The supernatant was stored at −20°C until further analysis. The protein content was measured using Bio-Rad DC protein assay (#500–0116; Promega, Madison, WI, USA). The luciferase activity was performed using a standard luciferase assay (#E4030; Promega) according to the manufacturer’s instructions and measured on a GloMax™ 20/20 luminometer (#E5311; Promega). For statistical evaluation, the Kruskall–Wallis test followed by a post

hoc test was used for comparisons between all groups in each experiment. A P-value ≤ 0·05 was considered significant. To investigate whether raloxifene can influence the induction phase of CIA, OVX DBA/1 mice were treated from 2 days pre-immunization until 10 days postimmunization with either raloxifene (60 µg/day), oestradiol (1 µg/day) or the Miglyol812 vehicle control (100 µl/day), Imatinib as described in Materials and methods. Arthritis scores were evaluated every other day after administration of the booster injection of CII on day 21. In this experiment raloxifene or oestradiol did not hamper the development of arthritis significantly, as measured by frequency (Fig. 1) and severity (data not shown) of arthritis. In addition, we found find protocol no differences in the serum levels of anti-CII antibodies, IL-6 or the cartilage degradation marker COMP

(Fig. 1). To investigate the anti-arthritic properties of raloxifene, female DBA/1-mice were ovariectomized Thiamet G or sham-operated, and CAIA was induced. Ten days prior to receiving the antibody cocktail, administration of raloxifene (60 µg/day), oestradiol (1 µg/day) or vehicle (Miglyol812, 100 µl/day) was started, and continued 5 days per week until termination of the experiment. Figure 2a shows that treatment with oestradiol resulted in a significantly later onset of disease compared to vehicle-treated OVX controls (P < 0·001 on day 7 and P < 0·01 on day 9). The presence of endogenous hormones (sham-operated

mice) also delayed the onset of arthritis (P < 0·01 on day 7), but this effect was not sustained. Raloxifene treatment did not result in delayed onset compared to vehicle controls. Figure 2b shows that oestradiol treatment resulted in less severe arthritic disease, and this effect was sustained throughout the experiment (P < 0·001 compared to vehicle-treated controls). There was no maintained difference in arthritic severity between the OVX and sham vehicle-treated groups, although the groups differed significantly (P < 0·05) on day 7. Raloxifene treatment did not alter disease frequency or severity significantly compared to OVX vehicle controls at any time-point. Histological examination of the paw sections (Fig. 2c and d) revealed the same degree of destruction in joints from OVX and sham-operated controls (median destruction scores of 5·2 and 6·0 of a maximum of 16, respectively).