Integrated genomic analyses of ovarian carcinoma.

A catalogue of molecular aberrations that cause ovarian cancer is critical for developing and deploying therapies that will improve patients' lives. The Cancer Genome Atlas project has analysed messenger RNA expression, microRNA expression, promoter methylation and DNA copy number in 489 high-grade serous ovarian adenocarcinomas and the DNA sequences of exons from coding genes in 316 of these tumours. Here we report that high-grade serous ovarian cancer is characterized by TP53 mutations in almost all tumours (96%); low prevalence but statistically recurrent somatic mutations in nine further genes including NF1, BRCA1, BRCA2, RB1 and CDK12; 113 significant focal DNA copy number aberrations; and promoter methylation events involving 168 genes. Analyses delineated four ovarian cancer transcriptional subtypes, three microRNA subtypes, four promoter methylation subtypes and a transcriptional signature associated with survival duration, and shed new light on the impact that tumours with BRCA1/2 (BRCA1 or BRCA2) and CCNE1 aberrations have on survival. Pathway analyses suggested that homologous recombination is defective in about half of the tumours analysed, and that NOTCH and FOXM1 signalling are involved in serous ovarian cancer pathophysiology.

Background

Ovarian cancer is the fifth leading cause of cancer death among women in the U.S., with 21,880 new cases and 13,850 deaths predicted for 2010[1]. Most deaths are of patients presenting with advanced stage, high grade serous ovarian cancer (HGS-OvCa)[2,3] (~70%). The standard of care is aggressive surgery followed by platinum/taxane chemotherapy. After therapy, platinum resistant cancer recurs in approximately 25% of patients within 6 months[4] and overall 5-year survival is 31%[5]. Approximately 13% of HGS-OvCa is attributable to germline mutations in BRCA1 or BRCA2[6,7], while a smaller percentage can be accounted for by other germline mutations. However, most ovarian cancer can be attributed to a growing number of somatic aberrations[8].

The lack of successful treatment strategies led TCGA to comprehensively measure genomic and epigenomic abnormalities on clinically annotated HGS-OvCa samples in order to identify molecular abnormalities that influence pathophysiology, affect outcome, and constitute therapeutic targets. Microarray analyses produced high resolution measurements of mRNA expression, microRNA expression, DNA copy number, and DNA promoter region methylation for 489 HGS-OvCa while massively parallel sequencing coupled with hybrid affinity capture[9,10] provided whole exome DNA sequence information for 316 of these samples.

Samples and clinical data

This report covers analysis of 489 clinically annotated stage II-IV HGS-OvCa and corresponding normal DNA (Methods S1, Table S1.1). Patients reflected the age at diagnosis, stage, tumor grade, and surgical outcome of individuals diagnosed with HGS-OvCa. Clinical data were current as of August 25, 2010. HGS-OvCa specimens were surgically resected before systemic treatment but all patients received a platinum agent and 94% received a taxane. The median progression-free and overall survival of the cohort is similar to previously published trials[11,12]. Twenty five percent of the patients remained free of disease and 45% were alive at the time of last follow-up, while 31% progressed within 6 months after completing platinum-based therapy. Median follow up was 30 months (range 0 to 179). Samples for TCGA analysis were selected to have > 70% tumor cell nuclei and < 20% necrosis.

Coordinated molecular analyses using multiple molecular assays at independent sites were carried out as listed in Table 1. Data are available at http://tcga.cancer.gov/dataportal in two tiers. Tier one datasets are openly available, while tier two datasets include clinical or genomic information that could identify an individual hence require qualification as described at http://tcga.cancer.gov/dataportal/data/access/closed/.

Table 1

Characterization platforms used and data produced

Data Type	Platforms	Cases	DataAvailability
DNA Sequence of exome	IlluminaGAIIxa,bABI SOLiDc	23680	ProtectedProtected
Mutations present in exome		316	Open
DNA copynumber/genotype	Agilent244Kd,eAgilent 415KdAgilent 1MeIllumina1MDUOfAffymetrixSNP6a	97304539535514	OpenOpenOpenProtectedProtected
mRNA expression profiling	AffymetrixU133AaAffymetrixExongAgilent 244Kh	516517540	OpenProtectedOpen
Integrated mRNAexpression		489	Open
miRNA expressionprofiling	Agilenth	541	Open
CpG DNA methylation	Illumina 27K1	519	Open
Integrative analysis		489	Open
Integrative analysis w/mutations		309	Open

Production Centers: Broad Institute, Washington University School of Medicine, Baylor College of Medicine, Harvard Medical School, Memorial Sloan-Kettering Cancer Center, HudsonAlpha Institute for Biotechnology, Lawrence Berkeley National Laboratory, University of North Carolina, University of Southern California.

Additional data are available for many of these data types at the TCGA DCC.

Mutation analysis

Exome capture and sequencing was performed on DNA isolated from 316 HGS-OvCa samples and matched normal samples for each individual (Methods S2). Capture reagents targeted ~180,000 exons from ~18,500 genes totaling ~33 megabases of non-redundant sequence. Massively parallel sequencing on the Illumina GAIIx platform (236 sample pairs) or ABI SOLiD 3 platform (80 sample pairs) yielded ~14 gigabases per sample (~9×109 bases total). On average, 76% of coding bases were covered in sufficient depth in both the tumor and matched normal samples to allow confident mutation detection (Methods S2, Figure S2.1). 19,356 somatic mutations (~61 per tumor) were annotated and classified in Table S2.1. Mutations that may be important in HGS-OvCa pathophysiology were identified by (a) searching for non-synonymous or splice site mutations present at significantly increased frequencies relative to background, (b) comparing mutations in this study to those in COSMIC and OMIM and (c) predicting impact on protein function.

Two different algorithms (Methods S2) identified 9 genes (Table 2) for which the number of non-synonymous or splice site mutations was significantly above that expected based on mutation distribution models. Consistent with published results[13], TP53 was mutated in 303 of 316 samples (283 by automated methods and 20 after manual review), BRCA1 and BRCA2 had germline mutations in 9% and 8% of cases, respectively, and both showed somatic mutations in an additional 3% of cases. Six other statistically recurrently mutated genes were identified; RB1, NF1, FAT3,CSMD3, GABRA6, and CDK12. CDK12 is involved in RNA splicing regulation[14] and was previously implicated in lung and large intestine tumors[15,16]. Five of the nine CDK12 mutations were either nonsense or indel, suggesting potential loss of function, while the four missense mutations (R882L, Y901C, K975E, and L996F) were clustered in its protein kinase domain. GABRA6 and FAT3 both appeared as significantly mutated but did not appear to be expressed in HGS-OvCa (Supplemental Figure S2.1) or fallopian tube tissue so it is less likely that mutation of these genes plays a significant role in HGS-OvCa.

Table 2

Significantly mutated genes in HGS-OvCa

Gene	Number ofMutations	Validated	Unvalidated
TP53	302	294	8
BRCA1	11	10	1
CSMD3	19	19	0
NF1	13	13	0
CDK12	9	9	0
FAT3	19	18	1
GABRA6	6	6	0
BRCA2	10	10	0
RB1	6	6	0

Validated mutations are those that have been confirmed with an independent assay. Most of them are validated using a second independent WGA sample from the same tumor. Unvalidated mutations have not been independently confirmed but have a high likelihood to be true mutations. An additional 25 mutations in TP53 were observed by hand curation.

Mutations from this study were compared to mutations in the COSMIC[17] and OMIM[18] databases to identify additional HGS-OvCa genes that are less commonly mutated. This yielded 477 and 211 matches respectively (Table S2.4) including mutations in BRAF (N581S), PIK3CA (E545K and H1047R), KRAS (G12D), and NRAS (Q61R). These mutations have been shown to exhibit transforming activity so we believe that these mutations are rare but important drivers in HGS-OvCa.

We combined evolutionary information from sequence alignments of protein families and whole vertebrate genomes, predicted local protein structure and selected human SwissProt protein features (Methods S3) to identify putative driver mutations using CHASM[19,20] after training on mutations in known oncogenes and tumor suppressors. CHASM identified 122 mis-sense mutations predicted to be oncogenic (Table S3.1). Mutation-driven changes in protein function were deduced from evolutionary information for all confirmed somatic missense mutations by comparing protein family sequence alignments and residue placement in known or homology-based three-dimensional protein structures using Mutation Assessor (Methods S4). Twenty-seven percent of missense mutations were predicted to impact protein function (Table S2.1).

Copy number analysis

Somatic copy number alterations (SCNAs) present in the 489 HGS-OvCa genomes were identified and compared with glioblastome multiforme data in Figure 1a. SCNAs were divided into regional aberrations that affected extended chromosome regions and smaller focal aberrations (Methods S5). A statistical analysis of regional aberrations (Methods S5)[21] identified 8 recurrent gains and 22 losses, all of which have been reported previously[22] (Figure 1b and Table S5.1). Five of the gains and 18 of the losses occurred in more than 50% of tumors.

Figure 1

Genome copy number abnormalities

(a) Copy-number profiles of 489 HGS-OvCa, compared to profiles of 197 glioblastoma multiforme (GBM) tumors[46]. Copy number increases (red) and decreases (blue) are plotted as a function of distance along the normal genome. (b) Significant, focally amplified (red) and deleted (blue) regions are plotted along the gnome. Annotations include the 20 most significant amplified and deleted regions, well-localized regions with 8 or fewer genes, and regions with known cancer genes or genes identified by genome-wide loss-of-function screens. The number of genes included in each region is given in brackets. (c) Significantly amplified (red) and deleted (blue) chromosome arms.

GISTIC[21,23] (Methods S5) was used to identify recurrent focal SCNAs. This yielded 63 regions of focal amplification (Figure 1c, Methods S5, Table S5.2) including 26 that encoded 8 or fewer genes. The most common focal amplifications encoded CCNE1, MYC, and MECOM (Figure 1c, Methods S5, Table S5.2) each highly amplified in greater than 20% of tumors. New tightly-localized amplification peaks in HGS-OvCa encoded the receptor for activated C-kinase, ZMYND8; the p53 target gene, IRF2BP2; the DNA-binding protein inhibitor, ID4; the embryonic development gene, PAX8; and the telomerase catalytic subunit, TERT. Three data sources: http://www.ingenuity.com/, http://clinicaltrials.gov and http://www.drugbank.ca were used to identify possible therapeutic inhibitors of amplified, over-expressed genes. This search identified 22 genes that are therapeutic targets including MECOM, MAPK1, CCNE1 and KRAS amplified in at least 10% of the cases (Table S5.3).

GISTIC also identified 50 focal deletions (Figure 1d). The known tumor suppressor genes PTEN, RB1, and NF1 were in regions of homozygous deletions in at least 2% of tumors. Importantly, RB1 and NF1 also were among the significantly mutated genes. One deletion contained only three genes, including the essential cell cycle control gene, CREBBP, which has 5 non-synonymous and 2 frameshift mutations.

mRNA and miRNA expression and DNA methylation analysis

Expression measurements for 11,864 genes from three different platforms (Agilent, Affymetrix HuEx, Affymetrix U133A) were combined for subtype identification and outcome prediction. Individual platform measurements suffered from limited, but statistically significant batch effects, whereas the combined data set did not (Methods S11, Figure S11.1). Analysis of the combined dataset identified ~1,500 intrinsically variable genes[24] (Methods S6) that were used for NMF consensus clustering. This analysis yielded four clusters (Methods S6, Figure 2a). Thesame analysis approach applied to a publicly available dataset from Tothill et al. [25], also yielded four clusters. Comparison of the Tothill and TCGA clusters showed a clear correlation (Methods S6, Figure S6.3). We therefore conclude that at least four robust expression subtypes exist in HGS-OvCa.

Figure 2

Gene and miRNA expression patterns of molecular subtype and outcome prediction in HGS-OvCa

(a) Tumors from TCGA and Tothill et al. separated into four clusters, based on gene expression. (b) Using a training dataset, a prognostic gene signature was defined and applied to a test dataset. (c) Kaplan-Meier analysis of four independent expression profile datasets, comparing survival for predicted higher risk versus lower risk patients. Univariate Cox p-value for risk index included. (d) Tumors separated into three clusters, based on miRNA expression, overlapping with gene-based clusters as indicated. (e) Differences in patient survival among the three miRNA-based clusters.

We termed the four HGS-OvCa subtypes Immunoreactive, Differentiated, Proliferative and Mesenchymal based on gene content in the clusters (Methods S6) and on previous observations[25]. T-cell chemokine ligands, CXCL11 and CXCL10, and the receptor, CXCR3, characterized the Immunoreactive subtype. High expression of transcription factors such as HMGA2 and SOX11, low expression of ovarian tumor markers (MUC1, MUC16) and high expression of proliferation markers such as MCM2 and PCNA defined the Proliferative subtype. The Differentiated subtype was associated with high expression of MUC16 and MUC1 and with expression of the secretory fallopian tube maker SLPI, suggesting a more mature stage of development. High expression of HOX genes and markers suggestive of increased stromal components such as for myofibroblasts (FAP) and microvascular pericytes (ANGPTL2, ANGPTL1) characterized the Mesenchymal subtype.

Elevated DNA methylation and reduced tumor expression implicated 168 genes as epigenetically silenced in HGS-OvCa compared to fallopian tube controls[26]. DNA methylation was correlated with reduced gene expression across all samples (Methods S7). AMT, CCL21 and SPARCL1 were noteworthy because they showed promoter hypermethylation in the vast majority of the tumors. Curiously, RAB25, previously reported to be amplified and over-expressed in ovarian cancer[27], also appeared to be epigenetically silenced in a subset of tumors. The BRCA1 promoter was hypermethylated and silenced in 56 of 489 (11.5%) tumors as previously reported (Figure S7.1) [28]. Consensus clustering of variable DNA methylation across tumors identified four subtypes (Methods S7, Figure S7.2) that were significantly associated with differences in age, BRCA inactivation events, and survival (Methods S7). However, the clusters demonstrated only modest stability.

Survival duration did not differ significantly for transcriptional subtypes in the TCGA dataset. The Proliferative group showed a decrease in the rate of MYC amplification and RB1 deletion, whereas the Immunoreactive subtype showed an increased frequency of 3q26.2 (MECOM) amplification (Table S6.2, Figure S6.4). A moderate, but significant overlap between the DNA methylation clusters and gene expression subtypes was noted (p<2.2*10−16, Chi-square test, Adjusted Rand Index = 0.07, Methods S7, Table S7.6).

A 193 gene transcriptional signature predictive of overall survival was defined using the integrated expression data set from 215 samples. After univariate Cox regression analysis, 108 genes were correlated with poor survival, and 85 were correlated with good survival (p-value cutoff of 0.01, Methods S6, Table S6.4). The predictive power was validated on an independent set of 255 TCGA samples as well as three independent expression data sets[25,29,30]. Each of the validation samples was assigned a prognostic gene score, reflecting the similarity between its expression profile and the prognostic gene signature[31] (Methods S6, Figure 2c). Kaplan-Meier survival analysis of this signature showed statistically significant association with survival in all validation data sets (Methods S6, Figure 2d).

NMF consensus clustering of miRNA expression data identified three subtypes (Figure S6.5). Interestingly, miRNA subtype 1 overlapped the mRNA Proliferative subtype and miRNA subtype 2 overlaped the mRNAMesenchymal subtype (Figure 2d). Survival duration differed significantly between iRNA subtypes with patients in miRNA subtype 1 tumors surviving significantly longer (Figure 2e).

Pathways influencing disease

Several analyses integrated data from the 316 fully analyzed cases to identify biology that contributes to HGS-OvCa. Analysis of the frequency with which known cancer-associated pathways harbored one or more mutations, copy number changes, or changes in gene expression showed that the RB1 and PI3K/RAS pathways were deregulated in 67% and 45% of cases, respectively (Figure 3A, Methods S8). A search for altered subnetworks in a large protein-protein interaction network[32] using HotNet[33] identified several known pathways (Methods S9) including the Notch signaling pathway, which was altered in 23% of HGS-OvCa samples (Figure 3B)[34].

Figure 3

Altered Pathways in HGS-OvCa

(a) The RB and PI3K/RAS pathways, identified by curated analysis and (b) NOTCH pathway, identified by HotNet analysis, are commonly altered. Alterations are defined by somatic mutations, DNA copy-number changes, or in some cases by significant up- or down-regulation compared to expression in diploid tumors. Alteration frequencies are in percentage of all cases; activated genes are red, inactivated genes are blue. (c) Genes in the HR pathway are altered in up to 49% of cases. Survival analysis of BRCA status shows divergent outcome for BRCA mutated cases (exhibiting better overall survival) than BRCA wild-type, and BRCA1 epigenetically silenced cases exhibiting worse survival. (d) The FOXM1 transcription factor network is activated in 87% of cases. Each gene is depicted as a multi-ring circle in which its copy number (outer ring) and gene expression (inner ring) are plotted such that each “spoke” in the ring represents a single patient sample, with samples sorted in increasing order of FOXM1 expression. Excitatory (red arrows) and inhibitory interactions (blue lines) were taken from the NCI Pathway Interaction Database. Dashed lines indicate transcriptional regulation.

Published studies have shown that cells with mutated or methylated BRCA1 or mutated BRCA2 have defective homologous recombination (HR) and are highly responsive to PARP inhibitors[35-37]. Figure 3C shows that 20% of HGS-OvCa have germline or somatic mutations in BRCA1/2, that 11% have lost BRCA1 expression through DNA hypermethylation and that epigenetic silencing of BRCA1 is mutually exclusive of BRCA1/2 mutations (P = 4.4×10−4, Fisher’s exact test). Univariate survival analysis of BRCA status (Figure 3C) showed better overall survival (OS) for BRCA mutated cases than BRCA wild-type cases. Interestingly, epigenetically silenced BRCA1 cases exhibited survival similar to BRCA1/2 WT HGS-OvCa (median OS 41.5 v. 41.9 months, P = 0.69, log-rank test, Methods S8, Figure S8.13B). This suggests that BRCA1 is inactivated by mutually exclusive genomic and epigenomic mechanisms and that patient survival depends on the mechanism of inactivation. Genomic alterations in other HR genes that might render cells sensitive to PARP inhibitors[38] (Methods S8, Figure S8.12) discovered in this study include amplification or mutation of EMSY (8%), focal deletion or mutation of PTEN (7%); hypermethylation of RAD51C (3%), mutation of ATM/ATR (2%), and mutation of Fanconi Anemia genes (5%). Overall, HR defects may be present in approximately half of HGS-OvCa, providing a rationale for clinical trials of PARP inhibitors targeting tumors these HR-related aberrations.

Comparison of the complete set of BRCA inactivation events to all recurrently altered copy number peaks revealed an unexpectedly low frequency of CCNE1 amplification in cases with BRCA inactivation (8% of BRCA altered cases had CCNE1 amplification v. 26% of BRCA wild type cases, FDR adjusted P = 0.0048). As previously reported[39], overall survival tended to be shorter for patients with CCNE1 amplification compared to all other cases (P = 0.072, log-rank test, Methods, S8 Figure S8.14A). However, no survival disadvantage for CCNE1-amplified cases (P = 0.24, log-rank test, Methods S8, Figure S8.14B) was apparent when looking only at BRCA wild-type cases, suggesting that the previously reported CCNE1 survival difference can be explained by the better survival of BRCA-mutated cases.

Finally, a probabilistic graphical model (PARADIGM[40]) searched for altered pathways in the NCI Pathway Interaction Database[41] identifying the FOXM1 transcription factor network (Figure 3d) as significantly altered in 87% of cases, Methods S10, Figures S10.1-3). FOXM1 and its proliferation-related target genes; AURB, CCNB1, BIRC5, CDC25, and PLK1, were consistently over-expressed but not altered by DNA copy number changes, indicative of transcriptional regulation. TP53 represses FOXM1 following DNA damage[42], suggesting that the high rate of TP53 mutation in HGS-OvCa contributes to FOXM1 overexpression. In other datasets, the FOXM1 pathway is significantly activated in tumors relative to adjacent epithelial tissue[43-45] (Methods S10, Figure S10.4) and is associated with HGS-OvCa (Methods S10, Figure S10.5)[22].

Discussion

This TCGA study provides the first large scale integrative view of the aberrations in HGS-OvCa. Overall, the mutational spectrum was surprisingly simple. Mutations in TP53 predominated, occurring in at least 96% of HGS-OvCa while BRCA1/2 were mutated in 22% of tumors due to a combination of germline and somatic mutations. Seven other significantly mutated genes were identified, but only in 2-6% of HGS-OvCa. In contrast, HGS-OvCa demonstrates a remarkable degree of genomic disarray. The frequent SCNAs are in striking contrast to previous TCGA findings with glioblastoma[46] where there were more recurrently mutated genes with far fewer chromosome arm-level or focal SCNAs (Figure 1A). A high prevalence of mutations and promoter methylation in putative DNA repair genes including HR components may explain the high prevalence of SCNAs. The mutation spectrum marks HGS-OvCa as completely distinct from other OvCa histological subtypes. For example, clear-cell OvCa have few TP53 mutations but have recurrent ARID1A and PIK3CA[47-49] mutations; endometrioid OvCa have frequent CTTNB1, ARID1A, and PIK3CA mutations and a lower rate of TP53[ while mucinous OvCa have prevalent KRAS mutations[50]. These differences between ovarian cancer subtypes likely reflect a combination of etiologic and lineage effects, and represent an opportunity to improve ovarian cancer outcomes through subtype-stratified care.

Identification of new therapeutic approaches is a central goal of the TCGA. The ~50% of HGS-OvCa with HR defects may benefit from PARP inhibitors. Beyond this, the commonly deregulated pathways, RB, RAS/PI3K, FOXM1, and NOTCH, provide opportunities for therapeutic attack. Finally, inhibitors already exist for 22 genes in regions of recurrent amplification (Methods S5, Table S5.3), warranting assessment in HGS-OvCa where the target genes are amplified. Overall, these discoveries set the stage for approaches to treatment of HGS-OvCa in which aberrant genes or networks are detected and targeted with therapies selected to be effective against these specific aberrations.

Methods Summary

All patient specimens were obtained under appropriate IRB consent. DNA and RNA were collected from samples using the Allprep kit (Qiagen). We used commercial technology for capture and sequencing of exomes from whole genome amplified tumor and normal DNAs. DNA sequences were aligned to Human NCBI build 36; duplicate reads were excluded from mutation calling. Validation of mutations occurred on a separate whole genome amplification of DNA from the same tumor. Data is submitted to dbGaP under accession number PHS000178. Significantly mutated genes were identified by comparing to expectation models based on the exact measured rates of specific sequence lesions. CHASM [20] and MutationAssessor (Methods S4) were used to identify functional mutations. GISTIC analysis of the CBS segmented Agilent 1M feature copy number data was used to identify recurrent peaks comparing to the results from the other platforms to identify likely platform specific artifacts. Consensus clustering approaches were used to analyze mRNA, miRNA, and methylation subtypes as well as predictors of outcome using previous approaches[46]. HotNet [33] was used to identify portions of the protein-protein interaction network that have more events than expected by chance. Networks that had a significant probability of being valid were evaluated for increased fraction of known annotations. PARADIGM[40] was used to estimate integrated pathway activity to identify portions of the network models differentially active in HGS-OvCa.