Ovarian cancer (OVC) is the fifth leading cause of cancer death among women in the United States, with 21,880 new cases diagnosed annually. This is a particularly deadly cancer, in part because it often goes undetected until it has spread to much of the peritoneal cavity. Indeed, 70% of the estimated 13,850 deaths attributed to ovarian cancer last year were of patients presenting with advanced-stage, high-grade serous ovarian carcinoma. The standard treatment is aggressive surgery followed by platinating chemotherapy; in 31% of patients, chemo-resistant tumors recur within six months. The overall five-year survival is 31%. In short, new treatments are desperately needed.
Two important studies shed new light on the genetic alterations driving ovarian cancer. In Nature, The Cancer Genome Atlas Research Network has published its integrated analysis of ovarian carcinoma. In PNAS, a group led by researchers at Dana-Farber Cancer Institute performed systematic loss-of-function studies in cell lines of ovarian and other cancers.
TCGA employed an arsenal of high-throughput technologies to systematically catalogue molecular aberrations in nearly 500 OVC cases:
- DNA copy number, assessed by high-density CGH and SNP arrays (489 cases)
- mRNA and miRNA expression, profiled on Affymetrix and Agilent arrays (489 cases)
- Promoter methylation, assessed on Illumina array (519 cases)
- Coding mutations, assessed by whole-exome sequencing on Illumina (236 cases) and ABI SOLiD (80 cases).
Altogether, these datasets enabled integrated copy-number/expression/methylation analyses of 489 cases and integrated copy-number/expression/methylation/mutation analyses in 316 cases.
Mutations in Ovarian Carcinoma
Exome sequencing targeted ~180,000 exons from ~18,500 genes totaling ~33 Mbp of non-redundant sequence, of which 76% were sufficiently covered for mutation detection (on average). A total of 19,356 mutations were catalogued, which works out to around 61 per tumor. Consistent with previous reports, the TP53 gene was mutated in almost all tumors (96%), while germline BRCA1 and BRCA2 mutations were present in 9% and 8% of cases, respectively. Some six additional genes were found to have statistically significant rates of mutation:
|Gene||NCBI ID||Ensembl ID||Description|
|CSMD3||114788||ENSG00000164796||CUB and Sushi multiple domains 3|
|GABRA6||2559||ENSG00000145863||gamma-aminobutyric acid (GABA) A receptor, alpha 6|
|FAT3||120114||ENSG00000165323||FAT tumor suppressor homolog 3 (Drosophila)|
|CRKRS||51755||ENSG00000167258||Cdc2-related kinase, arginine/serine-rich (CDK12)|
Some of these hits are unsurprising; NF1 and RB1 both encode well-known tumor suppressors. CRKRS, also known as CDK12, is an interesting novel hit; it encodes a cyclin-dependent kinase required for RNA splicing. Deletions on chromosome 17 in gastric cancer often produce rearrangements involving CDK12 and ERBB2. Comparisons of the TCGA-OV mutation dataset to COSMIC and OMIM databases yielded 477 and 201 matches, respectively, including mutations in well-known cancer genes BRAF, PIK3CA, KRAS, and NRAS.
Gene Expression Subtypes
There were ~1,500 intrinsically variable genes in HGS-OV cases, which clustered robustly into four expression subtypes:
- Immunoreactive, characterized by T-cell cytokine ligands (CXCL10, CXCL11) and a receptor (CXCR3).
- Proliferative, which showed high expression of transcription factors (HMGA2, SOX11) and proliferation markers (MCM2, PCNA)
- Differentiated, associated with expression of ovarian tumor markers MUC16 and MUC1.
- Mesenchymal, marked by high expression of HOX genes and stromal component markers.
Survival duration did not vary significantly between expression subtypes. There were some notable correlations with copy number events: MECOM amplification was more frequent in the immunoreactive subtype, while MYC and RB1 alterations were less common in the proliferative subtype. The former observation may have some therapeutic implications, as MECOM was one of a handful of amplified genes that are targeted by at least one drug compound.
DNA Methylation Patterns
DNA methylation was correlated with reduced gene expression across all samples. The promoters of three genes in particular (AMT, CCL21, and SPARCL1) were hyper-methylated (silenced) in most tumors. As previously reported, the BRCA1 promoter was epigenetically silenced in >10% of tumors. One surprising finding was that RAB25, which has been reported to be over-expressed in ovarian carcinomas, was epigenetically silenced in a subset of tumors.
Copy Number Alterations
GISTIC analysis of copy number data for 489 cases identified 63 regions of focal amplification. Three such amplifications (CCNE1, MYC, and MECOM) were present in at least 20% of cases. New tightly localized amplification peaks in ovarian cancer encoded the receptor for activated C-kinase (ZMYND8), a p53 target gene (IRF2BP2), a telomerase catalytic subunit (TERT), and, notably, an embryonic development gene (PAX8) that featured in the PNAS study. There were also 50 focal deletions; among these were known tumor-suppressor genes PTEN, RB1, and NF1.
Investigation of Genetic Vulnerabilities
Cheung and colleagues took a complementary approach. To identify genes essential for tumor proliferation and survival, they performed systematic RNAi screens of 11,194 genes in 102 cancer cell lines. Each cell line was infected in quadruplicate with a pool of lentivirus-delivered short hairpin RNA, and then assessed for the effect on proliferation.A weight-of-evidence statistic served to evaluate and rank the “essentiality” of each gene in each cell line.
BRAF and KRAS mutations, unsurprisingly, were ranked highly in cell lines harboring those mutations. Theirs was a striking effect, such that the top-scoring shRNAs for KRAS and BRAF easily distinguished cell lines that were mutated (mutant) or not (wild-type) for those genes. In PIK3CA-harboring cell lines, MTOR shRNAs scored highly, and could distinguish between PIK3CA-mutated and PIK3CA-wildtype tumors. This observation supports prior work that mTOR plays a role in PI3K signaling.
Tumor Lineage-Specific Dependencies
Next, the authors assessed all shRNAs for their ability to distinguish one specific tumor lineage (tumor type and cell type) from all others. Here, they build on recent work establishing that oncogenic transcription factors are often amplified, over-expressed, and essential in subsets of tumors from specific lineages (e.g. NKX2-1 in lung adenocarcinoma, MITF in melanoma, and SOX2 in squamous cell carcinoma). Therefore, the authors searched not only for shRNAs that distinguished specific lineages, but also those that were amplified or over-expressed in that tumor type.
- In colon cancer cell lines, KRAS, CTNNB1, and BRAFwere scored as essential, and KRAS and IGF1R were both essential and amplified. MYB and AXIN2 were scored as essential and also differentially expressed.
- In pancreatic cancer cell lines, KRAS stood out as essential, and SOX9 emerged as a lineage-specific dependency.
- In NSC lung cancer cell lines, NKX2-1 was essential, amplified, and over-expressed. There were seven genes essential and amplified, including CDK6. Interestingly, MAP2K4, an activator of JNK and p38, showed selective essentiality and expression.
Deeper Analysis of Ovarian Cancer
The authors next leveraged information from TCGA’s ovarian cancer study for a deeper analysis of gene essentiality. Among the 1,825 genes in amplified regions reported by TCGA, some 50 were deemed essential in the ovarian cancer cell lines. They included the known oncogene CCNE1, along with adaptor protein FRS2, RPTOR, and the PRKCE protein kinase.
One gene, PAX8, hit the trifecta. It was amplified in 16% of primary ovarian tumors. It was over-expressed in 21 of 25 ovarian cancer cell lines, and scored as “essential” by all three scoring methods the authors used. Further, when ovarian cancer cell lines were compared to other cell lines, PAX8 was the most-differentially-expressed gene. Suppression of PAX8 but had no effect on the proliferation of immortalized ovarian surface epithelial cells or 8 other cell lines that did not express PAX8. However, it induced apoptosis and reduced viability by >50% in 6/8 ovarian cell lines. Three of these had PAX8 amplifications, and the other three had PAX8 over-expression.
PAX8 encodes a lineage-specific transcription factor involved in the development of the thyroid, kidney, and female reproductive tract. In the latter, its expression is restricted to secretory cells of the fallopian tube epithelium, which (according to recent reports) is a point of origin for serous ovarian carcinoma. Previous studies have also shown that PAX8 is amplified in ovarian cancer; immunohistochemistry suggests that it is expressed in 90-100% of serous, clear-cell, and endometrioid subtypes.
Tumor Genomes and Genetic Vulnerabilities
I think that the TCGA study is important because it demonstrates the discovery power of integrated analysis – mutation, gene expression, copy number, and methylation combined. When all of these data types are put together, a comprehensive picture of the tumor emerges – the landscape of its genomic alterations and epigenetic changes, and the ensuing differential gene expression.
Combined with systematic functional studies, this atlas of ovarian cancer became even more powerful. A systematic and fairly unbiased analysis of genes essential for the tumor’s survival, reinforced with genomic data, yielded a possible weak point common to most cancers of that type, and offering a valuable starting point for new therapeutic approaches.