LOCALIZATION OF TUMOR NECROSIS FACTOR a IN SYNOVIAL TISSUES AND AT THE CARTILAGE-PANNUS JUNCTION IN PATIENTS WITH RHEUMATOID ARTHRITIS

Transcription

1 1125 LOCALIZATION OF TUMOR NECROSIS FACTOR a IN SYNOVIAL TISSUES AND AT THE CARTILAGE-PANNUS JUNCTION IN PATIES WITH RHEUMATOID ARTHRITIS C. Q. CHU, M. FIELD, M. FELDMANN, and R. N. MAIN1 Using immunoaffinity-purified polyclonal antihuman recombinant tumor necrosis factor a (TNFa) F(ab ), fragments and immunohistochemical techniques, the cells that make TNFa were localized in the inflamed synovial tissue of patients with rheumatoid arthritis (RA) and osteoarthritis (OA). Anti-TNFa antibody-stained cells were demonstrated in 9 of 11 RA and 2 of 4 OA but none of 5 normal synovial membranes examined. In RA, 2644% of the lining layer cells were positive for TNFa. In the interaggregate area, 1630% of the cells contained TNFa, often in a perivascular distribution, and up to 19% of the cells in lymphoid aggregates stained for TNFa. Some endothelial cells also stained with these antibodies. In OA tissues, the TNFacontaining cells were found predominantly in the deeper layer. Cells containing TNFa were also found at the cartilage-pannus junction in all 4 RA specimens examined. Double immunofluorescence analysis demonstrated that most TNFa-secreting cells in the RA synovial membrane expressed the monocyte/macrophage marker antigens CDllb and CD14, and a few expressed From the Division of Clinical Immunology, The Mathilda and Terence Kennedy Institute of Rheumatology, Hammersmith, and the Charing Cross Sunley Research Centre, London, England. Supported by the Arthritis and Rheumatism Council and by the Nuffield Foundation. Dr. Chu s work is supported by the Sino-British Friendship Scholarship Scheme. C. Q. Chu, MB BS: Research Fellow, Kennedy Institute of Rheumatology; M. Field, BSc, MD, MRCP: Lecturer, Kennedy Institute of Rheumatology; M. Feldmann, PhD, FRCPath: Professor, Charing Cross Sunley Research Centre; R. N. Maini, FRCP: Director, Kennedy Institute of Rheumatology. Address reprint requests to M. Field, BSc, MD, MRCP, Division of Clinical Immunology, The Mathilda and Terence Kennedy Institute of Rheumatology, 6 Bute Gardens, Hammersmith, London W6 7DW, UK. Submitted for publication August 28, 1990; accepted in revised form April 8, the T cell marker CD3. Our findings provide histologic evidence that TNFa is locally produced in the lining and deeper layers of the synovium by cells of the monocyte/ macrophage lineage, supporting its role in inflammation. Further, our findings demonstrate that TNFa is produced by cells at the cartilage-pannus junction, which could affect chondrocyte metabolism, leading to the cartilage degradation in RA. Human tumor necrosis factor a (TNFa) is a 17-kd nonglycosylated protein (1) originally defined by its ability to induce hemorrhagic tumor necrosis in animals (2). However, TNFa can also act on other cells (3) that regulate the immune and inflammatory responses, thereby causing damage to connective tissue. These properties suggest that TNFa may contribute to the pathogenesis of the synovitis and joint destruction in rheumatic diseases by stimulating fibroblast growth (4), inducing prostaglandin E, and collagenase release from synovial cells (5), inducing resorption of bone and cartilage (6,7), and inhibiting synthesis of proteoglycan in cartilage (8). In addition, it activates endothelial cells, which leads to induction of procoagulant activity (9) and increased adherence of granulocytes (10) and mononuclear cells (1 1). TNFa also regulates interleukin-1 (IL-1) production not only by endothelial cells (12) but also by mononuclear cells of the rheumatoid synovial membrane (13). Because both TNFa and IL-1 can induce synovitis and joint destruction (3), we have postulated that TNFa plays a pivotal role in perpetuating the disease process in rheumatoid arthritis (RA). TNFa has been detected in synovial fluid from patients with various arthritides (14-16), and messenger RNA for TNFa has been detected in synovial Arthritis and Rheumatism, Vol. 34, No. 9 (September 1991)

2 1126 CHU ET AL cells, which suggests that there is local production of the cytokine in inflamed tissues (17). After stimulation of peripheral blood mononuclear cells in vitro, macrophages are the major source of TNFa (18), but cultured T cells can also secrete TNFa after appropriate stimulation (19,20). In the inflamed synovial tissue, where macrophages and T cells are found, the localization and cellular source of TNFa remains obscure. Using immunoaffinity-purified polyclonal anti- TNFa F(ab ), fragments and an immunohistochemical technique, we localized TNFa-secreting cells in the synovial tissues of patients with RA and osteoarthritis (OA) and identified the macrophage as the major source of TNFa in RA synovial membranes. PATIES AND METHODS Recombinant cytokines. Recombinant human TNFa (rhutnfa) and recombinant human lymphotoxin (rhult) were obtained from Genentech (San Francisco, CA). Recombinant human IL-la (rhuil-la) was donated by Hoffmann-La Roche (Nutley, NJ), and rhuil-6 was a generous gift from Prof. Kishimoto (Osaka, Japan). Preparation of rabbit anti-tnfa antibody. New Zealand lop-eared rabbits were immunized subcutaneously with 30 pg of rhutnfa emulsified in an equal volume of Freund s complete adjuvant (Difco, Detroit, MI) and were given a booster injection with 30 pg of rhutnfa in Freund s incomplete adjuvant (Difco) at two 1-week intervals. Serum IgG was separated by staphylococcal-protein A (Pharmacia, Milton Keynes, Buckinghamshire, UK) chromatography and digested with pepsin (Sigma, Poole, Dorset, UK) for 20 hours at 37 C at an 1gG:pepsin ratio of 50: 1 in 0.1M sodium acetate buffer, ph 4.0. Intact IgG and Fc fragments were then removed by protein A chromatography. The anti-tnfa F(ab ), fragments were purified by elution from a column made with rhutnfa linked to cyanogen bromideeactivated Sepharose 4B (Pharmacia) using 3M guanidine (Sigma). These were then labeled with biotin (Pierce, Rockford, IL) as previously described (21). Normal rabbit F(ab ), fragments were obtained from normal rabbit IgG by pepsin digestion followed by Sephacryl S200 (Sigma) column separation. These fragments were also labeled with biotin (21). Confirmation of immunoreactivity. Immunoreactivity was verified by 2 methods. An enzyme-linked immunosorbent assay (ELISA) was used to demonstrate antibody specificity. Recombinant human TNFa or other cytokines were coated onto 96-well plates (Falcon, Oxnard, CA) at 1 pg/ml and stored at 4 C until used. The plate was washed, and nonspecific binding was prevented by incubating with 2% casein for 1 hour at 37 C. Rabbit serum, purified rabbit IgG, or immunopurified F(ab ), fragments labeled with biotin were added to the plates and incubated for 1 hour. The plate was washed, and bound antibodies were detected with anti-rabbit IgG conjugated with alkaline phosphatase (Sig- Table 1. Clinical, serologic, and therapeutic data of patients whose synovial membranes were examined for the presence of tumor necrosis factor a* Rheumatoid arthritis II Osteoarthritis Other inflammatory arthropathies Hemo- ESR Disease globin (mm/ RAHA duration Drug (gddl) hour) (titer) (years) therapy ,280 5,120 1, ,280 1,280 2, NR NR 5 Pred. None Pred. DP Pred., AZA Pred., MTX Pred. DP Pred. Pred., ssz Pred. * Synovium was obtained from the hip or knee joint at synovectomy or arthroplasty in patients I, 3-8, and 12-14, and from the knee joint at arthroscopy in all other patients. Patients 5 and 18 were rheumatoid factor positive (by latex fixation), but the rheumatoid arthritis hemagglutination (RAHA) titer was not significant (< 1: 160). All patients were taking nonsteroidal antiinflammatory drugs and/or a simple analgesic. ESR = erythrocyte sedimentation rate; Pred. = prednisolone; = not tested; DP = D-penicillamine; = negative; AZA = azathioprine; MTX = methotrexate; NR = no record; SSZ = sulfasalazine (Salazopyrine). ma), or avidin-alkaline phosphatase (Sigma) where appropriate, and incubated as described above. This was followed by a 1-hour incubation with p-nitrophenylphosphate (disodium salt; Sigma) substrate. The optical density was measured at 405 nm using a microplate reader. With the second technique, cells from the human T lymphoma cell line, HUT78, which secretes TNFa (20), were stimulated with phytohemagglutinin/phorbol myristate acetate (1 pg/ml and 50 nghl). Cells were harvested onto glass slides with a cytospin (Shandon Southern Instruments, Runcorn, Cheshire, UK), at 1,200 revolutions per minute, and then fixed in acetone:methanol (1:l) at -20 C. Subsequent procedures were performed at room temperature in a humidity chamber. Cells were incubated with 20% normal human serum (NHS) for 20 minutes to prevent nonspecific adherence. Biotinylated anti-tnfa F(ab ), fragments (5 pg/ml) were added, incubated for 60 minutes, and the slides

3 TNFa IN RA SYNOVIUM AND PANNUS 1127 A B C Figure 1. Distribution of tumor necrosis factor a (TNFatpositive cells in synovial membranes from patients with rheumatoid arthritis. Using anti-tnfa F(ab ), fragments, as described in Patients and Methods, TNFa-containing cells were found A, mainly in the lining cell layer, B, within cell aggregates, and C, in a perivascular distribution (upper arrow), with some endothelial cell staining (lower arrow). No cells stained with normal rabbit F(ab ), fragments (D). (Original magnification X 80 for A and D, and x 320 for B and C.:I D washed. Binding was detected with a streptavidinfluorescein conjugate (Amersham International, Little Chalfont, Buckinghamshire, UK). Washed cells were examined under ultraviolet light and photographed. The specificity of antibody binding to the cells and synovial sections was confirmed by: (i) overnight incubation of the anti-tnfa F(ab ), fragments with rhutnfa or rhuil-6 at 4 C before testing; (ii) passage of purified anti- TNFa F(ab ), fragments through cyanogen bromide- Sepharose 4B columns linked to rhutnfa or rhuil-la, and measurement of activity in the drop-through from each column; and (iii) assessment of biotinylated normal rabbit IgG F(ab ), fragments at equivalent protein concentration. Tissues from study patients. Synovial tissue was obtained from the hip or knee joint at arthroplasty or from the knee joint at arthroscopy from 11 patients with RA who fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1958 diagnostic criteria for RA (22). Tissue was also obtained from 4 patients with OA and from 4 patients with other inflammatory polyarthropathies, 2 of whom had arthropathy in association with hepatitis (1 of unknown cause and 1 with primary biliary cirrhosis), 1 of whom had juvenile chronic arthritis, and 1 of whom had arthritis following bacille Calmette-Guerin instillation into the bladder for treatment for carcinoma (23). Tissue samples were frozen and stored at -70 C until used. Table I gives some of the clinical features of the study patients. Normal synovial tissues were obtained from the knee joint of patients undergoing amputation for osteogenic sarcoma at a site distant from the knee joint (kindly provided by Dr. M. Bayliss of the Kennedy Institute). Tissue from the cartilage-pannus junction was taken from the joints obtained from 4 RA patients at arthroplasty. Samples were stored frozen at -70 C until used. Immunohistochemical techniques. Cryostat sections (5-44 were fixed and incubated with 20% NHS as described above. Sections to be stained with peroxidase were treated for 30 minutes with 0.3% hydrogen peroxide in methanol. Biotinylated anti-tnfa F(ab ), fragments at 5-10 pg/ml were incubated with the sections for 60 minutes, and the slides were washed. Binding was detected by streptavidinhorseradish peroxidase conjugate (Amersham International). Diaminobenzidine in phosphate buffered saline containing 0.03% hydrogen peroxide was subsequently added, and after staining, the sections were counterstained with hematoxylin and examined.

4 1128 CHU ET AL For double immunofluorescence staining, monoclonal cell marker antibodies, anti-cd3 (Becton Dickinson, Mountain View, CA) for T cells, and 63D3 and anti-cd14 (donated by Dr. P. C. L. Beverley, UCL, London) for macrophages, were detected with fluorescein-conjugated goat anti-mouse IgG (Sigma). Fluorescein-conjugated F(ab ), fragment of goat anti-human F(ab ), Ig (ICN, Lisle, IL) was applied to detect B cells and plasma cells containing immunoglobulin. In all these experiments, biotinylated anti- TNFa F(ab ), fragments were detected with Texas red linked to streptavidin (Amersham International). Statistical analysis. The numbers of TNFa-staining cells were graded by counting up to 200 cells in each area of the tissue examined. The data were analyzed using Wilcoxon s rank sum test. RESULTS Antibody reactivity and specificity. Rabbit anti- TNFa antibody bound specifically to rhutnfa in the ELISA, but failed to bind to rhuil-la, rhuil-6, and rhult. Immunopurified F(ab ), fragments produced a pattern of cytoplasmic staining on stimulated HUT78 cells; this was abolished by absorption of the antibody with rhutnfa linked to solid Sepharose, thus confirming antibody specificity. Normal rabbit F(ab ), fragments at equivalent concentration did not stain the HUT78 cells (results not shown). Localization of TNFa-containing cells in synovial tissue. Nine of the 11 RA patient synovial membranes contained cells positive for cytoplasmic staining with F(ab ), anti-tnfa antibodies. TNFa-containing cells were found mainly in the synovial lining layer (mean 37%) (Figure 1A and Table 2). Some TNFa-positive cells were also detected in the deeper interstitium, not only in cell aggregates, where 6% of the cells stained positive (Figure lb), but also in the interaggregate areas, where 10% of the cells stained positive, often in a perivascular distribution. In 3 specimens, endothelial cells in some of the blood vessels also stained with anti-tnfa F(ab ), fragments (Figure IC). TNFa-containing cells were also found at or near the cartilage-pannus junction in all 4 RA specimens examined (Figure 2A). In 3 sections, small fragments of bone were also present, and in 2 of these specimens, TNFa-containing cells were detected adjacent to these areas. Two of the 4 OA synovial membrane sections (Table 2) showed TNFa-containing cells in a similar distribution, but in lower proportion (mean 3.5%), in the lining layer cells (P < 0.05 versus RA group mean). In the synovial membranes from 4 patients with other inflammatory polyarthropathies (Table 2), 2 had cells that stained with the anti-tnfa F(ab ), antibodies. Both patients had longstanding disease and in both, the arthritis was associated with hepatitis (primary biliary cirrhosis in one and unknown etiology in the other). No cells in the 5 normal synovial membranes stained with the antibodies against TNFa (results not shown). Staining of the RA synovial membranes with anti-tnfa F(ab ), fragments was abolished by preincubation of the antibody with 50 pg/ml of rhutnfa, but not with 50 pg/ml of rhuil-6. In addition, binding was not observed after the antibody activity was removed with the TNFa-Sepharose column (Figure 2B); it was still present, however, after passage through an IL-la column (results not shown). Characterization of cells containing TNFa. Double-immunofluorescence using a combination of anti- TNFa F(ab ), antibodies and monoclonal cell marker antibodies showed that 70-90% of the TNFacontaining cells in the RA synovial membrane (Figure 3A) expressed the monocyte/macrophage marker Table 2. Percentages of cells staining for tumor necrosis factor a (TNFa) in the lining cell layer, interstitium, and aggregate areas of the synovial membrane* Lining Interstitium Aggregate Rheumatoid arthritis I NP NP NP Mean Osteoarthritis NP NP NP NP Mean Other inflammatory arthropathies NP NP NP NP Mean 7 10 * Synovium was obtained from the hip or knee joint at synovectomy or arthroplasty in patients 1, 3-8, and 12-14, and from the knee joint at arthroscopy in all other patients. Synovium from patients 2, 3, and 9 also showed TNFa in vascular endothelial cells. NP = not present; - = no staining with anti-tnfa F(ab ), antibodies.

5 TNFa IN RA SYNOVIUM AND PANNUS 1129 A Figure 2. Presence of tumor necrosis factor a (TNFa)-containing cells in the rheumatoid pannus at the interface of the cartilage-pannus junction (A). No staining was apparent in samples tested after the purified anti-tnfa F(ab ), fragments had been passaged through a column containing TNFa-Sepharose (B). Synovial tissue is on the left and cartilage is on the right of each figure. (Original magnification x 160.) B antigens detected by 63D3 and anti-cd14 (Figure 3B). Only 0-2.5% of these cells expressed CD3. None of the immunoglobulin-positive cells from the B cell lineage contained TNFa (Table 3). DISCUSSION To localize the cellular origin of TNFa in diseased synovial membranes, we generated monospe- A Figure 3. Double immunofluorescence analysis of rheumatoid synovial membranes, showing that streptavidin-texas red-positive cells, which contained tumor necrosis factor a (A), were also positive for monoclonal antibodies 63D3 and anti-cd14, using a fluorescein-conjugated antibody?gainst mouse IgG, which demonstrates macrophages (B). (Original magnification x 160.) B

6 1130 CHU ET AL Table 3. Cell phenotype of tumor necrosis factor a (TNFa)- containing cells in rheumatoid synovial membranes % of total TNFa- Cell phenotype Antibody positive cells Monocytehacrophage 63D3, acd T cells acd B cells aigg 0 cific polyclonal rabbit antibodies against rhutnfa. Our anti-tnfa antibodies showed no cross-reaction in an ELISA with rhuil-la and rhuil-6 (data not shown), and in particular, none with rhult, which shares 28% homology of amino acid sequence and has similar biological effects (18). F(ab ), fragments were made in order to prevent binding of rabbit antibodies to locally produced rheumatoid factors, and these fragments stained the cytoplasm of the TNFasecreting HUT78 cell line (data not shown). Specificity was established by showing that the staining reaction could be abolished by prior absorption with TNFa but not with IL-la. With the use of F(ab ), fragments of anti-tnfa antibodies, TNFa was found in cells throughout the RA synovial membrane. Although this could conceivably represent uptake of secreted TNFa by cell surface receptors, the pattern of staining in the synovial tissue cells (Figure 1A) was similar to the cytoplasmic pattern seen in the HUT78 cells. We conclude that the demonstration of TNFa in the intracellular compartment of RA synovial cells suggests local production of this cytokine. TNFa can be localized to 4 sites in inflamed synovial tissue. The lining layer is the major area in which TNFa-secreting cells can be detected; this finding is consistent with those of other recent studies (16,24). Production close to the joint space may be responsible for the presence of TNFa in the synovial fluid (14,15). Two cell types are found in the lining cell layer: one derived from macrophages and the other presumed to be of mesenchymal (fibroblast) origin (25). TNFa has been shown to stimulate fibroblast growth (4) and to induce collagenase and prostaglandin release from adherent synovial cells and fibroblasts (5). Hence, TNFa production in the lining cell layer may also be one of the factors responsible for the release of inflammatory mediators into the synovial space and the increase in the numbers of fibroblasts. Our studies have demonstrated TNFa-containing cells in the deeper areas of the synovium, frequently in a perivascular distribution in the intersti- tium. Here, TNFa could be sustaining activation of immune cells in the vicinity. In vitro studies have implicated TNFa in T cell activation (26) and B cell differentiation (27). Thus, in these aggregates, TNFa may contribute to continued T cell activation (28), and being in close proximity to appropriately primed B cells, TNFa may, in conjunction with other activation and differentiation factors, stimulate the B cell to differentiate and produce rheumatoid factors in vivo. These studies also show TNFa in the endothelial cells of some, but not all, blood vessels in the RA synovial membrane. Although TNFa has been shown to have diverse effects on endothelial cells, including the production of IL-l(9-12) and the increased expression of adhesion molecules (29), there is only limited evidence that TNFa can be made by endothelial cells (30). Nevertheless, immunohistochemical localization of TNFa to endothelial cells suggests that these cells can product TNFa. This feature further indicates involvement of TNFa in the regulation of cell adhesion prior to migration into diseased joints. Of potential relevance to joint destruction, TNFa was demonstrated at the interface of the distinct cartilage-pannus junction, the major target of joint damage in RA (31). Previous in vitro studies have demonstrated that TNFa has the potential to cause cartilage degradation (8). Our studies provide in vivo evidence linking TNFa with this target organ. Previous studies have not characterized the cellular source of TNFa production in rheumatoid tissues (17,32). After stimulation of peripheral blood lymphocytes with endotoxin in vitro (33), the macrophages are the major source of TNFa, but T cells can also secrete TNFa (19,20). Our results show that cells of the monocyte/macrophage lineage are the major source of TNFa in RA synovial membrane, but 2.5% of T cells also contained TNFa. Although we have not directly examined the possibility that synovial fibroblasts could secrete TNFa, some cells with a fibroblast morphology stained with anti-tnfa F(ab ), antibodies, and the mouse L cell fibroblast line has previously been shown to produce TNFa (34). Thus, fibroblasts in the rheumatoid synovial membrane may also secrete TNFa, which may explain why not all the TNFa-containing cells have been characterized with antibodies 63D3 and acd14 for macrophages, acd3 for T cells, and anti-immunoglobulin for B cells (Table 3). There are conflicting data concerning TNFa production in OA synovial tissues (16,24). Our study samples may not be truly representative of OA, in that 3 patients from whom tissues were obtained were

7 TNFa IN RA SYNOVIUM AND PANNUS 1131 undergoing knee/hip replacement surgery, and the fourth patient was undergoing arthroscopy because of a significant knee effusion. However, we confirmed the presence of TNFa (16) in 2 of the 4 samples and demonstrated that there is a lower frequency of positive cells in OA than in the RA tissues (P < 0.05). These cells are located mainly in the deeper interstitium in OA and have not been phenotypically characterized. They are also seen in a perivascular pattern, however, which suggests that they may be of similar phenotype as those seen in RA, but in contrast to RA, the cells are not so commonly located in the hypertrophied lining layer. Thus, the presence of TNFa is not specific for RA, since TNFa has been shown to be present in other inflammatory arthropathies. Contrary to other studies (16), we did not detect TNFa in normal synovial tissue; this may indicate that our technique is not sufficiently sensitive to detect very low amounts of TNFa. Only 9 of 11 RA synovial membranes showed TNFa-containing cells. Samples from 2 patients showed no staining with the anti-tnfa antibodies: one of them, patient 10, showed little evidence of active disease at arthroplasty, and in the other, patient 11, the active knee synovitis improved coincidentally after the arthroscopic biopsy and remained quiescent, although an episode of vasculitis subsequently developed. This absence of TNFa in patients with inactive disease and with rapidly resolving synovitis suggests an apparent inverse link between TNFa production and duration of symptoms in RA. Similarly, 2 patients with chronic nonrheumatoid inflammatory synovitis (patients 18 and 19) had TNFa in the lining cell layer and interstitium. The other patients with short-lived synovitis due to reactive arthritis (patient 16) and adult-onset juvenile arthritis (patient 17) had no TNFa detectable in the synovium, despite active clinical disease. This relationship between TNFa and chronicity of disease is supported by in vitro studies which have shown that anti-tnfa antibodies can reduce IL-1 production in RA synovial membrane mononuclear cells (13). Thus, TNFa may play a role in the perpetuation of chronic synovitis. The pathogenesis of rheumatoid synovitis is clearly a complex process. Current evidence suggests that important among the factors in the pathogenesis are immune complexes, IL-1, TNFa, and IL-6. Thus, immune complexes could stimulate monocytes to secrete IL-1 (35) and TNFa (36). The immune complexes (37), IL-1 (38), and TNFa in rheumatoid joints may, together, perpetuate synovitis by stimulating IL-6 syn- thesis, which if found in close proximity to plasma cells (39), could lead to autoantibody production by plasma cells. Subsequent iritraarticular formation of immune complexes could then stimulate cytokine production by the monocytes in the lining cell layer of the synovial membrane, thus perpetuating the chronic rheumatoid disease process. Our study demonstrates that immunohistologic studies are useful for localizing cytokines, and that information gained from these techniques may define further the role of cytokines in developing dynamic concepts of disease mechanisms. REFERE:NCES 1. Aggarwal BB, Kohr WJ, Hass PE, Moffat B, Spencer SA, Henzel WJ, Bringman TS, Nedwin GE, Goeddel DV, Harkins RN: Human tumor necrosis factor: production, purification and characterization. J Biol Chem 260: , Carswell EA, Old LJ, Kassel RL, Green S, Fiore N, Williamson B: An endotoxin-induced serum factor that causes necrosis of tumors. Proc Natl Acad Sci USA 72: , Le J, Viltek J: Biology of disease: tumor necrosis factor and interleukin- 1 : cytokines with multiple overlapping biological activities. Lab Invest 56:23&248, VilEek J, Palombella VJ, Henriksen-Destefano D, Swenson C, Feinman R, Hirai M, Tsujimoto M: Fibroblast growth enhancing activity of tumor necrosis factor and relationship to other polypeptide growth factors. J Exp Med 163: , Dayer JM, Beutler B, Cerami A: Cachectin/tumor necrosis factor stimulates collagenase and prostaglandin E, production by human synovial cells and dermal fibroblasts. J Exp Med 162: , Bertolini DR, Nedwin GE, Bringman TS, Smith DD, Mundy GR: Stimulation of bone and resorption and inhibition of bone formation in vitro by human tumour necrosis factors. Nature 319: , Saklatvala J, Pilsworth LMC, Sarsfield SJ, Gavrilovic J, Heath JK: Pig catablin is a form of interleukin-1: cartilage and bone resorb, fibroblasts make prostaglandin and collagenase, and thymocyte proliferation augmented in response to one protein. Biochem J 224: , Saklatvala J: Tumour necrosis factor-a stimulates resorption and inhibits synthesis of proteoglycan in cartilage. Nature , Bevilacqua MP, Pober JS, Majeau GR, Fiers W, Cotran RS, Gimbrone MA: Recombinant tumor necrosis factor induces procoagulant activity in cultured human endothelium: characterization and comparison with the actions of interleukin-1. Proc Natl Acad Sci USA 83: , Gamble JR, Harlan JM, Klebanoff SJ, Vadas MA:

8 1132 CHU ET AL Stimulation of the adherence of neutrophils to umbilical vein endothelium by human recombinant tumor necrosis factor. Proc Natl Acad Sci USA 82: , Cavender D, Saegusa Y, Ziff M: Stimulation of endothelial cell binding of lymphocytes by tumor necrosis factor. J Immunol 139: , Nawroth P, Bank I, Handley D, Cassimeris J, Chess L, Stern D: Tumor necrosis factor interacts with endothelial cell receptors to induce release of IL-1. J Exp Med , Brennan FM, Chantry D, Jackson A, Maini RN, Feldmann M: Inhibitory effect of TNF-a antibodies on synovial cell interleukin- 1 production in rheumatoid arthritis. Lancet II: , Di Giovine F, Francesco S, Nuki G, Duff GW: Tumour necrosis factor in synovial exudates. Ann Rheum Dis 47: , Saxne T, Palladino MA Jr, Heineglrd D, Tala1 N, Wollheim FA: Detection of tumor necrosis factor a but not tumor necrosis factor p in rheumatoid arthritis synovial fluid and serum. Arthritis Rheum 31: , Yocum DE, Esparza L, Dubry S, Benjamin JB, Voltz R, Scuderi P: Characteristics of tumor necrosis factor production in rheumatoid arthritis. Cell Immunol 122: , Buchan G, Barrett K, Turner M, Chantry D, Maini RN, Feldmann M: Interleukin- 1 and tumour necrosis factor mrna expression in rheumatoid arthritis: prolonged production of IL-la. Clin Exp Immunol73: , Pennica D, Nedwin GE, Hayflick JS, Seeburg PH, Derynck R, Palladino MA, Kohr WJ, Aggarwal BB, Goeddel DV: Human tumour necrosis factor: precursor structure, expression and homology to lymphotoxin. Nature 312:724729, Turner M, Londei M, Feldmann M: Human T cells from autoimmune and normal individuals can produce tumour necrosis factor. Eur J Immunol 17: , Steffen M, Ottmann OG, Moore MA: Simultaneous production of tumor necrosis factor-a and lymphotoxin by normal T cells after induction with IL-2 and anti-t3. J Immunol 140: , Zeheb R, Chang V, Orr GA: An analytical method for the selective retrieval of iminobiotin-derivatized plasma membrane proteins. Anal Biochem 129: , Ropes MW, Bennett GA, Cobb S, Jacox R, Jessar RA: 1958 revision of diagnostic criteria for rheumatoid arthritis. Bull Rheum Dis 9: , Hughes RA, Allard SA, Maini RN: Arthritis associated with adjuvant mycobacterial treatment for carcinoma of the bladder. Ann Rheum Dis 48:432434, Husby G, Williams RC Jr: Synovial localisation of tumour necrosis factor in patients with rheumatoid arthritis. J Autoimmun 1: , Revell PA: The synovial lining cells. Rheumatol Int 9:49-51, Scheurich P, Thoma B, Uger U, Pfizenmaier K: Immunoregulatory activity of recombinant human tumor necrosis factor (TNF)-a: induction of TNF receptors on human T cells and TNF-a-mediated enhancement of T cell responses. J Immunol 138: , Burmester GR, Locher P, Koch B, Winchester RJ, Dimitriu-Bona A, Kalden JR, Mohr W: The tissue architecture of synovial membranes in inflammatory and noninflammatory joint diseases. Rheumatol Int 3: , Kehrl JH, Miller A, Fauci AS: Effect of tumor necrosis factor-a on mitogen-activated human B cells. J Exp Med 166:78&791, Wellicome SM, Thornhill MH, Pitzalis C, Thomas DS, Lanchbury JSS, Panayi GS, Haskard DO: A monoclonal antibody that detects a novel antigen on endothelial cells that is induced by tumor necrosis factor, IL-1, or lipopolysaccharide. J Immunol 144: , Wilkinson LS, Cambridge G, Edwards JCW: Evidence for production of tumour necrosis factor alpha by endothelial cells (abstract). Br J Rheumatol 28 (suppl 2):129, Allard SA, Muirden KD, Camplejohn KL, Maini RN: Chondrocyte-derived cells and matrix at the rheumatoid cartilage-pannus junction identified with monoclonal antibodies. Rheumatol Int 7: , Brennan FM, Chantry D, Jackson AM, Maini RN, Feldmann M: Cytokine production in culture by cells isolated from the synovial membrane. J Autoimmun 2 (SUPPI): , Mannel DN, Moor RN, Merganhagen SE: Macrophage as a source of tumoricidal activity (tumor necrotizing factor). Infect Immun 30: , Rubin BY, Anderson SL, Sullivan SA, Williamson BD, Carswell EA, Old LJ: Nonhematopoietic cells selected for resistance to tumor necrosis factor produce tumor necrosis factor. J Exp Med 164: , Chantry D, Winearls CG, Maini RN, Feldmann M: Mechanism of immune complex-mediated damage: induction of interleukin- 1 by immune complexes and synergy with interferon-y and tumour necrosis factor-a. Eur J Immunol 19: , Vissers MC, Fantone JC, Wiggins R, Kunkel SL: Glomerular basement membrane containing immune complexes stimulate tumor necrosis factor and interleukin-1 production by human monocytes. Am J Pathol 134:1-6, Zvaifler NJ: The immunopathology of joint inflammation in rheumatoid arthritis. Adv Immunol 16: , Covington M, Field M, Chu CQ, Feldmann M, Maini RN: Interleukin 1 localisation in rheumatoid arthritis (abstract). Br J Rheumatol 28 (suppl 2):45, Field M, Chu CQ, Feldmann M, Maini RN: Interleukin-6 localisation in rheumatoid arthritis. Rheumatol Int (in press)