ARTHRITIS & RHEUMATISMVol. 62, No. 12, December 2010, pp 3806–3814DOI 10.1002/art.27715© 2010, American College of Rheumatology

Tumor Necrosis Factor Causes Persistent Sensitization of JointNociceptors to Mechanical Stimuli in Rats

Frank Richter,1 Gabriel Natura,1 Stefan Loser,1 Katarina Schmidt,1 Hanna Viisanen,2

and Hans-Georg Schaible1

Objective. During inflammation in the joint, nor-mal joint movements are usually painful. A neuronalmechanism for this form of mechanical hyperalgesia isthe persistent sensitization of joint nociceptors to me-chanical stimuli. Because tumor necrosis factor (TNF)is a major mediator of joint inflammation, we undertookthe present study both to explore the potential of TNF tosensitize joint nociceptors to mechanical stimuli and toaddress the cellular mechanism involved.

Methods. In anesthetized rats, action potentials(APs) were recorded from sensory nociceptive A� fibersand C fibers supplying the knee joint. We monitoredresponses to rotation of the knee joint at innocuous andnoxious intensities. TNF, etanercept, and a p38 inhibi-tor were injected into the knee joint, and the cyclooxy-genase (COX) inhibitor diclofenac was administeredintraperitoneally. APs were also recorded in isolatedcultured dorsal root ganglion (DRG) neurons in orderto test for changes in neuronal excitability induced byTNF.

Results. A single application of TNF into thenormal knee joint caused a significant persistent sensi-tization of nociceptive sensory fibers to mechanicalstimuli applied to the joint. This effect was dose depen-dent. It was prevented by coadministration of etanercept

or by an inhibitor of p38, and it was attenuated bysystemic application of a COX inhibitor. Patch clamprecordings from isolated DRG neurons showed a rapidincrease in neuronal excitability induced by TNF.

Conclusion. TNF can induce a long-lasting sensi-tization of joint nociceptors to mechanical stimuli andthus can induce long-lasting mechanical hyperalgesia injoints. TNF can act directly on neurons, underscoringits role as a sensitizing pain mediator.

Tumor necrosis factor (TNF) is a key proin-flammatory cytokine involved in the pathogenesis ofhuman joint diseases. This is particularly evident fromthe significant improvement of rheumatoid arthritisachieved with neutralization of TNF by biologic agentsin a large proportion of patients (1–5). A role of TNFhas also been discussed in osteoarthritis, which may becharacterized by synovial hyperplasia and synovitis (6).Pain is a major clinical burden of joint diseases. BecauseTNF acts on different cells such as synovial fibroblastsand immune cells (7), the question arises whether it mayalso act on nerve cells in the joint and thereby directlycontribute to the generation of pain. In fact, there isevidence that TNF has a neuronal target. TNF receptorI (TNFRI) and TNFRII were identified in subpopula-tions of dorsal root ganglion (DRG) neurons, the cellbodies of sensory afferent fibers (8–15). In experimentalarthritis, systemic application of TNF-neutralizingagents significantly reduced mechanical hyperalgesia ofthe inflamed joint within only 1 day or a few days, a timepoint at which inflammation was still well expressed(8,16). Injection of the TNF-neutralizing compoundetanercept into the cavity of an inflamed joint signifi-cantly reduced the responses of C fibers to rotation ofthe joint within 2 hours, showing that reduction ofnociceptive activity can occur much more rapidly thanreduction of inflammation per se (8). Finally, TNF can

Supported by DFG grant SCHA 404/13-1. Ms Viisanen’s workwas supported by the Finnish Cultural Foundation and the EmilAaltonen Foundation.

1Frank Richter, MD, Gabriel Natura, PhD, Stefan Loser,Katarina Schmidt, Hans-Georg Schaible, MD: Jena University Hospi-tal, Friedrich Schiller University of Jena, Jena, Germany; 2HannaViisanen (current address: Institute of Biomedicine/Physiology, Hel-sinki, Finland): University of Helsinki, Helsinki, Finland.

Address correspondence and reprint requests to Hans-GeorgSchaible, MD, Jena University Hospital, Friedrich Schiller Universityof Jena, Institute of Physiology I/Neurophysiology, Teichgraben 8,D-07740 Jena, Germany. E-mail: Hans-Georg.Schaible@mti.uni-jena.de.

Submitted for publication April 15, 2010; accepted in revisedform August 12, 2010.

3806

influence ion currents through the membrane of isolatedcultured DRG neurons in the absence of other cells (17).

These data suggest a contribution of TNF to jointpain by a neuronal target, but whether TNF is actuallyable to induce a state of sensitization to mechanicalstimuli in nociceptive joint afferent fibers has not beendirectly investigated. The sensitization of joint nocicep-tors to mechanical stimuli (18,19) and the ensuingactivation of the central nociceptive system (20,21) is amajor neuronal mechanism of mechanical hyperalgesiaduring joint inflammation, characterized by pain duringmovements and palpation. Until now, mechanical sensi-tization was mainly attributed to prostaglandins, brady-kinin, and other compounds (for review, see ref. 21), butthere has been little investigation of the role of cyto-kines. We recently found that a single injection of theproinflammatory cytokine interleukin-6 (IL-6), and par-ticularly a single injection of IL-6 together with solubleIL-6R (sIL-6R), into a normal knee joint caused a slowlyrising but persistent sensitization of C fibers to mechan-ical stimulation of the joint (22). This striking effectfostered the hypothesis that cytokines play an importantrole in the generation and maintenance of long-lastingmechanical hyperalgesia.

In the present study, we explored whether TNFinduces a persistent sensitization of nociceptive fibers ofthe normal joint to mechanical stimulation. Further-more, we addressed cellular mechanisms involved in theputative effect of TNF, namely, the involvement of p38,an important kinase in the TNF signaling pathway(23,24), and we tested whether cyclooxygenase (COX)inhibitors attenuate the putative sensitizing TNF effect,a question of clinical interest. In addition, using patchclamp recordings from isolated DRG neurons, we inves-tigated whether TNF can alter neuronal excitability invitro in the absence of other cells.

MATERIALS AND METHODS

Recording of afferent fiber responses in vivo. Theexperiments were approved by the government of Thuringen(registration nos. 02-07/04 and 02-013/08). Sixty adult maleWistar rats (300–460 gm) were anesthetized intraperitoneallywith 100 mg/kg sodium thiopentone (Trapanal; Altana). Sup-plemental doses (20 mg/kg) maintained areflexia. The tracheawas cannulated, and the animals breathed spontaneously dur-ing surgery and recordings. A gentle jet of oxygen was blowntoward the opening of the tracheal cannula. Mean arterialblood pressure was continuously monitored via a catheter inthe right carotid artery. The absence of blood pressure changesin response to noxious stimuli confirmed the sufficient depth ofanesthesia. Body temperature was kept at 37°C using afeedback-controlled temperature constanter (L/M-80; List).

After the measurements the rats were killed by intravenousadministration of sodium thiopentone.

For exposure of the right knee and recordings, the ratwas lying on its back. The skin was incised from the medial sideof the lower leg to the belly. The skin flaps were sewed to anoval metal ring fixed over the leg to form a pool which wasfilled with paraffin oil to prevent drying of the tissue. A specialfastener fixed the right femur. The right hind paw was fixed ina shoe-like holder that was connected to a force transducer andtorque meter that allowed rotation of the lower leg in the kneejoint, the main stimulus of the knee joint afferent fibers.

Action potentials (APs) were recorded from fibers ofthe medial articular nerve, which innervates the anteromedialside of the knee joint. The medial articular nerve joins thesaphenous nerve, which was dissected and cut near the inguinalregion, because nerve fibers that are further from the knee willremain in place during knee rotation, making recordings morestable. Thin bundles of the saphenous nerve were placed on asmall Perspex plate for further dissection with ultrafine watch-maker tweezers. To eliminate fiber activity from the lower leg,the saphenous nerve was cut distally to the entry point of themedial articular nerve.

Thin nerve filaments were placed on a platinum wireelectrode. The reference platinum wire electrode was con-nected to connective tissue. Amplified APs were continuouslymonitored on an oscilloscope to observe their shape and size.Filaments with either 1 or 2–3 fibers with discernible APs wereaccepted. The APs were also fed into a PC (interface cardDAB 1200; Microstar Laboratories) for off-line analysis withthe spike/spidi software package and construction of peristimu-lus time histograms.

Medial articular nerve afferent fibers in the saphenousnerve were searched by probing the anteromedial side of theknee with a blunt glass rod at moderate intensity. If APs wereelicited, the exact location and mechanical threshold of thereceptive field of this fiber were determined using calibratedvon Frey hairs (pressure range 16–512 mN). We only includedfibers that had a receptive field in the knee and responded tooutward rotation of the knee. Filaments with regular sponta-neous activity (typical for muscle spindles) were discarded.The conduction velocity of the nerve fibers (C �1.25 meters/second, A� 1.25–10 meters/second, A� �10 meters/second)was determined by electrical stimulation of the mechanicalreceptive field with a bipolar electrode (1–10V, 0.5 msec pulsewidth) and by dividing the distance between the receptive fieldand the electrode by the latency between the stimulus artifactand the evoked AP.

Responses of fibers were tested with a sequence ofmechanical stimuli. In each test block we recorded restingactivity for 1 minute, then we applied innocuous outwardtorque (20 mNm 3 times for 15 seconds each at intervals of 1minute), followed by noxious outward torque (40 mNm 3 timesfor 15 seconds each at intervals of 1 minute), as displayed inFigure 1B. Rotation to 20 mNm was considered innocuousbecause at this intensity the lower leg could be rotated to theend of the normal movement range without appreciable force.Rotation to 40 mNm was considered noxious because theserotations were performed against the resistance of the jointstructures. The interval between test blocks was 15 minutes.The first 4 test blocks defined the baseline. We then injected0.1 ml of the test substance into the knee joint cavity, and

SENSITIZATION OF JOINT NOCICEPTORS BY TNF 3807

another 12 test blocks were applied. In some experiments asecond injection was performed. APs were discriminated off-line according to shape and size and counted before and duringeach stimulus. For normalization we averaged all responses toeach type of stimulus preceding drug application (baseline)and subtracted these values from the responses to the mechan-ical stimuli after drug application. Furthermore, we averageddata from 3 subsequent stimulations (1 recording block) in therecording protocol to generate 1 data point in the diagrams.For most fibers the mechanical threshold in the receptive fieldwas measured initially and at the end of the experimentalprotocol.

The following compounds were injected into the kneejoint in a volume of 100 �l: TNF (Bachem), dissolved in 1%bovine serum albumin (BSA) in phosphate buffered saline(PBS) (doses 0.05–5 ng); etanercept (R&D Systems), dissolvedin 1% BSA in PBS (100 �g etanercept was administered eithertogether with TNF or 90 minutes after TNF); and the p38inhibitor SB203580 (Jena Bioscience), dissolved in 4% DMSOin isotonic NaCl and applied at a concentration of 0.26 mM (10�g) either alone or together with 5 ng TNF. In 10 experimentswe injected diclofenac intraperitoneally 2 hours before TNFwas injected into the knee joint. Doses were either 2 mg/kg or4 mg/kg.

Patch clamp recordings from cultured DRG neuronsin vitro. Male rats were killed with ether. DRGs were dissectedfrom all spinal segments and incubated at 37°C with 215units/mg type II collagenase (Gibco BRL) dissolved in Ham’sF-12 medium (Sigma) for 100 minutes. After washing 3 timeswith Ca2�- and Mg2�-free PBS, pH 7.4 (Gibco BRL), theDRGs were incubated for 11 minutes at 37°C in Dulbecco’smodified Eagle’s medium (DMEM; Sigma) containing 10,000units/ml trypsin (Sigma). The ganglia were then dissociatedinto single cells by gentle agitation and by triturating through

a fire-polished Pasteur pipette. The dispersed cells were col-lected by centrifugation (500g for 5 minutes), washed 3 times inDMEM, and centrifuged. The neuron pellets were suspendedin Ham’s F-12 medium containing 10–3M L-glutamine (Sigma),10% heat-inactivated horse serum (Gibco BRL), 100 units/mlpenicillin (Gibco BRL), 100 �g/ml streptomycin (Gibco BRL),and 10 ng/ml nerve growth factor (NGF 7S; BoehringerMannheim), centrifuged again, resuspended in culture me-dium, plated on 13-mm diameter glass cover slips precoatedwith poly-L-lysine (50 �g/ml; Sigma), and kept at 37°C in ahumidified incubator gassed with 3.5% or 5% CO2 in air andfed daily with Ham’s F-12 medium.

APs were recorded in the current-clamp mode fromsmall- and medium-sized DRG neurons (cell bodies of C andA� fibers) cultured for 12–30 hours. The bath was perfusedwith HEPES solution (control) (140 mM NaCl, 5 mM KCl,2 mM CaCl2, 2 mM MgCl2, 10 mM glucose, 10 mM HEPES,pH 7.4). Test compounds were added with an applicationsystem. The recording pipettes contained 140 mM KCl, 10 mMNaCl, 1 mM MgCl2, 0.5 mM CaCl2, 2 mM Na2-ATP, 5 mMEGTA, 10 mM HEPES, 10 mM sucrose, pH 7.2. We includedonly neurons with a membrane potential more negative than–45 mV. To assess neuronal excitability, APs were elicited bycurrent injection through the recording pipette. At the restingpotential, current was applied at amplitudes of 10–70 pA(in 10-pA steps, pulse duration 5 msec, interpulse interval2 seconds) until an AP with the typical overshoot was elicited.This protocol was repeated every 2 minutes before applicationof TNF (50 ng/ml) to the bath and then within 5–7 minutesafter application of TNF to the bath. In control recordingsstimulation was carried out repeatedly in the presence ofHEPES only.

In further experiments we measured the thresholdbefore and after administration of either diclofenac (1 �M) orindomethacin (1 �M) to the bath, and/or after coadministra-tion of diclofenac or indomethacin and TNF at differentintervals after administration of diclofenac or indomethacinalone (for timing, see Results). Data were analyzed usingpCLAMP 7.01 software (Axon Instruments) and Origin 7.5software (Microcal Software).

Statistical analysis. In in vivo fiber recordings, changesof responses within groups (before versus after treatment)were analyzed statistically using the Wilcoxon matched pairssigned rank test. We compared the mean of responses of thelast test block in the control period before injection with themean of responses in the last test block of each hour (or 30minutes in the case of 5 ng TNF) and determined the first testperiod in which a change from baseline was significant at P �0.05. In patch clamp recordings we compared samples ofneurons using the chi-square test. We tested whether theproportions of neurons with decreases of threshold versus nochanges or increases of threshold differed significantly (P �0.05) between samples.

RESULTS

Recordings in vivo were performed from 47 Cfibers (mean � SEM conduction velocity 0.9 � 0.03meters/second) and 34 A� fibers (mean � SEM conduc-

Figure 1. Receptive fields of joint afferent fibers of the medialarticular nerve and sensitizing effect of tumor necrosis factor(TNF). A, Centers of the receptive fields of C fibers and A� fibersin the joint capsule of the anteromedial aspect of the exposed kneejoint. Each ellipse displays the localization of the center of thereceptive field of 1 fiber. B, Specimen of a single C fiber showing anincrease of the responses to movements at 3 time points afterintraarticular injection of 5 ng TNF. The graphs show the numberof action potentials/second (APs/s) elicited by either innocuous out-ward rotation (Inn. OR) or noxious outward rotation (Nox. OR).

3808 RICHTER ET AL

tion velocity 4.9 � 0.38 meters/second) supplying theknee joint. Their receptive fields, identified by probingof the knee, were located in the joint capsule of themedial aspect of the joint (Figure 1A). Most neuronsshowed a weak response during a sustained innocuousoutward rotation to the end of the working range and astrong response to noxious outward rotation before anyknee injection (Figure 1B).

Effect of intraarticular TNF injection on re-sponses of joint afferent fibers. Figure 1B shows apersistent increase in the responses of a C fiber toinnocuous and noxious outward rotation of the joint 1, 2,and 3 hours after a single intraarticular injection of 5 ngTNF. In the whole sample of C fibers, the injection of0.05 ng TNF did not increase responses to innocuousoutward rotation (Figure 2A) or to noxious outwardrotation (Figure 2B). Responses remained near thebaseline level (set to 0) throughout 180 minutes. Incontrast, injection of 0.5 ng TNF caused a gradualincrease of the responses to innocuous and noxiousoutward rotation of the joint, and injection of 5 ng TNFcaused an even stronger and earlier increase of theresponses. Mechanical thresholds were significantly re-duced after administration of 5 ng TNF (on averagefrom 192 mN to 76 mN; n � 7) (P � 0.05 by Wilcoxonmatched pairs signed rank test). Three of 5 C fiberstested showed a reduction of threshold after administra-tion of 0.5 ng TNF, but none of 4 C fibers tested showeda decrease of threshold after administration of 0.05 ngTNF.

In A� fibers, responses to noxious outward rota-tion showed a significant small increase after adminis-tration of 5 ng TNF only (from 122.2 � 48.4 APs/15seconds to 190.7 � 50.6 APs/15 seconds; n � 7).Responses to innocuous outward rotation did notchange (P not significant by Wilcoxon matched pairssigned rank test).

Blockade of TNF effects by etanercept. In orderto test whether the TNF effect can be prevented orreversed by neutralization of TNF, we administeredetanercept into the joint either together with TNF or 90minutes after TNF administration. Figure 3A shows theeffect of 5 ng TNF alone (same curve as in Figure 2B)and the effects of the coinjection of 5 ng TNF with 100�g etanercept in C fibers. During coadministration ofboth compounds the responses of the C fibers did notincrease (nor did those of 5 A� fibers). Thus, etanerceptprevented the sensitization of the fibers by TNF.

In 3 A� fibers and in 2 C fibers we tested whetheretanercept reduces TNF-induced increases in responses.In the C fiber shown in Figure 3B and in the A� fiber

shown in Figure 3C, 5 ng TNF increased responses toinnocuous and noxious outward rotation, and adminis-tration of etanercept 90 minutes later reduced theenhanced responses. All 5 fibers tested showed the sameresponse pattern. Thus, etanercept reduced the TNF-induced increase of responses, whereas responses after

Figure 2. Sensitization of C fibers of the normal knee joint tomechanical stimulation by different concentrations of TNF injectedinto the joint cavity. A, Increase of responses to innocuous outwardrotation after injection of TNF in different doses. The baselineresponses before TNF injection are set to 0 (for the 0.05 ng TNFgroup, mean � SEM 9.8 � 5.6 APs/15 seconds; for the 0.5 ngTNF group, mean � SEM 5.0 � 2.0 APs/15 seconds; for the 5 ng TNFgroup, mean � SEM 32.3 � 13.6 APs/15 seconds). � � P � 0.05 versuspretreatment baseline (control), by Wilcoxon matched pairs signedrank test. B, Increase of responses to noxious outward rotation afterinjection of TNF in different doses. The baseline responses beforeTNF injection are set to 0 (for the 0.05 ng TNF group, mean � SEM82.8 � 24.1 APs/15 seconds; for the 0.5 ng TNF group, mean �SEM 59.9 � 12.0 APs/15 seconds; for the 5 ng TNF group, mean �SEM 194.4 � 62.8 APs/15 seconds). Arrows indicate the first intervalsin experimental groups in which the increase of responses frompretreatment baseline became statistically significant, staying signifi-cant thereafter. � � P � 0.05 versus pretreatment baseline, byWilcoxon matched pairs signed rank test. See Figure 1 for definitions.

SENSITIZATION OF JOINT NOCICEPTORS BY TNF 3809

administration of 5 ng TNF without etanercept re-mained high throughout 180 minutes (Figure 3A).

Blockade of the effect of TNF by an inhibitor ofp38. In isolated cultured DRG neurons of mice, TNFenhanced tetrodotoxin-resistant Na� currents (a cellularmechanism that may contribute to sensitization), andthis effect was blocked by an inhibitor of p38 (17). Herewe tested whether the increase in responses to mechan-ical stimulation in vivo is blocked by the p38 inhibitor

SB203580. Intraarticular application of the p38 inhibitoralone did not significantly change responses of C fibers toinnocuous and noxious outward rotation of the knee joint(Figure 4A). Coadministration of SB203580 with 5 ng TNFcaused a slight reduction of responses to noxious rotation,which did not reach statistical significance. A specimen ofa C fiber with responses to innocuous and noxious out-ward rotation before and 150 minutes after coapplica-tion is shown in Figure 4B. Mechanical thresholds in thereceptive fields did not change. Thus, the p38 inhibitorprevented the typical TNF-induced sensitization in Cfibers and in A� fibers (not shown).

Partial blockade of the effect of TNF by a COXinhibitor. In the next set of experiments we testedwhether the COX inhibitor diclofenac is able to reducethe sensitization by TNF. Either 2 mg/kg or 4 mg/kgdiclofenac was administered intraperitoneally 2 hoursbefore TNF was injected into the joint. Following ad-ministration of 2 mg/kg diclofenac, 5 ng intraarticular

Figure 4. Prevention of TNF-induced mechanical sensitization by thep38 MAPK inhibitor SB203580. A, Responses of C fibers of the kneejoint to noxious outward rotation after intraarticular injection ofSB203580 alone and after coadministration of 5 ng TNF andSB203580. Values are the mean � SEM. B, Specimen of a C fiber(conduction velocity 0.87 meters/second) showing responses to innoc-uous and noxious outward rotation before (left) and 150 minutesafter (right) coadministration of TNF and SB203580. See Figure 1 fordefinitions.

Figure 3. Prevention and reduction of TNF-induced mechanical sen-sitization by etanercept. A, Coadministration of 5 ng TNF and 100 �getanercept into the normal joint. There was no increase of responses ofC fibers to noxious outward rotation. The dashed line (effect of TNFalone) is the same as that shown in Figure 2B. Arrow indicates the firstinterval in the group treated with TNF alone in which the increase ofresponses from pretreatment baseline (control) became statisticallysignificant, staying significant thereafter. Values are the mean � SEM.� � P � 0.05 versus pretreatment baseline, by Wilcoxon matched pairssigned rank test. B, TNF-induced increase in the responses to innoc-uous and noxious outward rotation in a C fiber (conduction velocity0.33 meters/second) and reversal of this effect by intraarticular etan-ercept (administered 90 minutes after TNF). C, Reversal of TNF-induced mechanical sensitization by etanercept in an A� fiber (con-duction velocity 7 meters/second). See Figure 1 for definitions.

3810 RICHTER ET AL

TNF still induced a significant increase in the responsesof 6 C fibers to noxious pressure (Figure 5). However,after administration of 4 mg/kg diclofenac, 5 ng TNF didnot increase the responses to noxious pressure. Thus,diclofenac at the higher dose prevented the sensitizationof joint afferent fibers by TNF. Thresholds in thereceptive fields were only slightly and nonsignificantlyreduced in these experiments.

TNF effects on the excitability of isolated cul-tured DRG neurons in vitro. In the natural environment,effects of TNF on responses of nerve fibers may dependat least in part on the cooperation of nerve cells withother cells (i.e., the effects of TNF may be indirect).Therefore, we also tested whether TNF can actuallyenhance neuronal excitability in isolated DRG neurons.The use of isolated DRG neurons is a standard ap-proach in molecular pain research because the mem-brane of the cell body expresses all the molecules thatdetermine the function of these neurons (25–27). How-ever, even under in vivo conditions DRG neurons maybe a direct target of TNF because during the acute phaseof knee joint inflammation, macrophages invade thelumbar DRGs, and this effect is significantly correlatedwith the severity of hyperalgesia (28). A typical test forexcitability changes is to measure the amount of currentthat is necessary to evoke a full-blown AP (29). Figure

6A shows recordings from an isolated DRG neuron inwhich the effect of TNF was tested. Before administra-

Figure 5. Effect of the cyclooxygenase inhibitor diclofenac on TNF-induced mechanical sensitization of C fibers. There was a significantTNF-induced increase in the responses to noxious pressure afteradministration of 2 mg/kg diclofenac, but there was no TNF-inducedincrease in the responses after administration of 4 mg/kg diclofenac.For purposes of comparison, the effect of 5 ng TNF alone is displayed.Values are the mean � SEM. Arrows indicate the first intervals inwhich the increase of responses from pretreatment baseline (control)became statistically significant, staying significant thereafter. � � P �0.05 versus pretreatment baseline, by Wilcoxon matched pairs signedrank test. See Figure 1 for definitions.

Figure 6. Effect of TNF on the excitability of isolated cultured dorsalroot ganglion (DRG) neurons. A, Specimen showing the application ofcurrent injections of amplitudes of 10–60 pA and the elicitation ofAPs in a DRG neuron. The threshold for a full-blown AP beforebath application of TNF was 50 pA, but within 5 minutes after bathapplication of 50 ng/ml TNF, the first full-blown AP was alreadyelicited at a current of 40 pA. B, Threshold for the elicitation of APsin DRG neurons. Left, Threshold initially (control) and 5 minutesafter HEPES superfusion only (n � 16 neurons). Right, Thresholdinitially and 5 minutes after TNF application (n � 17 neurons). Therewas a significant difference in the proportion of neurons showing adecrease in threshold (� � P � 0.001 by chi-square test). C, Thresholdin DRG neurons before (control) and 5–7 minutes after bath applica-tion of either 1 �M indomethacin (Indom.) (n � 3 neurons) or 1 �Mdiclofenac (dicl.) (n � 18 neurons). D, Threshold in a subgroup ofDRG neurons (from DRG neurons in C), in which diclofenac orindomethacin was applied first and TNF was coadministered 5–7 minutesafter diclofenac or indomethacin. E, Threshold in DRG neurons afterbath application of diclofenac for �25 minutes and 5–7 minutes afteraddition of TNF to the bath solution. See Figure 1 for other definitions.

SENSITIZATION OF JOINT NOCICEPTORS BY TNF 3811

tion of TNF, currents up to 40 pA elicited only smalllocal subthreshold depolarizations, but at 50 pA and 60pA full-blown APs were evoked. Within 3–5 minutesafter bath application of TNF, the first full-blown APwas already elicited at a current of 40 pA, showing thatTNF lowered the threshold for elicitation of APs.

Figure 6B shows the effect of TNF on thresholdsfor elicitation of APs in the whole sample of DRGneurons tested. During the repeated testing of thresholdin the absence of TNF (only HEPES buffer was used;n � 16 neurons), the threshold remained unchanged oreven increased within 5 minutes. When TNF was admin-istered to the bath, the threshold was either unchangedor increased as in the control group (n � 7 neurons), butdecreased in a proportion of neurons (n � 10 neurons).Thus, in the sample of TNF-treated neurons, a signifi-cantly higher proportion (10 of 17) showed a reductionof threshold than in the sample of HEPES-treatedneurons (0 of 16) (P � 0.001 by chi-square test).

Because in the nerve fiber recordings in vivo thesensitization by TNF was reduced by diclofenac, wetested whether nonsteroidal antiinflammatory drugs(NSAIDs) reduce TNF effects on isolated DRG neu-rons. After bath application of indomethacin or diclofe-nac (n � 21 neurons), most neurons showed an increasein threshold (Figure 6C), and the proportion of neuronswith increases in threshold was significantly higher thanin the HEPES-treated group (P � 0.05 by chi-squaretest). Addition of TNF to the bath after previous short-term (5–7 minutes) bath application of NSAIDs (di-clofenac [n � 9 neurons], indomethacin [n � 3 neurons])reduced the threshold in 9 of 12 neurons tested (Figure6D). However, addition of TNF to the bath after previ-ous long-term (�25 minutes) bath application of diclofe-nac reduced the threshold in none of 9 neurons tested(Figure 6E). Thus, long-term exposure of DRG neuronsto diclofenac prevented TNF-mediated reduction of thethreshold.

DISCUSSION

The present study shows the potential of TNF toinduce persistent sensitization of joint afferent fibers tomechanical stimuli applied to the normal knee joint.After a single injection into the joint cavity, TNF causeda progressive increase in the responses to rotation of thejoint, which was dose dependent and persistent. TNF-induced sensitization was prevented and reversed bycoadministration of the TNF-neutralizing fusion proteinetanercept, prevented by an inhibitor of p38, and re-duced by pretreatment with COX inhibitors. TNF also

enhanced excitability in a proportion of isolated DRGneurons, showing direct effects of TNF on neurons.

The increase of mechanosensitivity in nociceptiveA� fibers and C fibers is likely to underlie the mechan-ical hyperalgesia observed in awake animals after TNFinjection into normal tissue (30,31). The long-termsensitization of nociceptive joint afferent fibers by asingle injection of either TNF or IL-6 (plus sIL-6R) (22)is notable because such an effect was not observed whenprostaglandins, bradykinin, serotonin, and histaminewere tested (18,21). One could argue that the sensitizingeffect of TNF instead represents an inflammatory reac-tion and that the effect is indirect. However, we believethat the TNF effect results from direct targeting of theneurons. First, approximately half of the small- andmedium-sized DRG neurons (the cell bodies of C fibersand A� fibers) express both TNFRI and TNFRII (8).Second, while TNF in low doses did not induce signifi-cant sensitization until �2 hours, higher doses inducedsensitization within 30 minutes, thus showing the possi-bility of a quite rapid sensitization. Third, bath appli-cation of TNF increased excitability in a proportion ofisolated DRG neurons within minutes. TNF also in-creased sodium currents in mouse DRG neurons withinminutes, and this was not observed in DRG neuronsfrom tnfr1–/– mice (17). The last of these data indi-cate rapid nongenomic effects of TNF on neurons,possibly by phosphorylation of sodium channels by p38MAPK (17).

The increase of mechanosensitivity by TNF wasprevented as well as reversed by administration ofetanercept into the joint. Thus, the previously observedreduction of responses to stimulation of an inflamedknee joint within 90 minutes of etanercept applicationinto the inflamed joint (8) may in fact show a significantcontribution of TNF to inflammation-evoked sensitiza-tion. The reversibility of TNF effects by etanercept is incontrast to the irreversibility of IL-6–induced sensitiza-tion by soluble gp130. This IL-6/sIL-6R–neutralizingcompound only prevented the increase of mechanosen-sitivity upon pretreatment but did not reverse the in-creased mechanosensitivity upon posttreatment (22).These data suggest that TNF-mediated effects on noci-ceptive neurons can be treated more effectively thanIL-6–mediated effects. Preliminary experiments supportthis hypothesis (Schaible H-G: unpublished observations).

TNF seems to be important in pain states ofdifferent etiology and pathophysiology. TNF is a medi-ator of neuropathic pain induced by nerve damage(11,13–15,32), and it has been suggested to have a role innoninflammatory fibromyalgia, because the application

3812 RICHTER ET AL

of TNF into the normal masseter muscle in rats evokeda long-lasting reduction of the mechanical threshold ofA� fibers in muscle (9). Notably, in the present experi-ments, C fibers responded to even lower TNF doses thandid A� fibers.

Knowledge about the molecular mechanisms ofTNF effects on neurons is fragmentary. In different celltypes TNF activates intracellular MAPKs such as p38MAPK, ERK-1/2, and JNK (6,23). The increase oftetrodotoxin-resistant Na� currents in isolated DRGneurons by TNF was mediated by the p38 pathway (17).Because these Na� currents are involved in the activa-tion of nociceptive afferent fibers by mechanical stimuli(27,33), we tested whether p38 was involved in thesensitization to mechanical stimuli. Indeed, the p38inhibitor SB203580 completely prevented the sensitiza-tion by TNF.

Interestingly, COX inhibitors at a dose of 4 mg/kgcompletely blocked the TNF-induced sensitization tomechanical stimuli, whereas with COX inhibitors inlower doses, TNF remained able to sensitize the jointafferent fibers. Thus, classic pain therapy may antago-nize TNF effects provided that the dose is sufficient.Preliminary data indicate that opioids do not preventsensitization (Richter F, Schaible H-G: unpublishedobservations). The recordings from isolated neuronsshow that NSAIDs and TNF influence excitability ofneurons in an opposite manner (i.e., NSAIDs decreaseexcitability, TNF increases excitability). After sufficientduration of bath application the NSAID prevented theTNF effect.

Several mechanisms may account for this inter-action. First, some DRG neurons express cyclooxygen-ases (34,35), and DRG cultures release small amounts ofprostaglandin E2 (PGE2) (Schaible H-G: unpublishedobservations). Because NSAIDs block the synthesis ofPGE2, a compound that enhances excitability of sensoryneurons (36,37), the NSAID may remove a mediatorthat acts synergistically with TNF in activating sensitiz-ing intracellular pathways such as the NF-�B pathway(38). Second, NSAIDs were shown to reduce Na�

currents in isolated DRG neurons by an unknownmechanism (39,40). Thus, the counteraction of TNFeffects by NSAIDs may also result from opposite effectsof NSAIDs and TNF on Na� currents. In favor of acritical role of Na� currents is the finding that inhibitionof p38 blocked the enhancement of tetrodotoxin-resistant Na� currents by TNF in isolated DRG neurons(17), as well as the increase in excitability of jointnociceptors to mechanical stimuli by TNF in the presentstudy.

In summary, this study shows that TNF has thepotential to cause long-lasting sensitization of jointnociceptors to mechanical stimuli. This effect may con-tribute significantly to mechanical hyperalgesia, a majorsymptom of joint diseases. The effect of TNF may resultat least in part from direct effects on sensory neurons.Thus, TNF acts as a pain mediator, and pain relief byTNF neutralization has a direct neuronal component.

ACKNOWLEDGMENTS

The authors thank Konstanze Ernst and Antje Wallnerfor excellent technical assistance.

AUTHOR CONTRIBUTIONS

All authors were involved in drafting the article or revising itcritically for important intellectual content, and all authors approvedthe final version to be published. Dr. Schaible had full access to all ofthe data in the study and takes responsibility for the integrity of thedata and the accuracy of the data analysis.Study conception and design. Richter, Natura, Loser, Schmidt,Viisanen, Schaible.Acquisition of data. Richter, Natura, Loser, Schmidt, Viisanen,Schaible.Analysis and interpretation of data. Richter, Natura, Loser, Schmidt,Viisanen, Schaible.

REFERENCES

1. Feldmann M, Maini RN. Anti-TNF� therapy of rheumatoidarthritis: what have we learned? Annu Rev Immunol 2001;19:163–96.

2. Lorenz HM, Kalden JR. Perspectives for TNF-�-targeting thera-pies. Arthritis Res 2002;4:S17–24.

3. Redlich K, Schett G, Steiner G, Hayer S, Wagner EF, SmolenJS. Rheumatoid arthritis therapy after tumor necrosis factorand interleukin-1 blockade [review]. Arthritis Rheum 2003;48:3308–19.

4. Smolen JS, Steiner G. Therapeutic strategies for rheumatoidarthritis. Nat Rev Drug Discov 2003;2:473–88.

5. Moller B, Villiger PM. Inhibition of IL-1, IL-6, and TNF� inimmune-mediated inflammatory diseases. Springer Semin Immu-nopathol 2006;27:391–408.

6. Bondeson J, Blom AB, Wainwright S, Hughes C, Caterson B, vanden Berg WB. The role of synovial macrophages and macrophage-produced mediators in driving inflammatory and destructive re-sponses in osteoarthritis [review]. Arthritis Rheum 2010;62:647–57.

7. Kinne RW, Stuhlmuller B, Burmester GR. Cells of the synoviumin rheumatoid arthritis: macrophages. Arthritis Res Ther 2007;9:224–40.

8. Boettger MK, Hensellek S, Richter F, Gajda M, Stockigt R,Segond von Banchet G, et al. Antinociceptive effects of tumornecrosis factor � neutralization in a rat model of antigen-inducedarthritis: evidence of a neuronal target. Arthritis Rheum 2008;58:2368–78.

9. Hakim AW, Dong XD, Svensson P, Kumar U, Cairns BE. TNF�mechanically sensitizes masseter muscle afferent fibers of malerats. J Neurophysiol 2009;102:1551–9.

10. Hensellek S, Brell P, Schaible HG, Brauer R, Segond von BanchetG. The cytokine TNF� increases the proportion of DRG neurones

SENSITIZATION OF JOINT NOCICEPTORS BY TNF 3813

expressing the TRPV1 receptor via the TNFR1 receptor and ERKactivation. Mol Cell Neurosci 2007;36:381–91.

11. Lee HL, Lee KM, Son SJ, Hwang SH, Cho HJ. Temporalexpression of cytokines and their receptors mRNAs in neuropathicpain model. Neuroreport 2004;15:2807–11.

12. Pollock J, McFarlane SM, Connell MC, Zehavi U, VandenabeeleP, MacEwan DJ, et al. TNF-� receptors simultaneously activateCa2� mobilisation and stress kinases in cultured sensory neurones.Neuropharmacology 2002;42:93–106.

13. Schafers M, Geis C, Brors D, Yaksh TL, Sommer C. Anterogradetransport of tumor necrosis factor-� in the intact and injured ratsciatic nerve. J Neurosci 2002;22:536–45.

14. Schafers M, Lee DH, Brors D, Yaksh TL, Sorkin LS. Increasedsensitivity of injured and adjacent uninjured rat primary sensoryneurons to exogenous tumour necrosis factor-� after spinal nerveligation. J Neurosci 2003;23:3028–38.

15. Shubayev VI, Myers RR. Axonal transport of TNF-� in painfulneuropathy: distribution of ligand tracer and TNF receptors.J Neuroimmunol 2001;114:48–56.

16. Inglis JJ, Notley CA, Essex D, Wilson AW, Feldmann M,Anand P, et al. Collagen-induced arthritis as a model of hyperal-gesia: functional and cellular analysis of the analgesic actions oftumor necrosis factor blockade. Arthritis Rheum 2007;56:4015–23.

17. Jin X, Gereau RW. Acute p38-mediated modulation of tetrodo-toxin-resistant sodium channels in mouse sensory neurons bytumor necrosis factor-�. J Neurosci 2006;26:246—55.

18. Schaible HG, Grubb BD. Afferent and spinal mechanisms of jointpain. Pain 1993;55:5–54.

19. McDougall J. Arthritis and pain: neurogenic origin of joint pain.Arthritis Res Ther 2006;8:220–9.

20. Fitzcharles MA, Almahrezi A, Shir Y. Pain: understanding andchallenges for the rheumatologist [review]. Arthritis Rheum 2005;52:3685–92.

21. Schaible HG, Schmelz M, Tegeder I. Pathophysiology and treat-ment of pain in joint disease. Adv Drug Deliv Rev 2006;58:323–42.

22. Brenn D, Richter F, Schaible HG. Sensitization of unmyelinatedsensory fibers of the joint nerve to mechanical stimuli by interleu-kin-6 in the rat: an inflammatory mechanism of joint pain.Arthritis Rheum 2007;56:351–9.

23. Kunisch E, Gandesiri M, Fuhrmann R, Roth A, Winter R, KinneRW. Predominant activation of MAP kinases and pro-destructive/proinflammatory features by TNF� in early-passage synovial fibro-blasts via TNF receptor-1: failure of p38 inhibition to suppressmatrix metalloproteinase-1 in rheumatoid arthritis. Ann RheumDis 2007;66:1043-51.

24. Waldburger JM, Boyle DL, Edgar M, Sorkin LS, Levine YA,Pavlov VA, et al. Spinal p38 MAP kinase regulates peripheralcholinergic outflow. Arthritis Rheum 2008;58:2919–21.

25. Basbaum AI, Bautista DM, Scherrer G, Julius D. Cellular andmolecular mechanisms of pain. Cell 2009;139:267–84.

26. Hucho T, Levine JD. Signaling pathways in sensitization: toward anociceptor cell biology. Neuron 2007;55:365–76.

27. Rush AM, Cummins TR, Waxman SG. Multiple sodium channelsand their roles in electrogenesis within dorsal root ganglionneurons. J Physiol 2007;579:1–14.

28. Segond von Banchet G, Boettger MK, Fischer N, Gajda M, BrauerR, Schaible HG. Experimental arthritis causes tumor necrosisfactor-� dependent infiltration of macrophages in rat dorsal rootganglia which correlates with pain-related behavior. Pain 2009;145:151–9.

29. Natura G, Segond von Banchet G, Schaible HG. Calcitoningene-related peptide enhances TTX-resistant sodium currents incultured dorsal root ganglion neurons from adult rats. Pain2005;116:194–204.

30. Cunha FQ, Poole S, Lorenzetti BB, Ferriera SH. The pivotal roleof tumour necrosis factor � in the development of inflammatoryhyperalgesia. Br J Pharmacol 1992;107:660–4.

31. Watkins LR, Wiertelak EP, Goehler LE, Smith KP, Martin D,Maier SF. Characterization of cytokine-induced hyperalgesia.Brain Res 1994;654:15–26.

32. Sommer C, Kress M. Recent findings on how proinflammatorycytokines cause pain: peripheral mechanisms in inflammatory andneuropathic hyperalgesia. Neurosci Lett 2004;361:184–7.

33. Akopian AN, Souslova V, England S, Okuse K, Ogata N, Ure J, etal. The tetrodotoxin-resistant sodium channel SNS has a special-ized function in pain pathways. Nat Neurosci 1999;2:541–8.

34. Maihofner C, Tegeder I, Euchenhofer C, deWitt D, Brune K,Bang R, et al. Localization and regulation of cyclooxygenase-1 and-2 and neuronal nitric oxide synthase in mouse spinal cord.Neuroscience 2000;101:1093–108.

35. Willingale HL, Gardiner NJ, McLymont N, Giblett S, Grubb BD.Prostanoids synthesized by cyclo-oxygenase isoforms in rat spinalcord and their contribution to the development of neuronalhyperexcitability. Br J Pharmacol 1997;122:1593–604.

36. England S, Bevan S, Docherty RJ. PGE2 modulates the tetrodotoxin-resistant sodium current in neonatal rat dorsal root ganglionneurones via the cyclic AMP–protein kinase A cascade. J Physiol1996;495:429–40.

37. Gold MS, Levine JD, Correa AM. Modulation of TTX-R INa byPKC and PKA and their role in PGE2-induced sensitization of ratsensory neurons in vitro. J Neurosci 1998;18:10345–55.

38. Yamamoto Y, Gaynor RB. Therapeutic potential of inhibition ofthe NF-�B pathway in the treatment of inflammation and cancer.J Clin Invest 2001;107:136–42.

39. Fei XW, Liu LY, Xu JG, Zhang ZH, Mei YA. The non-steroidalanti-inflammatory drug, diclofenac, inhibits Na� current in ratmyoblasts. Biochem Biophys Res Commun 2006;346:1275–83.

40. Lee HM, Kim HI, Shin YK, Lee CS, Park M, Song JH. Diclofenacinhibition of sodium currents in rat dorsal root ganglion neurons.Brain Res 2003;992:120–7.

3814 RICHTER ET AL