WASHINGTON—Genome-wide studies of head and neck squamous cell cancer (HNSCC) show that tumors driven by human papillomavirus (HPV) are molecularly distinct from other subtypes, Cancer Genome Atlas (TCGA) researchers reported here at the American Association of Cancer Research Annual Meeting (Abstract 1117).
Those data, together with data published online April 25 in Cancer Discovery (doi: 10.1158/2159-8290.CD-13-0103), also identify potential drug targets amplified by copy number changes and a molecular subtype of disease shared with some cervical and lung cancers.
The TCGA researchers used a variety of genomic techniques to study 279 tumors, including DNA sequencing, DNA copy number profiling, mRNA expression, promoter methylation, and protein arrays. The tumor samples were surgically resected from previously untreated patients. Eighty percent of the tumors came from patients with a history of smoking. Additionally, 29 tumors were HPV-positive.
“This is the first TCGA effort to characterize a viral-associated tumor,” said David N. Hayes, MD, MPH, Associate Professor at the University of North Carolina and a medical oncologist at the UNC Lineberger Comprehensive Cancer Center in Chapel Hill, who presented the data on behalf of the TCGA collaborators.
“This is a fascinating data set. The team has generated unbiased RNA transcript profiles, low pass whole genome sequencing from many of the virally infected patients, and whole genome sequences from selected patients.”
Confirm and Extend
The TCGA data confirm and extend earlier findings from two smaller studies, noted Mitchell Frederick, PhD, Associate Professor of Head and Neck Surgery at the University of Texas MD Anderson Cancer Center. Those studies (he was a coauthor for one of them) relied solely on genomic sequencing of tumor-normal pairs and lacked the breadth of analytical tools used in both the newly reported TCGA study and his follow-up study in Cancer Discovery.
“In the original papers [published in 2011], where we just looked at somatic mutations by sequencing, we didn't identify a lot of drivers that could be targeted—meaning things that were pushing the tumor to divide or survive,” he said in a follow-up telephone interview for this article. “We certainly identified a large number of tumor suppressor genes, but those are hard to target, so it was a bit frustrating for everybody.”
Copy Number Changes
That perspective has changed, though, with the additional genome analyses. For example, copy number analysis shows that there are a large number of deletions and amplifications in HNSCC. “If you mine these data for gains, you can develop the hypothesis that there may be additional drivers that are turned on in these tumors because there are extra copies of the genes and the genes are expressed at higher levels. Within those you find a lot of potentially targetable oncogenes,” Frederick said.
Some of the copy number changes are extraordinarily consistent across tumors, Hayes noted. “This has been shown before, but I think it is worth reinforcing, that in these independent samples of head and neck cancers, across patients, there is an incredible pressure to assume a very specific configuration of chromosome losses and gains.”
For example, the vast majority of the head and neck tumors, including many of the HPV-positive ones, carry extra copies of chromosome 3q, which includes three key genes: PIK3CA, SOX2, and TP63. Copy number changes in chromosomes 4, 5, and 8 also appear relatively consistent across the HPV-negative tumors.
When TCGA researchers aligned the head and neck copy number chromosome map with maps from 358 squamous cell lung cancers, released by TCGA last year (OT 10/25/12 issue), and from 114 cervical cancers, the consistency of the pattern is striking. (Other cancer types, including ovarian and glioblastoma, show very different patterns.)
“I think this must be important, and we must understand it,” Hayes said. “It is suggesting there is the possibility to start looking at cancer across tumor types as molecular patterns that can be leveraged and understood, both from biologic and therapeutic standpoints.”
But not all head and neck cancers resemble one another in terms of copy number changes. In the case of chromosome 7, which includes the EGFR gene, only a subset of the HPV-negative tumors have the amplification; most do not. Remarkably, there is a group of the tumors, perhaps 15 to 20 percent of the samples, that generally lack copy number changes altogether.
Commonly Mutated Genes
When the investigators looked at the individual genes most commonly mutated in head and neck cancers, they found some of the usual suspects as well as some that are less familiar in cancer. The most frequently mutated genes were TP53 (72% of samples), CDKN2A (22%), NOTCH1 (19%), and MLL2 (18%).
One of the genes that is less well known is FAT1, Hayes noted, explaining that it was recently described as a Wnt tumor suppressor gene. “It is mutated in 23 percent of squamous cell carcinomas of the head and neck,” he said. “This is going to transform our understanding of this disease.”
The high frequency of PIK3CA mutations (21% of samples) is also good news for patients, he said, because it is a “very druggable target,” with potential inhibitors already in clinical development.
“One of the hopes from these large genome projects is to come up with single gene events or single gene phenotypes or druggable genes, and I think there are many in the head and neck cancer data set,” Hayes said.
For example, NSD1, which encodes a histone methyltransferase, is mutated in about 10 percent of TCGA samples. Mutations in the gene were previously associated with a developmental syndrome and childhood acute myeloblastic leukemia. TCGA investigators found a near one-to-one correlation between NSD1 mutations and a subtype of HNSCC tumors that has very low methylation of CpG islands throughout the genome.
“So I think we have another single gene event from the cancer genome atlas.”
Cataloging gene alterations and expression changes is interesting, but the real goal of TCGA and other sequencing projects is to improve understanding of the biology and, ultimately, to improve patient care. With that in mind, both Hayes and Frederick emphasized the high frequency of alterations in the PI3 kinase pathway, including the PIK3CA mutations, with nearly 40 percent of HPV-positive tumors affected and nearly 50 percent of HPV-negative ones.
“That is a very near-term actionable observation,” Hayes said.
In HPV-negative tumors, the team found that 15 percent of the samples carried alterations of some type in one of the fibroblast growth factor receptor (FGFR) 1, 2, or 3 genes. “That is on the shortlist,” he said. “We need to evaluate the FGFR inhibitors in head and neck cancer.”
Similarly, cyclin D1 was amplified in 30 percent of HPV-negative tumors, which suggests that drugs targeting the pathway, such as CDK4/6 inhibitors, might be valuable in this setting.
One clinically important message comes from what the team did not find. Currently the only molecularly targeted therapy approved for head and neck cancer is cetuximab, which blocks EGFR. However, while it has been assumed by many clinicians and researchers that EGFR is mutated or altered in a large proportion of head and neck cancers, TCGA found it amplified or mutated in only 16 percent of the HPV-negative samples and six percent of the HPV-positive ones. EGFR changes are not reliable predictors of response to cetuximab in head and neck cancer patients, but Frederick says the rarity of EGFR alterations in HPV-positive patients suggests that these patients are unlikely to benefit from the drug.
One unusual subgroup of tumors, he said, is a class of HPV-negative tumors that have mutations in caspase 8, an enzyme involved in apoptosis. Many of those tumors also have alterations in HRAS, but are wild type for TP53 and p16. Moreover, these tumors encompass the group of tumors that have fewer mutations and copy number changes than other subtypes do.
The paucity of other changes suggests—though doesn't prove—that these tumors are heavily dependent on the HRAS and caspase 8 mutations and might, therefore, be particularly susceptible to inhibitors targeting these proteins or their pathways. “The caveat is that we need to develop therapies for targeting RAS and caspase 8 that we don't yet have in hand,” he said. Additionally, the caspase 8 changes are loss of function, which is a type of mutation that is harder to target than gain of function or over-expression changes.
Finally, he points out that while his team and other groups reported sequence data from head and neck cancers before TCGA, the larger sample size from TCGA provides for analysis that the smaller projects simply cannot do. “We saw NOTCH1 mutations in 15 percent of our samples; TCGA says 20 percent. But they also find other genes in the pathway that are affected, so the NOTCH pathway may actually be altered in 50 or 60 percent of the patients,” he said.
“You can glean information on the increased prevalence of these alterations by looking at pathways. That has been one of the strengths of the TCGA projects—assembling these pathways that might be targetable in some portion of the patients.”