Mouse anti-human Mcm2 monoclonal antibody (clone 46) was obtained from BD Transduction Laboratories (Lexington, KY, USA). Mouse anti-human Ki67 monoclonal antibody (clone MIB-1) was obtained from DAKO (Glostrup, Denmark). Affinity-purified rabbit polyclonal antibody against full-length hsgeminin protein was generated as described (Wharton et al, 2004). Anti-MEK5 polyclonal antibody was obtained from Stressgen Biotechnologies (San Diego, CA, USA), anti-ERK5 polyclonal and anti-HA monoclonal antibody from Santa Cruz Biotechnology (Santa Cruz, CA, USA), and anti-α-tubulin monoclonal antibody from Sigma-Aldrich (Gillingham, Dorset, UK). Anti-ERK5 polyclonal antibody for use in immunohistochemistry was a kind gift from Professor P Cohen (University of Dundee, UK) (Mody et al, 2001).

An ecdysone-inducible system expressing a constitutively activated mutant of MEK5 in EcR293 cells was used to determine the effects of increased MEK5/ERK5 signalling on the DNA RLP. Stable clone EcRD No. 16 has previously been shown to have negligible background and high Ponasterone A (PonA)-induced MEK5D expression (Mehta et al, 2003). Following induction of MEK5 expression with 5 μM PonA for 24 h, whole cell lysates were subjected to SDS–PAGE and immunoblot analysis. In addition to the ecdysone-inducible system, an ERK5 over-expressing PCa stable clone was generated. ERK5 was amplified from a full-length clone (Bouras et al, 2001) using oligonucleotides MP405 (GGATCCGCCGAGCCTCTGAAGGAGGAAG) and MP1310 (GCGGATCCTCAGGGGTCCTGGAGGTCAGG) with GC Rich PCR System (Roche, Lewes, East Sussex, UK). The resulting fragment was cloned into pCR2.1 (Invitrogen, Paisley, UK), sequenced, then subcloned into EGFP-C1 (Clontech, Palo Alto, CA, USA) to form EGFP-C1 ERK5. PC3 cells overexpressing ERK5 were generated by transfection with EGFP-C1 ERK5 using Superfect® Transfection Reagent (Qiagen, Crawley, UK) according to the manufacturer's recommendations. Following selection with geneticin (G418, 1 mg ml⁻¹; Sigma-Aldrich, Gillingham, Dorset, UK), colonies were left to form for 2–3 weeks. Control stable cell lines were similarly established with the vector (EGFP-C1) alone. Individual colonies were subcultured and screened for ERK5 expression by immunoblotting. Presence and subcellular localisation of ERK5 were also determined by fluorescent microscopy.

Cells were lysed directly on plates with 6 × SDS sample buffer containing 10% β-mercaptoethanol. Samples were denatured and proteins separated by SDS–PAGE, followed by transfer to nitrocellulose. Antibodies were used at the following dilutions: anti-HA, 1 : 500; anti-ERK5, 1 : 1000; Mcm2, 1 : 3000; Ki67, 1 : 2500; geminin, 1 : 1000; and anti-α tubulin, 1 : 2000. Horseradish peroxidase-conjugated secondary antibodies were applied (1 : 500) and detected using an enhanced chemiluminescence detection kit (ECL, Amersham, Bucks, UK).

PC3 cells stably overexpressing ERK5 (PC3-ERK5) and PC3 cells overexpressing the empty vector alone (PC3-EmptyVector) were seeded onto sterile coverslips in 6-well plates (Corning Incorporated, Corning Life Sciences, One The Valley Centre, Gordon Road, High Wycombe, Bucks, UK) at densities of 3 × 10⁴ cells/well. After 24 h, cells were washed and incubated in Basal Media overnight. Cells were stimulated with EGF for 30 min and immediately fixed with 100% methanol at −20°C for 30 min. To prevent non-specific staining, fixed cells were first blocked in 10% natural rabbit serum (DakoCytomation) for 30 min at RT. Blocking solution was subsequently removed and primary antibodies were added at the following dilutions: anti-ERK5, 1 : 100; anti-Mcm2, 1 : 200. Coverslips were incubated overnight in a humidified atmosphere at 4°C. Cells were washed (× 3) and incubated with rabbit anti-goat secondary antibody (1 : 250, DakoCytomation) and rabbit anti-mouse secondary antibody (1 : 250, DakoCytomation) for 1 h at RT in the dark. Vectashield with 4,6-diamidino-2-phenylindole (DAPI), for nuclear counterstain, was used to mount slides. Images of fixed cells were acquired with a Leica TCS SP2 UV laser-scanning microscope using a × 63 oil immersion lens (1.32 NA DIC). A series of 1 μm vertical optical sections through the entire thickness of the cells was used to produce a Z-series and all images were analysed using LCS 2.00.585 software. Controls, lacking primary antibody, failed to show any staining.

Ethical approval for tissue studies was obtained from the research ethics committee of the Freeman Hospital, Newcastle-upon-Tyne, UK. Consent for the use of surplus tissue obtained at the time of diagnosis was gained from patients undergoing transurethral resection of the prostate for bladder outflow obstruction. Diagnosis of PCa was made by histopathological examination. This cohort of patients was managed by surveillance and subsequent androgen suppression. Formalin-fixed and paraffin wax-embedded tissue blocks containing greater than 75% tumour involvement were selected for analysis, and tissue sections were cut at 4 μm thickness. The cohort included 58 PCa patients with a median age of 75 years (range 55–88). Included in the series were Gleason grades from 6 to 10 (Gleason 6: n=9 (16%), Gleason 7: n=7 (12%), Gleason 8: n=25 (43%), Gleason 9: n=12 (21%), Gleason 10: n=5 (9%)) and a range of clinical tumour (T) stages as determined by rectal examination (T1: n=13 (22%), T2: n=15 (26%), T3: n=17 (29%), T4: n=5 (9%)). Staging data were missing in eight (14%) patients. Nearly all patients (56/58) had a radionucleide bone scan at the time of diagnosis. The group included 28 (48%) patients with and 28 (48%) patients without evidence of bony metastasis. It should be noted that the patients in this cohort were recruited in the pre-PSA era, and thus PSA data were not available. At the time of our study, 45 (78%) patients had died and eight (14%) were alive. Survival times for five patients were unknown, and these patients were excluded from the survival analysis. In the survivors the follow-up ranged from 2 to 142 months, with a mean of 71.6 months (s.d.=55.4).

The association between all-cause mortality and marker expression was assessed using Kaplan–Meier curves and the log-rank test, using the ‘minimum P-value approach' with adjusted P-values and a significance level of 0.05 (Faraggi and Simon, 1996). A multivariate Cox proportional hazards model was fitted using a backwards stepwise procedure to investigate trends between marker expression and the standard prognostic variables Gleason grade, T stage, presence of bony metastases, and survival. Statistical significance of trends between markers and standard prognostic variables was assessed using the Mann–Whitney U and Jonckheere–Terpstra tests. In this secondary analysis, P-values of less than 0.01 were regarded as statistically significant. The Wilcoxon signed-rank test was used to compare overall labelling indices of Mcm2, geminin and Ki67. Analysis was performed using SPSS 12.0 for Windows (SPSS Inc., Chicago, IL, USA) and R version 2.0.

To investigate potential linkages between the MEK5/ERK5 signalling pathway and the downstream DNA replication (or origin) licensing pathway (RLP), we first used an ecdysone-inducible system to express a constitutively activated mutant of MEK5 (MEK5D) in EcR293 cells (Mehta et al, 2003). Clone EcRD No. 16 showed negligible background MEK5 expression (Mehta et al, 2003). Treatment of cells for 24 h with Ponasterone A (PonA) resulted in a two-fold increase in the percentage of cells in S phase (Figure 1A). MEK5D expression was potently induced and was associated with a shift in the size of the ERK5 band, in keeping with ERK5 phosphorylation (Figure 1B). ERK5 is the only known substrate of MEK5 and, like other MAPK family members, it is activated by phosphorylation (Abe et al, 1996). Induction of the MEK5/ERK5 signalling pathway was coupled to a marked increase in the expression of Mcm2 and geminin, and also to upregulation of the standard proliferation marker Ki67 (Figure 1C). Next we used PC3 cells stably overexpressing ERK5 (PC3-ERK5 cells) to confirm our findings in a PCa model. PC3 cells were shown to have a background level of endogenous ERK5 (Figure 2A). The ERK5 overexpressing clone produced much higher levels of EGFP-fused ERK5 and WST-1 cell proliferation assays showed a ∼3-fold increase in proliferation compared to empty vector and parental controls. In keeping with the observations in the inducible system, the ERK5 overexpressing PC3 line expressed higher levels of Mcm2, geminin and Ki67 protein (Figure 2B–D). Indirect immunofluorescence studies also showed increased Mcm2 expression in ERK5 overexpressing cells (Figure 2E–F), confirming the immunoblot data, which give a population mean, at the level of individual cells. Taken together, these data demonstrate linkage of the MEK5/ERK5 signalling pathway with the replication licensing system in human prostatic cancer cells.

To investigate coupling of MEK5/ERK5 signalling to the RLP in vivo and cell cycle kinetics of PCa, we analysed biopsy material from a cohort of PCa patients diagnosed after transurethral resection of the prostate (for clinical characteristics of the study cohort see Materials and Methods). Clinicopathological and immunohistochemical data were available for the entire cohort; however, five patients have been excluded from the survival analysis because survival times were unknown.

The pattern of RLF expression was first assessed in morphologically normal prostate tissue present in the biopsy material. Expression of Mcm2, geminin and Ki67 was extremely low (<2%), consistent with our previous finding that loss of proliferative capacity which accompanies differentiation is coupled to repression of origin licensing through down-regulation of the Mcm2–7 RLFs (Stoeber et al, 2001; Eward et al, 2004). In PCa, on the contrary, Mcm2 and geminin protein levels were high, indicative of cell cycle re-entry (Kingsbury et al, 2005).

Distribution of LIs, median and inter-quartile ranges for each marker are shown in Table 1 and Figure 3, and immunohistochemical staining of representative cases of Gleason grade 6, 8 and 10 PCa is illustrated in Figure 4. Mcm2 protein expression in these tumours was significantly greater than Ki67 expression, which was itself significantly higher than geminin expression [median: Mcm2, 36.7% Ki67, 18.1% geminin, 3.4% (P<0.0001, Wilcoxon signed-rank test)]. Furthermore Mcm2 expression was distributed over a far broader range than Ki67 or geminin (Mcm2: 2.7–100% Ki67: 2.2–77.8% geminin: 0.3–37.2%). The high percentage of Mcm2-expressing cells compared to Ki67 can be explained by the presence of replication licensed but non-cycling cells in addition to proliferating cells (Stoeber et al, 2001; Blow and Hodgson, 2002). The low growth fraction identified by geminin compared with Ki67 is due to geminin expression being restricted to S, G2 and M phase (Eward et al, 2004; Wharton et al, 2004). There was a clear association of increasing Mcm2, geminin and Ki67 LIs with increasing Gleason grade (Mcm2, P=0.0002; geminin, P=0.0003; Ki67, P=0.004; Table 1 and Figure 4). Figure 5 shows more clearly the mean and distribution of Mcm2 LIs for each grade. Although there is an increase in mean Mcm2_LI with increasing grade, notably a high Mcm2_LI does not preclude the possibility of low Gleason grade. A statistically significant link between increasing Gleason grade and Ki67_LI-geminin_LI (P=0.001) was also observed, indicating an increase in the population of cells entering G1 phase in more poorly differentiated tumours. There was no significant association between T stage and bony metastases with Mcm2, geminin or Ki67 expression.

We sought to identify the relationship between the MEK5/ERK5 growth signalling pathway and the RLFs Mcm2 and geminin, as well as the standard proliferation marker Ki67. Table 1 summarizes the links between these markers according to the degree of protein expression. The trends of increasing Mcm2, geminin and Ki67 expression with increased MEK5 expression are less marked. These weaker associations may reflect the fact that MEK5 lies upstream of ERK5 in the growth signalling pathway. We have previously stated that the RLP lies at a convergence point in growth signalling (Williams and Stoeber, 1999; Stoeber et al, 2001) and these data strongly support this hypothesis.

We have previously shown that RLF expression analysis can provide novel insights into the cell cycle kinetics of tumours. The Mcm2_LI minus Ki67_LI (Mcm2−Ki67) identifies the replication licensed but non-proliferating population of tumour cells, whereas the Ki67_LI minus geminin_LI (Ki67−geminin) represents the proportion of tumour cells in G1 phase (Stoeber et al, 2001, 2002; Dudderidge et al, 2005; Obermann et al, 2005). Geminin identifies the proportion of tumour cells traversing through S-G2-M phase, and therefore the geminin_LI to Ki67_LI ratio (geminin/Ki67) provides a measure of the relative length of G1 in dynamic cell populations, with higher geminin/Ki67 ratios corresponding to a shortening in the length of G1 phase (Eward et al, 2004; Wharton et al, 2004; Shetty et al, 2005). In this study, a statistically significant link between increasing Gleason grade and Ki67-geminin (P=0.001; Table 1) was observed, indicating an increase in the proportion of cells entering G1. Moreover, there was a significant relationship between increasing ERK5 protein expression and increasing Mcm2 (P=0.003) and Ki67 (P=0.009) expression (Table 1), recapitulating the data generated in the in vitro tissue culture model system (Figure 1). Furthermore, there was a statistically significant link between high ERK5 expression and increased Mcm2-Ki67 (P=0.01; Table 1), signifying an increase in the proportion of non-cycling tumour cells that are licensed for replication. We also observed a significant trend of increasing ERK5 expression with increasing geminin/Ki67 ratio, indicative of a relative shortening in the length of G1 phase and thus accelerated cell cycle progression (P=0.005; Table 1 and Figure 6).

The association between all-cause mortality and biomarker expression was assessed using Kaplan–Meier curves and the log-rank test, using the ‘minimum P-value approach' with adjusted P-values and a significance level of 0.05 (Faraggi and Simon, 1996). Survival analyses were performed for Mcm2, geminin and Ki67. A statistically significant association was found between lower Mcm2 protein expression and increased survival time (P=0.03; Figure 7A), and between lower Ki67 expression and increased survival time (P=0.002; Figure 7B). After adjustment of the P-value (Faraggi and Simon, 1996), the trend between lower geminin expression and increased survival time was not statistically significant (P=0.22; Figure 7C). The optimal cut points used were 44% for Mcm2, 25% for Ki67 and 9% for geminin. A Cox proportional hazards model was fitted to investigate trends between biomarker expression, the standard prognostic variables of Gleason grade, stage and survival. Increased expression of Mcm2 (P=0.005), geminin (P=0.04) and Ki67 (P=0.04) and the replication licensed but non-cycling cell population (Mcm2-Ki67; P=0.01) were all associated with significantly reduced survival time. However, only the strongest predictor of survival, Mcm2, was retained in the final model, as after adjustment for Mcm2 expression neither geminin, Ki67, Mcm2-Ki67, Ki67-geminin, geminin/Ki67 nor any of the other prognostic variables were significantly associated with survival time (hazard ratio for Mcm2=4.22; 95% CI (1.56, 11.45); P=0.005). Figure 7D shows the relationship between Mcm2 protein expression and survival. No significant relationships between Gleason grade and survival or stage and survival were identified. The presence of bony metastases was weakly associated with a reduction in survival time (hazard ratio 0.52; 95% CI (0.27, 1.00); P=0.05).