![Media: The Inside Scoop](https://webarchive.library.unt.edu/web/20090508012046im_/http://www.nidcr.nih.gov/NR/rdonlyres/8DEB8102-21F8-42A3-A71B-4A7FE2D120C1/0/MediaTheInsideScoop.gif)
January 2008
When discussing the TP53 gene, scientists describe it like a technician would a Swiss army knife. The gene encodes a multi-purpose protein that takes part in regulating the cell cycle, carrying out programmed cell death, initiating DNA repair, and regulating the transcription of a large cassette of genes that cells employ for various biological purposes. Given its many essential functions, P53 is frequently found inactivated in tumor cells, including those that cause most head and neck cancers, as a means to co-opt the cell cycle. In the December 20, 2007 issue of the New England Journal of Medicine, a team of NIDCR grantees and colleagues evaluated the prognostic value of TP53 mutations in 420 head-and-neck cancer patients treated with surgery only and whose survival was tracked for several years thereafter. Detecting TP53 alterations in the tumors of 53 percent of participants, the scientists found that collectively these mutations were associated with decreased overall survival. This was particularly so for a subset of TP53 mutations that affected the ability of its protein to function as a transcription factor. To hear more about this paper, the Inside Scoop spoke with Dr. Wayne Koch, the senior author on the paper and a scientist and head-and-neck cancer surgeon at Johns Hopkins University in Baltimore.
How did this study come about?
I work with a laboratory group at Johns Hopkins that for the last 15 years has sought to identify molecular changes in developing tumor cells. Our main goal is to use changes, commonly called “markers”, for clinical purposes, especially in learning to tailor treatment to target a patient’s specific tumor and thus improve their outcome. By the early 1990s, our group had found that TP53 is a good marker for head and neck cancer, and that led us to the next question: How technically robust and therapeutically informative would it be to look for cells with TP53 mutations in the margins, or a tumor’s outermost boundary? We organized a small pilot study, and it turned out to be a provocative little investigation. We had 35 patients, and half of them had cells with TP53 mutations in the margins. The presence of the TP53 mutation identified those cells as tumor cells, although the pathologist had found no tumor in the samples using a microscope. A third of those with mutations in the margins had local recurrence; while none without mutations had recurrence. We published the results in October 1993. But it was pointed out that, because the pilot study was small and statistically underpowered, we ought to do a larger study.
And that led to the current study?
Right. Dr. Arlene Forastiere is a medical oncologist at Johns Hopkins and an author on the paper. At the time, she also was the chair of the ECOG Head and Neck committee. ECOG stands for the Eastern Cooperative Oncology Group, and it is one of several cooperative cancer clinical trial networks around the country. Arlene thought this study would be a good one for ECOG. So, in the mid 1990s, we set out to collect 510 head and neck tumor samples.
Has it taken all of this time to collect and analyze the tumor samples?
Well no, the collection took place from 1997 until about 2003. Since then we have collected information about the clinical outcome of the patients, and have been working in the laboratory to complete the molecular analysis. We have not published the results of the margin analysis yet because of technical snags. In the mid to late 1990s, the available methods for margin analysis were extremely labor intensive. We realized that the technique we had used in our 1993 paper would never be reproducible on a mass scale in any kind of commercial or clinical setting. So that led to a series of bright ideas about what to do next. Each bright idea arrived with its own unique set of technical difficulties. In the meantime, we have compiled rigorous P53 mutation data for the tumors of 420 patients and followed all of the patients for at least three years. These two data sets allowed us to accomplish the second aim of the project, settling a longstanding controversy in the cancer literature: Does a tumor’s TP53 status have an impact on HNSCC patient survival?
Why the controversy?
If you flip through the literature, you’ll find most of the studies look at the levels of p53 protein instead of performing actual mutation analysis. They also tend to have a small number of cases and so, depending on which study you read, they come up with answers of yes, no, or maybe.
How then does this study rise above the previous methodological limitations?
I’d say the size, the rigorous technique to assess the mutation, and the length and the quality of the followup.
Let's break them down. What about the size of the study?
This investigation had 420 patients; most of the previously published prognostic studies included 100 patients or less. So the statistical power is significantly greater.
What about the rigorous technique?
We analyzed the state of the TP53 gene in each tumor using a commercially available P53 gene chip and a related analysis program. Both are state of the art and highly reproducible. But the larger issue here is this study looked at mutations, not protein.
Why's that?
Some people argue that it should be more clinically relevant to look at the protein because that’s the functional component in the cell. But, as it turns out, there are a number of instances in which protein staining does not reflect actual mutational status, and P53 is a good example.
You mean the mutation analysis could be more informative than the protein levels?
Right. Typically with P53, if the protein is seen in a cell, it’s judged to be due to a gene mutation. That’s because the wild type, or normal, protein only lasts in a cell for a few minutes. It’s never present in high enough concentrations to be picked up with immunocytochemistry staining, a standard technique to gauge protein levels. So, when staining is seen, it’s assumed that a mutation has aberrantly changed the folding of the protein and somehow stabilized it; therefore, we see p53 in this cell. In actuality, that’s a pretty good marker. But when you’re looking at an endpoint as messy as survival, if you misappropriate 10 or 15 percent of the cases into the wrong category, I think you end up muddying the waters. A relatively small but important survival difference can get lost in the final analysis. So, for our purposes in this study, we wanted to know the actual DNA sequence.
What about the length and quality of follow up? You already mentioned the three years of patient outcome data.
That’s where the co-operative group mechanism is really the key. It allowed us to collect a large number of samples from different ECOG centers and remove any suggestion of geographical or research bias. In this study, an independent group of statisticians sifted through the treatment-outcome data and then applied them to the molecular information from our lab. The study was conducted in a way that is highly credible scientifically, and its data should be highly believable.
What about the mutational classification scheme?
That’s an important point and outcome of the study. Luana Poeta, the lead author on the paper, devised a categorization scheme of “disruptive’ and “non disruptive” mutations in the P53 gene that framed our analysis. This simple, binary schema showed a correlation between disruptive mutations and decreased overall survival. I’m referring to relative risk here. In other words, we saw a decreased survival - in general - among those with disruptive mutations. Let me also quickly add, the relative risk was around two, and, from a clinical treatment perspective, that’s not such a big impact. But the median survival was 5.4 years for people with wild-type p53 in their tumor, compared to 3.2 years for those with mutation.
And yet, it's nevertheless meaningful?
We think so. The results for the impact of p53 were independent of all common factors known to be useful to categorize patients and predict outcome, including tumor site and stage. In fact, in a number of our sub-analyses, the prognostic power was a little stronger than all of the data lumped together. The bottom line here is there seems to be something that is very specific to the biological behavior of the tumor that’s predicted by the disruptive/non-disruptive schema.
Let's go back a second. How did the group develop the schema?
It was pretty straightforward. When you look into the possible spectrum of P53 mutations, there are hints in the scientific literature about the impact of amino acid substitutions to P53. These hints suggest that the disruptive impact will depend on how different the new amino acid is compared to the old one, and where in the molecule it resides.
As I understand it, you were most interested in the DNA- binding domain of P53 protein. That is, the part of the protein that is responsible for binding DNA and regulating the transcription of numerous genes and thus cellular processes. Is that correct?
That’s right. We categorized disruptive mutations as those located in the DNA-binding domain of the protein and thus impacting its ability to bind DNA. We were particularly interested in mutations in which a polarized, or charged, amino acid was switched to a non-charged, non-polarized amino acid or vice versa. As for the non-disruptive-mutations, they were defined as changes that fell outside the DNA-biding domain and which didn’t cause a switch in polarization.
So it's pretty basic?
Yes. You know, this categorization scheme is so old that it’s in my Lenninger biochemistry textbook that was published in the 1970s. At first blush, it seems too simple. But, we were pleasantly surprised to find that we could apply this simple scheme to effectively designate a mutation as good or bad.
What's interesting is you looked at several different tumor types that fall under the umbrella term "head and neck cancer," and the disruptive/non-disruptive distinction still had a prognostic effect.
Yeah, that’s where the size of the study and the power of the clinical follow up really helped.
There are other validated prognostic molecular markers in oncology. But none have yet changed clinical practice. What might be the clinical impact of this finding?
The short answer is: I don’t know. But I think there are two points that might be valid. I think TP53 status could be used, especially applying the disruptive and non-disruptive schema, to stratify and pick out people who are at higher risk of not responding to treatment. It could be one more tool in the metaphorical clinical toolbox.
What is in the toolbox right now for the prognosis of head and neck cancer?
The most obvious tool right now for head and neck cancer is the human papillomavirus, or HPV. If patients have an HPV-positive tumor, they tend to have a better prognosis and response to therapy. Overexpression of the epithelial growth factor receptor (EGFR) appears to be a valid prognostic biomarker. There also are several chromosomal breakpoints that may be pretty good markers in premalignant disease for progression to develop a tumor. But there aren’t very many other molecular markers that have rigorous data that show a prognostic impact. So this will certainly be one that I would include in a panel of markers that are informative about the biology of the tumor.
What's the second valid point?
Well, the data may also help promote new treatment approaches. In the cases where a disruptive mutation is found, it may be possible down the road either to put back the normal, wild-type gene or knock-out the mutant TP53. That might involve a small molecule strategy that could be used for targeted therapy.
What about the genetic epidemiology? You've made the distinction between disruptive and non-disruptive mutations. But within the disruptive category, different mutations could have different therapeutic and prognostic effects. Do you need to characterize individual disruptive mutations?
That’s a good question. I have a sneaking suspicion that our simple schematic has missed some of the bad behavers and allowed in some that aren’t so bad. What we’re going to do now with our data and our samples is create tissue microarrays and start to look at some of the co-factor markers up and downstream from P53 and correlate them with what we know in this group about the P53 alterations and outcome. One that’s obvious to look at is HPV. But we want start looking down into some of these pathways and see if we can refine the prognostic information a little bit better.
Thanks for your time.
Happy to talk to you.