besides L4 DRG. Dorsal root was teased and recorded. Representative AP firing from teased fibers by stimulating DRG (B) and sciatic nerve (C). Solid black arrows represent the start point of stimulation. Neuronal subtypes were distinguished by conduction velocity. (D) Minimum intensities with GFS to activate Aβ, Aδ, C, and motor fibers. Recordings were from 9 rats. (E) Conduction velocities of those recordings with GFS. (F) Comparison between averaged intensity to activate C-type neurons and 80% intensity to activate motor fibers on the same rat. (G) Minimum intensities with sciatic nerve stimulation to activate Aβ, Aδ, C, and motor fibers. Recordings were from 5 rats. (H) Conduction velocities of those recordings with sciatic nerve stimulation. (I) Minimum intensities with saphenous nerve stimulation to activate Aβ, Aδ, and C fibers. Recordings were from 5 rats. (J) Conduction velocities of those recordings with saphenous nerve stimulation. (K) Comparison of difference of intensities to activate Aβ and C with saphenous nerve, sciatic nerve and DRG stimulation. Red bar represents Mean. * P<0.05, *** P<0.001 by the Tukey test following one-way ANOVA.

The soma is the likely source of AP initiation during GFS

The findings above suggest that the increased excitability of the C-type neurons when field stimulation is applied to the DRG compared to the peripheral nerve is due to presence of the somata in the stimulation field. To test this, we examined if acutely dissociated somata, which lack axons, will replicate the above findings. We used field stimulation of DRG neurons loaded with the ratiometric Ca²⁺-sensitive fluorophore Fura-2. In this model, field stimulation produces R_340/380 transients in a saturating pattern of Ca²⁺ transients that reach a maximum despite increasing voltage (Fig. 3A), consistent with their generation requiring underlying AP firing, which is also supported by our prior finding that such transients are blocked by the sodium channel blocker lidocaine [18]. In turn, activation of voltage-gated calcium channels leads to intracellular Ca²⁺ accumulation, which used as an indicator of neuronal AP generation. Threshold values are the minimal voltage necessary to induce full-size Ca²⁺ transients. We observed moderate negative correlation (r=−0.45, P<0.001; n=172 cells from 5 rats) between cell size and the intensity of field stimulation needed to produce a Ca²⁺ transient (Fig. 3B). When these neurons are classified into large (>34 μm) vs. small diameter (≤34 μm), which represent predominantly fast-conducting non-nociceptive Aβ neurons vs. a mix of slower conducting Aδ nociceptive neurons and C-type nonmyelinated nociceptive neurons [19], we identify a pattern similar to GFS excitation of sensory neurons, with a 1.2-fold higher stimulating intensity necessary to activate small-sized dissociated neuronal somata (24.85±0.25 V, n=123 cells) than large neurons (21.76±0.7 V, n=49 cells; P<0.01 vs. small-sized neurons). To account for variations between cover slips, we calculated the ratio of averaged intensities necessary to activate large- vs. small-sized neurons for each cover slip, which averaged 1.15±0.05 fold, n=10 cover slips from 5 rats (Fig. 3C).

Figure 3.

Activation of dissociated neurons and in vivo intact DRG neurons by GFS with Ca²⁺ indicators. (A) Field stimulation (10Hz-0.5s) produces Ca²⁺ transients in DRG neurons loaded with the ratiometric Ca²⁺-sensitive fluorophore Fura-2 in a saturating pattern of Ca²⁺ transients that reach a maximum despite increasing voltage. The threshold (24V in this case) was defined as the minimal voltage necessary to induce full-size Ca²⁺ transient. Arrows indicate stimulation. (B) There is negative correlation (r²=0.20, P<0.001) between neuronal sizes and threshold intensities needed to activate the voltage-gated calcium channels. Red line is the linear regression line. n=172 neurons from 5 rats. (C) Ratios of intensities to activate small- and large-sized neurons on the same cover slip with in vitro field stimulation. Each data point represents the mean value for small neurons divided by the mean value of large neurons on a single slip. Red bar represent Mean for the group; per slip, n=8–16 small neurons and n=4–6 for large neurons. (D) GFS on in vivo intact DRG GCaMP-expressing neurons. Bright-field image with IR-LED (left), fluorescent image during stimulation (middle), and representative response trace (right) from cell #1. Numbers in the bright-field image represent cell recorded. Solid arrow represents GFS. (E) Summary of effects of GFS (20Hz-1s with intensity of 80% MT) on activation of DRG neurons. Data are shown for all fluorescent neurons in the examined fields regardless of the presence or absence of a response to stimulation. Red line is the linear regression line. n=66 total neurons from 5 rats.

Since dissociation may possibly disrupt neuronal properties, we also examined GFS activation of DRG neuronal somata in situ during live animal recording using GCaMP fluorescence increase as an indicator of neuronal activation. GCaMP has been used as an indicator of AP generation [14], although its ability to resolve individual AP events is not certain. Here, we use it to indicate the overall neuronal activity by GFS (Fig. 3D). The size of the Ca²⁺ transient response to field stimulation showed no relationship to cell diameter (r=−0.02, P=0.89, Fig 3E), indicating that different neuronal groups are equally affected by stimulation.

These findings together indicate that the intensities needed to activate the somata of different types of DRG neurons are similar, and that all sizes neurons can be activated with the 80% motor threshold, and further that the somata are the likely neuronal component stimulated by GFS.

APs initiated by GFS rapidly fail to propagate into the dorsal root

Since GFS initiates AP firing in most C-type and many Aẟ neuronal somata, the resulting AP trains would be expected to propagate to spinal cord and GFS should induce pain. However, we have previously shown that the sensory neuron T-junction, where the stem axon coming from the soma joins the central and peripheral axonal processes, filters the passage of AP trains in a use-dependent manner [20]. We therefore tested whether AP trains triggered in the somata by GFS can reach the dorsal root by recording AP firing in axon bundles teased from the dorsal root while applying GFS in the range of frequencies that showed effectiveness in reducing pain behavior (Fig. 1). GFS triggered AP firing that persisted in Aβ units for the full testing interval of 180s for stimulation at 10–20Hz (Fig. 4A–D). However, with stimulation at 50–100Hz, impulse propagation eventually failed in a subset of units after a variable time interval (Fig. 4D, E). A similar pattern was seen for Aδ units, although conduction failure happened more frequently and at lower frequencies (Fig. 4F), and failure occurred with a faster time course (Fig. 4G). For C-type units (Fig. 4H–J), firing failed within a few seconds at high frequencies and at 21.1±3.8s at 20Hz (Fig. 4H), with a pattern of progressively longer inter-spike intervals (reduced firing rates, Fig. 4J). Two of 26 units showed persistent firing (Fig. 4H). These data suggest that APs initiated in somata of DRG neurons cannot propagate to spinal cord with a frequency-dependent manner. It is interesting to note that GFS at 20Hz, the most commonly used frequency in clinical GFS application, resulted in near complete failure of impulse propagation in C-type units (Fig. 4H), failure of approximately half of Aδ units (Fig. 4F), and persisting firing of all Aβ units (Fig. 4D). Those recordings were from 32 naïve rats.

Figure 4.

Effects of GFS on AP propagation from soma to central. (A-C) Representative traces showing blocked APs from C-type not Aβ units during 20Hz GFS. (D) Effects of GFS with different frequencies on APs from Aβ units. (E) Comparison of the times to failure of AP propagation (i.e. time from beginning of GFS to the last induced AP) in Aβ units during GFS. (F) Effects of GFS with different frequencies on APs from Aδ units. (G) Comparison of onset of blockade of APs in Aδ units during GFS. (H) Effects of GFS with different frequencies on APs in C-type units. (I) Comparison of onset of blockade of APs from C-type units during GFS. (J) A representative pattern of progressively failed firing following 20Hz GFS. Those recordings were from 32 rats. Red bars represent Means. * P<0.05, ** P<0.01, *** P<0.001. In D, F, and H, Chi-Squared test was used to compare the incidence of blockade effects of GFS upon different groups and the post hoc correction for the P-value was multiplied by the total number of comparisons that were made. In G and I, one-way ANOVA followed by post hoc Tukey test.

GFS blocks propagation of AP trains from the sciatic nerve into the dorsal root

Rather than enhanced T-junction filtering, an alternative explanation for the failure of GFS-induced firing to reach the dorsal root could be failure of AP generation in the somata. To selectively examine the T-junction directly, we tested if GFS affects propagation of AP trains approaching the T-junction from the peripheral process, rather than from the stem axon, by stimulating activity in the sciatic nerve while recording from the dorsal root. As 20 Hz GFS is effective in clinical GFS and in rats (Fig. 1), we chose 20 Hz GFS for this purpose. Stimulation of the sciatic nerve at 5 Hz for 4s can consistently evoke 20 APs without failure in units recorded from the teased L4 dorsal root, including Aβ-, Aδ-, and C-types (Fig. 5A, B). However, when GFS is applied to the DRG (180s, 20Hz) in naïve rats (n=10), AP propagation to the dorsal root fails (recorded AP <20) in most C-type units and half of Aδ units (Fig. 5D–I). Similar effects were observed in rats with neuropathic pain following TNI (Fig. 5F–I, n=9 rats). When the GFS electrode was moved to a position adjacent to the spinal nerve rather than the DRG, 20Hz stimulation at the same intensity failed to activate C-type fibers and did not block AP propagation (recordings from 16 Aδ and 18 C-type units, n=4 rats). This indicates that GFS prevents AP traffic in the peripheral axonal process from entering the central process by a mechanism requiring stimulation of the sensory neuron somata, and implicates the T-junction as the site of GFS-induced propagation failure.

Figure 5.

Effects of 20Hz GFS on AP propagation by sciatic nerve stimulation. (A) The GFS lead was implanted besides L4 DRG. Dorsal root is teased and recorded. Representative traces show firings from Aδ and C-type units with DRG stimulation (B), sciatic nerve stimulation (C), and with both stimulations (D). (E) Sequence of events. 20Hz-4s sciatic nerve stimulations were given 30s before GFS, and 30s, 60s, 90s, 120s, 150s, and 180s after initiation of GFS. (F) Effects of GFS on recordings from sham surgery and rats with TNI. The incidence of full blockade of AP propagation were compared on different subtypes, Aβ (G), Aδ (H), and C (I). Those recordings were from 10 rats with sham injury and 9 rats with TNI. Arrows represent stimulation. 3 P<0.001 compared to data immediately before GFS by the Dunnett’s test following two-way repeated-measures ANOVA.

GFS regulates afferent traffic in C-type and Aδ units following nerve injury

Since GFS can reduce pain-related behavior triggered by normal nociception (Fig. 1), GFS may also block AP trains associated with allodynia and hyperalgesia following nerve injury, which is a possibility made more likely by our previous observation that nerve injury results in reduced T-junction filtering of C-type units [20]. We first measured the effect of GFS on the weakest of a series of von Frey filaments that can initiate AP firing recorded in the dorsal root (i.e. von Frey threshold) (Fig. 6A). We identified C-type (Fig. 6B) and Aδ units (Fig. 6C) that had receptive fields in the lateral side of hind paw plantar glabrous skin to be matched with recordings from rats with TNI (Fig. 6D). Because of interference by stimulation artifacts from the GFS, we compared thresholds at baseline (1 min before initiating GFS), and immediately after terminating GFS (which lasted 120 s). In rats with sham surgery, GFS did not change the von Frey threshold (Fig. 6B, C). In rats with TNI, the baseline von Frey threshold was reduced compared with that in sham rats, but this hypersensitivity was reversed by GFS (Fig. 6B, C). To further characterize the effect of GFS on sensory signal, we evaluated how GFS affects propagation of AP trains initiated by standardized cutaneous force applied using 29g standard von Frey (tip diameter 1.1mm) and 16g modified von Frey fiber (100μm tungsten tip), while recording single units from the dorsal root (Fig. 7A–D). Prior to GFS, C-type units, but not Aδ units, showed AP firing rates that were increased in TNI rats compared to sham animals during mechanical force application with both types of von Frey fiber (Fig. 7E, F, G, H). GFS reduced the rate of APs reaching the dorsal root recording site in C-type units, but not Aδ units, during mechanical force application with both types of von Frey fiber in both Sham and TNI rats (Fig. 7E, F, G, H), indicating amplified T-junction filtering immediately following GFS. Those recordings were from 10 rats with sham surgery and 10 rats with TNI surgery.

Figure 6.

Effects of 20Hz GFS on von Frey threshold to initiate AP firing by touching the receptive field. (A) The GFS lead was implanted besides L4 DRG. Dorsal root was teased and recorded. von Frey fibers were used to initiate AP firing. Effects of 20Hz GFS on threshold of AP firing by von Frey threshold in C-type fibers (B) and in Aδ (C). (D) Receptive fields of those recordings in the hind paw plantar glabrous skin. Those recordings were from 10 rats with sham surgery and 10 rats with TNI surgery. * P<0.05, ** P<0.01, *** P<0.001 by the Dunnett’s test following two-way repeated-measures ANOVA.

Figure 7.

Effects of 20Hz GFS on AP firing by mechanical stimulation. (A-D) Representative traces showing AP firing before and after GFS in rats with sham surgery and rats with TNI. Effects of 20Hz GFS on AP firing by blunt (E) and sharp (F) puncture in C-type fibers. Effects of 20Hz GFS on AP firing by blunt (G) and sharp (H) in Aδ fibers. Those recordings were from 10 rats with sham surgery and 10 rats with TNI surgery. Red bar represents Mean. * P<0.05, ** P<0.01 by the Dunnett’s test following two-way repeated-measures ANOVA.

GFS regulates spontaneous firing in C-type units following nerve injury

Our previous reports [46,62] have indicated that GFS can alleviate spontaneous pain behavior (conditioned place preference), which may be due to GFS limiting propagation of spontaneous firing that arises after nerve injury. The incidence of spontaneous firing in the L4 dorsal root after TNI was increased (Fig. 8A). Of those 14 units with spontaneous firing in rats with sham injury, 7 units are C-type, 4 units are Aẟ, and 3 units are Aβ. Of those 13 units with spontaneous firing in rats with TNI, 10 are C-type, 2 are Aẟ, and 1 is Aβ. Recordings immediately after GFS showed reduced firing rates of spontaneous firing in C-type units (Fig. 8B). Those recordings were from 6 rats with sham surgery and 5 rats with TNI surgery.

Figure 8.

Effects of 20Hz GFS on spontaneous firing after nerve injury. (A) Injury increases incidence of SA. ** P<0.01 by Chi-Squared test. (B) GFS reduces increased spontaneous firing rates in C-type units after injury. Those recordings were from 6 rats with sham surgery and 5 rats with TNI surgery. Red bar represents Mean. * P<0.05, ** P<0.01 by the Dunnett’s test following two-way repeated-measures ANOVA.

These in vivo observations suggest that excess afferent activity of C-type and Aδ units contributes to allodynia and hyperalgesia after nerve injury, and that GFS can enhance natural filtering function of the T-junction and return diminished T-junction filtering after injury to normal levels (Fig. 9).

Figure 9.

Proposed mechanism of GFS-induced augmented T-junction filtering. Field stimulation (1) induces depolarization of the somatic membrane that initiates APs (2), which travel down the stem axon to the T-junction (3), inducing use-dependent reduction of membrane excitability at the T-junction (4), which increases the probability of failure for both AP propagation from the stem axon to the central process fails, and for centripetal propagation of APs originating in the periphery (5).

Discussion

Electrical stimulation of segmental sensory neurons at the level of the DRG is an effective clinical treatment for pain, although the mechanism of its action is not established. We developed a preclinical model in which single level (L4) GFS provides analgesia against neuropathic pain [46,62], which makes exploration of the mechanism possible. Here, we used this model to test effects of GFS on generation and propagation of APs triggered in the sensory neuron axons and somata, and on natural activity originating in their peripheral terminals in rats with sham surgery and rats with nerve injury. Multiple mechanisms may contribute to the analgesia generated by GFS, including those based on activation of Aβ units, as occur during spinal cord stimulation [50]. However, GFS does not increase gamma-Aminobutyric acid (GABA) release in the dorsal horn of spinal cord [29], suggesting that mechanisms in the DRG may play a dominant role. Our findings support the view that an important component is contributed by a process in which GFS causes neuronal somata to generate APs that propagate to the T-junction, where use-dependent enhancement of the natural filtering function of the T-junction suppresses passage of impulses. This traffic originating in the somata rapidly elevates filtering to a level that prevents not only these impulse trains from continuing on to the spinal cord but, in addition, the T-junction is conditioned by this induced traffic such that APs from the peripheral process are also limited in their passage through the T-junction (Fig. 9).

Our first finding is that C-type units are readily activated by GFS (Fig. 2). This is in agreement with predictions from modeling the response of neurons to GFS [28], but disproves the prediction by others [24] that C-type units are unexcitable using clinical levels of stimulation intensities. Evidently, C-type neuronal somata are distinct from their axonal processes that are highly resistant to field stimulation. Sensitivity of C-type somata to activation by electrical field stimulation provides an opportunity to clinically modulate these nociceptive neurons. Aβ neurons are concurrently activated during GFS, but in the clinical setting, patients typically find the resulting paresthesia inoffensive and often desirable, possibly due to associating it with pain relief, or due to the analgesic effects of Aβ stimulation [43].

Neuronal branch points have been identified as key regulators of information integration in vertebrate central and peripheral neurons [5,45]. In sensory neurons, impedance mismatch at the T-junction results in low-pass filtering. APs that can propagate through T-junction have much slower repetition rates than the rates at which the soma and axon can be excited [20]. T-junction filtering is multidirectional [20,54], such that exposure of the T-junction to AP trains from the soma will limit the transmission of APs from the peripheral process to the central process, which we confirmed with recordings from the dorsal root. This block persisted after the termination of GFS for 20–30min, and C-type units were particularly sensitive. T-junction filtering is Ca²⁺-dependent [20], and the persistence and particular sensitivity of C-type units may be explained by our previous findings that Ca²⁺ transients generated by single APs in C-type neuronal somata decay more slowly and are much larger than those from Aβ-type neurons [22]. Furthermore, AP trains can produce a higher amplitude [Ca²⁺]_c signal by stacking of a sequence of [Ca²⁺]_c increments with each APs, and C-type neurons can do so at a much lower frequency (3–20Hz) than is needed by Aδ neurons (3–50Hz) or Aβ neurons (20–100Hz) [22]. If these observations in the somata reflect similar events at the T-junction, this may be the basis for our present findings that GFS blocks AP propagation in a frequency dependent manner that requires at least 50 Hz GFS in Aβ neurons, 10Hz in Aδ neurons, and 5–10Hz in C-type neurons (Fig. 4).

It is interesting to note that we found no differences between GFS at frequencies of 20Hz, 50Hz and 100Hz in their effectiveness in blocking behavioral responses to non-noxious stimuli of brush, von Frey, and acetone (Fig. 1), despite our other finding that 50Hz- and 100Hz-GFS block AP propagation in Aβ and Aẟ units more effectively than 20Hz-GFS. Since we observed no differences between frequencies in blockade AP propagation for C-type units (Fig. 4), this suggest that the analgesic effects of 20Hz-GFS are most attributable to blockade of C-type fibers. Consistent to this idea, it was reported that C-type fibers, not A fibers, convey low-threshold mechanoreceptor inputs by light touch such as brush to dorsal horn projection neurons [1], and that injury may induce mechanical hypersensitivity of nociceptors [11] or C-type low threshold mechanoreceptors [51]. Combined with our finding that the mechanical threshold for initiating AP firing in C-type fibers is reduced by injury (Fig. 6), these observations support the view that C-type fibers, not A fibers, are the dominant pathogenic neuronal type contributing to mechanical hypersensitivity after injury.

Activity-induced loss of membrane excitability may be explained by our prior findings that accumulated intracellular Ca²⁺ decreases membrane excitability by activating Ca²⁺-sensitive K⁺ channels [25,40], and that this accumulated intracellular Ca²⁺ in turn activates Ca²⁺-calmodulin kinase II [31], which amplifies Ca²⁺ influx, and retains its activated state long after intracellular Ca²⁺ levels return to baseline [57]. Similarly, modeling has suggested that action of GABA on T-junction membrane may inhibit AP propagation by amplifying T-junction filtering [10].

After peripheral nerve injury, resting [Ca²⁺]_c levels and depolarization-induced Ca²⁺ entry are decreased in axotomized neurons [17,18,22,27,41], which is accompanied by diminished T-junction filtering in C-type neurons [20]. The importance of the T-junction as the critical site for injury-induced hyperexcitability is highlighted by the observation that peak somatic firing rates are not affected by injury [20]. Our new data show that peripheral nerve injury results in elevated firing rates induced by cutaneous mechanical stimulation for units recorded from the dorsal root, reflecting reduced T-junction filtering. Upon initiation of GFS, the rate of APs arriving in the central process returns back to pre-injury levels (Fig. 7). This explains the prompt onset of analgesia in our behavioral experiments in animals with chronic pain after nerve injury [46,62]. Since these recorded units are responsive to tissue stimulation, they cannot themselves be axotomized, but may be hyperexcitable on the basis of inflammatory changes in the DRG that their somata share with axotomized neurons [33]. Since GFS normalizes the firing rates and reverses the lowered mechanical force threshold for unit activation, it is possible that loss of T-junction filtering may be a pathogenic component in the elevated excitability of these intact units adjacent to axotomized ones. We note, however, that the portion of the peripheral process distal to the T-junction may also show hyperexcitability for supramaximal stimulation [52]. Although this component cannot be explained by injury-induced loss of T-junction filtering, its contribution to the impulse trains can nonetheless be blocked by GFS.

We have previously shown that GFS can prevent neurogenic inflammation in peripheral tissues [47], so an additional contribution to analgesic effectiveness of GFS could be its blockade of efferent activity in C-type units that may otherwise release inflammatory peptides in the DRG. However, analgesia occurs within minutes of initiating GFS, so it unlikely that this is a dominant mechanism of immediate GFS action, although there may be a contribution during chronic use, as we showed in its prevention of joint inflammation [47].

Our electrophysiological findings showing reduced passage of AP trains into the central process during noxious peripheral stimulation in the sham injured control group (Fig. 7F, H) indicate that impulses triggered by normal nociception can also be regulated by GFS. This agrees with our behavioral findings in rats without nerve injury that GFS diminished sensitivity to mechanical and heat stimulation [62]. However, it remains to be determined whether GFS can be used clinically to treat acute pain as this is not currently an established indication and there are no published observations testing the efficacy of GFS against normal nociception in the clinical setting.

In summary, the novel findings of our present study indicate that GFS can activate all types of DRG neurons, and that AP trains induced by GFS cause elevated low-pass filtering of AP trains originating from the soma and from the peripheral axonal process. This in turn leads to transmission of fewer impulses to the spinal cord, resulting in analgesia (Fig. 9).

Supplementary Material

Supplementary figure 1

Supplementary Figure 1. Ultrastructure of recorded dorsal root bundles. (A) An Aβ is marked with open red arrow and an Aδ is marked with open yellow arrow. (B) A C-type fiber in a Remak bundle is marked with solid red star. (C) Summarized data from 3 teased dorsal root bundles.

Supplementary figure 2

Supplementary Figure 2. CVs are different when the fibers are stimulated at the sciatic nerve and DRG levels. (A) Single unit firings from dorsal root teased fiber, which can be activated by both DRG stimulation and sciatic nerve stimulation, showed different CVs. (B) Fold difference of CVs when stimulated at the sciatic nerve and DRG levels. *** P<0.001 by the paired t-test.

Supplementary Figure 3

Supplementary Figure 3. Threshold values are shown for response to von Frey mechanical stimuli before TNI and after injury but before each GFS treatment.