Ovarian cancer joins the genome atlas
In the effort to fight cancer, researchers look to the genome for answers. Tumor cells grow uncontrollably, driven by genomic alterations that may hold the keys to improved diagnosis and treatment for this deadly disease. In an attempt to reveal the genetic changes that drive cancer, an international consortium of researchers including scientists at the Broad Institute of MIT and Harvard and elsewhere has been working to characterize the genomes of several cancer types. The pilot phase of the project, known as The Cancer Genome Atlas (TCGA), is now complete with a new release of data on genetic changes underlying ovarian cancer, a disease that lacks successful treatments and is the fifth-leading cause of death due to cancer among women in the United States.
The largest and most comprehensive effort to sequence cancer genomes to date, the work not only uncovers new genetic links to ovarian cancer, but it also demonstrates the power of large-scale sequencing and analysis projects to unravel the complex sources of this disease. Further, the pipeline of experimental methods and analytical tools can now be used to explore the genetic basis of many more cancer types. The work appears in the June 30 issue of Nature.
In the project’s first pilot effort, completed in 2008, 600 genes were sequenced in samples from more than 200 patients with glioblastoma, a deadly brain cancer. The study also explored changes in activity levels, or expression, of genes and presence of extra or missing portions of DNA, known as copy number. The work successfully revealed new mutations and core pathways involved in the disease.
Two years ago, as the TCGA team planned the pilot phase’s next target, ovarian cancer, they decided to scale up an order of magnitude by sequencing several thousand genes. At the same time, the traditional method of sequencing, known as capillary electrophoresis, was being replaced by “next-generation” technologies that enabled massively parallel sequencing and the generation of volumes of data more cheaply and quickly.
“A remarkable thing about this project was this big transition,” said Stacey Gabriel, Head of Cancer and Medical Sequencing at the Broad and a lead scientist of the new study. New capabilities allowed the team to set its sights higher, aiming to sequence not just a few thousand genes, but all the protein-coding segments of the ovarian cancer genome, known as the exome, in hundreds of tumor samples, making this the largest cancer sequencing effort yet. “Now exome sequencing is fully in place due to hard work by people here at the Broad and other places,” she said.
DNA misspellings aren’t the only genomic modifications in tumor cells, so members of the TCGA team aim to generate comprehensive profiles of cancer genomes, as they did for glioblastoma. In addition to sequencing the exons, or protein-coding portions, of genes in 316 samples of high-grade serous ovarian tumors, along with matching normal tissue from the same patients, the team also gathered and analyzed other types of genomic data in 489 tumors, including the expression of messenger RNA and microRNA, levels of DNA methylation, and alterations in DNA copy number.
“The TCGA is a high-profile project because it brings together a really great community of everyone who’s using the latest tools to generate what’s the most comprehensive dataset for any cancer project,” said Gabriel.
The remarkable amount of data generated in the study was subjected to rigorous analytical methods, many developed here at the Broad. The scientists discovered that the tumor suppressor gene, p53, was mutated in 96 percent of the tumors. “It’s shocking that nearly all patients in the study have p53 mutations,” said Gaddy Getz, director of Cancer Genome Computational Analysis at the Broad and a leader of the study’s analytical effort. “In other tumor types, it’s not that high, but here, it’s basically all tumors.”
In addition to mutations in p53, nine other genes were mutated at much lower rates in tumors, including known cancer oncogenes NF1, BRCA1, BRCA2, RB1, and CDK12, a gene never before linked to ovarian cancer. “Many of the cancer genes are recurrent, known players,” said Getz. “But whenever we do these studies, we find some new players as well.” Inherited mutations in BRCA1 and 2 are known to play a role in predisposition for breast and ovarian cancer, but this study revealed that some ovarian tumor cells can acquire the mutations, too, suggesting a broader role for these genes in cancer.
Led by scientists at the Broad, the analysis of DNA copy number in the samples yielded some surprising results. In the previous pilot study of glioblastoma, a few chromosomes were either amplified (present in extra copies) or deleted in many samples, but most chromosomes were present in the normal amount. Analysis of the ovarian cancer samples painted a different picture, one of “genomic disarray.” Getz said, “Basically the whole genome was either amplified or deleted.” Either entire chromosomes or whole arms of the chromosomes were altered. The ovarian cancer genome also held dozens of regions of smaller amplifications or deletions, many of which contain known cancer genes. “It seems that ovarian cancer is very much driven by copy number,” explained Getz.
In an attempt to use genomic measurements to further classify these high-grade ovarian tumors, the research team analyzed messenger RNA and microRNA in the samples and revealed four stable subtypes with potential implications for prognosis. The data also showed that a significant portion of samples harbored genomic alterations in elements of known cancer pathways, such as the RB1 and PI3K/RAS pathways. Getz views this approach as complementary to the direct measurements of genetic changes. While individual genes contribute to the picture of a disease, taking a step back to view networks of genes can reveal larger patterns at work. “Pathway analysis is what I see as the route to getting to the additional driver genes,” he said. “We find those genes significant in their own right, by copy number or mutations or combining them. But this doesn’t give us all the genes that are important in cancer.”
The breadth of data also shed light on the role of BRCA DNA repair genes in ovarian cancer. One-third of samples harbored alterations in BRCA1 or 2, but the data showed that these changes – germline mutations, tumor cell mutations, or hypermethylation, which effectively silenced the gene – were mutually exclusive, with samples having only one type of BRCA alteration. For BRCA1, the manner of gene inactivation seemed to influence the patient’s survival.
In total, roughly half of the high-grade ovarian tumors displayed defects in homologous recombination, a DNA repair mechanism, through changes to BRCA genes and others, suggesting therapeutic potential for drugs that target this mechanism, such as PARP inhibitors. Common cancer pathways and gene copy number changes uncovered in the study also provide potential opportunities for targeted therapeutics.
Getz explained that only in a comprehensive, large-scale project like the TCGA can effects like the BRCA result be seen, with integrative analysis of several types of genomic data generated on the same samples. “I see this paper as a huge resource of very clean, high-quality data that could be used for building computational models of the relationships between different events, such as copy number events, expression levels, and mutation,” he said.
The data is also a starting point for scientists to conduct cellular experiments and explore the role of specific genes or pathways in ovarian cancer. Larger datasets and new sequencing technologies that look beyond the exome may reveal even more genes and pathways involved in the disease and help to build a more complete picture of its genetic basis.
The ovarian cancer study was a collaboration of many research centers in the TCGA network. The Broad was the only institute to serve as a Genome Characterization Center, led by Gabriel and Broad senior associate member Matthew Meyerson; a Genome Sequencing Center, led by Broad director Eric Lander; and a Genome Data Analysis Center, led by Getz and Broad senior associate member Lynda Chin.
Supporting the Broad’s heavy involvement in these activities required incredible teamwork by the institute’s Biological Samples Platform, Genome Sequencing Platform, and Genetic Analysis Platform, said Gabriel. Now that the pipeline of experimental and analytical methods is well established, the TCGA project is accelerating beyond the pilot phase, with the Broad and other institutes currently generating data on more than a dozen additional cancer types. Gabriel explained that these efforts will be completed more quickly, because they have refined methods for exome sequencing. “Using the exome [sequencing technology] was like building the airplane while you’re flying it,” she said. “We all had a feeling of that. But ultimately the project took flight and we did very well. We should all be really proud of the data.”
Other Broad researchers contributing to the work include Michael Lawrence, Kristian Cibulskis, Andrey Sivachenko, Douglas Voet, Kristin Ardlie, Roel Verhaak, Scott Carter, Craig Mermel, Gordon Saksena, Huy Nguyen, Rob Onofrio, D. Hubbard, Supriya Gupta, Andrew Crenshaw, Alexis Ramos, Rui Jing, Richard Park, Michael Noble, Carrie Sougnez, Wendy Winckler, Jane Wilkinson, Jennifer Baldwin, Toby Bloom, Tim Fennell, Juinhua Zhang, Hailei Zhang, CJ Wu, Sachet Shukla, Yingchun Liu, and Peter Park.