Antinociceptive effects of tumor necrosis factor α neutralization in a rat model of antigen-induced arthritis: Evidence of a neuronal target




The reduction of pain in the course of antiinflammatory therapy can result from an attenuation of the inflammatory process and/or from the neutralization of endogenous mediators of inflammation that act directly on nociceptive neurons. The purpose of this study was to investigate whether analgesic effects of the neutralization of tumor necrosis factor α (TNFα) are due to an attenuation of inflammation or whether direct neuronal effects significantly contribute to pain relief in the course of therapy.


Locomotor and pain-related behavior and histology were assessed in rats with chronic antigen-induced arthritis (AIA) in the knee joint, and the rats were treated with systemic saline, etanercept, or infliximab. The expression of TNF receptors (TNFRs) in dorsal root ganglia was measured using immunohistochemical analysis and polymerase chain reaction. Action potentials were recorded from afferent Aδ fibers and C fibers of the medial knee joint nerve, and etanercept and infliximab were injected intraarticularly into normal or inflamed knee joints (AIA or kaolin/carrageenan-induced inflammation).


In rats with AIA, both etanercept and infliximab significantly decreased inflammation-induced locomotor and pain-related behavior, while joint swelling was only weakly attenuated and histomorphology still revealed pronounced inflammation. A large proportion of dorsal root ganglion neurons showed TNFRI- and TNFRII-like immunoreactivity. Intraarticular injection of etanercept reduced the responses of joint afferents to mechanical stimulation of the inflamed joint starting 30 minutes after injection, but had no effect on responses to mechanical stimulation of the uninflamed joint.


Overall, these data show the pronounced antinociceptive effects of TNFα neutralization, thus suggesting that reduction of the effects of TNFα on pain fibers themselves significantly contributes to pain relief.

Tumor necrosis factor α (TNFα) is a key proinflammatory molecule in human rheumatoid arthritis (RA) and other chronic inflammatory diseases. In a large proportion of patients, neutralization of TNFα by biologic agents, such as etanercept and infliximab, significantly improves the clinical signs and the long-term outcome of RA (1–5).

A major presenting symptom of arthritis is pain. This is caused by the activation and sensitization of nociceptive nerve fibers (“pain fibers”) that supply the joint (6, 7), and the ensuing activation of the central nociceptive system (8, 9). Pain reduction in the course of antiinflammatory therapy might simply result from an attenuation of inflammatory processes. Alternatively, antiinflammatory compounds may neutralize endogenous mediators that cause ongoing pain and/or hyperalgesia (enhanced pain upon stimulation) by directly activating or sensitizing nociceptive neurons. Recently, TNFα has attracted considerable attention in pain research as a putative pain mediator, because the application of TNFα into healthy tissue induces mechanical and thermal hyperalgesia, i.e., increased responsiveness to noxious mechanical and thermal stimuli, in behavioral experiments (10–12). Direct neuronal effects of TNFα have also been shown. Short-term application of TNFα enhanced the heat-evoked release of calcitonin gene-related peptide from sensory nerve terminals in rat skin (13), and increased the nociception-related ion currents in isolated dorsal root ganglion (DRG) neurons, the somata of primary afferent neurons, at a latency of a few minutes (14). Concerning pathologic pain states, TNFα has mainly been implicated in the generation of neuropathic pain. Damaged nerve fibers seem to be activated by TNFα released from cells, including Schwann cells, at the lesion site (for review, see ref.15).

TNFα effects are mediated by 2 receptors, TNFRI and TNFRII (4). Several reports have described the expression of both TNFRI and TNFRII in rat DRG neurons (16–21), but other studies identified only TNFRI in neurons, while localizing TNFRII in non-neuronal cells in the DRG (22, 23). All of these findings support the hypothesis that treatment with etanercept or infliximab not only influences the inflammation process, but exerts antinociceptive effects by reducing direct pronociceptive actions of TNFα on nerve fibers.

In the present experiments, we addressed the role of TNFα in inflammatory joint pain using different approaches. In antigen-induced arthritis (AIA), a rat model of chronic arthritis, we tested whether long-term treatment with etanercept or infliximab attenuates inflammation-associated changes in locomotor and pain-related behavior and whether behavioral changes correlate with a reduction in the inflammation. AIA is an immune-mediated joint inflammation whose histopathology shows many similarities to human RA (24, 25). For experimentation, the AIA model has several advantages. In immunized rats, only the antigen-injected joint develops inflammation. AIA has an incidence of 100%, its acute phase starts within the first hours after antigen injection into the joint, and it spontaneously progresses into chronic inflammation with a homogeneous time course (24–27). We also assessed the expression of TNFRs in DRG neurons during the course of AIA. Furthermore, we made electrophysiologic recordings of sensory fibers supplying the knee joint in anesthetized rats with AIA or immunized controls, in naive rats, and in rats with an acute experimental knee inflammation, and tested whether intraarticularly (IA) injected etanercept or infliximab influenced the responses of joint afferents to mechanical stimulation on a short time scale. Preliminary data have been reported previously (28–30).


All experiments were approved by Thuringian state authorities and complied with European Union regulations (86/609/EEC) for the care and use of laboratory animals.

Induction of AIA.

Female Lewis rats (n = 125; ages 6–8 weeks, 160–180 gm) (Charles River, Sulzfeld, Germany) were used because Lewis rats are particularly susceptible to AIA (31). Lewis rats are also particularly suitable for behavioral experiments because they are quite tame. For immunization, 500 μg antigen (methylated bovine serum albumin [mBSA]; Sigma, Deisenhofen, Germany) in saline emulsified with 500 μl Freund's complete adjuvant (CFA; Sigma) supplemented with 2 mg/ml Mycobacterium tuberculosis strain H37Ra (Difco, Detroit, MI) was injected subcutaneously, and this injection was repeated 7 days later. After another 14 days, the rats were briefly anesthetized with 2.5% isoflurane, and a sterile mBSA solution (500 μg mBSA in 50 μl saline) was injected into the left knee joint cavity to induce monarticular AIA in 94 rats. Thirty-one rats served as “immunized controls” without induction of arthritis.

Histology and grading of arthritis.

Rats were deeply anesthetized with 120 mg/kg sodium thiopentone injected intraperitoneally (IP) (Trapanal; Byk Gulden, Konstanz, Germany) and perfused intracardially with heparin-enriched phosphate buffered saline (PBS) and 4% phosphate buffered formalin on day 3 or 21 after AIA induction. The knee joints were removed, skinned, postfixed in 4% formalin for at least 7 days, decalcified in 7% AlCl3 (in 2.1% HCl and 6% formic acid) for 48 hours, embedded in paraffin, cut into 5-μm–thick frontal sections, and stained with hematoxylin and eosin. Two independent observers (AW, MG), who were blinded with regard to the treatment, graded the sections (0 = no alterations, 1 = mild alterations, 2 = moderate alterations, and 3 = severe alterations). The acute inflammatory reaction was graded based on the amount of fibrin exudation, the relative number and density of granulocytes in the synovial membrane, and the size of the joint space; chronic inflammation was graded based on the relative number and density of infiltrating mononuclear leukocytes in the synovial membrane, the degree of synovial hyperplasia, the extent of infiltration, and fibrosis in the periarticular structures. Cartilage and bone destruction were also scored (0 = no erosion, 1 = erosion of <10%, 2 = erosion of 10–25%, 3 = erosion of 26–50%, and 4 = erosion of >50% in cross sections) (25, 27).

Treatment protocols and behavioral tests.

Three groups of 5 arthritic rats received short-term treatment with either etanercept, infliximab, or 0.9% NaCl solution and were killed 3 days after AIA induction. Another 3 groups of 10 rats received long-term treatment with these compounds and were killed 21 days after AIA induction. Etanercept (Wyeth, Münster, Germany), a recombinant TNF receptor (p75)–Fc fusion protein, was administered IP at a dose of 5.5 mg/kg; infliximab (Centocor, Malvern, PA), a monoclonal anti-human TNF antibody, was administered IP at a dose of 7 mg/kg (both in 300 μl saline). The dosages and frequency of injections (see Results) were chosen from experimental studies in which these compounds reduced neuropathic pain (20, 32–34). One investigator (MKB), who was blinded with regard to the respective treatments, performed all the tests. Five and 10 immunized rats without AIA were tested in parallel.

Locomotor behavior.

Animals were placed in a piece of cloth, and the hind paws were stained with liquid dye. Then the rats walked through a tunnel, leaving their paw prints on blotting paper (35), and several parameters were measured. In addition, a guarding score was assessed (0 = no guarding, 1 = transient guarding of the hind limb only after a brief manual compression of the knee elicited a withdrawal movement, 2 = persistent visible limping during walking, 3 = no use of the hind limb with the arthritic knee, and 4 = no walking) (27).

Pain-related behavior.

Mechanical hyperalgesia at both knee joints was tested with a dynamometer (Correx; Haag-Streit, Bern, Switzerland) (27). Animals were gently placed in a piece of cloth, which they usually entered due to their instinct to be in the dark. While the animals were held (not restrained) in this position, the dynamometer was placed on the knee (lateral side between the femur and the tibia) and pressure was increased at the knee until the rats withdrew their legs or vocalized. The weight force to elicit this response was measured in grams. To prevent tissue damage, a cutoff value of 250 gm was defined. If an animal was not calm before testing, it was returned to its cage and tested later.

Thermal hyperalgesia of the hind paws (as an indicator of “secondary” hyperalgesia, i.e., hyperalgesia remote from the inflamed tissue [6]) was assessed with an algesimeter (Ugo Basile, Comerio, Italy), as previously described (36). After the animals were acclimated to the testing device, 3 consecutive standardized heat stimuli were applied to the hind paws, with intervals of at least 2 minutes between stimuli. Mean latencies were used as a measure of the withdrawal threshold to heat. Stimuli were applied for a maximum of 20 seconds. Swelling was assessed by measuring the mediolateral diameter of each knee using Vernier calipers (Mitutoyo, Neuss, Germany).

For statistical analysis, group differences were assessed using multivariate analyses of variance (MANOVAs) followed by post hoc Bonferroni adjustments for all time points tested.

Immunohistochemical labeling of TNFRI and TNFRII in DRG sections.

Lewis rats (n = 38) at different stages of AIA (no treatment) were anesthetized using 120 mg/kg sodium thiopentone IP and perfused transcardially with PBS containing 0.1% sodium nitrite and 10,000 units/liter heparin, followed by 4% paraformaldehyde in a phosphate buffer. DRGs from segments L1–L5 from the left and right side were excised separately and fixed at 4°C in 4% paraformaldehyde in PBS for 24 hours, then watered, dehydrated with a graded ethanol series (50%, 70%, 96%, 100%, and 100%, each for 10 minutes) and xylene (100%, 10 minutes), and transferred to methylbenzoate (100%) overnight.

The ganglia were embedded in paraffin (Histosec; Merck, Darmstadt, Germany) and cut into 5-μm sections that were dewaxed and autoclaved for 15 minutes (120°C, 1 bar) in 0.1M citrate buffer (pH 6.0). Chilled and PBS-washed sections were incubated for 30 minutes in PBS containing Triton X-100 and 2% goat serum (Rockland, Gilbertsville, PA), and then incubated overnight at 4°C with the primary rabbit polyclonal antibodies diluted in PBS containing 1% Triton X-100 and 1% gelatin from coldwater fish skin. Antibody dilutions were 1:200 for anti-TNFRI (raised against amino acids 30–301 of human TNFRI; Santa Cruz Biotechnology, Santa Cruz, CA) and 1:1,000 for anti-TNFRII (raised against the C-terminus of mouse TNFRII; Lab Vision, Fremont, CA). Then PBS-washed sections were incubated for 40 minutes at 20°C in biotinylated secondary antibodies (1:200; Dako, Glostrup, Denmark). After 3 washes with PBS, the avidin–biotin–peroxidase complex (Vectastain Elite ABC kit; Vector, Burlingame, CA) was applied for 40 minutes. Sections from different experimental groups were developed simultaneously with Jenchrome px blue (MoBiTec, Göttingen, Germany). Control experiments were performed by omitting the primary antibodies.

From the ganglia of each side, time point, and animal, 12 sections were prepared. In every other section, the proportion of neuronal profiles with TNFRI- or TNFRII-like immunoreactivity was determined, and for each rat the average proportion of labeled neuronal profiles was calculated. At least 100 neuronal profiles per rat were counted. Only neuronal profiles with a visible nucleus were included.

Real-time polymerase chain reaction (PCR) experiments.

Thirty Lewis rats in different stages of AIA (no treatment) were killed with CO2, and DRGs from segments L1–L5 were removed, weighed, and then snap-frozen in liquid nitrogen. Total RNA was prepared using the Absolutely RNA Miniprep Kit (Stratagene, La Jolla, CA). After measuring RNA concentrations with a spectrometer (at 260 nm), the RNA was stored at −80°C. To confirm the integrity of the 18S and 28S ribosomal RNA, ethidium bromide staining in 1.5% denaturing agarose gel was performed. One microgram total RNA was obtained for reverse transcription using the RevertAid First Strand complementary DNA (cDNA) synthesis kit (Fermentas, Burlington, Ontario, Canada). Using the oligo(dT)18 primer, all messenger RNA (mRNA) templates were transcribed in cDNA. The final concentration was adjusted with distilled water to 15 ng/μl, relative to the applied total RNA quantity.

For real-time PCR, the following primers were designed using Primer3 software ( for β-actin (GenBank accession no. NM_031144), forward 5′-CATTGCTGACAGGATGCAGA-3′, reverse 5′-AGCCACCAATCCACACAGAG-3′, 107 bp; for TNFRI (GenBank accession no. BC086413), forward 5′-CCAGGAGAGGTGATTGTGGA-3′, reverse 5′-ACTGAGGAGGCCCTGAGAAG-3′, 93 bp; for TNFRII (GenBank accession no. NM_130426), forward 5′-AGCCTGTGGATGCTGAAGAA-3′, reverse 5′-AGCCTGTGGATGCTGAAGAA-3′, 221 bp.

For real-time PCR, the Platinum SYBR Green qPCR SuperMix-UDG kit (Invitrogen, Carlsbad, CA) was used. The amplification was performed in 23 μl master mix plus 2 μl cDNA template. A master mix was prepared using 12.5 μl SuperMix plus primers, fluorescein reference dye (final concentration 10 nmoles/liter; Invitrogen), and water. The PCR and data analysis were performed using the MyiQ Single-Color real-time PCR detection system and Optical System software, version 1.0 (Bio-Rad, Hercules, CA). The following PCR protocol was used: step 1, 5 minutes at 95°C, step 2, 30 seconds at 95°C, step 3, 20 seconds at 60°C, and step 4, 30 seconds at 72°C. Steps 2–5 were repeated 35 times and were followed by the melting point analysis of the PCR products. For size and quality of the PCR products, the PCR mixtures were analyzed using a 3% agarose gel. For size, a standard low-range DNA ladder (Fermentas) was used.

Data were quantified as described by Pfaffl (37). Data were initially expressed as a threshold cycle, and the PCR efficiency from each gene was calculated. All samples were run in triplicate for each gene and time point. The β-actin gene served as a reference standard. For all genes, mRNA values were expressed as relative changes versus samples taken from the ganglia of immunized control rats.

Electrophysiologic recordings from joint afferents.

Adult male Wistar rats (n = 30; ages 13–17 weeks, 300–460 gm) (supplied by the University of Jena, Jena, Germany) were used because the model of kaolin/carrageenan-induced inflammation is well established in these rats (6, 38). For comparison with behavioral experiments, further recordings were performed in adult female Lewis rats (n = 12; ages 12–14 weeks, 210–225 gm) (Charles River). Rats were anesthetized with 100 mg/kg sodium thiopentone IP; supplemental doses (20 mg/kg IP) were administered as necessary to maintain areflexia. The animals breathed spontaneously through the cannulated trachea during surgery and the recording protocol. Mean arterial blood pressure was continuously measured to control the depth of anesthesia. Body temperature was kept at 37°C using a feedback-controlled temperature constanter (L/M-80; List, Darmstadt, Germany). In the Lewis rats, recordings were performed either on day 3 of AIA or after immunization only. In 23 Wistar rats, acute inflammation was induced in the right knee joint 7–11 hours before recordings by injection into the joint cavity of a kaolin suspension (4%, 0.1 ml; Sigma) followed by a carrageenan solution (2%, 0.1 ml; Sigma). After recordings, the rats were killed by intravenous injection of sodium thiopentone.

For recordings, the knee was exposed. The femur was fixed with a fastener, and the right hind paw was fixed in a shoelike holder that was connected to a force transducer and torquemeter, which allowed for outward and inward rotation of the knee. We identified nerve fibers with receptive fields in the knee joint by mechanical stimulation with a blunt glass rod. According to their conduction velocities, fibers were classified as C (≤1.25 m/sec), Aδ (1.25–10 m/sec), and Aβ fibers (≥10 m/sec) (38). For this purpose, the mechanically identified receptive field was stimulated with a bipolar electrode (pulses of 1–10V and 0.5 m/sec duration), and the distance between the electrode and the recording site was measured. Fiber responses were repeatedly tested with a sequence of movements consisting of innocuous outward and inward torque (20 mNm, 15 seconds each) and noxious outward and inward torque (40 mNm, 15 seconds each). Stimulation blocks were started every 6 minutes. After establishing the baseline responses (the first 6–8 test blocks), 0.1 ml of the test substance (etanercept, infliximab, or vehicle) was injected into the knee joint cavity, and testing was continued. Doses of etanercept and infliximab were calculated from clinical case reports (39–41). The means of 3 subsequent test blocks before injection were compared with the means of subsequent test blocks after injection, and changes in responses within groups were tested for significance using Wilcoxon's matched pairs signed rank test. P values less than 0.05 were considered significant.


Treatment with etanercept and infliximab.

In order to assess the antinociceptive and antiinflammatory effects of TNFα-neutralizing drugs in the course of AIA, we performed a long-term behavioral study. Figure 1 shows the outcome of long-term treatment (21 days) in rats with AIA (ipsilateral) that received either etanercept or NaCl. The infliximab group (treated in parallel) is not shown in Figure 1 for clarity of presentation because many mean values were similar to the mean values in the etanercept-treated group. However, statistical testing included this group as well. All compounds were applied 6 hours, 1 day, 4 days, 8 days, 11 days, and 15 days after induction of AIA, and infliximab and NaCl were also administered on days 6 and 13 of AIA. For the initial time points, the behavioral testing was performed 6 hours after injection of the compounds. MANOVA for all 4 groups revealed significant differences between groups (F[7,105] = 5.553; P = 0.012).

Figure 1.

Outcome of treatment with etanercept or NaCl over 21 days in rats with antigen-induced arthritis (AIA). AIA was induced in the left knee joint. Another group of immunized rats, but without AIA and not treated, was included (immunized controls). All animals were tested twice during the immunization procedure and 1 day, 3 days, 7 days, 14 days, and 21 days after the induction of AIA. A, Knee joint swelling as an indicator of inflammation. B, Guarding score of locomotor behavior. C, Mechanical threshold (gm) for withdrawal response at the inflamed knee. D, Time (seconds) to paw withdrawal in response to noxious thermal stimulation of the inflamed side. Values in A–D are the mean ± SD. ∗ = P < 0.05; ∗∗ = P < 0.01; ∗∗∗ = P < 0.001, NaCl treatment group versus etanercept treatment group. BL = baseline. E, Typical paw print patterns of an immunized control rat, a rat 3 days after AIA induction, and a rat 7 days after AIA induction. Note the reduced distance between the left and right paws in the longitudinal direction and the increased angle between the left and right paws (measured from lines defined by the middle and hind phalanges of 2 consecutive paw prints) in the rat 7 days after AIA induction.

Rats with AIA exhibited pronounced swelling of the injected knee on days 1–3, which slowly subsided between days 7 and 21 (Figure 1A). Etanercept-treated rats (n = 10) showed significantly less swelling than NaCl-treated rats (n = 10) on days 3 (P = 0.003 with post hoc Bonferroni adjustment) and 7 (P = 0.034). The reduction of swelling in infliximab-treated rats (n = 10) did not reach significance (P = 0.086 on day 3).

Compared with immunized control rats, NaCl-treated rats with AIA showed reduced mechanical thresholds for withdrawal at the inflamed knee on days 1, 3, 7, 14 (P < 0.001), and 21 (P = 0.027) of AIA, indicating pronounced mechanical hyperalgesia at the inflamed knee joint (Figure 1C). Compared with NaCl-treated rats with AIA, etanercept-treated rats with AIA showed significantly higher mechanical thresholds on days 1 (P = 0.006), 3 (P < 0.001), 7 (P < 0.001), 14 (P < 0.001), and 21 (P = 0.02), and infliximab-treated rats with AIA showed higher mechanical thresholds on days 1 (P = 0.012), 3 (P < 0.001), 7 (P < 0.001), and 14 (P < 0.001). Thus, neutralization of TNFα with etanercept and infliximab significantly attenuated mechanical hyperalgesia at the inflamed knee. As in a previous study (27), the mechanical threshold also showed a reduction from 250 gm to 190 gm at the contralateral knee on days 1 and 3 of AIA, but this small and transient effect was not further studied. It may result from bilateral segmental changes of receptor expression in DRGs, which are, however, not associated with inflammatory changes in the contralateral knee (27, 42).

AIA in the knee also lowered the thermal threshold of the ipsilateral paw, which indicated secondary hyperalgesia remote from the inflamed site (Figure 1D). NaCl-treated rats with AIA showed reduced latencies for withdrawal of the leg with the inflamed knee on days 1, 3, 7 (P < 0.001), and 14 (P = 0.006) of AIA. Compared with NaCl-treated rats, etanercept-treated rats showed longer latencies on days 7 (P = 0.023) and 14 (P = 0.034). Infliximab also increased the latencies, but the effects were not significant. Thus, thermal hyperalgesia was significantly attenuated with etanercept, but less so, and nonsignificantly, with infliximab.

AIA had marked influences on locomotor behavior (see the guarding score in Figure 1B). Figure 1E shows typical paw print patterns from an immunized control rat (symmetric use of the hind limbs), a rat with 3-day AIA (no use of the hind limb with the arthritic knee), and a rat with 7-day AIA (asymmetric walking). Compared with NaCl-treated rats, etanercept-treated rats showed less guarding behavior, and this reduction was significant on days 3 (P < 0.001) and 14 (P < 0.001) of AIA (Figure 1B). A significant effect of infliximab was observed on days 3 (P = 0.011) and 14 (P = 0.002). In the paw print pattern, the distance between the prints of the left and the right paws in the longitudinal direction became greatly reduced in rats with AIA. Additionally, the angle between paws was increased during AIA. Both parameters were significantly improved with etanercept and infliximab on days 7, 14, and 21 (P < 0.05–0.001).

After completion of the treatment study, arthritis in the knee was graded histologically. In addition, we assessed inflammation on day 3 of AIA in other groups of rats that received short-term treatment (5 rats each in the NaCl, etanercept, and infliximab groups) for 3 days only. Behavioral data from these rats were similar to those in Figure 1. Figures 2A and B show severe inflammation in a knee joint on day 3 in an NaCl-treated rat with AIA. Overall, all rats with antigen injected into the knee showed severe acute inflammation on day 3 of AIA and mild acute inflammation on day 21 (Figure 2C). Signs of chronic inflammation were already visible on day 3 of AIA (Figure 2D). Neither the inflammation nor the destruction scores (Figure 2E) showed significant differences between the NaCl-, etanercept-, and infliximab-treated groups. Thus, the histomorphologic assessment of inflammation did not show a significant effect of TNFα neutralization, although swelling was slightly reduced in the early stages.

Figure 2.

Histologic scoring of knee joints of rats in different treatment groups, 3 days or 21 days after induction of antigen-induced arthritis (AIA). A, Representative histologic section of an inflamed knee joint from a rat with AIA, 3 days after NaCl treatment. Severe infiltration of inflammatory cells into the synovial tissue is seen, as well as pannus formation and cartilage and bone destruction (hematoxylin and eosin [H&E] stained; original magnification × 40). B, Higher-magnification view of the boxed area in A (H&E stained; original magnification × 100). C, Scores for acute inflammation. D, Scores for chronic inflammation. E, Scores for cartilage and bone destruction. Values are the mean and SD.

Assessment of TNFR-like immunoreactivity in DRGs.

In order to identify peripheral neurons as potential targets of TNFα-neutralizing drugs, the expression of TNFRs was assessed in DRG neurons using PCR and antibodies against TNFRI or TNFRII. Previous testing of the antibodies showed that the antibody to TNFRI stained DRG neurons in TNFRII-deficient (Tnfr2−/−) mice but not in TNFRI-deficient (Tnfr1−/−) mice, whereas the antibody to TNFRII stained neurons in Tnfr1−/− mice but not in Tnfr2−/− mice (16).

PCR analysis revealed the presence of both TNFRI and TNFRII in the DRGs. Figures 3A and B show the gel electrophoresis of TNFRI and TNFRII real- time PCR products of lumbar DRGs from representative rats. With antibodies, both TNFRI-like immunoreactivity (Figure 3C) and TNFRII-like immunoreactivity (Figure 3D) were visualized in small- to medium-sized DRG neurons, most of which are the somata of nociceptive afferent fibers, but not in large-sized neurons, which are the somata of low-threshold non-nociceptive fibers. Figure 3E shows the size distribution of the DRG neurons.

Figure 3.

Expression of tumor necrosis factor receptor I (TNFRI) and TNFRII in lumbar dorsal root ganglion (DRG) sections from normal rats and rats with antigen-induced arthritis (AIA). A and B, Gel electrophoresis of TNFRI and TNFRII real-time polymerase chain reaction products of lumbar DRGs from 3 representative rats. No amplification product was detected in the nonamplification control (NAC) or the nontemplate control (NTC). S = standard. C and D, DRG sections from untreated control rats, with neurons exhibiting TNFRI- and TNFRII-like immunoreactivity (neurons with dark-labeled cytoplasma [arrows]). Bars = 25 μm. E, Proportion of TNFRI-immunoreactive (top) and TNFRII-immunoreactive (bottom) neurons in lumbar DRG sections from untreated control rats, by neuron size. Black portions of the bars show the proportion of neurons with either TNFRI- or TNFRII-like immunoreactivity. F, Proportions of lumbar DRG neurons with TNFRI-like (left) and TNFRII-like (right) immunoreactivity in immunized rats (day 0) and rats with AIA. Values are the mean and SD. For TNFRI studies, n = 4 on day 0, n = 3 on day 1, n = 4 on days 3, 7, and 21; for TNFRII studies, n = 5 on days 0 and 1, n = 6 on day 3, n = 5 on days 7 and 21.

Proportions of neurons expressing either TNFRI- or TNFRII-like immunoreactivity did not significantly change in the course of AIA (Figure 3F). Real-time PCR analysis of DRGs from both sides revealed, on average, a <0.5-fold change on days 1 and 21 of AIA compared with immunized controls. Thus, TNFRI and TNFRII expression was stable during the course of AIA.

Electrophysiologic recordings from sensory fibers of the knee joint.

Since electrophysiologic recordings from afferent fibers are particularly suitable for monitoring peripheral effects of antinociceptive compounds (6, 43, 44), we recorded from identified joint afferents in anesthetized rats. In these experiments, etanercept and infliximab were injected once into the knee joint. Figure 4A shows typical action potentials in a small nerve fiber bundle during innocuous (left) and noxious (right) outward rotation of a knee joint with inflammation induced using kaolin/carrageenan. Because the fibers on the medial side of the knee joint respond mainly to outward rotation, only responses to this stimulus are further presented.

Figure 4.

Electrophysiologic recordings from C fibers of the medial articular nerve. A, Action potentials (APs) (amplification ∼10,000×) of a small fiber bundle during innocuous and noxious outward rotation of the inflamed knee joint. B, Responses to noxious (nox.) outward rotation and innocuous (innoc.) outward rotation (changes from baseline [0]) of the acutely inflamed knee joint after injection of etanercept into the joint cavity. Initial values were as follows: innocuous rotation 58 ± 8 APs/15 seconds; noxious rotation 244 ± 47 APs/15 seconds. K/C = kaolin/carrageenan-induced inflammation. C, Responses to noxious and innocuous outward rotation of the inflamed knee joint in rats with antigen-induced arthritis (AIA) after injection of etanercept into the joint cavity. Initial values were as follows: innocuous rotation 104 ± 13 APs/15 seconds; noxious rotation 246 ± 21 APs/15 seconds. In B and C, ∗ = first time point at which the reduction of responses became significant (P < 0.05 by Wilcoxon's matched pairs signed rank test). D, Changes in responses to noxious outward rotation of knees in rats in different experimental groups, 90–102 minutes after injection of the drug into the knee. Values are the mean ± SD. ∗ = P < 0.05 versus baseline. Et. Solv. = solvent of etanercept; Inf. = infliximab; Imm. = immunization.

In Wistar rats with acute inflammation induced by kaolin/carrageenan, administration of 100 μg etanercept progressively reduced responses in 8 of 9 C fibers to noxious outward rotation of the acutely inflamed knee joint as early as 30 minutes after injection (Figure 4B), and the reduction became significant 66 minutes after injection (P = 0.0117 by Wilcoxon's matched pairs signed rank test). Responses to innocuous outward rotation showed a transient and nonsignificant reduction. Responses of Aδ fibers were, on average, not influenced by etanercept (no change in 4 fibers, slight increase in 3 fibers, slight decrease in 2 fibers).

In Lewis rats with AIA (day 3), 100 μg etanercept reduced responses to both noxious and innocuous outward rotation of the inflamed knee (P = 0.0313 at 66 minutes for both innocuous and noxious rotation) (Figure 4C). A synopsis of the data from all groups is shown in Figure 4D. Responses to noxious outward rotation of the inflamed knee were reduced with etanercept, but not with the solvent of etanercept. Notably, etanercept did not reduce responses of C fibers in the normal joint or responses of C fibers in immunized rats. Neither IA injection of 200 μg infliximab into the inflamed knee nor IP administration of infliximab significantly reduced responses to rotation during the recording time, although there was a trend toward a decrease in the responses.


In the present study, systemic administration of the TNFα neutralizers etanercept and infliximab significantly attenuated inflammation-induced changes in locomotor and pain-related behavior in rats with AIA. These effects were accompanied by a slight but significant reduction in joint swelling at some time points, but not gross inflammatory changes. Both TNFRI and TNFRII were found to be expressed in a large proportion of DRG neurons during the course of the disease, indicating that primary afferent neurons may be a target of anti-TNF treatment. Intraarticular injection of etanercept reduced responses of joint afferents to rotation of the inflamed joint as soon as 30 minutes after injection, but had no effect on responses to rotation of the normal joint, thus showing antinociceptive effects of TNFα neutralization under inflammatory conditions on a short time scale.

AIA is characterized by both short-term and long-term inflammatory changes (see definition above). Detailed analysis of behavior revealed the most severe primary mechanical hyperalgesia at the knee (27) and secondary thermal hyperalgesia in the first 7 days after AIA induction, but some symptoms, such as knee hyperalgesia and gait abnormalities, persisted up to day 21. The significant reduction of these symptoms with etanercept and infliximab clearly shows that TNFα neutralization has a strong potential to reduce inflammation-evoked pain; the rats did not receive other analgesic compounds.

Notably, the highly significant antinociceptive effects were observed in the presence of marked inflammation (e.g., swelling was significantly reduced with etanercept on days 3 and 7 only, whereas the antinociceptive effects of etanercept were observed at all time points). Infliximab reduced nociception but the reduction of joint swelling was not significant. Histologic grading did not reveal normalization of gross pathology at times when antinociception with etanercept and infliximab was highly significant. In a model of CFA-induced paw inflammation, etanercept administered after onset of inflammation did not attenuate mechanical paw hyperalgesia, but it attenuated thermal paw hyperalgesia without altering gross inflammatory changes, thus indicating similar findings in a different model (23). These data suggest, therefore, that the antinociceptive effect of etanercept and infliximab at an early stage of AIA did not result from a reduction of inflammation per se. They do not exclude the possibility that, over a longer period, reduced progression of RA after neutralization of TNFα (1–5) may be of greater importance for pain relief. Since AIA shows a significant spontaneous attenuation within 21 days, it is difficult to identify treatment effects after a longer delay. Furthermore, we cannot exclude the possibility that either higher doses of etanercept or infliximab or pretreatment with these compounds would have had greater effects on inflammation. However, the present study clearly shows that TNFα neutralization can be antinociceptive at doses that do not abolish inflammation.

Two major findings from the present study strongly support the idea that the antinociceptive effect of etanercept results at least in part from an action on primary afferent neurons. First, the expression of TNFRI- and TNFRII-like immunoreactivity in a large proportion of small- and medium-sized DRG neurons throughout AIA suggests that neurons are a target of TNFα. Whether both TNFRI and TNFRII contribute to the neuronal effects of TNFα is unknown. Data from studies on cultured DRG neurons (14) and on neuropathic pain (45–47) suggest that primarily TNFRI is involved in the nociceptive effects of TNFα. Second, IA injection of etanercept rapidly reduced responses of joint afferents to movements of inflamed joints within 30 minutes, thus showing that neutralization of TNFα in the inflamed joint is antinociceptive on a short time basis. After administration of infliximab, we observed a trend toward a decrease in responses, but the effect did not become significant within the recording time.

The results from the electrophysiologic study provided several important insights (because infliximab had overall weaker effects than etanercept, we focused on etanercept in the recordings). First, etanercept reduced C fiber responses to mechanical stimulation in different rat strains and different models of inflammation, thus showing the general significance of its antinociceptive effect. Second, etanercept reduced only responses of C fibers to stimulation of the inflamed knee. In rats with AIA, even responses to innocuous rotation were reduced, which is consistent with the findings in the behavioral study. In contrast, responses to joint stimulation in normal rats and in immunized rats were not influenced. This pattern suggests that TNFα signaling is particularly important under inflammatory conditions. Presumably TNFα is acting together with other mediators that are present mainly during inflammation (9). Another consequence of interfering with TNFα effects at peripheral nociceptors could be a reduction of neurogenic inflammation, which is mediated by the release of neuropeptides from nerve endings. Perhaps such an effect is the reason for the attenuation of joint swelling.

The appearance of long-lasting thermal hyperalgesia at the paw in the course of AIA can be interpreted as secondary hyperalgesia. Hypersensitivity to noxious stimuli in an area remote from an inflamed or injured area is a well-known phenomenon, both in humans (9, 48) and in experimental animals (9). It is thought to be produced by a neuronal process called “central sensitization,” which displays the development of hyperexcitability of spinal cord neurons during a peripheral pathologic process (49). In fact, in the rat spinal cord many neurons with input from the knee also have receptive fields in the ankle and muscles of the hind limb. During the development of local inflammation in the knee, spinal neurons become more responsive to stimuli applied to the knee and to remote tissue (even including the paw), showing the development of spinal hyperexcitability (6). Neutralization of TNFα with systemic etanercept significantly reduced thermal hyperalgesia of the paw on days 7 and 14, indicating that secondary hyperalgesia is also attenuated. Whether this effect is entirely due to a reduction of input from the inflamed joint or whether TNFα neutralization at other sites, including the central nervous system, contributes to such effects will be addressed in future studies.

In conclusion, the present study demonstrates pronounced antinociceptive effects of TNFα neutralization in a model of knee inflammation. This effect resulted from a neuronal site of action rather than from a reduction of inflammation. However, TNFα neutralization did not fully reverse inflammation-evoked nociception. Therefore, additional treatment with other antinociceptive drugs is necessary to achieve better pain relief.


Dr. Schaible had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study design. Boettger, Richter, Segond von Banchet, Bräuer, Schaible.

Acquisition of data. Boettger, Hensellek, Richter, Gajda, Stöckigt, Segond von Banchet.

Analysis and interpretation of data. Boettger, Hensellek, Richter, Gajda, Segond von Banchet, Bräuer, Schaible.

Manuscript preparation. Boettger, Richter, Segond von Banchet, Schaible.

Statistical analysis. Boettger, Richter.


We thank Mrs. Gaby Cuny, Mrs. Konstanze Ernst, Mrs. Sabine Hitschke, Mrs. Cornelia Hüttich, and Mrs. Antje Wallner for excellent technical assistance, Dr. Anja Wölfert (AW), Mrs. Konstanze Weber, and Mr. Raphael Hilgenstock for help with the experiments, and Mrs. Claudia Hemmelmann for statistical advice.