The sub-classification of ovarian carcinomas, which is based on histopathological subtype and grade, is unable to adequately predict prognosis or response to treatment. Administration of adjuvant platinum/taxane chemotherapy is the standard treatment for ovarian carcinoma following debulking surgery [58
]. Approximately 70% of advanced-stage ovarian carcinomas, however, will recur with development, ultimately of platinum-resistant disease [59
]. A comparison of clinical outcomes between ovarian cancer patients with BRCA1 promoter hypermethylation to patients with BRCA1 mutations and wild-type BRCA1 genes demonstrated that patients with BRCA1 promoter hypermethylation had significantly shorter survival times compared to the other two groups [60
]. Ovarian carcinomas in patients with BRCA1, in contrast are associated with a favourable prognosis [23
]. This data suggests that different abnormalities in BRCA1 could be associated with different clinical outcomes and possibly distinct alterations in other underlying molecular abnormalities. For example, the colon cancers from patients with inherited mutations in DNA mismatch repair genes differ from sporadic cancers with microsatellite instability due to hypermethylation of the MLH1 promoters [62
], with respect to age of onset of disease, pathology, and molecular alterations [63
We evaluated 49 ovarian carcinomas and categorized them according to pathology and BRCA1 and BRCA2 status. We further evaluated 35 high grade serous/undifferentiated tumours that we divided into four groups based on BRCA1 mutation status, expression, and promoter hypermethylation. We observed increased positive p53 immunohistochemical staining, which correlated with negative p21 immunostaining, in cancers with epigenetic BRCA1 loss, compared to cancers with BRCA1 mutations and high grade serous/undifferentiated cancers without BRCA1 loss. p53 is a tumor suppressor that is involved in the progression of many cancers and is the most commonly mutated gene in ovarian carcinomas [37
]. Typically, mutations in p53 result in accumulation of p53 in the nucleus and the majority of cases with abundant p53 detectable by immunohistochemistry are p53 mutant [64
]. The p53 protein is an important mediator of apoptosis resulting from DNA damage, stress, or chemotherapy [66
]. p21 is a downstream effector of the cell cycle arrest function of p53 and is upregulated at the transcriptional level by wildtype but not mutated p53 [67
]. At present, p53 mutation status or expression is not used to guide clinical decisions [68
]. We have observed that p53 overexpression correlates, as expected, with loss of p21 expression. Furthermore the phenotype of p53 overexpression with loss of p21 is significantly more common is high grade serous/undifferentiated tumors with epigenetic loss of BRCA1 compared to high grade serous/undifferentiated tumors without loss of BRCA1.
In addition, we found that ovarian carcinomas with loss of BRCA1 through genetic events and those with BRCA1 loss through epigenetic events both have activation of the PI3K/AKT pathway, though the mechanism of activation is different. PI3K phosphorylates phosphatidylinositol lipids in response to activation by receptor tyrosine kinases [69
]. Its activity has been linked to proliferation, differentiation, cell adhesion, apoptosis, tumorigenesis, and angiogenesis [70
]. PTEN is a phosphatase whose activity counters PI3K. The serine/threonine kinase AKT is a downstream target of PI3K [71
] and the activity of one of its isoforms, AKT1, is elevated in ovarian carcinomas [72
]. Both decreased PTEN levels and amplification of PIK3CA lead to increased phosphorylation of AKT and it has been previously shown that ovarian cancers often have alterations in PI3K and PTEN [73
]. This is the first study, however, to report that decreased PTEN expression levels are associated with ovarian carcinomas carrying BRCA1 mutations while increased PI3KCA copy number is associated with ovarian carcinomas with epigenetic loss of BRCA1. It has been previously demonstrated in breast cancer and glioblastoma that PIK3CA mutations and PTEN loss are virtually mutually exclusive, suggesting that abrogation of just one of these proteins is sufficient for tumorigenesis [76
]. We observe a similar result in our ovarian serous cancer samples; it is likely that deregulation of this pathway, while critical for tumorigenesis, can be accomplished through alteration of a single key molecule at which point selective pressure is relieved for altering other proteins involved in this signalling pathway. Our findings in serous ovarian carcinoma, and the previous results from studies of breast cancer and glioblastoma, are in contrast to ovarian carcinoma of endometrioid type [78
], and endometrial cancer [79
], where PTEN mutations and PIK3CA mutations frequently co-exist. This is yet another example of molecular events during the genesis of ovarian cancer that show tumor cell type specificity, and reinforces the need to consider cell type differences in studies of ovarian cancer pathogenesis.
We would expect that cancers with activation of the PI3K/AKT pathway may not respond well to common chemotherapy, as overexpression of activated AKT decreases apoptosis induced by paclitaxel in ovarian cancer cells [81
] and introduction of the catalytic subunit of PI3K into ovarian cancer cells causes resistance to paclitaxel [82
]. In addition, the PI3K inhibitor, LY294002, has been shown to decrease growth of ovarian carcinoma and ascites formation in mouse xenograft models of ovarian carcinoma [83
]. As therapies continue to be developed that target the PI3K/AKT pathway, it will be essential to understand the molecular alterations that are affecting this pathway in different types of ovarian carcinomas.
The need for meaningful sub-classification of ovarian carcinoma is critical for improving the treatment and prognosis of patients. Though sub-classification may be done based on BRCA1 genetic testing, this cannot be done in a timely fashion such that it could be used to guide therapy of patients newly diagnosed with ovarian cancer. This is extremely important as patients must embark on therapy shortly after diagnosis. In addition, as new therapeutics are developed, rapid identification of appropriate patients will be necessary for clinical trials. Our results demonstrate that it may be possible to categorize patients based on rapid molecular tests to identify patients who are likely to harbour BRCA1 mutations. Negative BRCA1 immunohistochemical staining, decreased BRCA1 mRNA, lack of PI3K amplification, and absence of BRCA1 promoter hypermethylation is an example of a molecular profile that could be used to identify these patients. This would also allow more cost effective and efficient mutation screening in patients presenting with ovarian carcinoma.