Eur. J. Imrnunol. 1991. 21: 2575-2579

Regulation of GM-CSF production by TNF-a

Catherine Haworthno, Fionula M. BrennanO, David ChantryoA, Martin 'hrnero0, Ravinder N. Maini" and Marc Feldmannn

Expression of granulocyte-macrophage colony-stimulating factor in rheumatoid arthritis: regulation by tumor necrosis factor-a* Granulocyte-macrophage colony-stimulating factor (GM-CSF), in addition to being a growth factor for granulocytes and macrophages,is an activator of cells of the monocyte/macrophage lineage and induces HLA class I1 expression and cytokine synthesis in these target cells. Macrophage activation and class I1 expression are prominent features in rheumatoid arthritis (RA) joints, but the mechanism of their stimulation is not understood, since interferon-y, the major stimulus of class I1 expression, is not usually detectable at the protein level in synovial cell culture supernatants. We have, therefore, studied GM-CSF expression in cultures of cells derived from joints affected by F U and osteoarthritis (OA), and show that GM-CSF is produced spontaneously both by RA synovial cells and to a lesser extent by OA synovial cells in the absence of extrinsic stimuli. GM-CSFproductioncontinuesfor the 5-day duration of the culture period. Using neutralizing antibodies to tumor necrosis factor (TNF)-a we demonstrated that GM-CSF production in RA synovial cell cultures is dependent on the continued presence of active TNF-a. This result supports our concept that continued activation of the cytokine network is a marked feature of RA, and that TNF-a plays a pivotal role in this network, by regulating the production of other pro-inflammatory cytokines, such as interleukin 1, as demonstrated previously, and GM-CSE

Charing Cross Sunley Research Centreo, Kennedy Institute" and Department of Hematology, Charing Cross and Westminster Medical Schoolo, London

1 Introduction Rheumatoid arthritis (RA) is a chronic inflammatory disease with autoimmune features, which is primarily, but not exclusively, localized to synovial joints. These are infiltrated by lympho-hemopoietic cells: T cells, M a and plasma cells. Many of these cells show evidenw of activation. For example HLA class I1 expression, which is of critical importance for antigen presentation, is elevated on the M@,Tcells, and also on many of the resident cells such as endothelial cells and fibroblasts [l]. In the activated state both the infiltrating and resident cells are capable of producing large amounts of cytokines such as IL la, IL lp, TNF-a [2], IL6 [3] and IL8 [4]. Cytokines may be responsible for many manifestations of the disease process such as bone and cartilage destruction, which may be induced by IL 1and TNF-a [5-91, for the persistence of the inflammatory state, including HLA class II expression and

[I 95671

* A

This work was supported by the Arthritis and Rheumatism Council and Xenova plc. Present address: Departments of Virology and Molecular Biology, Salk Institute, La Jolla, San Diego, CA, USA.

Present address: Howard Hughes Medical Institute, University of Michigan, Medical Centre, Ann Arbor, MI, USA.

cOrrespOndence:

Catherine Haworth

2575

9

Charing

Sunley

Research Centre, Lurgan Avenue, Hammersmith, London W68LW, GB Abbreviations: M-CSF: Granulocyte-MQ-CSF m: a e u matoid arthritis OA: Osteoarthritis SF: Synovial fluid SM:

Synovial membrane 0 VCH Verlagsgesellschaft mbH, D-6940 Weinheim, 1991

for induction of other mediators of inflammation, e.g. PGE2 [lo]. Therefore, understanding the cytokine interactions in the rheumatoidjoint may facilitate the definition of targets for therapeutic intervention. The mechanisms of M a acivation and chronic class I1 expression in RA are not understood. It was expected from its general properties that IFN-y would be involved, but despite the presence of the mRNA, little protein is detectable [ll, 121. Granulocyte-Ma-CSF (GM-CSF) has been shown to induce class I1 expression on cells of the monocyte/M@ lineage [13-151, and the production of cytokines, such as IL 1,TNF-a, G-CSFand M-CSF, in these target cells [15-171. In addition it stimulates mature monocyte function and induces degranulation of neutrophils, which may lead to the release of collagenase and elastase [18].Thus, GM-CSF is a cytokine which could play an important role in the pathogenesis of RA. Indeed the presence of a CSFneutralizedby a polyclonal anti-GM-CSF antibody was identified in RA synovial fluid (SF) [19]. Furthermore, these workers showed the ability of RA synovial culture SN to induce class I1 expression in peripheral blood monocytes was neutralized by anti-GMCSF. In this report we have monitored the expression of GM-CSF in rheumatoid synovial tissue and compared its expression to synovial cells derived from osteoarthritis (OA), a non-autoimmune inflammatory joint disease. In a previous report [20], we studied the inter-relationship of cytokines produced spontaneously by RA synovial cells in culture using neutralizing antibodies and demon- specific strated a major role for TNF-a In the induction of IL I. TNF-a has also been shown to be a potent inducer of GM-CSF activity in a range of cell types [21]. In this report

+

0014-2980/91/1010-2575$3.50.25/0

2576

Eur. J. Immunol. 1991. 21: 2575-2579

C. Haworth, F. M. Brennan, D. Chantry et al.

we show that GM-CSF is produced spontaneously by RA synovial cells, and demonstrate that the presence of this growth factor, like IL 1, is dependent on TNF-a.

2 Materials and methods 2.1 Patients and tissue samples Patients with classical RA (defined by the revised ARA criteria [22]) and OAwere included in this study. SFexudate samples were aspirated from patients as part of their routine treatment and mononuclear cells isolated by Ficoll-Hypaque (Nyegard, Oslo, Norway) density centrifugation as previously described [20]. Mononuclearcells were also extracted from synovial membrane (SM) samples (knees and hip joints) obtained whilst patients were undergoing joint replacement. SM tissue was digested as described previously [20] using collagenase type IV and DNase I (Sigma, Poole, GB). Both SM and SF mononuclear cells are a heterogeneous mixture of cells representing the entire spectrum of infiltrating cells and connective tissue cells found in the synovial joint. In the RA joint the most abundant cells are Tcells and M a , whereas in OA samples, M a and non-hemopoietic fibroblastoid cells predominate [20]. To investigate GM-CSF production in culture, SM cells from both RA and OA synovial joints were cultured at 1 x lo6 cells/ml in 2 ml RPMI 1640 medium (Gibco, GrandIsland, NY)with 10% FCS (Gibco) for periods up to 5 days without exogenous stimulation, and SN were harvested and stored at -20 "C until analyzed. Due to the small size of the OA synovial joint samples the effect of polylconal anti-TNF-a antibody on GM-CSF production was only investigated in the RA synovial specimens. To assess the effect of TNF-a on GM-CSFproduction in RA synovial cultures, SM samples from seven RA patients were incubated with either control rabbit IgG or polyclonal antibody to TNF-a (rabbit anti-TNF-a, 15-1A, neutralizing titer 9 X lo5 unitdml, Genentech, South San Francisco, CA) sufficient to neutralize 500 units TNF-a [20]. In addition, in one RA SM sample there were sufficient cells to compare the effect of monoclonal antiTNF-a with polyclonal anti-TNF-a on GM-CSF production. Therefore, mAb (TNF-E Genentech) was added to the RA culture sufficient to neutralize 500 units TNF-a. In addition in the same sample the effect of anti-GM-CSFon TNF-a production was assessed. Polyclonal rabbit antiGM-CSF (Dr. E. Abney, Charing Cross Sunley Research Centre) was added to the cultures sufficient to neutralize 100 units GM-CSF, and TNF-a levels assayed on day 5 by a specific ELISA.

2.2 Assays for RNA Total cytoplasmicRNA was prepared from freshly isolated, or cultured mononuclear cells derived from RA and OA synovialjoint samples as previously described [2]. RNAwas transferred onto nitrocellulose filters using a Minifold I1 slot blot apparatus (Schleicher and Schuell, Dassel, FRG), and GM-CSF mRNA detected using the 550bp Eco RI/Nco I fragment of human GM-CSF (provided by Dr. G. Wong, Genetics Institute, Boston, MA), labeled by random oligopriming [23] to a sp. act. of 2 X 108-6 x 108 counts/pg DNA. Prior to use on slot blots this cDNA was

shown to recognize a 0.9-kb mRNA corresponding to the GM-CSF mRNA on Northern blots. Stripped blots were rehybridized with 708-bpl Pst 1-Dra 1 fragment of 7B6 (a control cDNA for which the cellular mRNA levels are constant in this assay system) as previously described [24]. Autoradiographs were scanned using the Joyce-Loebl (Gateshead, GB) Chromoscan 3 and absorbance corrected for RNA loading.

2.3 Assays for GM-CSF GM-CSF bioactivity was determined by the proliferationof the human megakaryoblastic Mo7E cell line [25] (Genetics Institute, Cambridge,MA),whichproliferate in response to GM-CSF and IL 3. Mo7E cells were seeded at 2 x 1O'Vwell in RPMI, 10% FCS in 96-well flat-bottom microtiter plates (Nunc, Roskilde, Denmark). Doubling dilutions of sample were added in triplicate to give a final volume of 200 pl, and cultured for 3 days with [3H]dThd added for the last 4 h. dThd uptake was determined by liquid scintillation counting. GM-CSF induced proliferation is linear over a range from 0.15-2.5 U/ml GM-CSF, and specificity of the assay for GM-CSF was confirmed by inclusion of a neutralizing rabbit polyclonal antibodies to GM-CSF (gift of Dr. E. Abney, CXSRC) and IL 3 (Genetics Institute). GM-CSF protein was also determined using a specific ELISA (IMRL, Darlinghurst, NSW, Australia).The lower limit of sensitivity of the assay was 100 pg/ml GM-CSE 2.4 Assays for TNF-a

TNF-a protein was determined by assaying SN in duplicate by ELISA as previously described [26]. Reagents used in theTNF-a assay were provided by B. Reed and Tony Chen (Genentech Inc.). Results are expressed as pg/ml derived from a standard curve using recombinant TNF-a (Genentech). The lower limit of sensitivity of the ELISA was 40 pg/ml for TNF-a.

3 Results 3.1 Expression of GM-CSF in RA and OA joint cell tissues We have previously shown [2] that slot blotting and hybridization is a rapid, sensitive technique to assay cytokine production from small tissue samples available from human disease sites such as RA joints. This method was used to assess samples of RA and OA joint tissue (SF and SM) for the expression of GM-CSF mRNA. GM-CSF mRNA was detected in freshly isolated samples of both SM and SF from RA and OA joints (Fig. 1, insert).The kinetics of GM-CSF mRNA expression was also investigated in RA SF mononuclear cell cultures, and was found to persist for the 72-h period tested (Fig. 1). GM-CSF bioactivity (unitdml) was determined by the MOTE bioassay in OA and RA synovial joint cell SN after culturingfor 2 and 5 days (Fig. 2). GM-CSFbioactivity was detected in RA and to a lesser extent in OA synovial cultures. GM-CSF levels increased with time, implying that

Eur. J. Immunol. 1991. 21: 2575-2579

Regulation of GM-CSF production by TNF-a

2577

there is ongoing synthesis by the interacting synovial cells. The SN-induced Mo7E proliferation was shown to be GM-CSF, as neutralizing anti-GM-CSFantibodies (but not antibodies to IL3) abolished proliferation (Fig. 2, insert). The presence of GM-CSFprotein in the culture SN was also confirmed using a specific ELISA (data not shown). 3.2 Dependence of GM-CSFproduction on TNF

I

I

24

48

I

72

Hours in culture Figure 1. GM-CSF mRNA is expressed spontaneously by freshly isolated RA and OA synovial cells and also in RA synovial joint mononuclear cell cultures. GM-CSF cytoplasmic mRNA levels were determined in freshly isolated synovial joint mononuclear cells (insert) obtained from patients with OA (slot 1,SM),and RA (slot 2, SM, slot 3, SF, slot 4, SM). Filters were hybridized with cDNA probe to GM-CSF (B) see Sect. 2.3, stripped and rehybridized with cell cycle invariaut probe 7B6 (A). GM-CSFcytoplasmic RNA levels were also determined in 5 RA synovial fluid joiat cells cultured from 24-72 h. RNA levels are expressed as arbitrary absorbance units and are corrected for the total RNA loaded

To investigate the role of TNF-a in GM-CSF regulation in RA synovial cells, neutralizing antibodies to TNF-a were added to synovial joint cell cultures and the effect on GM-CSF production determined. Consistent with the findings for untreated cultures (Fig. 2) GM-CSF levels in control cultures (rabbit IgG-treated) increased in culture. The addition of specific neutralizing antibodies to TNF-a significantly inhibited GM-CSF production in the RA synovial cell cultures, although this was not apparent until day 5 of culture (Fig. 3a).This effect of anti-TNF-a was also apparent at the mRNA level where a profound inhibition of GM-CSFmRNA expressionwas seen after 2 days of culture (Fig. 3b). The inhibitory effect of anti-TNF-a antibody on GM-CSF production was confirmed in seven RA synovial membrane cell cultured for 5 days (Irable 1).All RA control

0

0

0

0 0 0

0

n

r

Figure 2. GM-CSF protein is produced spontanously by cultured RA and OA RA synovial joint mononuclear cells. GM-CSF (unitdml) in synovial joint cell cultures was determined by the Mo7E bioassay as described in Sect. 2.3. SN from RA ( 0 , O ) and OA (m, 0 ) cultures were assayed after 2 and 5 days in culture. Specificity of the assay (insert) for GM-CSF is shown by the SN-induced Mo7E proliferation (open bar) being inhibited by neutralizing anti-GM-CSF antibody (hatched bar), but not by neutralizing antibodies to IL3 (stippled bar).

0

1 0

1

2

Days in culture

Figure3. GM-CSF production (mFWA and protein) by RA synovial cell cultures is inhibited with neutralizing antibodies to TNF-a. RA synovial joint cells were cultured for periods up to 5 days with (hatched bars) or without (open bars) neutralizing polyclonal antibodies to TNF-a (see Sect. 2.1). GM-CSF protein (unitshl) were determined in SN from RA synovial cells cultured for 2 and 5 days by the Mo7E bioassay (Fig. 3a).GM-CSFmRNA levels (Sect. 2) were also determined in RA synovial fluid mononuclear cell culture after 1and 2 days (Fig. 3b). mRNA expression (arbitrary absorbanceunits) were determined from cells cultured in of neutralizing rabbit the presence (0-0) or absence (El--0) antibodies to TNF-a.

2578

(IgG-treated) cultures had detectable levels of GM-CSF (range 0.6-64 units/ml) and immunoreactive TNF (range 60-1450 pg/ml). In contrast, the polyclonal anti-TNF-a antibody-treated cultures had sigificantly reduced levels of GM-CSF, and no detectable TNF-a. To examine the specificity of this inhibition an mAb toTNF-a was added to the RA synovial cell cultures. This antibody neutralizes TNF-a less efficiently than the polyclonal antibody in these cultures [20]. Consistent with this, it caused a less profound

Table 1. Anti-TNF-a inhibits GM-CSFproduction in RA synovial joint cells in cultures) Sample

Control culture

no.

Sm 220 Sm316 Sm 400

Sm 427 Sm 295 Sm 394 Sm355

Eur. J. Immunol. 1991. 21: 2575-2579

C. Haworth, F. M. Brennan, D. Chantry et al.

Polyclonal anti-TNF-a antibody-treated culture GM-CSF TNF-a (Ulml) (pglml)

GM-CSF

TNF-U

(U/ml)

(pghl)

9.0 fA.0

I150 *lo 60

~0.1 1.8 1.o

160

4.0 0.4

I .3 8.8 5.6 0.6 34.0

.urn

90 120