Though endometrial cancer is one of the most common malignancies and in most cases responds well to treatment, scientists have been challenged to understand exactly why some patients have poorer outcomes than others. Now, a recent landmark genomic, transcriptomic, and proteomic characterization of the disease has produced insights into its biology that could provide answers to that question. The findings also suggest that adding molecular testing to the standard diagnostic work-up for endometrial cancer could change treatment options for at least some women. In addition, the research, conducted by a team of at least 70 scientists under the auspices of The Cancer Genome Atlas (TCGA) program, points the way to the future of cancer diagnostics.

"Our findings highlight the benefit of digging deeper to find the genetic drivers of cancer growth. We are entering an era when tumors can be evaluated from a genomics standpoint, not just by looking at cancer cells under a microscope," said study co-leader Elaine Mardis, PhD, a professor of genetics and of molecular microbiology and co-director of the Genome Institute at Washington University in St. Louis. "This more comprehensive approach provides a clearer picture of the way particular endometrial cancers will behave and will be important to gynecological oncologists who treat this disease."

Characterizing Endometrial Cancer

The analysis, published in Nature, is one of the latest in a series of comprehensive characterizations by the TCGA, an ambitious project led by the National Institutes of Health to map genomic changes in major types of cancer (See Box, below). There are about 50,000 endometrial cancer cases in the U.S. each year, making it the fourth most common cancer in American women. Endometrioid (type I) cases generally are linked to excess estrogen and obesity and are highly curable. Serous (type II) cancer tends to occur in older women and has less favorable outcomes, with stage 4 tumors having a 16% 5-year survival rate from time of diagnosis. Currently, disease typing based on histology drives treatment for the disease. After surgery, patients with endometrioid tumors often receive adjuvant radiotherapy, whereas those with serous tumors typically receive chemotherapy alone or chemotherapy and radiotherapy.

The researchers collected tumor samples from 373 women with endometrial cancer who had not yet been treated for their disease, including 307 endometrioid, 53 serous, and 13 mixed histology cases (Nature 2013;497:67–73). The majority of cases —69%—were grade 1 based on histology according to American Joint Committee on Cancer staging. Of the remainder, 7% were grade 2, 20% grade 3, and 5% grade 4. The patients' median age was 63 years and they were followed a median of 32 months.

The Path to Comprehensive Cancer Genomics
Unique Collaboration Targets 20 Tumor Types

The Cancer Genome Atlas’s (TCGA) groundbreaking analysis of endometrial cancer is one of many similarly notable studies the consortium already has completed, with more in the works. Jointly funded by the National Cancer Institute and National Human Genome Research Institute, the TCGA started in 2006 with a pilot project to characterize glioblastoma and ovarian cancer. Based on the success of these efforts, in 2009 TCGA set a high target to map the genomes of at least 20 cancers over the next 5 years. With at least $250 million in funding and a research network of more than 150 investigators and better than two dozen institutions nationwide, the TCGA is well on its way to achieving that goal.

As of July 2013, the TCGA Research Network has generated data and published analyses on glioblastoma multiforme, ovarian serous adenocarcinoma, colorectal adenocarcinoma, lung squamous cell carcinoma, invasive breast cancer, acute myeloid leukemia, endometrial cancer and kidney cancer, with others in the publication pipeline.

TCGA leaders chose cancers for this effort based on each disease’s poor prognosis and overall public health impact, along with the availability of high-quality tissue samples, among other criteria. The goal has been to analyze at least 500 samples of each tumor type.

A tenet of TCGA since its inception has been to make data from each cancer characterization rapidly available to the research community. The goals of doing so include accelerating discovery, building on TCGA findings, and ultimately, improving cancer diagnosis, treatment, and prevention. Data for at least 20 cancer types or subtypes are freely available online on the TCGA Data Portal and CGHub.

The impressive output of TCGA has been possible largely due to its model of team science, suggested Harold Varmus, MD, director of the National Cancer Institute. “The Cancer Genome Atlas’s multidimensional approach to collecting genomic data, including clinical and pathology information, have made these findings possible,” he said. “Without the integrated characterization of so many tumor samples, correlations between histology and genomic data may not have been observed or potential clinical outcomes identified.”

To a person, TCGA researchers who spoke with CLN emphasized how much their shared expertise and perspectives have advanced their work. Andrew Cherniack, PhD, a research scientist at the Broad Institute in Boston, seemed to capture the sentiments of his colleagues. “The rewards are just that we have so many different people involved and looking at the data, and coming up with new and different ways of seeing it. This just couldn’t be done by a single group alone,” he said. “We all have our own biases and ways of doing things, and it’s been exciting to see the way we feed off each other. I’ve seen this not only in our endometrial cancer project, but in so many things coming out of TCGA. These projects will have a big impact on the oncology community.”

Redefining Disease Classifications

Different TCGA partnering sites performed different aspects of the analysis, which included exome and whole genome sequencing, microsatellite instability testing, RNA and microRNA sequencing, DNA methylation and copy number analysis, and proteomic analysis using reverse phase protein arrays. This comprehensive analysis and data integration—the most extensive on record for endometrial cancer—opened several promising avenues for further research. Most notably, the investigators for the first time described four genomic subtypes of the disease. Up until now, endometrial cancer has only been described as endometrioid (type 1) or serous (type II).

Overall, the investigators identified 48 genes with differential mutation frequencies and distinct progression-free survival profiles across the four subtypes. "Since these were all endometrial cancer one would expect similar genetic alterations would occur, and the evolution of the cancers would be similar. However, we found that they actually are extremely diverse, not only clinically different but also with clear evidence that their genetic evolution is completely different," said Sandra Orsulic, PhD, director of women's cancer biology and associate professor of obstetrics and gynecology at Cedars-Sinai Medical Center in Los Angeles. "One subset has totally different genetic alterations or is genomically stable; another is completely unstable. We don't know what causes that, and our understanding of tumor evolution at this point is still a black box."

POLE and More

One subtype, POLE ultramutated, had an unusually high rate of mutations, especially in the POLE gene, which is involved in DNA replication and repair. Three quarters of this subtype also had hotspot mutations in POLE at Pro286Arg and Val411leu. This subtype tended to have improved progression-free survival in comparison to the other subgroups.

"These POLE-mutated samples have 100 times the mutations of any of the other tumors and they all seem to have a very good outcome. This suggests the very high mutational burden is so much that the tumor cells can't overcome that insult and don't have the capacity to metastasize or recur. That's a hypothesis that needs to be tested, but it's exciting that we may have an interesting biologic subgroup that clinically does very well," said the study's co-leader, Douglas Levine, MD, a gynecologic oncologist and head of the Gynecology Research Laboratory at Memorial Sloan-Kettering Cancer Center in New York City. "Theoretically this is a subgroup where we can recognize these patients do well, have a low risk of recurrence and therefore do not need any additional treatments."

The researchers found that a second subtype, hypermutated microsatellite instability, also had a high rate of mutations, albeit considerably lower than POLE ultramutated subtypes, with 18 x 10-6 mutations per Mb versus 232 x 10-6 mutations per Mb. These cases tended to have MLH1 promoter hypermethylation and decreased MLH1 mRNA expression, but a striking paucity of POLE mutations.

"It's been known for a number of years that a subset of endometrial tumors have what is referred to as microsatellite instability, which is a surrogate for high mutation frequency in the tumor cells. However, this instability reflects genetic changes in only a subset of the genes that are capable of causing this high mutation. This is clinically very important because it turns out that for those patients whose tumors have very high levels of mutations, long term outcomes are excellent," said Raju Kucherlapati, PhD, Paul C. Cabot professor of genetics and a professor of medicine at Harvard Medical School in Boston.

The third subtype, copy number low, had the greatest microsatellite stability of the four, along with a low number of copy number alterations. This subtype had a high rate of mutations in CTNNB1, a gene crucial to maintaining the linings of organs.

When Histology Is Not Enough

The fourth subtype, copy number high, was mostly serous (type II) cancers. More than 90% of these cases had TP53 mutations versus 11.4% in predominantly endometrioid (type I) subtypes, along with increased TP53 protein expression. It had the lowest mutation frequency of all subtypes—2.3x10-6 mutations per Mb—and was associated with poor prognoses.

Although this subtype mostly was comprised of serous tumors, the researchers found that about one-quarter of tumors identified by histology as high-grade endometrioid had a nearly identical molecular profile. This finding could have important consequences both for the diagnostic workup of these patients and for their treatment options.

"While some tumors were histologically classified as endometrioids, which is a less aggressive tumor type, if you look at their molecular, copy number, and mutational data, they looked almost exactly like the serous type tumors," said Andrew Cherniack, PhD, a research scientist at the Broad Institute in Boston, who was involved in the project's copy number analysis. "I think this suggests that there are molecular tests that could be developed that may be simpler or less subjective than current histological tests, and they'll be able to classify tumors into the different categories."

Levine emphasized the potential treatment implications of this discovery. "If confirmed that these patients have a high likelihood of recurrence, then they'll likely need more aggressive treatment," he suggested. "Oftentimes endometrioid tumors are treated with radiation therapy, whereas serous tumors are given chemotherapy. But we may be able to determine that this particular group of endometrioid tumors actually do worse and may be more appropriately treated with chemotherapy rather than radiation therapy, and that they truly do need treatment. Otherwise, they typically would not receive this more aggressive treatment. That would change our clinical approach in these patients." Levine's lab is in the process of studying this question in a separate trial under the auspices of the Gynecologic Oncology Group.

Cross-Cancer Commonalities

Even as the researchers made connections within types of endometrial cancers, their analysis also brought to light molecular commonalities between serous endometrial tumors and subtypes of high-grade serous ovarian cancer and basel-like breast cancer associated with more aggressive disease and poorer outcomes. The three share both a high frequency of TP53 mutations—ranging from 84% in basel-like breast cancer to 96% in ovarian cancer—and a very low frequency of PTEN mutations, less than 2% for all three. Some DNA methylation clusters also are similar, along with shared regions of amplification. At the same time, the investigators identified molecular features shared between endometrioid tumors and colorectal cancer, most notably a high frequency of microsatellite instability and POLE mutations. In addition, 93% of endometrioid tumors have mutations that suggest they might respond to PI(3)K/AKT targeted therapy. The researchers also identified another important pathway in endometrial cancer, RTK/RAS/ß-catenin.

These findings not only could lead to further breakthroughs in understanding cancer biology but also could have more immediate treatment implications. For example, women with serous tumors with ERBB2 alterations might be treated with existing HER2 inhibitors such as trastuzumab or lapatinib currently given to women with breast cancer who have the same mutation.

Levine cautioned, however, about jumping to the conclusion that common mutation profiles among different tumor types necessarily would lead to treatment successes. "Since these drugs are not FDA-approved for endometrial cancer it's often difficult for a doctor simply to prescribe a drug for a patient simply because they have a mutation," he explained. "The problem is we have no idea if it works. We know that every drug that targets a particular mutation doesn't work in a tumor that carries that mutation. So we have to be able to differentiate responders from non-responders, and that will take us a little bit of time to figure out."

Orsulic also emphasized the discovery aspects of the research. "As with any good science, it raises more questions than it gives answers," she said. "One of our major findings is that we presently are wrong in the way we categorize tumors, and this of course has implications for the way patients are treated. This is nice evidence that there is a genetic profile that fits three tumor types—uterine serous, ovarian serous, and basal-like breast carcinomas—that we previously thought of as separate entities, and that they perhaps should be thought of now in the same way."

In addition to Levine's research with the Gynecologic Oncology Group, TCGA's investigation of endometrial cancer continues, both to accrue up to 500 samples and to complete whole genome sequencing on 50. According to Kucherlapati, the group decided to publish results from 373 samples because they were so significant. After the team completes the remaining analysis it will decide whether to publish any additional findings or merely to make the data publicly available, he added.

The Clinical Lab Connection

If the TCGA's findings seem far-removed from clinical practice and the work of clinical laboratories, Kucherlapati was not alone in suggesting otherwise. "Determining all these different types of genetic variations eventually will have to be done in diagnostic labs. While some would argue that that is not done today, soon every newly diagnosed tumor will have to be tested for all the genetic and genomic changes. So that means there is actually a great opportunity for laboratorians to make it happen," he said.
Cherniack also emphasized the connection between the TCGA's results and the evolving landscape of lab practice. "Molecular tests will be really important, and our paper points to their utility," he said. "They're coming into fruition and already are being done by several labs. This is something different diagnostic companies are taking advantage of, too. So it's really the future of oncology."