The cells were solubilised at 4��C in lysis buffer (Tris-base 50mM, pH 7.6, 2mM EDTA, 100mM NaCl, 1% Nonidet P-40, 1mM PMSF, 2��gml?1 aprotinin, 2��gml?1 20S proteasome inhibitor pepstatin, 2��gml?1 antipain) for 60min. Cell lysate was centrifuged at 15000r.p.m. for 20min at 4��C and aliquots of supernatants were used to measure protein concentration. Samples were boiled for 3min in SDS-sample buffer (50mM Tris-base, pH 6.8, 2% SDS, 100mM dithiothreitol, 10% glycerol and 0.025% ��-mercaptoethanol) and separated on 15% SDS�Cpolyacrylamide gel electrophoresis (SDS�CPAGE). Proteins were then transferred onto Immobilon-P membrane (Millipore, Bedford, MA, USA) and blots were probed with an anti-p21ras (1:500) or anti-p21rhoA antibody (1:1000), and detected with the use of horseradish peroxidase-conjugated secondary antibody (dilution, 1:10000).

The membranes were then exposed to Kodak X-Omat AR film, and film densities were quantified as described in ��Analysis of data’. To evaluate the effect of fluvastatin and gemcitabine on p42MAPK/ERK2 phosphorylation, MIAPaCa-2 cells were treated with fluvastatin (2�C10��M) or gemcitabine (1�C50nM) for 72h alone or in combination with mevalonic acid 100��M. Cells were solubilised in lysis buffer with the protein phosphatase inhibitors sodium metavanadate and sodium fluoride (200��M each) for 45min at 4��C, and then centrifuged at 4��C for 20min at 15000r.p.m. The supernatant was boiled for 3min in SDS-sample buffer and separated on 11% SDS�CPAGE. Proteins were transferred onto Immobilon-P membrane and probed with anti-p42MAPK/ERK2 antibody (1:1000), and detected as described above.

Immunocytochemistry Cells were seeded onto chamberslides (Nalge Nunc, Naperville, IL, USA) at a density of 104 per 150��l; after treatment with fluvastatin at 2��M for 48h or vehicle, cells were washed in PBS and fixed with 1% neutral buffered formalin (10min) at 4��C. The specimens were then permeabilised with a 10-min exposure to a solution of 0.2% Triton X-100�CPBS. The samples were subsequently treated with a solution of 3% H2O2 for 5min to quench endogenous peroxidase activity, and nonspecific reactivity was blocked with 5% swine serum for 20min at 37��C. The samples were then incubated overnight at 4��C in a humidified chamber with the mouse anti-human-p21rhoA or anti-human-Ha-p21ras antibody (1:20�C1:50 in 0.1% BSA-PBS).

The detection protocol was carried out using biotinylated secondary antibodies and a streptavidin�Cperoxidase complex (LSAB kit, Dako). The reaction was developed by incubating samples in the substrate�Cchromogen solution AV-951 (1mgml?1 3,3��-diaminobenzidine tetrahydrochloride containing 0.02% H2O2) for 5min in the dark. Finally, the slides were mounted with Universal Mount and observed with a DMRB Leica microscope equipped with a �� 100 oil immersion lens. Each treatment was performed in three wells and observations were made from 10 fields from each well.