Is There Life Beyond Histochemical Staging? by Robert Warren, MD

SLIDES & TRANSCRIPTS
Tuesday, February 15, 2000

Is There Life Beyond Histochemical Staging?
Robert Warren, MD

Slide 1:

DR. WARREN: Thank you, Carolyn.

Wow, I am exhausted. I am not sure what to say now because it certainly seems depressing.

May I have the first slide?

Slide 2:

What I have to say is mostly a reiteration of what you have heard and, I am sure, for the whole conference, will be a recurring theme, and that is can we do better at identifying patients, for example, in Stage II colon cancer and colorectal cancer who will do poorly and should be the target of novel therapies or in the case of Stage III cancers a subset of patients who will do well and maybe don't need 5-FU or those patients who will do poorly in spite of current adjuvant chemotherapy.

May I have the next slide, please?

Slide 3:

I certainly have to thank Dr. Hamilton for really covering a great part of the subject of at least genotypic markers of outcome of colorectal cancer.

What I will do in my talk is for a brief period of time focus on where the field is going, at least in terms of markers of gene expression, either at the protein or RNA level. Is there any information there that can help guide us and how can we sort these issues out.

The slide you would be seeing right now, if the slide machine worked, would show a review of the literature, over 100 papers looking at more than 40 markers of outcome. Each of these markers, and I don't know how many bins this would make but a lot, has been shown in one or another published studies to be predictive of outcome in colorectal cancer.

Certainly 18q, LOH at H and Q is important but in trying to look at and make sense out of this it really becomes a very challenging problem, and I am very interested in what the next speaker has to say about the statistical analysis.

One of the problems with interpreting all these studies is the end points vary from one study to another. Dr. Hamilton talked about overall survival and the 18q LOH. What about disease-free survival? Is there a relationship between a particular marker and local versus distant metastasis, and in patients with advanced disease do markers predict response to therapy in addition to these other end points.

Slide 4:

Some of the impediments in interpreting the results from many of these studies are that most of them are retrospective. Some are in fact prospective, but the minority, most studies lump colon and rectal cancer together and that doesn't make biological sense to most of us.

In very few studies are the markers looked at in a stage-specific way. Usually investigators group Stage II and Stage III patients together and the duration of follow-up is very variable, and with the statistical analysis we will hear more about it.

Most studies simply do univariate analysis, and it is very rare for one marker to be judged against either the standard pathological markers that Carolyn talked about or other molecular markers and most of the studies, particularly the retrospective studies, are really underpowered to say very much.

Slide 5:

But if you take that group of 40 or so markers and think about them as predictors of metastasis, you can cluster them according to at least the theoretical behavior that a cancer cell has to go through in going from a primary site to a metastasis.

I will talk about proliferation in a moment, but of course, cells have to extravasate and ultimately once they arrest in the target organ invade the tissue and so proteases are important in colorectal cancer. A large number of proteases have been shown to be prognostic, plasminogen, a whole variety of matrix metalloproteinases, and there are in fact 20 of them and so it makes the study of those very difficult.

The UPA receptor is also important. In terms of arrest in the target organ antigens, protein or carbohydrate antigens on the surface of the cancer cell are likely to be important and for example, DCC is probably an adhesion molecule that may have to do with extravasation, but the sialated Lewis X and Lewis A carbohydrate antigens which form mucins on the surface of most colorectal cancer cells have been shown to be predictive of metastasis, and then finally, of course, angiogenesis in many tumors seems to be predictive in the case of colorectal cancer of both microvascular density and in particular level of expression of estrin and fetal growth factor seems to be predictive.

Slide 6:

If you look at these so-called "clusters" of markers they fall into some logical categories, markers that predict proliferation, markers that regulate progression through the cell cycle such as the p53, these two cyclin kinase inhibitors, p21 and p27. Dr. Hamilton talked about at least 17p LOH which was the site for the p53 gene.

Slide 7:

There have been at least a dozen studies in the literature looking at the expression of the relationship between either p53 expression or 17p LOH in outcome.

This is a study from the Giack group at USC showing that in patients with nuclear overexpression of p53, the recurrence rate for Stage II colon cancer was 70 percent at 3 years, whereas for patients with no p53 nuclear overexpression, based on Kaplan Meier, the recurrence rate was 23 percent at 3 years, and that was highly significant.

Slide 8:

There have been a number of other studies including Dr. Hamilton's that really have not shown this to be the case. Part of the problem in using immunohistochemistry as a surrogate for mutations in p53 has to do with the fact that some of the studies will use paraffin, others frozen sections. There is a variety of antibodies out there to use for staining for p53. Antigen retrieval methods certainly vary from investigator to investigator. Experience of the rater can differ markedly and, finally, generally the studies will dichotomize continuous variables and say that there is either high expression or low expression, meaning there is overexpression or not whereas it may be, in fact, more important to analyze p53 as a continuous variable.

Slide 9:

What about p27? That perhaps of all the factors looked at in colorectal cancer may have the highest predictive value.

Slide 10:

This is a study by Max Loda, who is now at the Dana Farber looking at the level of expression by immunohistochemistry of this cyclin pentakinase inhibitor in Stage II cancer. Patients fell into three categories, those patients where more than 50 percent of the cells stained positive for p27 and this is the survival in that group of Stage II patients, those patients where between 1 and 50 percent of the cells in a combination of sections were positive for p27, and here is their survival and those patients who had no p27 staining at all, and here is their survival, all Dukes' B patients, and of course, very strongly significant as you can imagine

Slide 11:

and when you look at risk ratio if you compare those who have those patients with greater than 50 percent of the cells showing p27 versus those that are absent, the risk ratio is 32. It is huge.

Slide 12:

Dr. Hamilton also talked about Kim Jessupis study, and I won't reiterate that, using the immunohistochemistry for DCC.

Slide 13:

One of the questions, of course, that arises in examining these markers is can markers of outcome suggest new targets for therapies. If proteases are important in progression of colorectal cancer, do protease inhibitors have a role and I won't talk about this in detail other than while it is logical, it is important to point out a very recent development where an MMP inhibitor developed by Beyer was found in a study of lung cancer to have an adverse effect on outcome. In fact, Bayer withdrew all studies with that protease inhibitor.

What about cell cycle inhibitors? One of the approaches we have taken in our laboratory based on the p27 data is to try to restore high levels of p27 in tumors that have low levels and with that approach, with the gene transfer approach using adenovirus this seems to be a very promising approach in animal models at least.

You will hear more, I think, today from Lee Ellis about angiogenesis inhibitors. There are two molecules out there specifically targeted against VEGF. One is a humanized monoclonal antibody against the ligand. The other is an inhibitor of the kinase activity of one of the VEGF receptors.

At UCSF we have undertaken a series of studies now in Phase II trying to restore wild-type p53 function in patients with colon cancer and the liver using adenovirus expressing wild-type p53 in an effort to enhance the kind of chemosensitivity that Dr. Hamilton showed was lost with mutations in p53.

Slide 14:

What about other markers? One I want to show a little bit of data on, although I am sure you will hear more from Peter Danenberg, is thymidylate synthase since this represents a distinctly different marker, that is a marker that would predict response to a specific type of chemotherapy, in this case 4,0-pyrimidines.

Slide 15:

Again, this is data from USC showing that in patients with TS levels here by immunohistochemistry, although the same group has shown the same data with TS measured by RTPCR the chance of recurrence is 71 percent, whereas those patients with low TS levels, the chance for recurrence is approximately one-third of that, again, very highly statistically significant. But these same investigators showed that TS and p53 are, in fact, related.

Slide 16:

If you take tumor cells in culture that harbor mutations in the p53 and put in wild-type p53, you down regulate TS levels and in human tumors of those patients with high TS staining most of them showed nuclear overexpression suggesting mutations in p53 whereas the low TS staining tumors rarely had p53 mutations.

Slide 17:

So clearly one prognostic marker is likely to interact with another. The approach we have taken in CALGB is to look at a panel of markers in two studies. One is a study that Mike O'Connell talked about, a randomized trial in Stage II cancers of the colon, comparing observation alone versus the monoclonal antibody 171A and secondly in Len Saltzis study, and both of these are intergroup studies, comparing 5-FU, leukovorin versus 5-FU, leukovorin, CPT-11 as an adjuvant for Stage III colorectal cancer.

These two studies involve about 3500 patients. We will be looking at a group of collaborators within CALGB, we will be looking at -- this shouldn't be DCC. It should be 18q LOH; p53 we actually will be doing SSCP and sequencing in these 3000 patients to determine whether specific mutations in p53 are important. We will be looking at p21 by immunohistochemistry as well as p27, TS levels in those patients in this protocol treated with chemotherapy or topoisomerase-1 levels as well, microsatellite instability which Monica Bertagnolli will be doing, VEGF microvascular density, and Mayo Clinic will be doing flow cytometry.

We have chosen this panel of markers based on the strength of most of those retrospective studies that I showed you, but our hope will be to use this as a point of departure. We expect that other markers will develop over time

Slide 18:

and our plan is to be prepared for that.

One thing we will be doing, we hope, is to generate tissue arrays.

Slide 19:

In this schematic plugs are taken.

Slide 20:

I guess this will be very hard to see. I will just cut to the chase. Plugs are taken, very, very narrow diameter plugs of paraffin blocks are taken from a large number of patients, and those plugs are combined into a single paraffin block. So a single paraffin block might contain hundreds, up to hundreds of plugs from different patients, and under high power they are all stained with H&E to confirm that there is cancer, and consequently a single thin section, 5-micron section from a plug may allow you to do immunohistochemistry for example, for thymidylate synthase on two or three hundred patients all at once, and so we think that that sort of tissue array approach will allow us to have a storehouse, a tissue bank for looking at other markers by immunohistochemistry, but these tissue arrays also can be used for in situ hybridization looking for gene expression.

Slide 21:

They can also be used to look at gene copy number using fluorescence in situ hybridization.

Slide 22:

I think I will skip this section on expression arrays. We may want to talk about that through the discussion and just summarize as follows: There are numerous markers which show promise as you have heard from other investigators but large prospective studies are needed. Rational biological therapy based on genotype or phenotype of individual markers is the goal, and we really believe the technology needs to be incorporated into prospective cooperative group clinical trials.

Thank you.

(Applause.)