Human ovarian cancer cell lines, OVCAR3, SK-OV-3, and SNU119, were obtained from the Korean Cell Line Bank and grown in monolayer cultures in RPMI 1640 supplemented with fetal bovine serum (10% v/v).

Primary ovarian cancer samples were collected from the Seoul National University Hospital from patients undergoing cytoreductive surgery. Tumour tissue was dissected into a 100 mm Petri dish containing serum-free RPMI medium supplemented with trypsin. After 30 min at 37°C, the cells were resuspended and maintained in RPMI 1640 supplemented with fetal bovine serum (10% v/v).

Tissue specimens for immunohistochemistry were obtained from 160 epithelial ovarian carcinomas by primary cytoreductive surgery between 1998 and 2005. All samples were taken from primary ovarian lesions in each patient. The disease stage of each sample was determined according to the International Federation of Gynecology and Obstetrics (FIGO) criteria. Of the patients, 82 experienced a relapse and 37 died of ovarian cancer. Ten normal ovarian tissue samples were obtained as controls. Formalin-fixed, paraffin-embedded ovarian carcinoma tissues were used for histological evaluation. All haematoxylin and eosin (H & E)-stained sections were reviewed by gynaecological pathologists. The study was approved by the Institutional Review Board.

Protein concentrations were measured with the Bradford protein assay before analysis. Total cell lysate (30 μg of protein) was separated by SDS–polyacrylamide gel electrophoresis and subjected to western blot analysis. Anti-BubR1 antibody (BD, San Jose, CA, USA) was used at a dilution of 1 : 1000 and anti-Aurora A antibody (Transgenic, Japan) at a dilution of 1 : 300.

Cells grown on cover slips were fixed with 4% paraformaldehyde, then permeabilised with 0.5% Triton X-100 phosphate-buffered saline (0.5% PBS-T). They were then incubated in blocking solution (20% goat serum in 0.1% PBS–T) and probed with anti-BubR1 antibody or anti-Aurora A antibodies. The cells were mounted with Vectashield containing 4′,6-diamino-2-phenylindole (DAPI) (Vector Laboratories, Burlingame, CA, USA). Images were acquired on the DeltaVision microscope (Applied Precision, Seattle, WA, USA) as a series of 0.4-μm-thick sections and merged.

Spreads of metaphases were described previously with slight modifications (Patel et al, 1998; Lee et al, 1999).

Paraffin blocks from ovarian carcinoma were cut at 4 μm adjacent to H & E sections. Samples were deparaffinised in xylene, rehydrated with graded ethanol, and washed in distilled water. The sections were then placed in 10 mM citrate buffer (pH 6.0) and boiled in a microwave for epitope retrieval. Endogenous peroxidase activity was quenched by incubating tissue sections in 3% H₂O₂ for 10 min. After the blocking procedure with 20% goat serum and 1% bovine serum albumin (BSA) in PBS at room temperature for 1 h, sections were incubated with a primary antibody against BubR1 (mouse monoclonal, BD) at a dilution of 1 : 300 or Aurora A (rabbit polyclonal, Transgenic) at a dilution of 1 : 250 in a humidifying chamber at 4°C overnight. They were then washed in PBS for 5 min at room temperature, subsequently stained by the labelled streptavidin biotin (LSAB) method using a Dako LSAB kit (Dako, Glostrup, Denmark), and visualised using 3,3′-diaminobenzidine. The sections were then counterstained with haematoxylin.

To check the nuclear staining of Aurora A, antigen retrieval in 20 mM Tris–EDTA buffer (pH 9.0) was also performed in 80 samples using antibodies from Novocastra (NCL-L-AK2, New Castle, UK) at a dilution of 1 : 50 and following the methods of Burum-Auensen et al (Burum-Auensen et al, 2007a). The results were similar to the antigen retrieval procedure using citrate (pH 6.0) and Tris–EDTA (pH 9.0) buffer described above (Supplementary Figure). Slides were subjected to the same staining procedure without the addition of primary antibodies for negative controls. Microscopic fields from each stained section were randomly sampled.

In most cases, BubR1 and Aurora A were stained in the cytoplasm in interphase. However, a few specimens revealed nuclear staining of Aurora A, especially when Tris–EDTA buffer (pH 9.0) was used. The percentage of positive staining distribution was recorded and scored as follows: a score of 0 for staining <5%, 1 for 6–25% staining, 2 for 26–50% staining, 3 for 51–75% staining, and 4 for staining >75%. The staining intensity was scored as follows: a score of 0 for absent tumour cell staining, 1 for weak staining (equivocal to normal epithelium), 2 for moderate staining, and 3 for strong staining. For BubR1 and Aurora A, the results for intensity and distribution were summed and total score was assigned as follows: sum of 0–1, score 0; sum of 2–3, score 1; sum of 4–5, score 2; sum of 6–7, score 3.

The relationships between categorical variables were assessed using the χ² test. Recurrence-free survival (RFS) was estimated using the Kaplan–Meier method, and the differences in survival were compared using the log-rank test. Multivariate analysis was carried out using the Cox regression method. A P-value of <0.05 was considered to be statistically significant. Data were analysed using SPSS software, version 12.0 (SPSS Inc., Chicago, IL, USA).

We used western analysis to determine whether the two proteins are expressed in OVCAR3, SK-OV-3, and SNU119 ovarian cancer cells. The results show that all three cancer cell lines, which were randomly picked, express BubR1 and Aurora A to varying degrees (Figure 1A). As accurate mitosis is guaranteed by the proper localisation of BubR1 at kinetochores and of Aurora A at centrosomes, we examined the levels and localisation of BubR1 and Aurora A by immunofluorescence microscopy. As revealed in Figure 1B, all three cell lines analysed displayed defective Aurora A or BubR1 immunostaining at appropriate locations. Furthermore, the patterns of BubR1 and Aurora A staining were heterogeneous from cell to cell in the same cell line (data not shown).

Chromosome instability is accompanied by a heterogeneity of chromosome numbers in individual cells. Therefore, cytogenetic analysis of a statistically significant number of cells is the most direct measure of assessing aneuploidy. To confirm that these three cell lines exhibit CIN, we analysed metaphase chromosome spreads of 10 different cells from each line. The number of chromosomes in all cancer cell lines (OVCAR3, SK-OV-3, and SNU119) was highly variable in 10 different metaphase chromosome spreads (Figure 1C). This result is consistent with information available from the American Culture Type Collection. A representative chromosome spread revealed that one of the cell lines analysed, OVCAR3, exhibited aberrant chromosome fusions, as well as aneuploidy (Figure 1D).

As immunofluorescence assay (IFA) in ovarian cancer cell lines depicted a varying degree of BubR1 or Aurora A levels and subcellular localisation in mitosis (Figure 1B), we next asked whether primary tumour cells exhibit similar results. The result showed that the primary tumour cells also displayed mis-localisations and heterogeneity of BubR1 or Aurora A levels in IFA (Figure 1E).

Immunohistochemistry staining was conducted on 170 paraffin-embedded samples: 160 ovarian cancer and 10 normal ovary tissues. The age of cancer patients ranged from 18 to 79 years (median, 51.5 years). Table 1 describes the clinicopathological characteristics of 160 cancer patients. Except for 10 patients, all the cancer patients were at stage Ic or higher and hence underwent primary debulking surgery, followed by taxane and platinum chemotherapy. Serous cystadenocarcinoma (57.5%) was the most common histology. Cytoreductive surgery was optimal (⩽1 cm residual disease) in 51.9%. More than two-thirds of patients had a complete response to first-line treatment, yet half of them relapsed (Table 1).

Of the 160 ovarian cancer samples, a quarter of them (24.2%) were scored 0 for BubR1 immunostaining and about 30% scored 0 for Aurora A, whereas normal specimens scored 0 in all cases (Table 2, Figure 2). The relationship between levels of BubR1 and the clinicopathological parameters of 160 tumours is shown in Table 3. Positive BubR1 expression was associated with advanced stage (P=0.001), serous histology (P=0.008), high grade (P=0.001), and residual tumour (P=0.021). In contrast, no significant correlation was observed between the expression of BubR1 and Aurora A. In univariate survival analysis, BubR1 immunoreactivity was a significant prognostic factor for RFS (P<0.001), whereas the Aurora A expression was not (P=0.633). Patients with immunoreactive BubR1 (score 1, 2, 3) had a median RFS of 27 months (range, 3–48 months), whereas those with a negative expression of BubR1 (score 0) had a median RFS of 83 months (range, 2–115 months). Figure 3 shows survival curves stratified by BubR1 and Aurora A expression. Multivariate survival analysis (Table 4) indicated that the BubR1 level was an independent prognostic factor for predicting recurrence of ovarian cancer patients.