Differential binding of chemokines to macrophages and neutrophils in the human inflamed synovium

Tissue samples were obtained from patients with RA (n = 6) who fulfilled the American Rheumatism Association criteria for RA (Supplementary Table 1). Synovia were taken from these subjects at total knee replacement (patients 1, 3–5) or knee synovectomy operations (patients 2 and 6). Non-RA patients (n = 6) had knee joint symptoms for suspected articular cartilage, meniscal or anterior cruciate ligament damage (patients 7–12). Knee joint synovial biopsies were obtained from these non-RA individuals, with informed consent, at arthroscopy. For RA and non-RA patients, samples were taken from the suprapatel-lar pouch and medial gutter, which is reported to provide representative sampling of synovial membrane pathology [18].

The method of Hub and Rot [19] was followed, with minor modifications (see Supplementary Material).

Representative sections of synovia that had been incubated with ¹²⁵I-labeled chemokines were immunostained for CD68 as a marker for macrophages. Slides were deparafinized, rehydrated and treated with 1% pronase for seven minutes at room temperature. Following blocking of endogenous peroxidase, sections were treated with normal horse serum and incubated with 2.5 μg/ml anti-CD68 monoclonal antibody (PG-M1, Dako A/S, Glostrup, Denmark) in PBS/normal horse serum. The bound antibody was detected using biotinylated second antibody, avidin-peroxidase conjugate and diaminobenzidine substrate (Vectastain ABC Elite Kit; Vector Labs, Burlingame, USA). Slides were then coated with emulsion, exposed, developed and stained in haematoxylin, as for in situ binding autoradiography.

Neutrophils were identified by their characteristic multi-lobed nuclei in haematoxylin and eosin stained sections.

Synovial samples that had been examined for chemokine binding were also sectioned and incubated with rabbit polyclonal antibodies to the carboxy termini of CXCR1 and CXCR2 (both from Santa Cruz Biotechnology, Santa Cruz, USA) [20]. Antibody binding was detected using an immunoperoxidase method (Vectastain ABC Elite Kit; Vector Labs, Burlingame, USA). See Supplementary Material for more detailed methods.

See Supplementary Material.

In all six RA subjects the histological appearance of synovia showed classic inflammatory pathology with mononuclear cell infiltrates and a thickened intimal layer. In two of these individuals a minority of synovial samples also showed microscopically normal regions with minimal infiltration of inflammatory cells and an intimal layer of normal thickness (one to two cells thick). The synovia of the six non-RA patients showed inflamed and noninflamed regions. These inflamed regions displayed intimal layer thickening and some infiltration of mononuclear cells, although not as severe as in RA samples, and noninflamed sites showed a normal intimal layer and no leukocyte infiltrates.

To examine in situ binding of chemokines, synovia were treated with ¹²⁵I-CXCL8, ¹²⁵I-CCL2, ¹²⁵I-CCL3 and ¹²⁵I-CCL5 followed by autoradiogaphy. For direct comparison, each chemokine was added at the same specific radioactivity and concentration, and autoradiography was performed for an identical exposure time. All four chemokines bound to synovial tissue cells in the subintima and the intima in every RA and non-RA individual (Supplementary Fig. 1a for CCL5). Radiolabelled chemokine binding was absent in synovia treated with 1000-fold excess homologous non-radiolabelled chemokine, indicating that the binding sites were specific (Supplementary Fig. 1b). Histological examination indicated that macrophages accounted for a high proportion of the chemokine binding, so to confirm this observation, representative sections were immunostained with CD68 as a marker of monocytes/macrophages, then processed for autoradiogaphy. All four chemokines bound to CD68⁺ cells in the subintima in every RA and non-RA synovial sample. No major differences in chemokine binding were apparent between the six RA patients, despite variation in disease duration and drug treatment (Supplementary Table 1). The positive cells were located close to blood vessels in leukocyte infiltrates and non infiltrated sites, and in more remote sites within the subintimal stroma (Fig. 1). In addition CD68⁺ macrophages bound all four chemokines in the intimal layer of inflamed (Fig. 1) and noninflamed samples. In control experiments there was no immunostaining when isotype-matched control mouse IgG was used instead of the anti-CD68 antibody.

The number of intimal cells that bound chemokines were counted and the results for RA and non-RA patients were expressed as means (Table 1). There was a significant increase of chemokine-positive cells in inflamed intima compared to noninflamed intima, for both RA and non-RA synovia. The number of positive cells varied between chemokines and in RA tissue the rank order was, CCL3 > CCL5 > CCL2 > CXCL8. In non-RA tissue the order was, CCL3 > CCL5 > CXCL8 > CCL2. The number of CCL3 and CCL5 positive intimal cells did not differ between RA and non-RA tissue. However, the number of intimal cells binding CCL2 in RA was greater than non-RA synovium. In addition, there were more cells binding CXCL8 in inflamed, non-RA intima compared to inflamed RA tissue.

Data were also expressed as the percentage of intimal cells binding chemokines (Supplementary Table 2). Similar results to data expressed as the number of cells binding chemokines were obtained (Table 1 and Supplementary Table 2). In both RA and non-RA synovia there was a significantly higher percentage of positive cells in inflamed compared to noninflamed intima. The rank order of chemokine binding was almost identical whether the data were expressed as percentage or number of cells binding chemokines. In the cases of CCL2, CCL3 and CCL5 there were no differences in the percentages of positive cells between RA and non-RA tissue. A higher percentage of cells, however, bound CXCL8 in non-RA inflamed intima compared to RA inflamed intima.

Neutrophils were present in the synovia of four out of the six RA patients, but they were less abundant than mononuclear cells. Neutrophils were absent in non-rheumatoid synovia. Neutrophils were positive for ¹²⁵I-CXCL8 binding (Fig. 2) which was displaced in the presence of 1000-fold excess non-radiolabelled chemokine. These cells bound IL-8 in all locations within the synovium: within blood vessels; close to blood vessels in the extravascular tissue; and close to (or at) the intimal surface. Neutrophils did not show any signal for binding CCL2, CCL3 and CCL5 (Fig. 2). No major differences in neutrophil chemokine binding were apparent between the four patients who had these cells in their synovia.

To define the molecular identity of the ¹²⁵I-CXCL8 binding sites on synovial macrophages and neutrophils, immunohistochemistry was performed to examine expression of CXCR1 and CXCR2 by these cells. Sections from the same blocks of synovia used for radiolabelled chemokine binding were treated with antibodies to CXCR1, CXCR2 and CD68. Samples from every RA and non-RA patient showed expression of both CXCR1 and CXCR2 by macrophages in the subintima and intima (Supplementary Fig. 2 and Fig. 3). Similarly, in RA samples that possessed neutrophils, immunoreative CXCR1 and CXCR2 was displayed by these cells throughout the subintimal and intimal layers; the staining intensity for CXCR2 was more strong than for CXCR1 (Fig. 3c). Sections treated with rabbit IgG instead of CXCR1 or CXCR2 were negative for macrophage and neutrophil staining (Fig. 3d).

Synovia were cut into approximately 2 mm³ pieces. Four samples were placed in tubes containing 0.1 to 4.0 ng of recombinant human ¹²⁵I-labeled (rh ¹²⁵I)-CXCL8 (M_r = 8400), rh ¹²⁵I-CCL5 (M_r = 8000), rh ¹²⁵I-CCL2 (M_r = 8800) or rh ¹²⁵ I-CCL3 (M_r = 8000) (262 μCi/μg; NEN, Boston, USA) in 400 μl Hank's balanced salt solution containing 0.1% bovine serum albumin and 20 mM HEPES. The samples were then incubated for one hour at room temperature on a platform shaker [19]. In these experiments, synovial samples from each patient were treated with each of the four chemokines. Initial experiments showed that binding sites were saturable over this concentration range and the optimum binding/background ratio was obtained using 0.8 ng of chemokine; this amount was used in all the subsequent experiments. Control tubes contained 1000-fold excess of the corresponding unlabeled recombinant human chemokine (molecular mass as for radiolabelled chemokines) (PeproTech EC, London, UK; CXCL8 was kindly donated from Novartis Research Institute, Vienna, Austria). The tissue was washed four times for 10 min in 2 ml Hank's balanced salt solution on a platform shaker, then fixed overnight in 4% paraformaldehyde. It was then dehydrated in alcohol, embedded in fibrowax and sectioned at 5 μm thickness. Sections were deparafinized, rehydrated, air-dried and coated with Amersham LM1 emulsion (Amersham Pharmacia Biotech, Little Chalfont, UK). The slides were exposed at 4°C for two weeks, developed and stained in haematoxylin and eosin.

Sections were deparafinized, rehydrated and in the case of CXCR1 pre-treated with 1% pronase for 10 min to unmask epitopes. Following blocking of endogenous peroxidase with 0.3% hydrogen peroxide in methanol for 30 min, slides were rinsed in PBS and treated with a 1/60 dilution of normal goat serum for 30 min. Sections were rinsed in PBS, incubated with CXCR1 or CXCR2 antibodies (at 1/50 and 1/500 dilution respectively) or rabbit IgG (at the same concentration as primary antibodies) for 30 min. After washing with PBS, the antibodies were detected using reagents in the Vectastain ABC Elite Kit (Vector Labs, Burlingame, USA); the sections were treated with biotinylated goat anti-rabbit IgG (at 1/200 dilution) for 30 min, rinsed with PBS, incubated with streptavidin-conjugated peroxidase for 30 min, washed with PBS, then treated with the chromogen 3,3'-diaminobenzidine for 5 min. Sections were washed twice in PBS then water, dried and counter-stained in Mayer's haematoxylin. All steps in the method were carried out at room temperature.

From these data, the number and percentage of positive cells in the intimal samples were obtained and expressed per 500 μm of intima. Statistical analyses were by Student's t-test, The result was significant when P < 0.05.