All reagents used in cell culture were obtained from Sigma (Poole, UK) unless otherwise stated. Fibroblasts were isolated from biopsies of matched skin, bone marrow and synovium removed during total knee arthroplasty 23 from consenting patients who fulfilled the American College of Rheumatology (formally the American Rheumatism Association) criteria for rheumatoid arthritis (RA; n = 6) and osteoarthritis (OA; n = 3). Fibroblasts were isolated by mechanical digestion of tissue followed by dissociation in 5 mmol/l EDTA for 2 hours. Dissociated tissue was then washed and transferred to a culture flask. The fibroblasts were then cultured through to a maximum of seven passages in complete fibroblast medium consisting of RPMI-1640, 1% (vol/vol) non-essential amino acids, 1% penicillin/streptomycin, 1% sodium pyruvate, 2 mmol/l glutamine and 20% heat-inactivated foetal bovine serum (Labtech International, Sussex, UK) 23. Fibroblasts were treated with 10 ng/ml IL-1, tumour necrosis factor (TNF)-α, IL-4 and interferon-γ (R&D Systems, Abingdon, UK) or 100 nmol/l cortisol or cortisone for 24 hours before harvesting.

Cryostat tissue sections of synovial tissue from patients with RA were analyzed using goat polyclonal antibodies to 11β-HSD1 (The Binding Site, Birmingham, UK) with anti-sheep/goat biotin AB360 as secondary antibody (The Binding Site). Immunohistochemistry was also carried using polyclonal antisera to the following: the endothelial marker von Willebrand Factor (vWF; Dako, Ely, UK) with a secondary anti-rabbit AMCA antiserum (Jackson ImmunoResearch Laboratories, Inc., West Grove, PA, USA); the fibroblast marker ASO2 (CD90) with a secondary anti-mouse IgG₁ Alexa 633 antiserum (Invitrogen, Paisley, UK); and the UCHT-1 (CD3) antiserum (gift from Peter Beverly, UCL, London, UK) with secondary anti-mouse IgG_2b TRITC (Southern Biotech, Birmingham, USA). Immunohistochemical analyses and subsequent image visualization were carried out as described previously 24.

RNA was extracted from cultured fibroblasts using the single-step extraction method (TRI Reagent, Sigma, Poole, UK). Aliquots (1 μg) of RNA were then reverse transcribed using random hexamers in a 20 μl volume, as stated in the manufacturer's protocol (Promega, Madison, WI, USA) 25.

Expression of mRNA for 11β-HSD1, H6PDH, GRα, GRβ, IL-6, CCAAT/enhancer binding protein (C/EBP)α and C/EBPβ was assessed using real-time PCR in an ABI 7500 system (Applied Biosytems, Warrington, UK). The reactions were performed in 25 μl aliquots on a 96 well optical reaction plate (Sigma). Primers for 18S were used in conjunction with our target primers as an internal reference. Reactions contained TaqMan universal PCR master mix (Applied Biosytems), 900 nmol primers, 100–200 nmol TaqMan probe and 25–50 ng cDNA. Target genes probes were labelled with FAM and 18S probes were labelled with VIC. Reactions occurred as follows: 50°C for 2 minutes, 95°C for 10 minutes, 44 cycles of 95°C for 15 seconds and 60°C for 1 minutes. Data were obtained as Ct values (the cycle number at which logarithmic PCR plots cross a calculated threshold line) according to the manufacturer's guidelines (Applied Biosytems), and used to determine ΔCt values (Ct of target gene - Ct of housekeeping gene) as raw data for gene expression (high ΔCt = low gene expression). The fold change in gene expression was determined by subtracting ΔCt values for treated cells from their respective control samples. The resulting ΔΔCt values were then used to calculate fold change in gene expression according to the expression 2^ΔΔCt. Probe and primer sequences used are summarized in Table 1.

Confluent cells in 12-well plates were cultured with 1 ml complete fibroblast medium, containing 100 nmol/l cortisone (to measure reductase activity) or cortisol (dehydrogenase activity) along with tritiated tracer for 18 hours. Steroids were extracted from growth medium using dichloromethane (5–10 vol) and separated by thin-layer chromatography (TLC) using ethanol:chloroform (8:92) as the mobile phase. TLC plates were analyzed using a Bioscan imaging detector (Bioscan, Washington, DC, USA) and the fractional conversion of steroids was calculated. In each case total protein concentration was assessed using a 96-well plate assay kit (Bio-Rad laboratories Inc., Hercules, CA, USA). Results were expressed as pmol product/hour per mg protein. All experiments were carried out in triplicate.

Cortisol levels in supernatants from cultured cells were measured using a commercially available sandwich ELISA in accordance with the manufacturer's instructions (R&D systems, Abingdon, UK). Data were expressed as pg cortisol/mg protein. Soluble IL-6 in supernatents from cultured cells was measured using a commercially available sandwich ELISA in accordance with the manufacturer's instructions (BD Biosciences Pharmingen, San Diego, CA, USA). Data were expressed as pg IL-6/mg protein.

Data are reported as mean ± standard deviation (SD) of replicate mean values for separate patient cell cultures. However, because of interindividual variation, data in Figures 1 and 2 are shown as representative means (±SD) from single patient cell cultures. In these experiments assays were repeated at least twice using different cell preparations, with similar results. Regression analysis was performed using Microsoft Excel 2003. One-way analysis of variance was performed using SPSS Data Editor (SPSS Inc., Santa Clara, CA, USA).

We first examined whether synovial fibroblasts in situ expressed 11β-HSD1. Confocal immunohistochemistry (Figure 3) indicated that 11β-HSD1 was expressed by a variety of cells types in rheumatoid synovial tissue, including leukocytes such as T cells and macrophages. However, it was only weakly expressed by dendritic cells. The enzyme was abundantly expressed by CD90-positive fibroblasts and by endothelial cells (identified by von Willebrand factor staining).

To investigate the significance of 11β-HSD1 expression in fibroblasts, further studies were carried out using primary cultures of fibroblasts isolated from matched synovium, bone marrow and skin. Analysis of mRNA from these cells indicated that expression of 11β-HSD1 was greatest in synovial fibroblasts (ΔCt = 20.5 ± 2.2), followed by dermal (ΔCt = 23.3 ± 1.8) and bone marrow fibroblasts (ΔCt = 24.2 ± 0.9; Figure 4). In all three cell types expression of 11β-HSD1 was upregulated after treatment with TNF-α with the synovial fibroblasts (ΔCt = 17.7 ± 1.7) continuing to show the highest levels of mRNA relative to dermal (ΔCt = 18.9 ± 1.4) and bone marrow fibroblasts (ΔCt = 20.9 ± 1.2). These tissue-specific variations in baseline expression and potent inducibility by TNF-α were not observed for other components of glucocorticoid metabolism and signalling. For example, the NADPH-generating enzyme H6PDH and GRα exhibited similar levels of expression in the different fibroblast types, and this was not significantly affected by treatment with TNF-α (Figure 4). The nonfunctional GRβ was only weakly expressed in basal fibroblasts compared with GRα (ΔCt = 27.9 ± 2.1 in vehicle-treated versus 14.7 ± 1.9 for GRα) and was also unaffected by TNF-α treatment. Comparison of cells isolated from patients with RA (n = 6) with those from OA patients (n = 3) indicated that there was no significant difference in expression of 11β-HSD1, H6PDH, GRα, or GRβ mRNA in either vehicle or TNF-α-treated fibroblasts between the two disease types (data not shown).

Similar induction of 11β-HSD1 mRNA was seen with IL-1 in bone marrow (37-fold increase for IL-1 treatment and 8-fold increase for TNF-α treatment relative to vehicle), dermal (14-fold and 4-fold increase) and synovial (31-fold and 7-fold) fibroblasts (Figure 5). By contrast, classical T-helper-1 and T-helper-2 cytokines such as interferon-γ and IL-4 and had no significant effect on 11β-HSD1 expression in any of the fibroblasts.

The site-specific differences in fibroblastic 11β-HSD1 mRNA expression and differential induction by IL-1 and TNF-α were biologically relevant because they translated into differences in enzyme activity (Figure 1). Measurement of ³H-cortisone and ³H-cortisol metabolism by scanning analysis of TLC indicated predominant reductase activity in all three cell types, being highest in synovial fibroblasts, followed by bone marrow fibroblasts, and lowest in dermal fibroblasts (Figure 1a). Treatment with IL-1 (10 ng/ml, 24 hours) increased cortisol production in all three cell types at an order of magnitude that was similar to that seen for mRNA data (Figure 5). Further experiments were also carried out to confirm the predominant reductase activity (cortisone to cortisol conversion) of 11β-HSD1 in human fibroblasts. ELISA data to assess the absolute levels of cortisol in cell culture supernatants revealed a pattern of 11β-HSD1 activity similar to that determined by TLC analysis. Cortisol accumulation following 24 hours incubation with 100 nmol/l cortisone was again significantly higher in synovial fibroblasts and increased still further in cells that were pretreated with TNF-α (Figure 1b).

Previous studies have shown that members of the C/EBP family of transcription factors, namely C/EBPα and C/EBPβ, are potent regulators of 11β-HSD1 expression in hepatocytes 26. A similar association between 11β-HSD1 and C/EBPα and C/EBPβ was also observed in fibroblasts (Figure 6a), although levels of C/EBPβ were significantly higher than the α isoform (higher ΔCt value = lower mRNA level). Both C/EBPα and C/EBPβ exhibited tissue-specific variations in expression (Figure 6b): C/EBPα was more strongly expressed in synovial fibroblasts, followed by bone marrow and then dermal fibroblasts, whereas the reverse pattern was observed for C/EBPβ.

The functional significance of tissue-specific variations in glucocorticoid metabolism was investigated by determining whether increased 11β-HSD1 activity was able to mediate autocrine regulation of the cytokine IL-6 in each of the different fibroblast types (Figure 2). Cortisol suppressed IL-6 mRNA levels in dermal (86%; P < 0.01), bone marrow (50%; P < 0.01) and synovial fibroblasts (87%; P < 0.01), indicating that all three cells had a similar capacity for GR-mediated responsiveness (Figure 2a). However, only synovial fibroblasts suppressed IL-6 mRNA expression when treated with the inactive glucocorticoid cortisone. This was completely reversed in the presence of the 11β-HSD inhibitor glycyrrhetinic acid, suggesting that cortisone achieves its effects on IL-6 via autocrine activation to cortisol (Figure 2a). Analysis of IL-6 protein secretion by ELISA in synovial fibroblasts confirmed that the findings observed at the RNA level were also reflected at the protein level. Because of variability in baseline IL-6 levels in different patients, data shown in Figure 2b were derived from a representative culture of RA fibroblasts. Similar results were observed in three sets of cells, and analysis of percentage change in IL-6 levels from these different cell preparations confirmed that in synovial fibroblasts both cortisol and cortisone were able to suppress IL-6 levels, with the effect of cortisone being abrogated by glycyrrhetinic acid (data not shown).