Skip to Main Content
Text size: SmallMediumLargeExtra-Large

The Exome Factor

October 2011

In the August 26 issue of the journal Science, a team of NIDCR-supported scientists provides one of the most comprehensive analyses yet of the genetic landscape that underlies head and neck squamous cell carcinoma, or HNSCC, the most common form of head and neck cancer. The data help to pin down that HNSCC, although spoken in the singular, is actually plural. The condition represents dozens of molecular conditions, each driven by a unique acquired pattern of cancer-causing gene alterations. These gene alterations, like blinking red lights on a control panel, shut down or amplify regulatory signals within the cell, accelerating its growth cycle, and giving rise to a tumor. Interestingly, a companion study, also published in the August 26 issue of Science and partially NIH funded, reported similar results in HNSCC. Thus, each study already has confirmed much of the other’s data, a crucial next step in the scientific process. Taken together, the results suggest that the reclassification of these tumors based on their molecular characteristics is starting to come into technological reach as a key first step in establishing personalized medicine.

Jeffrey Myers, M.D., Ph.D.

The paper also offers another important storyline involving the Oral Cancer Genome Project, a collaborative research effort to define the genetic changes that underlie these cancers. This project supports this important undertaking and is one of NIDCR’s signature initiatives funded through the American Recovery and Reinvestment Act (ARRA). These data are a direct result of this investment.

Science Spotlight recently caught up with Dr. Jeffrey Myers, a surgical oncologist and researcher at M. D. Anderson in Houston, who is one of the senior authors on the paper. He was in Washington, D. C. to attend a meeting on translational science sponsored by the NIH’s National Cancer Institute. As Myers stressed, the paper was truly a team effort and collaborative research at its best with equal contributions from his colleagues at The Johns Hopkins University, Baylor College of Medicine, and M. D. Anderson Cancer Center.

Scientists often say new technology and discovery go hand in hand. Your paper certainly fits the bill. So let’s start with the technology that enabled the discoveries. You and your colleagues utilized what’s referred to as next-generation sequencing technology. Which next-generation approach did you use?

DNA molecules being scanned, showing massively parallel /next-generation sequencing

We performed what’s called exome sequencing. The exome is the complete set of exons, or protein-coding sequences within the human genome. The exome comprises only about 1 percent of human DNA.

But for cancer studies, that 1 percent is the extremely high value content.

That’s right. At its root, cancer is a genetic disease. That makes the non-coding DNA superfluous at first pass. It’s a little like rooting through the Sunday New York Times to get your hands on the crossword puzzle. The other stuff may ultimately be of tremendous interest, but the preferred starting point is the exome.

What makes exome sequencing next generation?

The precision and efficiency. Traditional sequencing strategies slog through everything. Next generation sequencing enables the parallel sequencing of thousands of gene fragments that can be processed simultaneously. We also now have the technology to do the sequencing equivalent of a keyword search – “exon.” That allowed us to capture the complete exome and drill down from there to look for mutations and generate our data. As a result, the time and expense of analyzing each tumor are reduced dramatically. The lower costs and increased efficiency make it feasible to analyze more tumors. The more samples studied, the greater the statistical power and strength of the data.

Exome image

Exome image: represents transition from primary sequence data (left) through the "brain filtering" and computer processing of bioinformatics (right).

Photo Credit: Richard F. Wintle, The Centre for Applied Genomics, Toronto

How many samples did you study?

We analyzed 32 primary tumors and sequenced more than 18,000 genes in total. Notably, 30 of the 32 tumors were biopsied from surgical patients who had yet to receive chemotherapy. As you may know, chemotherapy typically alters a tumor’s biochemical behavior. These drugs, after all, are designed to kill them. But our data derive largely from tumors that act as they would naturally in the body, and that’s very high quality information. Once the most frequent mutations in HNSCC were identified, we used more traditional sequencing approaches to sequence the candidate genes in an additional 88 tumor samples to validate our initial findings. This approach of using a discovery set of tumors and a scaled-up validation set has been developed by our colleagues from Johns Hopkins and was utilized successfully in this collaborative study.

I should quickly note that in the same issue of Science is a companion paper from Stransky et al. that largely employed an exome sequencing strategy on 74 primary HNSCC tumors. For me, this was a best-case scenario: two studies, two complementary sequencing approaches. They validate each other’s findings in most instances, which is a key aspect of the scientific process. Just as importantly, they move the field forward, and everybody wins.

Let’s turn to the discoveries. The findings present a mix of something old and something new. What about the old?

We detected many previously identified genes involved in HNSCC, and that’s a good thing. It confirms that we haven’t been wandering blindly for the past few decades pursuing false leads. For example, TP53 has been one of the most intensively studied genes in head and neck cancer research. It encodes the multi-purpose protein p53, which helps to manage the cell cycle, carry out programmed cell death, initiate DNA repair, and regulate the transcription of a large cassette of essential genes. Given its many key functions, p53 is frequently found inactivated in HNSCC tumor cells to enable their aberrant growth.

Or so the traditional sequencing methods have indicated.

That’s exactly right. We wiped the slate clean and applied a new and more powerful sequencing approach. In doing so, we looked anew across the genome without any pre-existing biases. We let the technology lead the way. Thirty-two tumors later, TP53 emerged as the most commonly mutated gene. In the Stansky et al. paper, TP53 also came out on top. So, that was reassuring. It tells us that we can’t think enough about p53. We need to continue to optimize our understanding of p53 to help our patients.

What about the new?

Immunofluorescent staining of the receptor NOTCH-1 (red)

The biggest find is the NOTCH1 gene. Twenty-one of the 120 primary tumors had mutations in one or both copies of the gene. That made NOTCH1 the second most frequently altered gene after TP53. The NOTCH1 discovery was completely unexpected. It’s also a totally different kind of discovery, and one that was validated in the Stansky et al. paper.

How is the discovery different?

NOTCH1 has been described previously in oncology, primarily as playing a role in the blood cancer acute T-cell lymphoblastic leukemia. This is the first time that it has been reported to be altered at such a high frequency in a solid tumor. We found a total of 28 NOTCH1 mutations, and seven of the 21 patients had two independent mutations in the gene.

And traditional sequencing approaches hadn’t previously detected NOTCH1 alterations?

Correct. What’s really interesting is the gene’s protein product sits partly inside and partly outside the cell. In leukemia, the research focus to date has been largely on the region inside the cell. There, NOTCH1 functions as an oncogene that, like a needle stuck on a turntable, keeps replaying the same pro-growth message to the cell. But in HNSCC, we noticed gene mutations that alter portions of the protein outside the cell. The alterations likely deactivate the protein and its signal. In this context, NOTCH1 acts as a tumor suppressor. That is, it puts the brakes on some aspect of cellular growth, and tumor cells selectively silence NOTCH1 to turn its red light to green. And biologically, we already may know why in a general sense. In epithelial tissues, i.e., skin cells like those in the oral mucosa, NOTCH1 is considered important for cell differentiation.

Another interesting finding is that the bulk of activating mutations involve tumor suppressor genes.

Well, it’s interesting. I’m attending a meeting today called NCI Translates. Dr. Bert Vogelstein, one of the fathers of cancer genomics, just spoke to us. He presented a list of 10 characteristics of cancer genomes, which showed that very few tumors have more than one activating oncogene. The mutations primarily involve tumor suppressor genes. Tumor suppressor genes are largely equal opportunity targets for inactivation, and that makes tumors by nature quite heterogeneous.

Is that surprising?

No, I don’t think so. Cancers long have been defined by their presumed site of origin, and we talk about them in the singular. Head and neck cancer. Breast cancer. Prostate cancer. But that’s too simplistic. What researchers are finding is every cancer type really represents a spectrum of distinct molecular diseases. Or, throwing the site of origin out of the mix for a second, they represent unique types of molecular malfunctions that can arise not only in the prostate, but potentially in other tissues and organs.

By analogy, when the car doesn’t start in the morning, we don’t say, “Oh darn, I have a Toyota problem.” We narrow it down to the specific dysfunction underneath the hood. That problem could arise in any make and model of car.

I agree. I just attended a presentation, and the speaker summarized his work assembling gene-expression profiles of pancreatic tumors. He clustered the gene profiles, subdivided them into four general categories, and studied each group’s response to treatment. Interestingly, when he and his colleagues applied the gene-expression methods to cancers of the breast, they discovered a similar four-part clustering. What’s more, when they combined the pancreas profiles with their comparable breast profiles, each group responded similarly to treatment. His experience and those of others suggests that in the future we will reclassify tumors according to their underlying molecular profiles. For example, a suspicious lesion in the mouth might be described as having a class E biomarker profile, not solely as an oral squamous cell carcinoma.

And this paper helps to move the field further down that road in the head and neck?

That’s exactly right. Between both Science papers, we have comprehensive exome data on about 115 tumors. In the near future, the NCI’s The Cancer Genome Atlas (TCGA) will add significantly to that number. At that point, we’ll be able to get into more specifics. In the meantime, there are additional technological platforms, or perspectives, from which to study head and neck cancers. The exome data are invaluable. But they represent just one perspective.

What are some of the other needed platforms?

An important one is copy number variation. That means duplications or deletions of one or more sections of the genome. It would be like repeating the same word two or three times in a row in a sentence. We know such editorial mistakes occur in HNSCC, and these changes can be highly informative to characterize a tumor’s subsequent behavior and susceptibility to treatment. Another important platform involves epigenetics, or factors other than gene mutations that affect gene expression. An example is DNA methylation. It is a common biological mechanism that tumors can exploit to shut down a tumor suppressor gene. So we need to know when and where DNA methylation occurs. Bottom line, all of these platforms will give us a comprehensive, three-dimensional view of the genome. Right now, I think we have a nice two-dimensional perspective, which is definitely a giant step forward, but we need to go further to appreciate the big picture.

What will this big-picture view mean for patients down the road?

Three things come readily to mind. The first is improved diagnostics. As I’ve mentioned, we can’t keep lumping cancer patients together based on broad and insufficient diagnostic criteria. We need to reclassify tumors based on their distinctive molecular features.

Is this already occurring?

Yes, it is starting to happen in head and neck cancer. For example, I mentioned a moment ago that about half of all HNSCC’s have mutated copies of TP53. However, for a subset of patients with tumors caused specifically by the herpes simplex virus (HPV), we don’t find mutations in the gene. The p53 protein is indeed inactivated in these patients, but it turns out to be via another mechanism. HPV contains a protein that directly binds p53 and inactivates it. This is important diagnostic information and leads to my second point.

And that is?

As the molecular diagnostics improve, treatment will become more targeted to the specific glitches driving the development of an individual tumor. Let me give you another example. As we noted in our paper, nearly 90 percent of the non-HPV tumors harbored inactivating mutations in tumor suppressor genes, while only about 18 percent had activating mutations in a bona fide oncogene. This is critical information. Why? Most molecularly targeted treatments now under development take aim at oncogenes. But if oncogenes do not always drive the bus, so to speak, and multiple tumor suppressors are inactivated and already off the bus, what can we reliably target? It’s a critical question, and that’s why we need to assemble a more comprehensive three-dimensional view of cancer biology.

Are these additional analyses already getting off the ground?

They definitely are, and people already are talking about the next logical steps to pull together all of the data. That brings me to my third point. I think sufficient progress already has been made scientifically to translate fairly rapidly the most compelling clinical leads. I just came from a session on P13 kinase and efforts to develop drugs that inhibit this enzyme. Well, P13 kinase is altered in about 5 percent of HNSCC cases. That means we can build on this research. We don’t have to reinvent the wheel. We can identify people who have a P13 kinase alteration, and specifically test the safety and efficacy of these inhibitors in a more precise, tailored way. I think the next several years will be an extremely interesting and productive period that will bring a more personalized touch in cancer diagnostics and care.

Finally, the Science paper involves a unique collaboration. Initially, NIDCR issued two individual grants – one to the group at Johns Hopkins; the other to your group and colleagues at Baylor College of Medicine. All parties involved eventually joined forces. Was it a best case scenario?

Absolutely. The collaboration was tremendously productive. It also shows just how beneficial big research consortia can be to generate the large amounts of high-quality biological data that are now within our technological reach. Moving forward, I think we need to consider how best to assemble these teams to maximize their productivity. By that, I mean future consortia might require the combined expertise of six, eight, or 15 institutions, not only the three that were in play in our paper.

That also raises another key issue for young faculty and their institutions: How do you apportion the credit? We have about 30 authors on our paper, and there’s only one first and last author. Well actually, we had co-first and last authors in our study. Nevertheless, people tend to notice just the order of names. It’s ingrained. I’ve been on a promotion-and-tenure committee at our institution, and I can tell you that’s what they notice.

But how do you change the culture?

Maybe instead of focusing on authorship, the senior scientists could sign off on the percentage of the project that each author fulfilled. Then again, that might miss the point. We need to approach our challenges more like a team of engineers. For example, who built the wing of the last airplane that you flew on? A team of 70 people at Boeing. They designed it. They did the calculations. Nobody knows who they are. And yet, they work together, and they don’t want anyone to crash. So, maybe that’s the better way to frame the issue.

Thanks for taking the time to talk about the paper and collaboration.

Share This Page

GooglePlusExternal link – please review our disclaimer

LinkedInExternal link – please review our disclaimer


This page last updated: February 26, 2014