After preclinical evaluation of potential new drugs for rheumatoid arthritis (RA), clinical trials in humans are required to assess efficacy and safety. Improved therapy, ethical aspects and the growing number of novel therapeutic agents have increased the need for trials generating a higher density of data in smaller groups of patients. Limited numbers of patients and a shorter study period could be used in early development to assess the effect of a therapeutic intervention on biomarkers that are potentially predictive of clinical efficacy.^1,2 The aim of this study was to examine the utility of real-time quantitative PCR (Q-PCR) as a method to detect changes in gene expression in a small proof-of-concept clinical trial on small synovial tissue samples. Patients with RA were treated with a well-known effective treatment, and real-time Q-PCR technique was used to quantify gene expression in serial synovial biopsy specimens obtained after active treatment with prednisolone compared with placebo.

PATIENTS, MATERIALS AND METHODS

Patients

Twenty-one patients, aged 18–85 years, with RA according to the 1987 criteria of the American College of Rheumatology were included in the study.³ All had active disease at enrolment, defined by the presence of ⩾6 tender and ⩾6 swollen joints out of the 28 joints assessed. In addition, patients had one of the following: erythrocyte sedimentation rate ⩾28 mm/h, serum levels of C reactive protein ⩾1.5 mg/dl or morning stiffness ⩾45 min. They had to be stable for at least 28 days on disease modifying antirheumatic drugs.

Study design

Following informed consent, eligible patients were randomised to receive either oral prednisolone 60 mg daily for the first week, followed by 40 mg daily during the second week (comparable to the dosages used in the COBRA study,⁴ n = 10), or matching placebo for 2 weeks (n = 11) in a 1:1 ratio.⁵ Clinical assessment included Disease Activity Score 28 (DAS28). An independent assessor performed the clinical evaluation.

Arthroscopy

All patients underwent synovial tissue sampling at the Academical Medical Center, University of Amsterdam, Amsterdam, The Netherlands, by needle arthroscopy of an actively inflamed joint (knee, ankle or wrist), performed under local anaesthesia before and after 14 days of treatment, as described previously in detail.^6,7 During each procedure, biopsy specimens were taken from six or more sites of the joint to minimise sampling error.^8,9 The biopsy specimens were snap frozen in liquid nitrogen and stored until RNA isolation.

RNA isolation and complementary DNA synthesis

RNA from synovium and peripheral blood monocytes (PBMCs) was isolated using RNAStat-60 (TelTest, Friendswood, Texas, USA). PBMC were stimulated in vitro with Concanavalin A (Sigma Chemical Co, St Louis, Missori, USA) for 24 h to induce mRNA transcription. The RNA content was determined with RiboGreen (Molecular Probes, Eugene, Oregon, USA), and up to 500 ng per reaction was reverse-transcribed in a final volume of 50 μl. The resulting synovial complementary DNA (cDNA) was stored at −80°C. PBMC-cDNA was diluted in fourfold steps to yield a set of standards representing mRNA acquired from between 24 and 100 000 cells.

Real-time Q-PCR

Q-PCR was performed as described previously in detail.⁸ mRNA levels were quantified using the GeneAmp 5700 Sequence Detection System (Applied Biosystems, Foster City, California, USA). Predeveloped sequence detection reagents specific for human tumour necrosis factor (TNF) α, IL1β and IL8, including forward and reverse primers as well as a fluorogenic TaqMan FAM/TAMRA-labelled hybridisation probe, were used. Matrix metalloproteinase (MMP) 1 was amplified with forward primer: TTT CAT TTC TGT TTT CTG GCC A; reverse primer: CAT CTC TGT CGG CAA ATT CGT and detected byprobe: 6FAM-AAC TGC CAA ATC GGC TTG AAG CTG CT-TAMRA. The fluorescent signal was plotted versus cycle number, and the threshold cycle (C(t), the cycle number at which an increase above background fluorescence could be reliably detected) was determined for each sample.

In order to relate message levels for the cytokines, MMP1 and glyceraldehyde-3-phosphate dehydrogenase to known standard, fourfold dilutions of PBMC-cDNA were included. Standard curves were generated by linear regression using log(C(t)) versus log(cell number). The PBMC equivalent (cell equivalent (CE)) number for synovial samples was calculated from C(t) values using the PBMC standard curve. Data are expressed as the ratio between inflammatory mediator-CE and glyceraldehyde-3-phosphate dehydrogenase (GADPH) CE, yielding the relative expression unit. Each PCR run also included non-template controls containing all reagents except cDNA. These controls generated a C(t) >40 (ie, message below detection level) in all experiments.

Statistical analysis

An analysis of covariance was used to control for initial differences in the values of the baseline samples in this small group of samples. The model fitted included terms for treatment as a fixed effect and the baseline measurement as a covariate, with the aim to estimate a treatment difference. The geometric mean was taken instead of the mean or median, because the data were skewed to the right. A normal distribution was reached after transforming the data to a logarithmic scale. The geometric mean was obtained by taking the exponent of the mean of the transformed data.

RESULTS

Demographic features

Eight male and 13 female patients were included into the trial. Of these, 10 received prednisolone and 11 received placebo treatment. Demographic features were comparable in both groups. The mean age was 49.4 (range 32–63) years for patients in the prednisolone group and 55.7 (range 37–69) years in the placebo group. The median disease duration was 17 (range 4–107) and 22 (range 5–92) months for the prednisolone and placebo group, respectively. Use of disease-modifying anti-rheumatic drugs was the same in both groups. All patients had active disease with comparable disease activity scores DAS28 in both groups: 6.27 (range 4.59–7.98) for the prednisolone group and 5.98 (range 4.56–7.59) for the placebo group.

Clinical results

A strong positive effect of oral prednisolone treatment on the DAS28 (SD) was found, with a decrease of 2 units (from 6.27 (0.95) to 4.11 (1.43)). The mean DAS28 of the placebo group did not change after treatment (5.98 (0.99) before vs 5.68 (1.31) after treatment).

Cytokine mRNA expression

The mRNA yield of the samples was 5.1 (1.13) μg. Table 1 shows the mean (SD) relative expression units before and after treatment.

The effect of prednisolone treatment on the cytokine mRNA expression of IL1β and TNFα in the synovial tissue showed a trend towards reduction after treatment, although confidence intervals included the value for no effect (fig 1 A,B). Prednisolone markedly decreased the expression levels of IL8 and MMP1 mRNA compared with placebo (fig 1C,D).

DISCUSSION

Biomarker-based clinical trials are becoming increasingly important as a means to evaluate the efficacy of novel therapeutic agents for RA. While peripheral blood is readily available for analysis, the target organ, (the synovium), likely provides the most accurate assessment of disease activity. As a result, several investigators have focused on the analysis of serial synovial tissue samples.^10–15 Among the techniques that have been developed, immunohistochemistry to evaluate cell populations or protein expression in the inflamed synovium correlates well with subsequent changes in clinical disease activity.^5,16

Whereas protein is most relevant to function, analysis of RNA transcript levels may provide additional evidence of drug efficacy or information on the mechanism of action in proof-of-concept clinical trials. This study was designed to determine the utility of a Q-PCR technique, developed specifically for the analysis of small synovial biopsy samples,⁹ to measure mRNA expression of genes implicated in the pathogenesis of RA after oral prednisolone or placebo treatment. A marked clinical effect was seen in patients treated with prednisolone, but not in the placebo group.

Expression of two biomarkers (IL8- and MMP1-mRNA) significantly decreased in the corticosteroid-treated patients compared with those treated with placebo. A trend was also observed in IL1β and TNFα gene expression in the prednisolone group. This difference did not reach statistical significance, possibly because of a relatively small sample size or relatively low transcript abundance. The more prominent effect observed in MMP1 is especially interesting, because its expression represents the integration of multiple proinflammatory cytokines, including IL1 and TNFα, which regulate this family of proteases.

This is the first study demonstrating that changes in expression of genes determined by Q-PCR in serial synovial biopsies can differentiate between effective and ineffective treatment. Synovial mRNA levels of IL8 and MMP1 in particular could be useful as biomarkers in clinical proof-of-concept studies, especially in combination with immunohistochemistry to quantify the synovial cell infiltrate. More recent developments in synovial biopsy techniques that obtain larger specimens might also improve the power of cytokine mRNA expression for relatively low abundance transcript species.