Cytotoxic T cells modulate inflammation and endogenous opioid analgesia in chronic arthritis

Female BALB/c mice were purchased from Charles River (Sulzfeld, Germany) and Janvier (France). Mice were used at 10–14-weeks of age. All animal studies were approved by the state ethics committee (Landesamt für Gesundheit und Soziales, Berlin, Germany) and performed according to institutional and state guidelines, under specific pathogen-free conditions.

BALB/c mice were immunized by subcutaneous (s.c.) injection of 100 μg methylated bovine serum albumin (mBSA, Sigma-Aldrich, Schnelldorf, Germany) in 50 μl PBS emulsified with 50 μl complete Freund’s adjuvant (CFA, Sigma-Aldrich), followed 1 week later by a second s.c. immunization with mBSA plus bovine type II collagen in incomplete CFA [9]. One week later, mice were immunized s.c. with 50 μg mBSA and 100 μg bovine collagen type II (mdbioproducts, Zurich, Switzerland) in 50 μl PBS emulsified with 50 μl Freund’s incomplete adjuvant (IFA, Sigma-Aldrich). In parallel to each immunization step, 200 ng of Bordetella pertussis toxin (PTx, Calbiochem, La Jolla, CA, USA) were injected intraperitoneally (i.p.). Fourteen days later, on day 0, arthritis was induced by intraarticular (i.a.) injection of 50 μg mBSA dissolved in 20 μl PBS into one knee joint (ipsilateral) during brief isoflurane (Abbot, Wiesbaden-Delkenheim, Germany) anesthesia. The same volume of the vehicle (PBS) was injected into the contralateral knee. An “induction-only” control group of animals were not immunized with s.c. mBSA, but instead received s.c. PBS emulsified with adjuvant. On day 0, mBSA was injected into the ipsilateral and PBS into the contralateral knee joint.

To deplete cytotoxic T cells (CTL), ACIA mice received i.p. injections of 90 μg anti-CD8 [25] (Leinco Technologies, Inc., St. Louis, MO, USA), diluted in 200 μl sterile saline. The antibody was injected every 9–10 days starting 3 days before i.a. mBSA injection. Control animals received the isotype-matched control antibody (Leinco Technologies, Inc.). The amount of antibody was titrated so that depletion would last until the end of the study period (i.e., 60 days after arthritis induction).

Animals were sacrificed and both knee joints were fixed in 4% buffered formaldehyde, decalcified with EDTA, and embedded in paraffin. Serial sections (4–5 μm thick) were cut and stained with hematoxylin and eosin (H&E) for microscopic evaluation. Scoring of the knee sections was performed in a blinded manner in three to four sections per knee joint. Two different scores were used, one for the severity of inflammation and one for the degree of destruction. Inflammation was scored by the degree of synovial hyperplasia (0–3 points), the amount of mononuclear cells infiltrating the synovial membrane (0–3 points), and fibrosis (0–3 points). Destruction of the joint was scored according to the level of cartilage and bone destruction (0–3 points) and pannus formation (0–3 points). Zero points in any of the categories corresponded to the absence of the parameter, 3 points indicated the strongest pathological phenotype, and points for each parameter were added up to the overall score. The experimenter was not aware of the preceding treatment of the animals.

Local endogenous opioid-mediated pain control was evaluated by injecting i.a. naloxone-methiodide (NLXM) (1 μg/10 μl) under isoflurane anesthesia. Control mice received solvent. The effect on heat sensitivity was assessed at 15 and 30 min after drug application using the Hargreaves test.

For the analysis of serum components, mice were bled retro-orbitally at regular intervals. Analysis of antigen-specific IgG titers or cytokine levels in the serum was performed with capture ELISA. To measure mBSA-specific IgG, immunostep microtiter plates (Nunc, Langenselbold, Germany) were coated with 2 μg per well mBSA in PBS and washed and incubated with diluted serum. Bound serum antibodies were detected using horseradish peroxidase-conjugated rabbit anti-mouse serum (Southern Biotech, Eeling, Germany). For the detection of ACPA, the anti-MCV (mutated citrullinatedvimentin) ELISA (OrgentecDiagnostica, Mainz, Germany) was used. Serum TNFα level was determined with the mouse TNFα DuoSet ELISA (R&D systems). To detect the IL-17A level in the serum, the purified TC11-18H10.1 antibody (BioLegend) served as capture antibody, in conjunction with the biotinylated TC11-8H4 (BioLegend) as detection antibody and recombinant murine interleukin 17 (IL-17) (BioLegend).

Ipsilateral knee joints from CTL-depleted and isotype-treated arthritic mice were dissected in the late chronic phase, i.e., >day 60. Knee joints were cut in half and subjected to enzyme digestion using RMPI-1640 medium supplemented with 3 mg/ml hyaluronidase and 1 mg/ml collagenase II (both Sigma-Aldrich) for 30 min at room temperature. The knee joint was grinded, and tissue homogenates were filtered twice using 70- and 40-μm cell strainers, respectively. Cell suspensions were washed in pure RMPI medium, spun down for 10 min at ×350g, and resuspended in 250 μl of pure medium. Cells were kept for 1.5 h at 37 °C under agitation; the supernatants were harvested after spinning and kept at −80 °C. The amounts of liberated opioid peptides were determined using a Methionine-enkephalin radioimmunoassay (RIA, Bachem, Peninsula Laboratories, USA) and a fluorescent human/rat beta-endorphin enzyme immunoassay (EIA, Phoenix Peptides) as previously described [29]. The EIA kit is validated by the manufacturer to be 100% cross reactive for murine endorphin. For RIA, 100 μl of supernatants were analyzed in duplicates. The remaining supernatants were analyzed in duplicate using EIA. Both assays are suitable to detect opioid peptides at the pg-level.

All data were documented using Microsoft Excel 2003 for Windows (Microsoft Corporation). GraphPad Prism Version 5.01 for Windows (GraphPad Software, Inc.) software was used for statistical analyses. Data were analyzed for equal variance and for normal distribution using the D’Agostino & Pearson omnibus normality test or Kolmogorov-Smirnov test. Normally distributed independent data were analyzed by the Student t test and nonparametric independent data by the Mann-Whitney U test if only two groups were compared. To compare two groups of normally distributed dependent data, the paired t test was used. In case of nonparametric dependent data, the Wilcoxon test was applied. Multiple measurements of normally distributed data were analyzed by the one-way repeated measurements ANOVA or by the Kruskal-Wallis test in case of not normally distributed data. Post hoc comparisons were performed by Bonferroni’s multiple comparison test for normally distributed data and by Dunn’s method for not normally distributed data. Two-way repeated measurements ANOVA was used for two factor analyses followed by Bonferroni’s multiple comparison test. Statistical significance was usually considered if P < 0.05. For the analysis of repeated measurements, a Bonferroni correction was performed in MS Excel to calculate t values with a stepwise bisection of alpha starting with 0.05 according to the equation t = TINV (alpha; degrees of freedom). The residual degrees of freedom (DF), as calculated in the test by GraphPad Prism V5, was used for these calculations. Then, t values obtained from the Bonferroni multiple comparison tests were compared to the corrected t values. If t _{(Bonferroni’s multiple comparison test)} was larger or equal to the corrected t value, the differences were considered significant. Correlations of data were performed using the Spearman correlation.

Once the mice had reached the chronic phase of arthritis, i.e., day 60 after induction, their joints were analyzed histologically by hematoxylin and eosin (H&E) staining.

In addition to the comparison of ipsilateral (i.a. mBSA induction) vs. contralateral (i.a. PBS) joints of the same animals, we also examined a control group (induction-only) that was not immunized s.c. but had only received i.a. mBSA (tissue sections Fig. 1a–d). As expected, the induction-only mice had significantly lower inflammation scores when compared with ACIA mice (Fig. 1e, left panel). The incidence of scores ≥4 was 100% in ACIA mice, but less than 50% in induction-only mice. Higher scores (≥6) were observed in 45% of the ACIA mice (N = 11) but only in 11% of induction-only animals (N = 9).

Next, we compared ipsilateral with contralateral (i.a. PBS) knee joints. The latter also developed inflammation in later stages of the disease. Ipsilateral joints showed higher severity scores than contralateral ones, but these differences were not statistically significant.

Scores for joint destruction were highest for the ipsilateral knees of ACIA mice (Fig. 1e, right panel). Scores ≥4 were observed in 42% of the ACIA mice (N = 7) but only in 14% of induction-only animals (N = 6).

Throughout the acute and chronic phases of joint inflammation (Fig. 2), ACIA mice showed significantly lower mechanical thresholds in the von Frey test in ipsilateral (i.a. mBSA) compared to contralateral (i.a. PBS) hind limbs (Fig. 2a). Animals that were not previously immunized, but received only i.a. ipsilateral mBSA (induction-only), showed a similar difference during the acute phase of inflammation; however, this effect decreased as the inflammation subsided (Fig. 2a).

As shown in Fig. 2b, in ACIA mice, the ipsilateral paw withdrawal latencies to heat were significantly lower than the contralateral ones at all time points. In induction-only animals, they were also significantly lower during early joint inflammation, but not at later time points (days 23–56).

The paw withdrawal latencies elicited by heat significantly decreased with increasing severity and destruction scores (Fig. 2c). In contrast, paw withdrawal thresholds elicited by mechanical stimuli did not show a correlation with parameters of inflammation or destruction (Fig. 2d).

In summary, the ACIA protocol induced hyperalgesia and allodynia for at least 60 days after arthritis induction and appears therefore suitable to study chronic pain conditions in arthritis.

To investigate the role of endogenous opioids, we examined the effect of the opioid receptor antagonist NLXM. This compound does not cross the blood brain barrier [30] so that centrally mediated effects can be excluded.

ACIA and non-immunized mice received ipsilateral i.a. NLXM in the chronic stage of arthritis (>day 60). Following NLXM treatment, induction-only mice showed no differences in paw withdrawal latencies to mechanical stimulation between ipsi- and contralateral sides, whereas paw withdrawal latencies to heat were significantly reduced in ipsilateral limbs of ACIA mice as compared to baseline thresholds (Fig. 3). Thus, endogenous opioids seem to decrease heat hyperalgesia in the chronic stage of our arthritis model.

The role of CD8⁺ T cells in chronic arthritis is controversial [31, 32]. We depleted these cells 3 days before i.a. injection of mBSA with a CD8-specific antibody and repeated this treatment every 9 days. Depletion efficacy was continuously surveyed by counting CD8⁺ cells in the peripheral blood by flow cytometry and was compared to mice receiving an isotype-matched control antibody. Animals receiving the anti-CD8 antibody displayed very low numbers of CD8⁺ T cells in comparison to mice treated with the control antibody at all time points (Fig. 4a). There were no significant differences in baseline nociceptive thresholds between CD8⁺-depleted and control animals (Fig. 4b, c).

Both groups received i.a. NLXM either 16 or 60 days after arthritis induction (Fig. 5a). At both time points, arthritic mice treated with isotype-control antibody showed significantly decreased thermal thresholds following NLXM injection, indicating exacerbated hyperalgesia. In CD8⁺-depleted animals, however, NLXM had no effect. These findings suggest an involvement of CD8⁺ cells in the fine tuning of inflammatory pain by the release of endogenous opioids at both stages. The amount of Met-enkephalin released from joint cells of CD8⁺-depleted animals was significantly lower than in isotype control-treated animals (Fig. 5b left panel), while cellular contents showed no statistically significant differences (Fig. 5b right panel). The amount of released beta-endorphin did not differ between the treatment groups (Fig. 5c).

In accordance with our previous results, both control antibody-treated as well as CD8⁺-depleted mice developed chronic arthritis in the later observation phases (Fig. 6a, b). Mean scores of inflammation (Fig. 6a) or destruction (Fig. 6b) did not differ statistically between the groups. Also, serum levels of arthritis markers like ACPA or antibodies against the inducing agent mBSA were not different between CD8⁺-depleted and control animals (Fig. 6c). However, serum levels of the pro-inflammatory cytokines TNFα and IL-17 were significantly increased in CD8⁺-depleted mice in the chronic stages of the disease (Fig. 6d).