Abstract
Objectives . Despite recent major advances in the understanding of the pathogenesis of rheumatoid arthritis, with tumour necrosis factor-alpha (TNFα) established as a major therapeutic target, comparatively little is known about the mediators involved in the destructive and inflammatory pathways in osteoarthritis (OA). Recently, it has become appreciated that an inflammatory synovitis contributes not only to the signs and symptoms of osteoarthritis, but also to disease progression. Here, we use high-efficiency adenoviral gene transfer to investigate the role of the transcription factor nuclear factor-κB (NFκB) in regulating inflammatory and destructive mediators in the late stage OA synovium.
Methods . Infection with reporter adenoviruses transferring the β-galactosidase or green fluorescent protein genes verified that OA synovial cells could be infected (>95%). Adenovirus transferring the inhibitory subunit IκBα inhibited NFκB. The production of a whole array of pro- and anti-inflammatory cytokines and mediators, and several matrix metalloproteinases and their main inhibitor, was measured by enzyme-linked immunosorbent assay.
Results . The spontaneous production of macrophage-produced pro-inflammatory cytokines varied: TNFα was modestly inhibited by IκBα overexpression, but interleukin (IL)-1 was unaffected. Both IL-6 and IL-8 were potently inhibited, as were granulocyte–macrophage colony stimulating factor and oncostatin M. Anti-inflammatory mediators like IL-10, the IL-1 receptor antagonist and the p55 soluble TNF receptor were unaffected. Matrix metalloproteinases 1, 3, 9 and 13 were potently inhibited by IκBα overexpression, but not their main inhibitor tissue inhibitor of metalloproteinase-1.
Conclusions . The OA synovium is a highly inflammatory environment, with spontaneous production of many cytokines and matrix metalloproteinases. Inhibition of NFκB may have a beneficial effect on the balance between pro-inflammatory cytokines and anti-inflammatory mediators, and between destructive metalloproteinases and their main inhibitor.
While there has been very marked progress in the field of rheumatoid arthritis (RA), with advances in molecular pathogenesis leading to anti-tumour necrosis factor-alpha (anti-TNFα) therapy gaining prominence [ 1 , 2 ], much less is known about the most common joint disease of all, osteoarthritis (OA). OA is a complex disease, and neither its aetiology nor its pathology is understood. It is usually not directly related to trauma and ‘wear and tear’, and has a spontaneous onset. Unlike RA, the clinical progress of OA is usually indolent, and the disease is asymptomatic in its early stages, making an understanding of its later clinical stages, when treatment is possible, of importance.
Some of the matrix metalloproteinases (MMPs) that have degradative effects on the extracellular matrix may be co-factors or disease mediators in OA [ 3 , 4 ]. The collagenases MMP-1 and MMP-13 are capable of cleaving collagen type II, and MMP-3 is active against other components of the extracellular matrix, such as fibronectin and laminin. Although there has been some interest in MMP inhibitors as therapeutic agents in this disease, there is no clinical trial showing any clear benefit of such approaches, and hence the importance of these molecular pathways is not established.
There is a growing body of evidence that synovial inflammation is implicated in many of the signs and symptoms of OA, including joint swelling and effusion [ 5 , 6 ]. Synovitis in OA is not an innocent bystander, but contributes to disease progression, as judged by the correlation between biological markers of inflammation, like C-reactive protein and cartilage oligomeric protein, with the progression of structural changes in OA [ 7 , 8 ]. The overproduction of cytokines and growth factors from the inflamed synovium is likely to play an important role in the pathophysiology of OA. The low-grade OA synovitis is itself cytokine driven [ 9–11 ], although the levels of pro-inflammatory cytokines are lower than in RA. Some groups have proposed IL-6 or IL-8 as important players in OA pathogenesis [ 11 ], others have suggested oncostatin M [ 12 , 13 ] or nitric oxide [ 14 ] as important co-factors or disease mediators. In particular, the pro-inflammatory cytokines TNFα and interleukin (IL)-1, both of importance in RA, have been suggested as key players in OA pathogenesis [ 5 , 6 , 9 , 10 ], in both synovial inflammation and in activation of chondrocytes. These cytokines can stimulate their own production and induce synovial cells and chondrocytes to produce IL-6, IL-8 and leucocyte inhibitory factor, as well as to stimulate protease and prostaglandin production. The hypothesis that TNFα and IL-1 are key mediators of articular cartilage destruction has raised the possibility of anti-cytokine therapy in OA, or the design of specific disease-modifying osteoarthritic drugs [ 6 ], but to our knowledge none have yet been tested in large-scale, randomized clinical trials.
Using a technique of adenoviral infection that infects in excess of 95% of normal human macrophages, we found that an adenoviral construct transferring the endogenous inhibitor IκBα (AdvIκBα) inhibited TNFα production by 70% [ 15 ]. While the pro-inflammatory cytokines IL-1, IL-6 and IL-8 were also NFκB dependent, anti-inflammatory mediators like IL-10 and the IL-1 receptor antagonist were not [ 16 ]. We have established novel techniques for the use of adenoviral infection as a tool to study a mixture of different subsets of cells, like RA synovial cell co-cultures [ 17 ] or sarcoid lung cells [ 18 ], using fluorescent antibodies against specific surface markers, like CD3 and CD14, to assess the infectibility of selected cell populations. After it had been ascertained that more than 90% of all cell types (macrophages, T cells and fibroblasts) in RA synovial cell co-cultures could be infected with adenovirus, and that the AdvIκBα adenovirus was functional in this tissue, it was possible to study the effect of IκBα overexpression on not only the balance between pro- and anti-inflammatory cytokines in the inflamed joint tissue, but also on the balance between destructive metalloproteinases and their inhibitors [ 17 ]. While the spontaneous production of TNFα, IL-1β, IL-6 and IL-8 was inhibited, there was no effect on IL-10, IL-11 and the IL-1 receptor antagonist. Similarly, the destructive metalloproteinases MMP-1 (collagenase-1), MMP-3 (stromelysin-1) and MMP-13 (collagenase-3) were all NFκB dependent, but not their inhibitor tissue inhibitor of metalloproteinase-1 (TIMP-1), a finding later reproduced in various models of fibroblasts [ 19–21 ] and chondrosarcoma cells [ 20 , 22 , 23 ]. It thus appears that one of the pathophysiological roles of NFκB in RA is to selectively ‘turn on’ an inflammatory and tissue destructive response.
In this paper, we have used this adenoviral gene transfer technique to study molecular pathways and potential therapeutic targets further into another important disease, OA. The aim was to study the role of NFκB on the balance of pro-inflammatory cytokines and anti-inflammatory mediators, and of destructive MMPs and their inhibitor TIMP-1, in the OA synovium, using synovial tissue obtained from patients with late stage OA undergoing knee or hip arthroplasty.
Materials and methods
Cells
Synovium from patients with OA undergoing joint surgery was cut into small pieces with sharp scissors, and digested with collagenase and deoxyribonuclease (DNase). Sufficient numbers of cells [(5–50) × 10 6 per specimen] could be obtained by these means. The optimum conditions of digestion used 1 mgsol;ml collagenase for 2 h, since a higher concentration or a 4-h digestion period leads to macrophage apoptosis, as judged by the subsequent production of IL-1 and TNFα. Fluorescence-activated cell sorter (FACS) studies show that these OA synovial cells are mainly fibroblast-like synoviocytes, with 3–8% macrophages, less than 1% neutrophils and less than 0.1% T cells (not shown). The resulting total cell mixture was cultured at 37°C in RPMI medium 1640 with 25 m m HEPES and 2 m m l -glutamine, supplemented with 5% heat-inactivated fetal bovine serum. Ethical approval for this study was granted by the Gwent Research Ethics Committee and the ethical committee of the Bro Taf Health Authority.
Adenoviral vectors
Recombinant adenoviral vectors encoding Escherichia coli β-galactosidase (Advbgal) or having no insert (Adv0) were provided by Drs A. Byrnes and M. Wood (Oxford, UK). An adenovirus encoding porcine IκBα with a CMV promoter and a nuclear localization sequence (AdvIκBα), and an adenovirus encoding the green fluorescent protein (AdvGFP) were provided by Dr R. de Martin (Vienna, Austria). These were all first-generation, E1- and E3-deleted, serotype 5 adenoviruses. They were propagated in the 293 human embryonic kidney cell line and purified by ultracentrifugation through two caesium chloride gradients [ 24 ]. The titres of viral stocks were determined through a plaque assay on 293 cells, as described [ 24 ]. All viruses used were plaque purified from a master stock, in order to prevent contamination with wild-type adenovirus.
Analysis of infectibility
For infectibility experiments, cells were plated on 12-well plates in 0.5 ml serum-free RPMI 1640 at 1 million cells per well. After being allowed to adhere, they were either left uninfected, or infected with Advbgal or Adv0 at a variable multiplicity of infection, to find out the optimal viral titre for later experiments. After 2 h, the supernatants were removed and replaced with 1.0 ml RPMI 1640 supplemented with 5% heat inactivated fetal bovine serum. Cell were trypsinized off the plates 48 h after infection, spun down and washed in FACS staining solution as described in [ 25 ]. Each batch of uninfected, Adv0-infected or Advbgal-infected cells was then resuspended in 25 ml of staining solution, and incubated with 125 ngsol;ml of anti-CD3 PerCP and 500 ng of anti-CD14 PE (both from Becton Dickinson), for use in identifying the T-cell and macrophage populations, respectively, in a total volume of 45 ml for 45 min at 4°C. They were then incubated at 37°C for 10 min, before 50 ml of a 2 m m solution of Fluorescein di-(β- d -galactopyranoside) (Sigma) was added for 1 min. Addition of excess (10×) ice-cold staining solution was used to stop the reaction. Cell fluorescence was analysed by FACS as described [ 26 ]. In a parallel series of experiments, infection of OA synovial cells with either Adv0 or AdvGFP was performed under the same conditions to verify the above findings.
Western blotting
In these experiments, three batches of 5 × 10 6 cells each were either left uninfected or infected with Adv0 or AdvIκBα. After 2 days, cytosolic and nuclear extracts were prepared as described [ 27 ] and proteins were separated by sodium dodecyl sulphate polyacrylamide-gel electrophoresis (SDS-PAGE) on a 10% (wsol;v) polyacrylamide gel, followed by electrotransfer onto nitrocellulose membranes. IκBα and the p42sol;44 MAP kinases were detected by using antibodies purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA).
Electrophoretic mobility shift studies
Nuclear extracts (20 mg) prepared as in [ 27 ] were mixed with 5× binding buffer prepared as described [ 25 ], and the volume was brought up to 20 ml with distilled H 2 O. After 20 min at room temperature, 5 × 10 4 countssol;min of a double-stranded oligonucleotide probe for NF-κB was added. After 20 min on a shaker, the resulting mixture was run on a pre-electrophoresed native (0.09 m Tris borate, 2 m m EDTA, pH 8.0; TBE) polyacrylamide gel for 90 min at 200 V. The gel was fixed, dried and autoradiographed by exposure to Hyperfilm MP (Amersham, UK).
Analysis of cytokines and metalloproteinases
In experiments concerning cytokine and metalloproteinase synthesis, 2 million OA synovial cells were plated on a 12-well plate in 1 ml serum-free RPMI 1640. After being allowed to adhere, they were either left uninfected or infected with AdvIκBα or Adv0 at a multiplicity of infection of 30:1, as suggested by previous infectibility experiments. After 2 h, the supernatants were removed, non-adherent cells spun down, and reintroduced to the culture in 1 ml RPMI 1640 supplemented with 5% heat-inactivated fetal bovine serum. After incubation for 48 h, the supernatants were taken off and analysed for TNFα, IL-1β, IL-6, IL-8, IL-10, IL-11, granulocyte–macrophage colony-stimulating factor (GM-CSF), oncostatin M, monocyte chemoattractant protein-1 (MCP-1), the IL-1 receptor antagonist and the p55 and p75 soluble TNF receptors by enzyme-linked immunosorbent assay (ELISA). The production of MMP-1, MMP-3, MMP-9, MMP-13 and TIMP-1 was analysed by ELISA kits purchased from Amersham (Little Chalfont, Bucks, UK). In all these experiments, care was taken to analyse a sufficient number of patients ( n = 6–12) to achieve sufficient strength in statistical testing, due to the not inconsiderable intra-individual variation ( Table 1 ).
Effect of the AdvIκBα adenovirus on cytokine and MMP production in OA synovial cells
| Control | Adv0 | AdvIκBα | ||||
|---|---|---|---|---|---|---|
| Mean | s . e . | Mean | s . e . | Mean | s . e . | |
| TNFα ( n = 10) | 2255 | 1047 | 3271 | 1589 | 1965 | 762 |
| IL-1β ( n = 10) | 352 | 127 | 457 | 175 | 522 | 229 |
| IL-10 ( n = 8) | 624 | 60 | 590 | 44 | 610 | 59 |
| IL-6 ( n = 11) | 2 386 175 | 657 835 | 2 817 464 | 977 188 | 526 235 | 211 226 |
| IL-8 ( n = 11) | 1 312 722 | 271 219 | 1 421 977 | 467 809 | 654 054 | 168 996 |
| MCP-1 ( n = 7) | 3702 | 678 | 4264 | 1064 | 888 | 515 |
| Oncostatin M ( n = 8) | 311 | 124 | 250 | 107 | 13 | 3 |
| GM-CSF ( n = 8) | 1078 | 364 | 979 | 296 | 223 | 85 |
| TNFR1 ( n = 8) | 159 | 35 | 137 | 39 | 152 | 58 |
| TNFR2 ( n = 8) | 663 | 340 | 670 | 402 | 290 | 190 |
| IL-1ra ( n = 8) | 17 229 | 10 186 | 16 706 | 10 882 | 12 373 | 8212 |
| IL-11 ( n = 8) | 3233 | 1412 | 1917 | 686 | 214 | 126 |
| MMP-1 (ngsol;ml) ( n = 11) | 6461 | 1428 | 7675 | 1504 | 1791 | 363 |
| MMP-3 (ngsol;ml) ( n = 7) | 967 129 | 318 739 | 929 543 | 280 256 | 364 157 | 242 836 |
| MMP-9 ( n = 7) | 59603 | 21857 | 49650 | 17751 | 9307 | 3781 |
| MMP-13 ( n = 7) | 11274 | 5466 | 7980 | 4018 | 1478 | 715 |
| TIMP-1 ( n = 7) | 14627 | 4677 | 15087 | 4836 | 16000 | 4043 |
| Control | Adv0 | AdvIκBα | ||||
|---|---|---|---|---|---|---|
| Mean | s . e . | Mean | s . e . | Mean | s . e . | |
| TNFα ( n = 10) | 2255 | 1047 | 3271 | 1589 | 1965 | 762 |
| IL-1β ( n = 10) | 352 | 127 | 457 | 175 | 522 | 229 |
| IL-10 ( n = 8) | 624 | 60 | 590 | 44 | 610 | 59 |
| IL-6 ( n = 11) | 2 386 175 | 657 835 | 2 817 464 | 977 188 | 526 235 | 211 226 |
| IL-8 ( n = 11) | 1 312 722 | 271 219 | 1 421 977 | 467 809 | 654 054 | 168 996 |
| MCP-1 ( n = 7) | 3702 | 678 | 4264 | 1064 | 888 | 515 |
| Oncostatin M ( n = 8) | 311 | 124 | 250 | 107 | 13 | 3 |
| GM-CSF ( n = 8) | 1078 | 364 | 979 | 296 | 223 | 85 |
| TNFR1 ( n = 8) | 159 | 35 | 137 | 39 | 152 | 58 |
| TNFR2 ( n = 8) | 663 | 340 | 670 | 402 | 290 | 190 |
| IL-1ra ( n = 8) | 17 229 | 10 186 | 16 706 | 10 882 | 12 373 | 8212 |
| IL-11 ( n = 8) | 3233 | 1412 | 1917 | 686 | 214 | 126 |
| MMP-1 (ngsol;ml) ( n = 11) | 6461 | 1428 | 7675 | 1504 | 1791 | 363 |
| MMP-3 (ngsol;ml) ( n = 7) | 967 129 | 318 739 | 929 543 | 280 256 | 364 157 | 242 836 |
| MMP-9 ( n = 7) | 59603 | 21857 | 49650 | 17751 | 9307 | 3781 |
| MMP-13 ( n = 7) | 11274 | 5466 | 7980 | 4018 | 1478 | 715 |
| TIMP-1 ( n = 7) | 14627 | 4677 | 15087 | 4836 | 16000 | 4043 |
Values given in pgsol;ml except for MMP-1, -3 (ngsol;ml) and MMP-9, -13 and TIMP-1 (pgsol;mg protein).
Effect of the AdvIκBα adenovirus on cytokine and MMP production in OA synovial cells
| Control | Adv0 | AdvIκBα | ||||
|---|---|---|---|---|---|---|
| Mean | s . e . | Mean | s . e . | Mean | s . e . | |
| TNFα ( n = 10) | 2255 | 1047 | 3271 | 1589 | 1965 | 762 |
| IL-1β ( n = 10) | 352 | 127 | 457 | 175 | 522 | 229 |
| IL-10 ( n = 8) | 624 | 60 | 590 | 44 | 610 | 59 |
| IL-6 ( n = 11) | 2 386 175 | 657 835 | 2 817 464 | 977 188 | 526 235 | 211 226 |
| IL-8 ( n = 11) | 1 312 722 | 271 219 | 1 421 977 | 467 809 | 654 054 | 168 996 |
| MCP-1 ( n = 7) | 3702 | 678 | 4264 | 1064 | 888 | 515 |
| Oncostatin M ( n = 8) | 311 | 124 | 250 | 107 | 13 | 3 |
| GM-CSF ( n = 8) | 1078 | 364 | 979 | 296 | 223 | 85 |
| TNFR1 ( n = 8) | 159 | 35 | 137 | 39 | 152 | 58 |
| TNFR2 ( n = 8) | 663 | 340 | 670 | 402 | 290 | 190 |
| IL-1ra ( n = 8) | 17 229 | 10 186 | 16 706 | 10 882 | 12 373 | 8212 |
| IL-11 ( n = 8) | 3233 | 1412 | 1917 | 686 | 214 | 126 |
| MMP-1 (ngsol;ml) ( n = 11) | 6461 | 1428 | 7675 | 1504 | 1791 | 363 |
| MMP-3 (ngsol;ml) ( n = 7) | 967 129 | 318 739 | 929 543 | 280 256 | 364 157 | 242 836 |
| MMP-9 ( n = 7) | 59603 | 21857 | 49650 | 17751 | 9307 | 3781 |
| MMP-13 ( n = 7) | 11274 | 5466 | 7980 | 4018 | 1478 | 715 |
| TIMP-1 ( n = 7) | 14627 | 4677 | 15087 | 4836 | 16000 | 4043 |
| Control | Adv0 | AdvIκBα | ||||
|---|---|---|---|---|---|---|
| Mean | s . e . | Mean | s . e . | Mean | s . e . | |
| TNFα ( n = 10) | 2255 | 1047 | 3271 | 1589 | 1965 | 762 |
| IL-1β ( n = 10) | 352 | 127 | 457 | 175 | 522 | 229 |
| IL-10 ( n = 8) | 624 | 60 | 590 | 44 | 610 | 59 |
| IL-6 ( n = 11) | 2 386 175 | 657 835 | 2 817 464 | 977 188 | 526 235 | 211 226 |
| IL-8 ( n = 11) | 1 312 722 | 271 219 | 1 421 977 | 467 809 | 654 054 | 168 996 |
| MCP-1 ( n = 7) | 3702 | 678 | 4264 | 1064 | 888 | 515 |
| Oncostatin M ( n = 8) | 311 | 124 | 250 | 107 | 13 | 3 |
| GM-CSF ( n = 8) | 1078 | 364 | 979 | 296 | 223 | 85 |
| TNFR1 ( n = 8) | 159 | 35 | 137 | 39 | 152 | 58 |
| TNFR2 ( n = 8) | 663 | 340 | 670 | 402 | 290 | 190 |
| IL-1ra ( n = 8) | 17 229 | 10 186 | 16 706 | 10 882 | 12 373 | 8212 |
| IL-11 ( n = 8) | 3233 | 1412 | 1917 | 686 | 214 | 126 |
| MMP-1 (ngsol;ml) ( n = 11) | 6461 | 1428 | 7675 | 1504 | 1791 | 363 |
| MMP-3 (ngsol;ml) ( n = 7) | 967 129 | 318 739 | 929 543 | 280 256 | 364 157 | 242 836 |
| MMP-9 ( n = 7) | 59603 | 21857 | 49650 | 17751 | 9307 | 3781 |
| MMP-13 ( n = 7) | 11274 | 5466 | 7980 | 4018 | 1478 | 715 |
| TIMP-1 ( n = 7) | 14627 | 4677 | 15087 | 4836 | 16000 | 4043 |
Values given in pgsol;ml except for MMP-1, -3 (ngsol;ml) and MMP-9, -13 and TIMP-1 (pgsol;mg protein).
Analysis of apoptosis
To rule out that apoptosis-mediated cell depletion interfered with the results, cells were routinely checked by microscopy and through the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assay. In addition, a series of experiments were performed using an assay for nuclear fragmentation [ 28 ]. One million OA synovial cells were plated on 12-well plates and either left untreated or infected with 30:1 of Adv0 or AdvIκBα. Addition of sodium azide (200 m m for 2 h) or cycloheximide (2 mgsol;ml) and TNFα (20 ngsol;ml) was used as a positive control. After 48 h, cells were stained for 30 min in 1 ml of a hypotonic fluorochrome solution (50 mgsol;ml of propidium iodide in 0.1% sodium citrate plus 0.1% Triton X-100) and the resulting propidium iodide-stained nuclei were analysed by flow cytometry.
Caspase-3 is an intracellular protein that becomes activated during the cascade of events associated with apoptosis, making it suitable for analysis as an apoptosis marker [ 29 ]. A colorimetric assay for caspase-3 enzymatic activity was purchased from R&D Systems (Minneapolis, MN, USA), and used according to the manufacturer's instructions. Another set of experiments were performed using the early apoptosis marker annexin V, using a kit with propidium iodide and annexin V-FITC purchased from R&D Systems and flow cytometry.
To rule out specific apoptosis and cell death of macrophages in cells that had been infected with 30:1 of Adv0 or AdvIκBα, cultures were checked for CD14 and CD68 positivity using anti-CD14-PE and anti-CD68-PE antibodies purchased from BD-Pharmingen (San Diego, CA, USA) and flow cytometry after 48 h of incubation.
Statistical testing
In all statistical testing, a one-sided, paired-comparisons Student's t -test was used on the original data ( Table 1 ).
Results
Efficient adenoviral gene transfer into OA synovial cells
The infectibility of OA synovial cells was investigated using the Advbgal adenovirus. Since previous data [ 17 , 20 ] indicated that RA synovial cells and human skin fibroblasts could be infected with 30–40:1 of adenovirus, titres in the range 10–100:1 were considered of interest. About 95% of OA synovial cells were infected by a multiplicity of infection of 30 Advbgal plaque-forming units per cell, and that increasing the viral titre to 100:1 did not increase the percentage of infected cells ( Fig. 1 A). This finding was reproduced using AdvGFP, and it was observed that a titre of 30:1 infected the vast majority (>95%) of OA synovial cells ( Fig. 1 B).
In excess of 95% of OA synovial cells can be infected with adenovirus. OA synovial cells were either left uninfected or infected with various titres of an adenovirus with no insert (Adv0) or an adenovirus encoding E. coli β-galactosidase (Advbgal). Cells were harvested 48 h after infection, washed and incubated in 37°C FACS staining solution for 10 min before addition of fluorescein-di-(β- d )-galactopyranoside for 1 min. Reaction was stopped by addition of 10× ice-cold staining solution, and cell fluorescence from Adv0-infected (grey area) and Advbgal-infected (black line) cells was analysed by FACS. In (A) results are shown for the entire mixture of cells, while in (C), OA synovial macrophages have been identified using gating for sizesol;granularity and an anti-CD14-PE antibody. A representative of three experiments. In (B), OA synovial cells were infected with 30:1 of an adenovirus encoding the green fluorescent protein (AdvGFP) (upper panels) or an adenovirus with no insert (lower panels), and transgene expression was assessed by UV light microscopy after 48 h, with light microscopy confirming roughly equivalent cell numbers. A representative of three experiments.
FACS analysis using the fluorescent antibodies anti-CD3 PerCP and anti-CD14 PE to detect T cells and macrophages showed that the late stage OA synovium mainly contains synoviocytes (90–95%), with 3–8% macrophages and less than 0.1% T cells. Whereas the percentage of T cells was too low to allow efficient FACS analysis of infectibility, it was possible to pre-incubate the cells with anti-CD14 PE or CD68 PE to define the macrophage population, using double gating for fluorescence (CD14 positivity) and forwardsol;side scatter characteristics. It was observed that 30:1 of Advbgal infected ∼95% of the OA synovial macrophages, and increasing titres did not improve the percentage of infected cells ( Fig. 1 C).
Analysis of IκBα overexpression and NFκB function
Infection of OA synovial cells with AdvIκBα resulted in cytosolic overexpression of IκBα, whereas infection with Adv0 had no such effect ( Fig. 2 ). Equivalent amounts of protein were loaded on each track, and reprobing with a p42sol;44 MAP kinase antibody was performed to demonstrate this. There were signs of constitutive NFκB activity on bandshift in uninfected and Adv0-infected cells, but not in AdvIκBα-infected cells ( Fig. 2 ). This demonstrates that the AdvIκBα adenovirus is functional in OA synovial cells, at a titre of 30:1. The NFκB activity observed could well be related to up-regulation of the cells in the OA synovium secondary to synovial inflammation, although it is impossible to rule out that other factors, like enzymatic digestion, exposure to fetal calf serum and tissue culture, might have played a part in activating the cells.
The AdvIκBα adenovirus is functional in OA synovial cells. Infection of OA synovial cells with 30:1 of an adenovirus encoding porcine IκBα (AdvIκBα), but not 30:1 of Adv0, causes cytosolic overexpression of the IκBα molecule (upper panel), while not affecting the p42sol;p44 MAP kinases that were used as a control (middle panel). Nuclear extracts from the same experiments showed that 30:1 of AdvIκBα, but not Adv0, caused inhibition of NFκB activity (lower panel).
Variable effects of IκBα overexpression on cytokine production
Late stage OA synovial cell cultures produce significant amounts of the macrophage-produced, pro-inflammatory cytokines TNFα, IL-1β and oncostatin M. There is less production of the anti-inflammatory cytokine IL-10, but this cytokine is still detectable ( Table 1 ). In OA synovial co-cultures depleted of macrophages using CD14-conjugated magnetic beads, none of these cytokines could be detected (not shown). The spontaneous production of TNFα was moderately inhibited (20–30%; P <0.05) by IκBα overexpression, while there was no significant effect on the spontaneous production of IL-1β ( Fig. 3 ). In contrast, the macrophage-produced pro-inflammatory mediator oncostatin M, considered important in OA [ 13 ], is clearly ( P <0.001) NFκB dependent ( Fig. 4 ). The spontaneous production of IL-10 from OA synovial cells is unaffected by IκBα overexpression ( Fig. 3 ), in agreement with data from both human macrophages and RA synovial cells that IL-10 production is NFκB independent and that this cytokine lacks a functional NFκB element on its promoter [ 16 , 17 ].
TNFα, but not IL-1β or IL-10, is NFκB dependent in the OA synovium. OA synovial cells were either left uninfected or infected with 30:1 of either AdvIκBα or Adv0, and the spontaneous production of various cytokines and mediators was measured by ELISA. The production of the macrophage-produced cytokines TNFα, IL-1β and IL-10 is given as a percentage of the production from uninfected cells ( n = 6–12).
IL-6, IL-8 and oncostatin M are NFκB dependent in the OA synovium. OA synovial cells were either left uninfected or infected with 30:1 of either AdvIκBα or Adv0, and the production of various cytokines and mediators was measured by ELISA. The spontaneous production of various pro-inflammatory cytokines (IL-6, IL-8, MCP-1, oncostatin M and GM-CSF) is given as a percentage of the production from uninfected cells ( n = 6–12).
There was considerable production of several other pro-inflammatory cytokines from the OA synovial cells ( Table 1 ): IL-6, IL-8, GM-CSF and MCP-1. Data from experiments in which synovial cell populations were separated using magnetic beads suggest that these are produced both from macrophages and from fibroblast-like synoviocytes (not shown). It was observed that IL-6 was very potently ( P <0.001) inhibited by IκBα overexpression (80%; Fig. 4 ). This agrees with data from normal human macrophages and fibroblasts, as well as from RA synovial cells, that IL-6 is a cytokine that is strongly dependent on transcriptional regulation by NFκB. Another cytokine that has been reported to be NFκB dependent in various systems is IL-8. In OA synovial cells, the spontaneous production of IL-8 is moderately (50%) down-regulated by IκBα overexpression ( Fig. 4 ), again a situation resembling that in RA [ 17 ]. The pro-inflammatory chemokine MCP-1 was significantly ( P <0.001) inhibited by IκBα overexpression, as was GM-CSF ( Fig. 4 ).
OA synovial cells also spontaneously produce several anti-inflammatory mediators. As in RA, there was no significant effect of IκBα overexpression on the production of the IL-1 receptor antagonist. In contrast, IL-11 was significantly ( P <0.001) inhibited ( Fig. 5 ). Interestingly, there was a selective effect on the spontaneous production of the soluble TNF receptors: the p75 receptor was significantly ( P <0.01) inhibited, but not the p55 receptor ( Fig. 5 ). As judged from OA synovial co-cultures depleted of macrophages, both these soluble TNF receptors are produced both by macrophages and by synovial fibroblasts (not shown).
IL-1ra and the p55 soluble TNF receptor are NFκB independent in the OA synovium. OA synovial cells were either left uninfected or infected with 30:1 of either AdvIκBα or Adv0, and the production of cytokines and mediators was measured by ELISA. The spontaneous production of various anti-inflammatory cytokines (IL-11, IL-1ra, p55 and p75 soluble TNF receptors) are given as a percentage of the production from uninfected cells ( n = 6–12).
IκBα overexpression reduces matrix metalloproteinase production
IκBα overexpression potently inhibited ( P <0.001) the spontaneous production of several MMPs from OA synovial cells ( Fig. 6 ). As previously observed in RA [ 17 ], both MMP-1 and MMP-3 were NFκB dependent, as were MMP-9 and MMP-13. In contrast, the spontaneous production of the inducible inhibitor of theses MMPs, TIMP-1, was not significantly affected by IκBα overexpression ( Fig. 6 ).
Several matrix metalloproteinases are NFκB dependent in OA. OA synovial cells were either left uninfected or infected with 30:1 of either AdvIκBα or Adv0, and the spontaneous production of various matrix metalloproteinases (MMP-1, -3, -9, -13 and TIMP-1) was measured by ELISA. The production of each MMP is given as percentage of the production from uninfected cells ( n = 6–12).
Lack of apoptosis after AdvIκBα infection
Some studies have suggested that NFκB inhibition may induce apoptosis within the infected cell population [ 30,31 ]. It is thus important to rule out that the inhibition of various cytokines or MMPs observed in this paper is secondary to apoptosis-mediated cell depletion.
Several lines of evidence indicate that this was not the case. First, there is no reduction of cell numbers in AdvIκBα-infected cultures as compared with uninfected and Adv0-infected cells, nor any indication of increased cell death as judged by microscopy or by the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide assay or the alamar Blue assay (Biosource). Nor was there any reduction of the levels of a cytosolic control protein (p42sol;44 MAP kinases; Fig. 2 ) or the mRNA levels of a housekeeping gene (GAPDH; not shown). An assay for a late apoptosis marker, DNA fragmentation [ 28 ] showed that infection of OA synovial cells with Adv0 or AdvIκBα did not increase the percentage of apoptotic cells ( Fig. 7 A–D) Nor was there increased activity of the caspase-3 protein ( Fig. 7 E), arguing against cell depletion caused by apoptosis playing a role in these experiments.
The AdvIκBα adenovirus does not cause apoptosis. One million OA synovial cells were plated on 12-well plates and either left untreated (A) or infected with 30:1 of Adv0 (B) or AdvIκBα (C). Addition of sodium azide (200 m m for 2 h) was used as a positive control (D). After 48 h, cells were stained for 30 min in 1 ml of a hypotonic fluorochrome solution (50 mgsol;ml of propidium iodide in 0.1% sodium citrate plus 0.1% Triton X-100) and the resulting propidium iodide-stained nuclei were analysed by flow cytometry. In (E), cells plated and infected in the same manner were studied for caspase-3 activity, with the addition of cycloheximide (2 mgsol;ml) and TNFα (20 ngsol;ml) for 16 h used as a positive control. This is a representative of three independent experiments.
In experiments involving the early apoptosis marker annexin V, 15–35% of cells were annexin positive in uninfected OA synovial cell co-cultures, as a result of the naturally occurring apoptosis in the OA synovium, as well as a consequence of the digestion of the tissue sample, and culture for 2 days at a relatively high density of cells ( Fig. 8 A). There was no significant increase in annexin V positivity in cells infected with either of the adenoviruses used, as judged by three independent experiments.
The AdvIκBα adenovirus does not cause increased annexin V positivity, or specific macrophage apoptosis. Two million OA synovial cells were plated on 6-well plates and either left uninfected or infected with 30:1 of Adv0 or AdvIκBα. After 48 h, cells were stained with propidium iodide and annexin V-FITC, and annexin V positivity was analysed using flow cytometry (A). Other aliquots of the cells were stained with an anti-CD14-PE antibody to detect the presence of macrophages, and a gate was established that contained less than 0.5% of unstained cells in either of the three populations. Using this gate, this experiment showed 5.3, 6.2 and 7.2% macrophages in the uninfected, Adv0-infected and AdvIκBα-infected cells, respectively. A representative of three independent experiments.
It was also considered important to rule out that NFκB inhibition led to selective apoptosis of macrophages, by means of comparing percentages of macrophages in uninfected, Adv0-infected and AdvIκBα-infected cultures, using flow cytometry with either anti-CD14-PE or anti-CD68-PE as macrophage markers. Three independent experiments, of which a representative is reproduced here ( Fig. 8 B), showed macrophage percentages of between 3 and 8%, with no reduction in macrophage numbers observed in AdvIκBα-infected cultures.
Discussion
Adenoviral gene transfer is a valuable tool for studying intracellular signal transduction in cell lines or in cultured primary cells permitting kinetic analysis [ 15–17 , 20 , 32 ]. Adenoviral gene transfer can also be used in mixed populations of cells, using fluorescent antibodies to surface markers to detect the infectibility of the various constituents. This makes it possible to assess the function of a certain signal transduction step in the diseased tissue itself. For example, we have used adenoviral gene transfer to study signalling in RA synovial cells [ 17 ], in sarcoid lung cells [ 18 ] and in murine colitis gut cells [ 33 ]. This paper extends this technology to late stage OA synovial cells. Given that both mature macrophages [ 16 ] and human skin fibroblasts [ 20 ] were infectible, as were RA macrophages and fibroblast-like synoviocytes [ 17 ], it was anticipated that the OA synoviocytes and macrophages could also be infected with relatively modest titres of adenovirus. Due to the scarcity (<0.1%) of OA synovial T cells, it was not possible to study this particular cell population.
We confirmed that late stage OA synovium is a highly inflammatory environment, as suggested by other groups [ 34 , 35 ] reporting prominent features of inflammation, synovial cell proliferation and angiogenesis in the OA synovium. There is also abundant production of multiple pro- and anti-inflammatory cytokines in these OA synovial cell cultures ( Table 1 ). An unexpected observation was that in the OA synovium the spontaneous production of TNFα was only moderately reduced by IκBα overexpression, and that IL-1β was entirely unaffected, since TNFα and IL-1β have functional NFκB elements on their promoters, and have been demonstrated to be NFκB dependent in RA cells and in various macrophage models [ 15–17 , 36–38 ]. It is known, however, that in human macrophages, there are both NFκB-dependent and NFκB-independent ways of inducing TNFα and IL-1β. Whereas some stimuli, like lipopolysaccharide (LPS) and phorbol myristate acetate (PMA), act via NFκB, others, like zymosan and CD45 ligation, do not [ 16 , 39 ]. The data presented here would suggest that the former pathways are predominant in RA and both are active in the OA synovium. It is interesting that in LPS-stimulated human macrophages and in RA synovial cells, the production of TNFα is independent of the IκB kinase 2, although the production of IL-1β is not [ 21 ]. It may also be speculated that the induction of TNFα and IL-1β relies on the translocation of the NFκB family members p50 and p65 to the nucleus and transactivation of the Ets transcription factor ESE-1 via a high-affinity NFκB binding site. It has been noted that expression of ESE-1 is higher in RA synovial tissue than in OA [ 40 ].
In RA and in various fibroblast models, the induction of IL-6 and IL-8 is NFκB dependent [ 16 , 20 , 41 ]. This paper shows that this also is the case in the OA synovium, and that IL-6 and MCP-1 in particular are very tightly regulated by this transcription factor. Earlier data with regard to the possible NFκB regulation of GM-CSF are contradictory [ 42 , 43 ], but our data clearly indicate that this cytokine is NFκB dependent in the OA synovium. Of particular importance is that the mediator oncostatin M, produced mainly by macrophages in the OA synovium, is strongly NFκB dependent ( Fig. 4 ). With regard to anti-inflammatory mediators, the IL-1 receptor antagonist turned out to be NFκB independent, in accordance with data from human macrophages and from RA synovial cells [ 16 , 17 ]. Interestingly, there was a selective effect on the soluble TNF receptors, with the inducible p75 receptor being NFκB dependent ( P <0.01), while the p55 constitutive receptor was not. This agrees with data concerning the influence of NFκB activation on the expression of these receptors on the cell surface [ 44 , 45 ], as well as studies on the production of the p75 receptor in macrophages and RA synovial cells [ 16 , 17 ].
It is well appreciated today that many MMPs, including the important collagenases MMP-1 and MMP-13, are regulated through the interaction of several transcription factors, including AP-1, NFκB and Ets-1 [ 22 , 23 , 46 ]. There is known to be cell type and stimulus specificity, as well as complex interactions between these transcription factors [ 46 ]. In various fibroblast and chondrocyte models [ 19 , 20 ], and also in RA synovial cells, [ 17 , 21 ], MMP-1, -3, -9 and -13 are all NFκB dependent. In the OA synovium, the spontaneous production of all these MMPs, but not their main inhibitor TIMP-1, is also NFκB dependent ( Fig. 6 ). This finding indicates that in the RA and OA synovium one of the important functions of NFκB activation is to specifically turn on these destructive enzymes. Both in RA and in OA, the inhibition of MMP production induced by IκBα overexpression also occurs when cells are treated with excess of a soluble TNF receptor-Ig fusion protein or an anti-IL-1 antibody to specifically neutralize the effect of these cytokines, and is thus likely to represent a direct effect on the MMP promoter (data not shown).
The OA synovium is an inflammatory environment, and with increased apoptosis of the fibroblast-like synoviocytes [ 40 ]. In HeLa cells, cultured fibroblasts and other cell lines, inhibition of NFκB leads to apoptosis when the cells are stimulated with TNFα [ 30 , 31 , 47 ]. This is not the case in differentiated myeloid cells like mature human macrophages and dendritic cells, however [ 15 , 48 ]. The results reported in this paper cannot be explained through apoptosis-mediated cell depletion of either macrophages or fibroblast-like synoviocytes, since some macrophage-produced cytokines (TNFα, oncostatin M) were inhibited, while others (IL-1β, IL-10) were unaffected. Similarly, some mediators produced mainly from fibroblast-like synoviocytes, like IL-6 and the MMPs, were potently inhibited, while others (TIMP-1) were unaffected. Nor did several different assays for early or late apoptosis markers support the notion that there is increased apoptosis in the cultures infected with the Adv IκBα adenovirus, as compared with uninfected or Adv0-infected cells ( Figs 7 and 8 ). In particular, there was no selective loss of macrophages in Adv IκBα-infected cultures.
This paper demonstrates that late stage OA synovium is a highly inflammatory environment, with abundant production of many inflammatory and destructive mediators. The spontaneous production of many of these mediators, including TNFα, IL-6, IL-8, MCP-1, GM-CSF and oncostatin M, is NFκB dependent in OA. Some key anti-inflammatory mediators, including IL-10, the IL-1 receptor antagonist and the p55 soluble TNF receptor, are NFκB independent, meaning that inhibition of this transcription factor would have a positive, anti-inflammatory influence on the balance between pro- and anti-inflammatory pathways in OA. This is in keeping with the TNF independence of IL-1 bioactivity in OA, in contrast to the situation in RA [ 49 ]. While there are many similarities with the situation in the RA synovium, like the NFκB dependence of IL-6 and IL-8 and the NFκB independence of IL-10 and IL-1ra, there are also some differences, mainly the NFκB dependence of the p75 soluble TNF receptor and the NFκB independence of IL-1β in OA. The regulation of MMPs in the OA synovium greatly resembles the situation in RA: MMP-1, -3, -9 and -13 are all NFκB dependent, while their inhibitor TIMP-1 is not; again this provides a positive balance. NFκB has been recognized as a potential therapeutic target in RA [ 17 , 50 ] and work is ongoing to generate small-molecule inhibitors of this transcription factor, although systemic therapy may be unsafe due to possible hepatic cell apoptosis and consequent hepatotoxicity, or the risk of infection. In OA, much more work will be needed to evaluate whether the synovial inflammation has a clear relation to cartilage breakdown and bone changes, before the potential of NFκB as a therapeutic target in OA can be seriously discussed.
Acknowledgements
This work has been supported by the Arthritis Research Campaign (UK), grant ID W0596 and 14570.
The authors have declared no conflicts of interest.
Comments