Processing noxious information at the subnucleus reticularis dorsalis (SRD) of anesthetized cats: Wind-up mechanisms

www.elsevier.com/locate/pain

Pain 140 (2008) 190–208

Processing noxious information at the subnucleus reticularisdorsalis (SRD) of anesthetized cats: Wind-up mechanisms

Cristina Soto 1, Francisco J. Martın-Cora, Roberto Leiras, Patricia Velo, Antonio Canedo *

Department of Physiology, Faculty of Medicine, University of Santiago de Compostela, 15706 Santiago de Compostela, Spain

Received 20 February 2008; received in revised form 24 May 2008; accepted 12 August 2008

Abstract

With the exception of one monkey’s study, where wind-up was not reported, electrophysiological data from SRD neurons wereobtained in rodents where they show wind-up. This work was designed to examine the response properties of SRD neurons in anes-thetized cats to study how general the data from rats may be. Since cat’s SRD cells showed wind-up, its underlying mechanisms wereapproached, an issue not previously addressed at supraspinal level. Electrical stimulation, extracellular (combined with microion-tophoresis) and intracellular techniques revealed that Ad information reaches the SRD via the ventrolateral cord, whereas C infor-mation preferentially follows a dorsal route. Wind-up was usually generated by spinal and peripheral stimulation, but it was alsoevoked either by stimulating the nucleus reticularis gigantocellularis (NRGc), even after spinal cord section and bilateral full thick-ness removal of the cerebral cortex, or by applying microiontophoretic pulses of L-glutamate at 0.3–1 Hz. Wind-up relied on afferentrepetitive activity gradually depolarizing the SRD neurons leading 3–4.5 Hz subthreshold membrane rhythmic activity to threshold.Riluzole retarded wind-up generation and decreased the number of spikes per stimulus during wind-up. GABA or glycine abolishedspontaneous and sensory-evoked activity and bicuculline, but not strychnine, increased spontaneous and stimulus-evoked activity.These results demonstrate that wind-up at the SRD is not merely the reflection of spinal wind-up, but (i) can be locally generated, (ii)is partially dependent upon persistent sodium currents, and (iii) is under the modulation of a tonic GABAa-dependent inhibitiondecreasing SRD excitability. Injury and/or inflammation producing tonic C-fiber activation will surpass tonic inhibition generatingwind-up.� 2008 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.

Keywords: Subthreshold oscillations; Microiontophoresis; Ad and C pathways; Medullary reticular formation; Intracutaneous and spinal stimul-ation

1. Introduction

The SRD is located in the caudal medulla oblongataventral to the dorsal column nuclei in the rat [41,58],rabbit [38] and monkey [64], extends rostro-caudallyfrom the area postrema to the spinomedullary junction,and lies medial to the pars caudalis of the spinal trigem-inal nucleus. In the rat, both superficial I–II and deepdorsal horn laminae are reciprocally connected with

0304-3959/$34.00 � 2008 International Association for the Study of Pain. P

doi:10.1016/j.pain.2008.08.007

* Corresponding author. Tel.: +34 981582658; fax: +34 981574145.E-mail address: [email protected] (A. Canedo).

1 Present address: University of Southern California, 3641 WattWay, HNB 328, Los Angeles, CA 90089-2520, USA.

the SRD [1,3,11,35,46,54,61,65], suggesting that theSRD might be implicated in a spino-reticulo-spinal net-work modulating the ascending noxious information.

The SRD neurons of rats and monkeys [62–64] havelarge noxious receptive fields all over the body. In thecat, however, no report was specifically addressed tostudy the SRD. Nevertheless, some data from the med-ullary caudal reticular formation allow deducing someproperties of feline’s SRD neurons. It was reported thatneurons in the caudal bulbar reticular formation ofanesthetized and decerebrated cats had large noxiousreceptive fields, some of which comprised the face andthe body [15], and that there is a somatotopic represen-tation in the cat’s caudal medullary brain stem [26]:

ublished by Elsevier B.V. All rights reserved.

C. Soto et al. / Pain 140 (2008) 190–208 191

nociceptive segmental neurons were located medially inrelation to those having convergent receptive fields fromthe body and face.

Suprathreshold intracutaneous electrical stimulationinduces wind-up in rat’s SRD neurons [63]. Wind-upwas first described in the spinal cord [39] and consistsof a gradual build-up in neuronal excitability pro-duced by low-frequency (0.3–2 Hz) repetitive stimula-tion of C-fibers exclusively. In the spinal cord, thedorsal horn multimodal neurons in the deeper laminaeshow the most pronounced wind-up [51] relying onNMDA and neurokinin receptors [20,22,55,56] as wellas on intrinsic membrane properties [25,40,44]. In theSRD, the mechanisms underlying wind-up have notyet been addressed.

This work aimed to analyze cat’s neurons with seg-mental noxious receptive fields located ventral to thedorsal column nuclei (SRD neurons) to study, (i) theresponses elicited by peripheral and spinal cord electricalstimulation. Stimulation of the dorsolateral and of theventrolateral spinal cord also served to antidromicallydistinguish spinally projecting (SPr) from non-spinallyprojecting (nSPr) SRD neurons; (ii) whether cat’s SRDneurons show wind-up and, if so, approach its underly-ing mechanisms; and (iii) the effects produced by themicroiontophoretic ejection of various presumed neuro-transmitters and their antagonists on the spontaneousand stimulus-evoked activity of SRD neurons.

2. Methods

2.1. General

All procedures conformed to the Spanish Physiologi-cal Society, the International Council for LaboratoryAnimal Science, the European Union Council Directive(86/609/EEC), and were approved by the UniversityAnimal Care Committee. Successful electrophysiologi-cal experiments were performed on 53 anesthetizedand paralyzed cats of either sex weighing 2–4.5 kg. Sur-gical anesthesia was induced with ketamine HCl (10–20 mg/kg im) and continued with a-chloralose (60 mg/kg iv, n = 35 animals) or sodium pentobarbital(35 mg/kg iv, 18 cats). Since rat’s nociceptive thalamicventrobasal neurons showed depressive and unmaskingreceptive field effects under a-chloralose anesthesia[27], a-chloralose and sodium pentobarbital were usedto evaluate their effects on the receptive field propertiesof feline’s SRD cells. There was no evidence that choiceof anesthetic affected receptive field size or modality ofresponses (noxious vs. non-noxious). Additional dosesof anesthesia (1/2 of a full dose) were regularly adminis-tered every 4–5 h. The depth of anesthesia was evaluatedby continuously monitoring the heart rate (maintainedaround 120 beats/min), the electrocorticogram (ECoG;digitally filtered at a frequency band-pass of 1–20 to

100 Hz) and by observing the state of the pupil. High-amplitude and low-frequency electrocorticographicwaves were taken as sign of adequate anesthesia, anddilated pupils or pupils reacting rapidly to electricalstimuli were considered to reflect inadequate anesthesiain which case a supplementary half of a full dose ofeither anesthetic was immediately injected.

Tracheal and venous cannulae were inserted; the ani-mal was positioned in a stereotaxic frame and artificiallyventilated. A bilateral pneumothorax was routinely per-formed to minimize pulsatile movements, the expiredCO2 was maintained at 4 ± 0.3%, a pH-balanced solu-tion of 5% glucose in physiological saline was continu-ously infused (4 ml/h iv), and the temperature wasmaintained near 37.5 �C via a DC heating pad undercontrol of a rectal thermoprobe. Warm agar (4% in0.9% saline) was added to all exposed tissues to increasestability and avoid desiccation. The animals were sus-pended with clamps attached to the lumbar and thoracicvertebrae, the foramen magnum was exposed, and theposterior arch of the atlas and the occipital bone wereresected to uncover the cerebellar vermis. The duraand arachnoid were then removed to permit the inser-tion of recording electrodes from the obex to about4 mm caudal to it.

2.2. Electrical stimulation

Bipolar stimulating electrodes, 2 mm inter-tip separa-tion, were placed at C2 in the ipsilateral dorsal spinalcord (PQ, posterior quadrant) and in the ipsilateral ven-trolateral spinal cord (AQ, anterolateral quadrant) atC3. The PQ electrodes were medio-laterally oriented sothat the dorsal column and the dorsolateral fasciculus(DLF) could be stimulated. Additional bipolar stimulat-ing electrodes were stereotaxically positioned [10,48,52]in the ipsilateral nucleus reticularis gigantocellularis(NRGc: P6–P8, 1–1.5 mm lateral to the midline, 3–5 mm deep from the floor of the fourth ventricle), astructure in the rostro-ventro-medial medulla that mod-ulates the activity of dorsal horn noxious ascendingactivity [9,66]. Intracutaneous electrical stimuli wereapplied through pairs of steel needle electrodes spaced6–10 mm apart, which were routinely inserted into thecentral footpad of each limb. Two additional stimulat-ing electrodes separated 300–350 mm were also placedin the tail to have an estimation of the conduction speedof peripheral fibers activating SRD neurons. Rectangu-lar pulses of 0.05–0.15 ms duration and up to a maxi-mum of 1 mA intensity were applied to the centralstructures and from 5 to 10 mA intracutaneously. Elec-trical stimuli were isolated from the ground providingconstant current to the preparation by using a transis-torized optically coupled unit (PSIU6, Grass Instru-ments, Quincy, MA, USA) placed as close to thepreparation as possible and connected in series with

192 C. Soto et al. / Pain 140 (2008) 190–208

the output of a dual pulse digital stimulator (S8800,Grass Instruments) and the stimulating electrodes.

2.3. Extracellular recording and iontophoresis

In a first series of animals (n = 12), the recording ofsingle neurons was accomplished via tungsten elec-trodes (12 MX resistance), while in a second set (28animals) five-barreled pipettes were attached to amechanical microdrive and used for extracellularrecording and iontophoresis. Standard controls forpH and current balancing were performed. The centerbarrel of each five-barrel electrode was filled with 3 MNaCl for recording, one barrel was used for currentbalancing, and each of the remaining three barrelswas filled with a solution of one of the followingdrugs: GABA (1 M, aqueous solution brought to pH4 with HCl), bicuculline (BiCu) methiodide (20 mM,GABAa receptor antagonist, dissolved in 0.1 N HCland brought to pH 4 with 0.1 N NaOH), glycine(1 M in aqueous solution adjusted to pH 3.5 withHCl), strychnine (10 mM, glycine receptor antagonist,aqueous solution adjusted to pH 6.5 with NaOH),D-2-amino-5-phosphonovaleric acid (AP5; 50 mM,NMDA receptor antagonist, aqueous solutionadjusted to pH 8 with NaOH), 6-cyano-7-nitroquinox-aline-2,3-dione (CNQX; 1 mM, non-NMDA receptorantagonist, aqueous solution adjusted to pH 8 withNaOH), L-glutamate (100 mM, aqueous solutionadjusted to pH 8 with NaOH), or the neuroprotectiveagent 2-amino-6-trifluoro-methoxy-benzothiazole (Ril-uzole, 30 mM, dissolved in a 3:1 aqueous solution ofDMSO which was adjusted to pH 4 with HCl). Whenejecting riluzole, the balancing pipette was filled withits vehicle at pH 4, and the effect of vehicle ejectionwas tested alone at the currents also used for riluzoleejection. No effect on the SRD neurons was detectedwhen using the vehicle alone. Furthermore, this samevehicle was also sporadically used for current balanc-ing when ejecting glutamate, which increased the firingof SRD neurons, and thus the decreasing effectinduced by riluzole ejection (see Section 3) cannot beattributed to its vehicle.

Ejecting DC currents (negative for AP5, CNQXand glutamate; positive for the others) were in therange of 10–100 nA for AP5, of 5–30 nA for riluzole,of 5–40 nA for glutamate and glycine, and of 10–50 nA for CNQX, GABA and BiCu. The drugs weretypically administered in 5–10 nA steps of increasingcurrent, while monitoring changes in spontaneous aswell as in sensory-and-electrically evoked activity.The currents used for BiCu and strychnine wereselected by measuring the current level necessary toreverse the effect of iontophoresed GABA and glycine,respectively. The currents for riluzole were alwayskept below 30 nA. When not in use, each drug barrel

was subjected to retaining 5–15 nA current of theappropriate polarity to prevent unwanted diffusionfrom the pipette.

The effects produced by intracutaneous and cervicalspinal cord electrical stimulation were averaged for30–100 trials in control conditions, again during ejec-tion of drugs, and finally after recovery. Stability wasascertained by comparing responses collected duringthe first and the last set of trials, and only cells thatshowed recovery of responses above 80% of controlwere included in the study. Post-stimulus histogramswere constructed from the accumulated number ofall spikes in the response for the total number ofsweeps. Neuronal injury discharges as well as pre-sumed fiber recordings (exclusively positive dischargesof 60.5 ms duration) were disregarded. Stimulationthrough PQ and AQ electrodes served as huntingstimuli. The mean of measures in control vs. in thepresence of drugs as well as the average conductionvelocities were compared using paired or unpairedStudent’s t-tests. One-way repeated measures analysisof variance was used to analyze wind-up data gener-ated by ejecting glutamate at increasing currents.P < 0.05 was regarded as statistically significant. Dataare presented as mean ± SEM. Single tracings areshown throughout.

Recordings were stored unfiltered in magnetic tape(sampling frequency 15–25 kHz) and were digitizedoff-line for analyses through an analog to digital inter-face (CED 1401 PLUS, Cambridge, UK). CED spike2software was used to process the neuronal activity.Multiple spikes, if present, were isolated on the basisof spike shape parameters from the amplified unpro-cessed neuronal signals using the spike shape templatefunctions in spike2 v.5. To ensure that the same neu-rons were evaluated before and after drug ejection, thetemplate generated from recordings before drug appli-cation was used for spike sorting in all consecutiverecordings. The unitary nature of the activity selectedfor processing was confirmed by autocorrelation anal-ysis (2- or 5-ms bin width). Spikes were considered assingle units when their action potentials maintained aconstant shape and when there was a clear gap, corre-sponding to the refractory period, spanning zero timeon the autocorrelogram. The anterior–posterior andmedial–lateral positions with respect to the obex aswell as the depth of neurons from the dorsal surfacewere monitored (see Fig. 1A and C).

Following pilot experiments, the criteria for select-ing a neuron were (i) its location ventral to the cune-ate nucleus (Fig. 1A), in a region where antidromicresponses to PQ stimulation were present (Fig. 1B)(AQ stimulation did not generate antidromicresponses), and showing (ii) wide peripheral receptivefields to 5–10 mA intracutaneous electrical stimulation(Fig. 1D).

Fig. 1. Sampling criteria and distribution of sampled cells. (A) Neurobiotin injection (signaled by an arrow at left) where the cell illustrated in (B)was encountered marked some cells, magnified at right. Cu, cuneate nucleus; Gr, gracile nucleus; LRt, lateral reticular nucleus; IO, inferior olive; Pyr,pyramidal tract; Sol, nucleus tractus solitarius; Sp5C, spinal trigeminal nucleus pars caudalis; SRD, subnucleus reticularis dorsalis; SRV, subnucleusreticularis ventralis. (B) An antidromic spike evoked by posterior quadrant electrical stimulation (PQ, the arrows point to the stimulus artifacts)collided with a synaptically driven spike (arrowhead). (C) Distribution of spinally projecting (SPr) and non-spinally projecting (nSPr) cells taking theobex and the dorsal surface as references. AP, anterior–posterior; DV, dorsal–ventral; ML, medial–lateral. (D) Responses of the silent cell shown in(B) to intracutaneous electrical stimulation of the four central pads (stimulation occurred at time 0). CFP, contralateral forepad; CHP, contralateralhindpad; IFP, ipsilateral forepad; IHP, ipsilateral hindpad.

C. Soto et al. / Pain 140 (2008) 190–208 193

194 C. Soto et al. / Pain 140 (2008) 190–208

2.4. Wind-up protocol

To develop wind-up, sequences of 15–65 stimuli atthe receptive field and/or the PQ were routinely deliv-ered at 1 Hz (see for example, A and B in Fig. 2). Eachstimulus was a single shock of 0.05–1 ms duration. Thedifferent sequences were applied at P5 min intervals toavoid cumulative effects. Post-stimulus histograms werebuilt and since no evoked responses were observed atlatencies compatible with Ab-fiber activation, the Ad-like and C-like fiber-evoked responses to PQ stimulationwere taken as the shortest latency action potentialswithin a time frame of 0.76–3.8 and 7.6 to P46 ms,respectively. This corresponds to conduction velocitiesof 60.5–3 m/s for the C-like volley and of 6–30 m/sfor the Ad-like volley, considering that the activatedafferents in the PQ made monosynaptic connectionswith the responding SRD neurons (mean average dis-tance of 23 mm from the stimulating site in the cordto the SRD). Evidence for first-latency monosynapticinfluence came from frequency-following data: first-latency Ad-like and C-like responses followed up to50–60 and 40–50 Hz repetitive stimulation, respectively,with modest jitter (<0.1 ms for Ad-like responses and<1 ms for C responses). The ranges in conduction veloc-ities were chosen according to the latencies observed fol-lowing two-point electrical stimulation at the tail. Theremaining neuronal response was taken as the neuronalpost-discharge. The determination of Ad-like and/or C-like responses was based on the correlation between pre-dicted and actual first-latency response latencies.Graphs were plotted using Ad-like and C-like fiber com-ponents of the response. Responses were quantified asspike counts. Once wind-up was established, stimulationwas continued at 1 Hz up to 400 stimuli in a number ofcases to evaluate the effects of AP5 and/or riluzolemicroiontophoretic ejection on wind-up maintenance(see Section 3.6).

In three animals, the spinal cord was fully sectionedat C2 and the cerebral cortex aspirated bilaterally fullthickness to ascertain whether wind-up was still presentwhen electrically stimulating through a bipolar electrodestereotaxically placed in the ipsilateral NRGc.

2.5. Intracellular recording

Intracellular data were obtained under a-chloraloseanesthesia from 10 different animals from those usedfor extracellular recording. Sharp micropipettes filledwith 2.5 M K+-acetate (15–35 MX resistance measuredin the neural tissue) were used. Descending of pipetteswas accomplished using a motorized driver (model662, David Kopf, Tujunja, CA, USA). After passingthrough the dorsal column nuclei, where the neuronshad easily identifiable cutaneous receptive fieldsresponding to light touch and brushing the skin [16], a

neuronal silence was always present before enteringthe SRD. At this stage, the tissue was covered withwarm agar and, when solidified, the electrode was fur-ther advanced and the search for SRD cells began. Elec-trode resistance was continuously checked by observingthe changes in voltage produced by current pulses usingan Axoclamp-2B intracellular amplifier in the bridgemode (Molecular Devices, Sunnyvale, CA, USA). Thebridge was balanced and the capacitance neutralizationadjusted to give the fastest step response to a currentpulse.

Electrical stimulation was applied to the PQ, whilethe electrode was advanced in 1- to 2-lm steps. A Dig-idata 1322A (Molecular Devices) interfaced to a PC dig-itized the signal online at 20–30 kHz. pClamp 9 software(Molecular Devices) was used for data acquisition andanalysis. The signals were also digitized at 20 kHzthrough a four-channel digitizing unit (CDAT4, CygnusTechnology, Delaware Water Gap, PA, USA) andstored unfiltered on videotape for off-line analysis.

2.6. Histology

At the end of each experiment, positive current(20 lA for 20 s) was passed through the stimulating elec-trodes to mark their tip positions by electrolytic lesions.The animals were perfused with normal saline followedby 4% paraformaldehyde. The neural tissues of interestwere removed and post-fixed. Transverse 50-lm frozensections were cut, serially mounted, stained with cresylviolet or neutral red, and the locations of recordingand stimulating sites determined.

2.7. Immunohistochemistry

In order to locate the SRD neurons of interest, inpilot experiments (3 animals), single-barrel electrodesfilled with 1.5% neurobiotin dissolved in a solution con-taining 0.05 M Tris buffer and 0.5 M NaCl (pH 7.4)were used to mark the sites ventral to the dorsal columnnuclei, where extracellularly recorded neurons with seg-mental noxious receptive fields were encountered (seeFig. 1A). The filled pipettes had resistances of 4–8 MX. After a cell was identified that responded to nox-ious stimuli applied to the body, except for the head,neurobiotin was ejected, in a region where antidromicresponses were also observed, by applying 200 nA ofpositive current pulses (200 ms on, 200 ms off) for60 min. After neurobiotin ejection, animals survivedfor 2–4 h and were then perfused through the ascendingaorta with 10 mM PBS (phosphate buffered saline; pH7.4) and heparin, followed by a fixative solution contain-ing 4% paraformaldehyde and 0.5% glutaraldehyde in10 mM PBS. The tissue containing the SRD was post-fixed in the same fixative overnight at room temperatureand then sectioned with a cryostat in the coronal plane

Fig. 2. The SRD cells responded to afferent repetitive stimulation generating wind-up followed by post-discharges even after sectioning the spinalcord. (A) A neuron showing wind-up to PQ repetitive electrical stimulation. (B) A different cell showing wind-up to repetitive CFP electricalstimulation (arrows point to the stimulus artifacts). The numbers at the left of the single sweeps in (A) and (B) indicate the stimulus number. (C)Repetitive electrical stimulation of the ipsilateral nucleus reticularis gigantocellularis (NRGc) still generated wind-up after cerebral cortex fullthickness removal and section of the spinal cord. Three different cells are shown. (D) The SRD neurons generated post-discharge activity after low(top) and high (bottom) frequency afferent stimulation. Stimulus artifacts marked by asterisks (upper) and a horizontal line (bottom). (E) Twodifferent SRD neurons (1 and 2) spontaneously active discharged rhythmic groups of spikes at low-frequencies as shown by the autocorrelograms atright.

C. Soto et al. / Pain 140 (2008) 190–208 195

Table 1Cervical (C2) posterior quadrant (PQ) stimulation

Cells C-volley Ad-volley Ad + C-volleys

SPr 74 66 3 5nSPr 76 48 14 14Total 150 114 17 19

Seventy PQ non-responsive cells were also encountered.

Table 2PQ and cervical (C3) anterolateral quadrant (AQ) stimulation

Cells C-volley Ad-volley Ad + C-volleys

PQ AQ PQ AQ PQ AQ PQ AQ

SPr 36 – 30 – 1 – 3 –nSPr 45 81 20 5 7 54 8 12Total 81 81 50 5 8 54 11 12

Eleven non-responsive cells, spontaneously active, were also found.

196 C. Soto et al. / Pain 140 (2008) 190–208

at a thickness of 80–100 lm. Sections were processedusing the procedure described by Huang et al. [29] forin vivo studies.

3. Results

3.1. Selection of recording and location of sampled

neurons

During pilot experiments, we tested diverse somato-sensory stimuli to ascertain whether the SRD neuronsresponded to different sensory stimuli. None of the59 U tested responded to proprioceptive (passivemovement of joints and muscle palpation) or cutane-ous (air puffs and gentle touching and brushing theskin) stimulation. Pinching and squeezing the skin allover the body drove most of the tested units (51/59)and, thereafter, bipolar stimulating electrodes wereroutinely placed in all four central pads to drive theSRD cells (Fig. 1D) and to have an indication of theirreceptive field size, in accordance with their responsesor lack of responses to stimulation of all four, or less,pads (see Section 3.4). Accordingly, the following cri-teria were further used for sampling SRD neurons: (i)location of neurons ventral to the dorsal columnnuclei (Fig. 1A and C), (ii) presence of cells respond-ing antidromically to PQ stimulation (a collision testis shown in B of Fig. 1), and (iii) wide noxious recep-tive fields, exemplified by the data shown in Fig. 1Dillustrating the responses of a SRD neuron to supra-maximal electrical stimulation applied to the centralfour pads. SRD cells responding to both non-noxiousand noxious stimuli were not encountered within theSRD (see below).

Both classes of SRD neurons, SPr and nSPr,appeared intermingled throughout the sampled region(from the obex to 4 mm posterior to it), peaked innumber at 2 mm lateral with respect to the midline,2 mm posterior to the obex and at a depth of 2000–4500 lm from the surface (Fig. 1C). Most of the cells(71% SPr; 81% nSPr) were encountered at depthsfrom 3000 to 4500 lm from the surface. Goingmedio-laterally, the nociceptive neurons switched theirreceptive fields from segmental (1–3 mm lateral to themidline) to segmental–trigeminal (3–3.5 mm lateral tothe midline: receptive fields on the body and the face)to receptive fields in the face only (more than 3.5 mmlateral to the midline). The transition region wasexcluded from this study, and thus the lateral bound-ary of the SRD was arbitrarily taken as 3 mm lateralfrom the midline.

3.2. Data set

A total of 312 well-isolated neuronal units were testedto PQ stimulation and 102 of them were also tested to

AQ stimulation. A total of 231/312 (�74%) of the testedneurons responded to PQ and 110 (�48%) of them sentan axon to the spinal cord (responded antidromically toPQ stimulation) (Tables 1 and 2). A total of 81/102(�79%) of the tested cells to PQ and AQ stimulationresponded to both stimuli (Table 2) with 36 (44.4%) ofthe responsive neurons sending a descending axonthrough the PQ (probably via the DLF).

3.3. Background activity

A total of 153 of 231 (66.2%) cells responding to PQstimulation were silent (80% SPr cells; 60.3% nSPr cells).The spontaneously active neurons showed 0.5–5 Hzrhythmic activity (18/78 cells; 12 SPr), 8–20 Hz rhythmicactivity (10/78 cells; 5 SPr), random activity (16 of 78cells; 8 SPr), or grouped spikes at 6300 Hz during 1–10 s separated by 1.5–20 s of silenced activity (34/78 or43.6% of cells; all nSPr neurons; see Fig. 2E).

A total of 31 of 91 tested cells that did not respond toPQ stimulation were silent and driven by suprathresholdintracutaneous stimuli. A total of 11 of 81 cells that didnot respond to PQ or to AQ stimulation were spontane-ously active and unresponsive to peripheral cutaneous,proprioceptive or noxious stimuli. Therefore, the silentcells not responding to PQ or to AQ stimulation, ifany, could not be detected by the procedures followedin this study.

The intracellular recordings (all under chloraloseanesthesia) permitted us to ascertain subthreshold activ-ity showing that 8/25 of nSPr cells presented slow(<1 Hz) rhythmic subthreshold activity, and that otherdifferent 6/25 nSPr neurons had coupled slow (about0.5 Hz) and delta (1–4 Hz) rhythmicities. On the con-trary, all SPr cells presenting slow oscillatory activity(10/23) had subthreshold oscillations within the deltarange.

C. Soto et al. / Pain 140 (2008) 190–208 197

In no case the SRD oscillatory activity could be cor-related with the electrocorticographic activity recordedfrom the lateral tip of the contralateral cruciate sulcus.

3.4. Responses to intracutaneous electrical stimulation

A total of 184 SRD cells were tested to stimulation ofall four pads; 55 of them were also tested to tail stimu-lation. These neurons were grouped in two classes, oneconsisting of 80 cells (43.5%) responding to all fourpads, and another comprising 104 cells responding to1–3 pads. A total of 25 cells from the first class were alsotested to tail stimulation, responding all of them. On thecontrary, only 12 of 30 tested cells (40%) pertaining tothe second class responded to tail stimulation. The cellsresponding to all four pads received a greater percentageof ‘‘Ad + C” (40/80:50%) as well as ‘‘only-C” (20/80:25%) inputs than the neurons with smaller receptivefields (Ad + C = 22/104:21%; C = 15/104:14.4%), whichreceived mostly ‘‘Ad-only” input (69/104:66.3%). Also,the conduction velocity of the inputs to the cellsresponding to all four pads was within the slower range(C = 0.4–1 m/s; Ad = 6–12 m/s) for both classes of nox-ious fibers. There was then a clear relationship betweenreceptive field size, convergence of inputs and afferentconduction velocity: the slower noxious ‘‘Ad” and ‘‘C”

inputs converged on SRD neurons with the larger recep-tive fields. Although Ad and C-fibers terminate in thespinal cord, we refer to Ad and C-fiber afferents to theSRD when the latency of the responses to stimulationis compatible with Ad-and-C-fiber conduction velocities,respectively. Since we did not systematically map thereceptive field size all over the body, a comparison withthe rat’s data of Villanueva et al. [63] cannot be made. Ingeneral, the first latency from the contralateral pads wassmaller than that from the ipsilateral ones. The averageconduction speeds for first-latency inputs from the fourpads, measured in cells responding to stimulation of allfour pads, were 22.4 ± 0.9 and 20.5 ± 1.02 m/s for thecontralateral and the ipsilateral forepads, respectively(exact P value = 0.006; paired Student’s t test); and26.09 ± 0.9 and 24.7 ± 1.2 m/s for the contralateraland the ipsilateral hindpads, respectively (no significantdifference). The later response latencies were much morevariable within and between cells than the first-latencyresponses. The first latencies to both forepads were notsignificantly different as were not to both hindpads.

The great majority of the convergent neurons affectedby inputs from the pads (34/40; 85%) showed an earlypeak and two late peaks. The early peak and the firstlate peak did not show wind-up while in most occasions(27/34; 79.4%) the second late peak did (Fig. 2B). On thecontrary, all SRD neurons receiving late-only inputshowed a unique peak and also generally responded tosuprathreshold intracutaneous repetitive stimulationgenerating wind-up (28/35:80%) (Fig. 2A).

3.5. Responses to cervical spinal cord stimulation

Electrical stimulation of the PQ white matter acti-vated 150/220 SRD cells that also responded to intracu-taneous stimulation within the noxious range. The greatmajority of these PQ-responsive cells showed latenciescompatible with ‘‘C” fiber activation (114/150:76%),whereas only 17/150 (11.3%) responded to ‘‘Ad-only”

input and 19/150 (12.6%) were activated by both ‘‘Ad”

and ‘‘C” inputs (Table 1). The fact that PQ stimulationproduced mostly ‘‘C” responses despite both ‘‘Ad” and‘‘C” responses being evoked by intracutaneous stimula-tion (Fig. 2A and B), together with the fact that 70 of220 tested cells (31.8%) did not respond to PQ stimula-tion, led us to test whether Ad information affecting theSRD ascends mostly through the AQ (probably in theventrolateral spinothalamic tract), and thus a greaterpercentage of ‘‘Ad” responses would be observed whenstimulating the AQ. It is clear from Table 2 that thisassumption appears to be correct since most of theeffects evoked by AQ stimulation were ‘‘Ad-only”

responses (54/81:66.6%) which summed with 12/81 cellsthat showed ‘‘Ad + C” responses giving a total of 66/81:81.5% of the SRD cells receiving ‘‘Ad” informationvia the AQ. The data of Table 2 also show that noSRD antidromic responses were evoked from the AQ.

The intracellular recordings demonstrated that theeffect of PQ stimulation (the AQ was not stimulated inthese experiments) on SRD cells was not only excitatorybut also inhibitory, excitatory–inhibitory or inhibitory–excitatory. The examples in Fig. 3 are from six differentcells and show that the Ad fiber-induced excitation wasusually followed by inhibition (Fig. 3A–C) that in turncould be followed (Fig. 3C) or not by a post-inhibitoryexcitation. The C-like excitation could be preceded(Fig. 3D) or not (Fig. 3E) by Ad-like excitation. In nocase the C-like-induced excitation, when present alone,was followed by inhibition. In two instances, Ad-likeand C-like excitatory responses appeared in the absenceof excitatory or inhibitory postsynaptic potentials evenat different membrane potentials (Fig. 3D). Inhibitionwas the shortest latency response to PQ stimulation infive cells (Fig. 3F) in which excitatory responsesappeared during the repolarizing phase and that couldbe due to C-like synaptic activation and/or to post-inhibitory rebounds (see in Fig. 3F the short-amplituderesponses resembling low threshold calcium spikes butthat could also reflect excitatory postsynaptic poten-tials). All these five SRD cells (2 SPr) responded toPQ stimulation generating a fast hyperpolarization(3 ± 0.15 ms latency) followed by a slower one(5 ± 0.2 ms latency) (see upper inset in Fig. 3F), indicat-ing that two different postsynaptic inhibitory mecha-nisms may be present. It appears clearly from thesedata that the synaptic behavior of the SRD cells is farfrom being simple, and that PQ stimulation can produce

Fig. 3. Examples of superimposed intracellular responses to PQ stimulation. The lower record in (F) is expanded at the upper inset to clearlydistinguish a fast and a slow hyperpolarization. The PQ stimuli are marked by asterisks (see text for details).

198 C. Soto et al. / Pain 140 (2008) 190–208

a collection of distinct responses that were present inboth SPr and nSPr cells thus pointing to a complex cir-cuitry network.

3.6. Wind-up

A total of 102/164 (62.2%) SRD cells (all under a-chloralose anesthesia) showed temporal summation torepetitive electrical stimulation of at least one of the fol-lowing sites: PQ, the pads or the NRGc. To ascertainwhether the wind-up phenomenon is established in thedorsal spinal cord and transmitted to the SRD, NRGcstimulation was applied after spinal cord section;wind-up was still evoked in the SRD. Furthermore,the establishment of wind-up in the SRD was indepen-dent of an intact cerebral cortex since NRGc stimulationstill produced wind-up after bilateral full thicknessremoval of the cerebral cortex (Fig. 2C).

Because the preceding data suggest that wind-upmight be generated within the SRD, the following stepwas aimed to ascertain the local mechanisms causingthe phenomenon. Since NMDA ligand receptors playa role in wind-up generation in the rat’s dorsal spinal

cord [20,22,28], we first tested whether microiontopho-retic ejection of the NMDA antagonist AP5 modifiedwind-up. The results were negative in all 26 cells tested(10 SPr). Secondly, we reasoned that since the neuronalmembrane needs to be depolarized to relieve the Mg2+

block of the NMDA channel, AP5 might be more effec-tive when ejected once wind-up was established and thuswould decrease the number of spikes/stimulus duringwind-up maintenance. The obtained results were incon-sistent since the previous assumption occurred only in 6of 25 cells (24%) tested (Fig. 4A and B), and ejection ofAP5 failed to modify the number of spikes/stimulus dur-ing wind-up persistence in other 14 (56% of tested) dif-ferent neurons (Fig. 4C). Further, AP5 ejectionincreased the number of spikes/stimulus in a differentset of five (20%) cells (see below).

The following step consisted in testing whether gluta-mate ejection at 0.3–1 Hz produced wind-up. Wind-upwas generated by ejecting glutamate with current pulsesof 15–30 nA in all tested cells (n = 36, 20 SPr) (Fig. 5A1and D; F2,87 = 185.9, P < 0.001); higher currents pro-duced an elevated number of spikes from the second-third pulse with small, non-significant, variations all

Fig. 4. AP5 treatment during wind-up produced different results. (A) Once wind-up was attained (control), ejection of AP5 (middle; 20 nA) clearlydecreased the number of spikes/stimulus with respect to control and recovery (15 min after ejection of AP5 was discontinued). (B) AP5 significantlydecreased the mean number of spikes/stimulus during wind-up with respect to control (*P value < 0.002; paired Student’s t test; n = 6 cells from fourcats). (C) Upper: 20 consecutive PQ stimuli (numbers 5 and 19 expanded at right) fired two different neurons (the smaller cell is signaled byarrowheads at right) with two latency peaks each. Bottom: the second latency peak showed wind-up in both cells, but AP5 ejection at 60 nA duringwind-up was ineffective.

C. Soto et al. / Pain 140 (2008) 190–208 199

along the ejecting period (Fig. 5A2). Gradually increas-ing uninterrupted glutamate ejection at 30 nA and above

generated grouped spikes separated by silenced activity(Fig. 5B and C); these glutamate-induced grouped

Fig. 5. Low-current ejection of glutamate-induced wind-up. (A) Two different silent SDR spinally projecting (SPr) cells. Iontophoretic pulses ofglutamate at �15 nA-induced wind-up (A1), whereas at �40 nA pulses (A2) did not. (B) Gradually increasing continuous glutamate ejection up to�30 nA generated grouped spikes and post-discharges in a silent SPr neuron. (C) The activity in grouped spikes separated by silences induced bycontinuous ejection of glutamate at �50 nA in a nSPr cell was abolished by simultaneous ejection of CNQX at �50 nA. (D) Glutamate generatedwind-up when ejected in the range of �15 to �30 nA. Mean values + SEM (n = 8 cells from four cats) are represented. A fourth order polynomialbest fitted the data. (E) After glutamate pulse-ejection-induced wind-up (E1), AP5 was concurrently and continuously ejected with graduallyincreasing currents from �15 nA (E2) to �100 nA (E4). AP5 increased the number of spikes per glutamate pulse, and a reset was produced at the endof AP5 ejection (E4). Pulses of AP5 alone did fire the neuron without inducing wind-up (E5). (F) Graph showing the mean + SEM (n = 5 cells fromthree cats) number of spikes per each AP5 pulse at �100 nA.

200 C. Soto et al. / Pain 140 (2008) 190–208

C. Soto et al. / Pain 140 (2008) 190–208 201

spikes were generated at a low-frequency (<1 Hz) andwere abolished by concurrent ejection of CNQX(Fig. 5C).

We noticed in five silent cells at rest (3 SPr) that oncewind-up was established by glutamate ejection, uninter-rupted concurrent ejection of AP5 at 15, 50 and 100 nA(Fig. 5E2–E4) further increased the number of spikes/stimulus with relation to glutamate-alone ejection.Simultaneous ejection of glutamate and the solutionused to dissolve AP5 at 15, 50 and 100 nA currentsdid not vary the glutamate responses (not shown), whichdemonstrates that the effect was in fact due to AP5application. On ceasing the concurrent AP5 ejection,glutamate-alone ejection not only produced lesserspikes/stimulus, but also led to a resetting of the gluta-mate response (Fig. 5E4). Finally, ejecting pulses ofAP5 alone at 0.3–1 Hz did fire these five silent cells(Fig. 5E5), but did not generate wind-up (Fig. 5F).

These data discard a role for NMDA ligand channelsin the generation of SRD wind-up in general, and areinconclusive about whether they have a role in wind-up maintenance.

The intracellular recordings gave us a new cue toapproach the mechanisms underlying wind-up at theSRD. We observed that both SPr and nSPr well-impaledSRD neurons that generated wind-up upon PQ andperipheral stimulation had subthreshold oscillatoryactivity within the delta range (3.5–4 Hz) at rest, andthat repetitive C-like fiber activation gradually depolar-ized the cellular membrane so that these oscillationsreached threshold leading to wind-up. Fig. 6 showsexamples from two different silent cells at rest; note thatthe cell in Fig. 6A generated post-discharges during 38 sat slow (0.5 Hz) and delta (4 Hz) rhythmicities as illus-trated in the autocorrelogram (AUTO). These data sug-gest that subthreshold oscillatory activity may underliewind-up at the SRD. One of the candidates to cause sub-threshold oscillations is the persistent sodium currentwhich is decreased by low concentrations of riluzole[47,57,59]. Therefore, we microiontophoretically ejectedriluzole with small (<30 nA) currents to test its effects onwind-up generation and maintenance. The obtainedresults show that riluzole delayed wind-up generation(Fig. 7A–C) and also reduced the number of spikes/stimulus during wind-up (Fig. 7A, C and D). The effectof riluzole was specific over the C-fiber-like laterresponses showing wind-up, leaving the first late peakunaffected, when present, as well as the Ad-likeresponse, if present, as it is illustrated in Fig. 7A, wherethe first-latency response remained unaffected during ril-uzole ejection. Overall, riluzole delayed wind-up genera-tion by 15 ± 3 s (n = 12), and reduced the number ofspikes/stimulus during wind-up by 6 ± 2.1 spikes(n = 12), with respect to control. Finally, some SRDcells showed regular wind-up resets during long-lastingstimulation (Fig. 7D); riluzole did not influence such

resetting although it decreased the number of spikes/stimulus between resets and delayed the subsequentwind-up.

These data strongly suggest that persistent sodiumcurrents partially underlie wind-up generation withinthe SRD.

3.7. GABA and glycine

PQ stimulation induced inhibitory post-synapticpotentials in some SRD neurons (Fig. 3) suggesting thatinhibition may contribute to shape the nuclear activity.Iontophoretically applied GABA or glycine abolishedspontaneous, when present, and stimulus-evoked activ-ity in all cells tested (n = 26; 12 SPr; Fig. 8A and B).The potency of glycine was higher than that of GABA;in many instances abolition of the retention current forglycine was enough to suppress the spontaneous and/orevoked activity of SRD cells. The effects of GABA andglycine were reversed by concurrent application of BiCuand strychnine, respectively (Fig. 8A and B), indicatingthe specificity of both neurotransmitters. In addition,iontophoresis of BiCu alone increased not only the num-ber of spikes/stimulus by 35 ± 8.6% with respect to con-trol (Fig. 8C2), but also the spontaneous, when present,and post-discharge (Fig. 8D4) activity by 42 ± 8.2% rel-ative to control in all cells tested (n = 12; 4 SPr). Theseeffects of BiCu were reversed by simultaneously ionto-phoresing GABA (Fig. 8C3). Iontophoresis of strych-nine alone did not significantly modify spontaneous,post-discharge or stimulus-evoked activity (Fig. 8D).These data suggest a tonic GABAa-induced inhibitionand point to the presence of GABAa and strychnine-sensitive glycine receptors at the feline’s SRD.

4. Discussion

This study shows that (1) the SRD cells of cats havesimilar location and properties as those of rats [63] andmonkeys [64], receive solely noxious information andhave large segmental receptive fields; (2) the afferentAd-like-input reaches the SRD through the AQ, proba-bly via the spinothalamic tract, whereas C-like-inputruns in the PQ, probably in the DLF; (3) the SRDdescending axons travel in the PQ, probably in theDLF; (4) wind-up was generated by repetitive afferentactivation that gradually depolarized the SRD neuronsleading subthreshold delta oscillations to threshold; (5)riluzole delayed wind-up which (6) was still generatedby electrically stimulating the ipsilateral NRGc afterspinal cord section and cerebral cortical extirpation;and (7) wind-up was also generated by pulses of ionto-phoresed glutamate.

Two classes of SRD cells were found in rats anesthe-tized with halothane and a mixture of N2O/O2, oneresponding to innocuous and noxious stimuli and

Fig. 6. PQ stimulation produced wind-up by gradual membrane depolarization. (A) A silent SRD cell showing subthreshold oscillations within thedelta range that was gradually depolarized upon 1 Hz PQ stimulation (asterisks) that generated wind-up. Resting membrane potential is indicated bya dashed line. The portion of recording marked with one horizontal line is expanded below (A1). The post-discharge of this cell (A2) presented tworhythms at 0.5 and 4 Hz as shown in the autocorrelogram (AUTO). (B) Single sweeps showing the responses to different PQ stimulus numbers(marked at right) of a different silent SRD neuron. The subthreshold oscillations progressively reached threshold and generate wind-up. Note that theshorter-latency response did not show wind-up.

202 C. Soto et al. / Pain 140 (2008) 190–208

Fig. 7. Wind-up partially depended on a riluzole-sensitive mechanism. (A and B) Ejection of riluzole for 5 min with 20 nA (A) and 15 nA (B) currentsretarded wind-up. (C) Mean number of spikes/stimulus in control and during riluzole ejection (left, n = 8 cells from four cats). Riluzole ejectionsignificantly decreased the mean number of spikes/stimulus (right; *P < 0.0001, paired Student’s t test; n = 16 cells from seven cats). (D) Long-lastingPQ stimulation produced regular wind-up resetting. Riluzole treatment did not modify the resetting (left, n = 6 cells from three cats) but significantlydecreased the mean number of spikes/stimulus during wind-up (right; *P < 0.001, paired Student’s t test; n = 14 cells from seven cats).

C. Soto et al. / Pain 140 (2008) 190–208 203

another responding exclusively to noxious stimuli [63].We did not observe SRD neurons responding to innoc-uous mechanical stimuli. This discrepancy is probablynot linked to the anesthetics but may represent a speciesdifference since the SRD neurons of monkeys anesthe-tized with a-chloralose plus pentobarbital also showednoxious and innocuous convergence [64].

4.1. Ascending Ad and C pathways and descending SRD

output

The great majority of PQ axons directed to the SRDwere slow conducting and their activation generated

wind-up in SRD cells; that is, they had C-like properties.On the contrary, the AQ-induced response latencieswere compatible with Ad-fiber-like speeds. In felines,noxious information is transmitted supraspinallythrough the DLF and the AQ [37], paths that are usedby C and Ad information, respectively, to render theSRD neurons nociceptive-specific (NS) and thus beingprobably inappropriate to play a discriminatory role.The majority of spinal NS neurons are located in dorsalhorn lamina I, receive Ad and C inputs [67] and projectrostrally mostly in the DLF [4,5,30], although noxiousDLF fibers do not appear to constitute a true dorsalspinothalamic tract since they terminate primarily in

Fig. 8. GABA and glycine ejection abolished the PQ-induced responses. (A and B) GABA (A2) or glycine (B2) treatment annulled the PQ-evokedresponses (control; A1 and B1), effect that was reversed by simultaneously ejecting BiCu (A3) or strychnine (B3), respectively. (C and D) The numberof spikes/stimulus to PQ stimulation was increased after BiCu injection (C2) but not after strychnine ejection (D2). BiCu treatment also increased thepost-discharge activity (D4).

204 C. Soto et al. / Pain 140 (2008) 190–208

the brainstem [33]. Fibers ascending in the dorsal funic-ulus were described in the rat [2] but not in felines. Other

fibers ascending in the PQ such as post-synaptic axonsthat run in the dorsal columns as well as spinocervical

C. Soto et al. / Pain 140 (2008) 190–208 205

and spinomesencephalic tracts may also send afferents tothe cat’s SRD, although they do not appear to influencethe rat’s SRD [12].

Descending axons from the SRD run probably in theDLF since SPr neurons responded antidromically to PQbut not to AQ stimulation. Recent studies indicate thatpersistent pain is linked to an enhanced activation ofdescending modulatory networks that in normal condi-tions produce a net spinal inhibition. Descending path-ways, however, can inhibit or excite spinal dorsal hornneurons [8,43]. A spino-SRD-spinal loop [36] is subsidi-ary to other networks [32,53] contributing to regulatenoxious transmission in the dorsal horn. Descendingfacilitation can overcome inhibition contributing to sus-tain persistent pain in processes associated with second-ary hyperalgesia or nerve injury [43]. It has been recentlyshown that On-and-Off cells of the rostro-ventro-medialmedulla are sensitized after peripheral nerve injuryresponding to non-noxious cutaneous stimuli after, butnot before, the injury [17]. Whether SRD cells are alsosensitized after injury/inflammation is an open question,but it is very probable since persistent pain usuallyresults in peripheral and/or dorsal horn sensitizationleading to tonic ascendant activity affecting the SRD.In fact, reciprocal connections between the SRD andlamina I dorsal horn neurons are present in the rat [1–3], and lamina I cells expressing the neurokinin-1 recep-tor for substance P are necessary for the development ofhyperalgesia and dorsal horn neuronal sensitization [31].Further, chronic pain states appear to be establishedand maintained by lamina I neurons through supraspi-nal structures that, in turn, modulate dorsal horn cellu-lar excitability via descending pathways [14,53,56]. TheSRD would play a pronociceptive role [36] through alamina I–SRD–lamina I reverberating circuitry helpingto maintain chronic pain in pathological conditions.

In addition to a pronociceptive function, the SRDalso exerts inhibitory [13] and excitatory [23] influenceson deep dorsal horn multimodal neurons respondingto noxious and non-noxious stimuli. Superficial anddeep dorsal horn multimodal neurons as well as NS cellsof various spinal segments are inhibited by noxiousstimuli applied outside their receptive fields (‘‘diffusenoxious inhibitory controls or (DNICs)” [34]). DNICsaffect wind-up in spinal multimodal cells [50], are trig-gered by activation of Ad and C-fibers and are mediatedthrough the SRD [13,60]. In the rat, the ascending path-way triggering DNIC runs in the AQ, whereas thedescending route travels in the DLF [60]; if this werealso true in cats, DNICs would be triggered by Ad fibersin this species.

4.2. GABA and glycine

Upon PQ stimulation, inhibitory synaptic potentialswere recorded (Fig. 3), indicating that inhibition may

shape SRD neuronal activity. Further, ejection ofGABA or glycine canceled the activity of all tested cells(Fig. 8A and B). Whether these inhibitory neurotrans-mitters are released from intranuclear or extranuclearsources is unknown as are their functional roles. Giventhe large receptive fields of SRD cells, inhibition doesnot probably serve to shape the receptive fields butwould intervene in pronociceptive and DNICs mecha-nisms by modulating the activity of SPr cells.

Bicuculline, but not strychnine, ejection increased thespontaneous and post-discharge activity (Fig. 8C andD), indicating the presence of a tonic GABAa-depen-dent inhibition at the SRD. Tonic inhibition regulatesneuronal excitability in a variety of central neurons con-trolling network activity by GABAergic mechanisms(for review see [24]). A tonic GABAa-mediated conduc-tance can enhance burst firing by increasing membranehyperpolarization [18]. GABA will produce a shuntinginhibition decreasing the voltage change induced by agiven stimulus and thus decreasing SRD excitability.Injury and/or inflammation will produce tonic afferentC-fiber activation surpassing tonic inhibition and gener-ating wind-up.

4.3. Wind-up

In the spinal cord, wind-up relies not only on synapticactivity involving NMDA and neurokinin receptors[6,28], but also on intrinsic post-synaptic properties[25,40,49]. Our intracellular data show that PQ stimula-tion depolarized SRD cells leading subthreshold oscilla-tions to progressively reach threshold and producingwind-up (Fig. 6). Whether this is auxiliary to otheramplifying synaptic and/or intrinsic mechanisms is anopen issue. In situations in which afferent C-fibers aretonically active, the persistent temporal summation(wind-up) in SPr neurons will contribute to maintainabnormal discharge in dorsal horn cells.

AP5 ejection did not modify wind-up generation and,during wind-up, decreased the activity of some cells (6/25) and increased the activity of others (5/25) suggestingthat (i) the membrane depolarization necessary to gener-ate wind-up was insufficient to open NMDA channels inthe AP5-unaffected cells, (ii) once depolarizationrelieved NMDA channels from their Mg2+ blockade,AP5 decreased stimulus-evoked activity during wind-up (6/25 cells), and (iii) NMDA may exert an inhibitoryrole at presynaptic level [7], and blockade of presynapticNMDA channels by AP5 increased the cellular activityduring wind-up (5/25 cells).

Low-current ejection of riluzole delayed wind-up anddecreased the number of spikes/stimulus during wind-up(Fig. 7). In vitro experiments showed that low concentra-tions of riluzole blockade or reduce the persistentsodium current (INaP) in a variety of systems[47,57,59] slightly attenuating transient sodium currents

206 C. Soto et al. / Pain 140 (2008) 190–208

[45] while leaving action potential generation largelyunaffected [19,21,42,45,47,57,59,69] (see also Fig. 7B).One role of INaP may be to amplify synaptic responsespromoting wind-up and bursting activity. Our results doindicate that a riluzole-sensitive mechanism amplifiesexcitatory synaptic drive enhancing bursting activity inthe cat’s SRD. Riluzole-insensitive mechanisms to com-plement synaptic amplification, if any, remain to bedetermined.

Most of the previous ‘‘mechanism based” painresearch has been done on rodents with the discoveredmechanisms extrapolated to humans and oftentimesused as the basis to test new pain therapies. Thisstudy showing that wind-up can be supraspinally gen-erated even in the absence of ascending nociceptivefibers and partially relying on oscillatory activity reso-nating with afferent activity, a mechanism notreported in rodents, argues in favor of pain researchin different animal species. If wind-up reveals someof the mechanisms underlying dorsal horn neuronalsensitization [68] then it could also reveal supraspinalsensitization. In humans, a symptom of chronic painafter tissue or nerve injury is the abnormal temporalsummation (wind-up) of pain that can be regardedas one of the several initiators of central enhancedresponsiveness of noxious neurons [68]. An increasedknowledge of the mechanisms underlying wind-upand its role in chronic pain is essential to develop amore effective pain therapy. Abnormal temporal sum-mation of noxious signals within the SRD, linked ornot to a loss in descending inhibition, would increasenoxious ascending transmission mostly at the lamina Idorsal horn. The development of novel therapies totreat neuropathic pain relies on experimental models,and a major challenge is to improve translationalresearch from mechanisms to clinical manifestationsin order to get better treatments.

Acknowledgements

This work and P.V. were supported by grants fromthe Spanish Ministerio de Ciencia y Tecnologıa(BFU2006-06598), and the Xunta de Galicia. F.M.C.is a postdoctoral fellow supported by the Ramon y Cajalprogram. C.S. and R.L. were supported by the SpanishFPI program. We express our gratitude to Adriana Pau-los for her help with the histology. The authors declarethat they have no conflicts of interest.

References

[1] Almeida A, Tavares I, Lima D, Coimbra A. Descending projec-tions from the medullary dorsal reticular nucleus make synapticcontacts with spinal cord lamina I cells projecting to that nucleus:

an electron microscopic tracer study in the rat. Neuroscience1993;55:1093–106.

[2] Almeida A, Tavares I, Lima D. Projection sites of superficial ordeep dorsal horn in the dorsal reticular nucleus. Neuroreport1995;6:1245–8.

[3] Almeida A, Tavares I, Lima D. Reciprocal connections betweenthe medullary dorsal reticular nucleus and the spinal dorsal hornin the rat. Eur J Pain 2000;4:373–87.

[4] Apkarian AV, Hodge Jr CJ. The primate spinothalamic pathwaysII. The cells of origin of the dorsolateral and ventral spinotha-lamic pathways. J Comp Neurol 1989;288:474–92.

[5] Apkarian AV, Stevens RT, Hodge CJ. Funicular location ofascending axons of lamina I cells in the cat spinal cord. Brain Res1985;334:160–4.

[6] Baranauskas G, Nistri A. Sensitization of pain pathways in thespinal cord: cellular mechanisms. Prog Neurobiol 1998;54:349–65.

[7] Bardoni R, Torsney C, Tong C-K, Prandini M, MacDermott AB.Presynaptic NMDA receptors modulate glutamate release fromprimary sensory neurons in rat spinal dorsal horn. J Neurosci2004;17:2774–81.

[8] Basbaum AI, Fields HL. Endogenous pain control systems:brainstem spinal pathways and endorphin circuitry. Annu RevNeurosci 1984;7:309–38.

[9] Basbaum AI, Clanton CH, Fields HL. Three bulbospinal path-ways from the rostral medulla of the cat: an autoradiographicstudy of pain modulating systems. J Comp Neurol1978;178:209–24.

[10] Berman AL. The brain stem of the cat. A cytoarchitectonic atlaswith stereotaxic coordinates. Madison: The University of Wis-consin Press; 1968.

[11] Bernard JF, Villanueva L, Carroue J, Le Bars D. Efferentprojections from the subnucleus reticularis dorsalis (SRD): aPhaseolus vulgaris leucoagglutinin study in the rat. Neurosci Lett1990;122:257–62.

[12] Bing Z, Villanueva L, Le Bars D. Ascending pathways in thespinal cord involved in the activation of subnucleus reticularisdorsalis neurons in the medulla of the rat. J Neurophysiol1990;63:424–38.

[13] Bouhassira D, Villanueva L, Bing Z, Le Bars D. Involvement ofthe subnucleus reticularis dorsalis in diffuse noxious inhibitorycontrols in the rat. Brain Res 1992;595:353–7.

[14] Budai D, Sergey G, Khasabov SG, Mantyh PW, Simone DA.NK-1 receptors modulate the excitability of ON cells in the rostralventromedial medulla. J Neurophysiol 2007;97:1388–95.

[15] Burton H. Somatic sensory properties of caudal bulbar reticularneurons in the cat (Felis domestica). Brain Res 1968;11:357–72.

[16] Canedo A, Aguilar J. Spatial and cortical influences exerted oncuneothalamic and thalamocortical neurons of the cat. Eur JNeurosci 2000;12:2515–33.

[17] Carlson JD, Maire JJ, Martenson ME, Heinricher MM. Sensiti-zation of pain-modulating neurons in the rostral ventromedialmedulla after peripheral nerve injury. J Neurosci2007;27:13222–31.

[18] Cope DW, Hughes SW, Crunelli V. GABAa receptor-mediatedtonic inhibition in thalamic neurons. J Neurosci2005;25:11553–63.

[19] Cramer NP, Li Y, Keller A. The whisking rhythm generator: anovel mammalian network for the generation of movement. JNeurophysiol 2007;97:2148–58.

[20] Davies SN, Lodge D. Evidence for involvement of N-methylas-partate receptors in ‘‘wind-up” of class 2 neurones in the dorsalhorn of the rat. Brain Res 1987;424:402–6.

[21] Del Negro CA, Morgado-Valle C, Feldman JL. Respiratoryrhythm: an emergent network property? Neuron 2002;34:821–30.

[22] Dickenson AH, Sullivan AF. Evidence for a role of the NMDAreceptors in the frequency dependent potentiation of dorsal horn

C. Soto et al. / Pain 140 (2008) 190–208 207

nociceptive neurones following C fibre stimulation. Neurophar-macology 1987;26:1235–8.

[23] Dugast C, Almeida A, Lima D. The medullary dorsalreticular nucleus enhances the responsiveness of spinal noci-ceptive neurons to peripheral stimulation in the rat. Eur JNeurosci 2003;18:580–8.

[24] Farrant M, Nusser Z. Variations on an inhibitory theme: phasicand tonic activation of GABAa receptors. Nat Rev Neurosci2005;6:215–29.

[25] Fossat P, Sibon I, Le Masson G, Landry M, Nagy F. L-typecalcium channels and NMDA receptors: a determinant duo forshort-term nociceptive plasticity. Eur J Neurosci 2007;25:127–35.

[26] Fujino Y, Koyama N, Yokota T. Differential distribution of threetypes of nociceptive neurons within the caudal bulbar reticularformation in the cat. Brain Res 1996;715:225–9.

[27] Guilbaud G, Peschanski M, Gautron M. Functional changes inventrobasal thalamic neurones responsive to noxious and non-noxious cutaneous stimuli after chloralose treatment: new evi-dence for the presence of pre-existing ‘‘silent connections” in theadult nervous system? Pain 1981;11:9–19.

[28] Herrero JF, Laird JMA, Lopez-Garcıa JA. Wind-up of spinalcord neurones and pain sensation: much ado about something?Prog Neurobiol 2000;61:169–203.

[29] Huang Q, Zhou D, DiFiglia M. Neurobiotin, a useful neuro-anatomical tracer, retrograde and transneuronal tract-tracing andfor in vitro labelling of neurons. J Neurosci Methods 1992;41:31–43.

[30] Jones MW, Hodge CJ, Apkarian AV, Stevens RTA. Dorsolateralspinothalamic pathway in the cat. Brain Res 1985;335:188–93.

[31] Khasabov SG, Rogers SD, Ghilardi JR, Peters CM, Mantyh PW,Simone DA. Spinal neurons that possess the substance P receptorare required for the development of central sensitization. JNeurosci 2002;22:9086–98.

[32] Khasabov SG, Ghilardi JR, Mantyh PW, Donald A, Simone DA.Spinal neurons that express NK-1 receptors modulate descendingcontrols that project through the dorsolateral funiculus. J Neu-rophysiol 2005;93:998–1006.

[33] Klop EM, Mouton LJ, Hulsebosch R, Boers J, Holstege G. In catfour times as many lamina I neurons project to the parabrachialnuclei and twice as many to the periaqueductal gray as to thethalamus. Neuroscience 2005;134:189–97.

[34] Le Bars D, Dickenson AH, Besson JM. Diffuse noxious inhibitorycontrols (DNIC). Part I: Effects on dorsal horn convergentneurons in the rat. Pain 1979;6:283–304.

[35] Lima D. A spinomedullary projection terminating in the dorsalreticular nucleus of the rat. Neuroscience 1979;34:577–89.

[36] Lima D. A spinomedullary projection terminating in the dorsalreticular nucleus of the rat. Neuroscience 1990;34:577–89.

[37] Martin RJ, Apkarian AV, Hodge Jr CJ. Ventrolateral anddorsolateral ascending spinal cord pathway influence on thalamicnociception in cat. J Neurophysiol 1990;64:1400–12.

[38] Meesen H, Olszewski J. A cytoarchitectonic atlas of the rhomb-encephalon of the rabbit. Basel: Karger; 1949.

[39] Mendell LM, Wall PD. Responses of single dorsal cells toperipheral cutaneous unmyelinated fibers. Nature 1965;206:97–9.

[40] Morisset V, Nagy F. Plateau potential-dependent wind-up of theresponse to primary afferent stimuli in rat dorsal horn neurons.Eur J Neurosci 2000;12:3087–95.

[41] Newman DB. Distinguishing rat brainstem reticulospinal nucleiby their neuronal morphology. I: Medullary nuclei. J Hirnforsch1985;26:187–226.

[42] Pace RW, Mackay DD, Feldman JL, Del Negro CA. Role ofpersistent sodium current in mouse preBotzinger complexneurons and respiratory rhythm generation. J Physiol2007;580:485–96.

[43] Porreca F, Ossipov MH, Gebhart GF. Chronic pain and medullardescending facilitation. Trends Neurosci 2002;25:319–25.

[44] Price DD, Hull CD, Buchwald NA. Intracellular responses ofdorsal horn cells to cutaneous and sural nerve A and C fiberstimuli. Exp Neurol 1971;33:291–309.

[45] Ptak K, Zummo GG, Alheid GF, Tkatch T, Surmeier DJ,McCrimmon DR. Sodium currents in medullary neurons isolatedfrom the pre-Botzinger complex region. J Neurosci2005;25:5159–70.

[46] Raboisson P, Dallel R, Bernard JF, Le Bars D, Villanueva L.Organization of efferent projections from the spinal cervicalenlargement to the medullary subnucleus reticularis dorsalis andthe adjacent cuneate nucleus: a PHA-L study in the rat. J CompNeurol 1996;367:503–17.

[47] Reboreda A, Sanchez E, Romero M, Lamas JA. Intrinsicspontaneous activity and subthreshold oscillations in neurons ofthe rat dorsal column nuclei in culture. J Physiol2003;551:191–205.

[48] Reinoso-Suarez F. Topographischer Hirnatlas der Katze furexperimental-physiologische Untersuchungen. Darmstadt: Her-ausgegeben von E. Merck AG; 1961.

[49] Russo RE, Hounsgaard RJ. Short-term plasticity in turtle dorsalhorn neurons mediated by L-type Ca2+ channels. Neuroscience1994;61:191–7.

[50] Schouenborg J, Dickenson AH. The effects of a distant noxiousstimulation on A and C fibre evoked flexion reflexes and neuronalactivity in the dorsal horn of the rat. Brain Res 1985;328:23–32.

[51] Schouenborg J, Sjolund BH. Activity evoked by A- and C-afferentfibers in the rat dorsal horn neurons and its relation to a flexionreflex. J Neurophysiol 1983;50:1108–21.

[52] Snider RS, Niemer WT. A stereotaxic atlas of the cat brain. Chi-cago: The University of Chicago Press; 1961.

[53] Suzuki R, Morcuende S, Webber M, Hunt S, Dickenson A.Superficial NK1 expressing neurons control spinal excitability byactivation of descending pathways. Nat Neurosci 2002;5:1319–26.

[54] Tavares I, Lima D. Descending projections from the caudalmedulla oblongata to the superficial or deep dorsal horn of the ratspinal cord. Exp Brain Res 1994;99:455–63.

[55] Thompson SWN, Dray A, Urban L. Injury-induced plasticity ofspinal reflex activity: NK1 neurokinin receptor activation andenhanced A- and C-fiber mediated responses in the rat spinal cordin vitro. J Neurosci 1994;14:3672–87.

[56] Todd AJ, McGill MM, Shehab SA. Neurokinin 1 receptorexpression by neurons in lamina I, III and IV of the rat spinaldorsal horn that project to the brain stem. Eur J Neurosci2000;12:689–700.

[57] Urbani A, Belluzzi O. Riluzole inhibits the persistent sodiumcurrent in mammalian CNS neurons. Eur J Neurosci2000;12:3567–74.

[58] Valverde F. Reticular formation of the pons and medullaoblongata. A Golgi study. J Comp Neurol 1961;116:71–99.

[59] Van Drongelen W, Koch H, Elsen FP, Lee HC, Mrejeru A, DorenE, et al. Role of persistent sodium current in bursting activity ofmouse neocortical networks in vitro. J Neurophysiol2006;96:2564–77.

[60] Villanueva L, Le Bars D. The activation of bulbo-spinal controlsby peripheral nociceptive inputs: diffuse noxious inhibitorycontrols. Biol Res 1995;28:113–25.

[61] Villanueva L, Bernard JF, Le Bars D. Distribution of spinal cordprojections from the medullary subnucleus reticularis dorsalis andthe adjacent cuneate nucleus: a Phaseolus vulgaris-leucoagglutininstudy in the rat. J Comp Neurol 1995;353:11–32.

[62] Villanueva L, Bing Z, Bouhassira D, Le Bars D. Encoding ofelectrical, thermal and mechanical noxious stimuli by subnucleusreticularis dorsalis neurons in the rat medulla. J Neurophysiol1989;61:391–402.

[63] Villanueva L, Bouhassira D, Bing Z, Le Bars D. Convergence ofheterotopic nociceptive information onto subnucleus reticularis

208 C. Soto et al. / Pain 140 (2008) 190–208

dorsalis neurons in the rat medulla. J Neurophysiol1988;60:980–1009.

[64] Villanueva L, Cliffer KD, Sorkin D, Le Bars D, Willis WD.Convergenge of heterotopic nociceptive information onto neuronsof the caudal medullary reticular formation in the monkey. JNeurophysiol 1990;63:1118–27.

[65] Villanueva L, De Pommery J, Menetrey D, Le Bars D. Spinalafferent projections to subnucleus reticularis dorsalis in the rat.Neurosci Lett 1991;134:98–102.

[66] Wei F, Dubner R, Ren K. Nucleus reticularis gigantocellularisand nucleus raphe magnus in the brain stem exert opposite effects

on behavioral hyperalgesia and spinal Fos protein expression afterperipheral inflammation. Pain 1999;80:127–41.

[67] Willis WD, Coggeshall RE. Sensory mechanisms of the spinalcord. Primary afferent neurons and the spinal dorsal horn. NewYork: Kluwer Academic/Plenum Publishers; 2004.

[68] Woolf CJ. Windup and central sensitisation are not equivalent.Pain 1996;66:105–8.

[69] Wu N, Enomoto A, Tanaka S, Hsiao CF, Nykamp DQ,Izhikevich E, et al. Persistent sodium currents in mesencephalicv neurons participate in burst generation and control of mem-brane excitability. J Neurophysiol 2005;93:2710–22.