SLIDES & TRANSCRIPTS
Tuesday, February 15, 2000

Present Status Markers
Stanley Hamilton, MD

DR. MAYER: Thank you, Eric. In Eric's first slide buried in that list of authors was Stan Hamilton back in 1987, and Stan has been a pioneer in this area as well. Stan recently moved from Johns Hopkins to M.D. Anderson where he is Chairman of the Department of Pathology, and he is going to speak to us now on markers in colorectal cancer.

DR. HAMILTON: Thank you very much, Bob, and I want to thank Eric for that very nice introduction that is going to make my presentation much easier to do.

I think it is worth while when we address the issue of markers to realize what a tall order this is. The problem is not that we don't have markers, but we have too many markers, and not many of them are very good, and one of the real challenges that we face is trying to bring the utilization of the advancing knowledge in both the clinical and molecular area to impact on patient management.

This is Cuthbert Dukes, pathologist at St. Mark's Hospital in London. It is sobering to realize that the seminal publications on the staging of colorectal cancer are now almost 70 years old,

TOP

and that the mainstay of determining prognosis and therapeutic intervention in patients with colorectal cancer relates to the pathologic staging of the disease. Here is the stylized Dukes' classification that has now been around for almost seven decades.

The key issue to point out in this is that pathology staging actually does pretty well in predicting outcome of patients with disease at the very early phase of Stage I or Dukes' A disease and Stage IV or disseminated disease. The problem is that the majority of patients with colorectal cancer have Stage II or Dukes' B disease or Stage III or Dukes' C disease with metastases, and in particular if you consider the group with Stage III disease prognosis in these patients represents a coin flip, a 50 percent 5-year survival rate in most studies which is essentially the same as flipping a coin and trying to determine what the outcome is for an individual patient who is sitting in a physician's office and needs therapy.

TOP

There have been advances in the pathologic evaluation that have identified a whole variety of different pathologic markers that impact within these staging categories, and I am not going to go through a litany of these. They are available in a number of reviews in the literature.

One of the ones that can be identified relatively easily is vascular invasion. Here is a small vein in the perirectal soft tissue in a patient with rectal cancer showing involvement by cancer, and a number of studies on the basis of stage comparisons of this sort of histologic finding is associated with decreased survival rate, and again, there are a whole variety of these, perineural invasion, differentiation, and you can go through a laundry list of ones that have been identified.

TOP

Where we potentially have the opportunity to make the most progress, however, is in trying to translate basic research opportunities into clinical application. As Eric has pointed out there are three major pathways that are involved in the genesis of colorectal cancer.

He has described for you very nicely the APC pathway which leads in the progression step to a variety of different mutations including chromosomal instability with loss of allelic areas throughout the genome at a very high rate in this subset of tumors.

The second pathway is the microsatellite instability pathway, the one that occurs in patients HNPCC with a whole group of different target genes and a very low rate of allelic loss in contrast to this group, and finally as Eric mentioned the very recently identified methylator pathway in which there is extensive transcriptional silencing of genes throughout the genome. When the HMLH1 gene is silenced as Eric has indicated it kicks these tumors over into the microsatellite instability pathway but there is obviously a large variety of other genes unrelated to mismatch repair which also get silenced in this pathway and are probably also important, and this pathway has only been recognized for a relatively brief period of time. The clinical correlations with it are just beginning to unfold.

TOP

The real question is can any of these be moved into trying to improve the understanding and management of patients? Two major areas obviously come to mind, and these include prognostic markers, that is, looking at the natural history of patients with colorectal cancer as to being able to stratify those that are going to do well or badly and secondly the area of predictive markers and looking at response to therapy.

In the area of prognostic markers there are two major areas that are being addressed at the molecular level. The first is that of improved staging with the evaluation of micrometastasis by various techniques or looking at sites that usually are not the location of clinically significant metastases, such as the bone marrow as surrogates for disseminated disease, and I am not going to touch on that further because there is going to be another session following this that is going to address this issue.

The other major area is that of looking at the metastatic phenotype, that is, trying to look at the collection of alterations that occur in the tumor cells to be able to identify those that are associated with the likelihood of dissemination of the cancer or alternatively the absence of successful metastatic disease.

TOP

What I was going to show at this point was to begin to look at some of the alterations that have been identified. As Eric indicated we began a number of years ago looking at a variety of the molecular alterations that are beginning to appear in the understanding of the genesis of colorectal cancer to see if any of these could be used at the clinical level as ways of trying to impact on the understanding and the staging of patients. In particular, we looked at the chromosome 18q allelic loss initially published in 1994, now with a number of confirmatory, although some dissenting, papers in the literature that have shown association of loss of this chromosomal area with adverse outcome in patients with Stage II and Stage III colon cancer. There are a variety of ways to try to skin a cat, and one of the messages that has to come loud and clear through looking at markers is the difficulty inherent in trying to be able to do in a satisfactory fashion the evaluation of markers, not only to understand them at the molecular level but also to put them into clinical practice.

The gene target that we think is important on the long arm of chromosome 18q is the DCC gene, and this was in a study done by Kim Jessup when he was still in Boston, now in San Antonio, in which he was looking at expression of the DCC protein in a group of Stage II and III cancers and showed by immunohistochemical techniques, with the correlate, with the molecular status, that the presence of DCC protein in Stage II patients and Stage III patients was associated with improved survival compared to those same stage patients that had lost that particular protein expression.

TOP

The second area of interest has been in the microsatellite instability pathway. As Eric has shown you, there is now recognized this particular molecular pathway that has been associated initially with HNPCC, and now is recognized as occurring in sporadic patients.

Again, this highlights one of the problems in trying to define markers at the clinical level. It took an NCI-sponsored workshop in December about two years ago to get the groups of investigators who were working on this area into a room to try to identify what the criteria were going to be for characterizing microsatellite tumors at the clinical level. I must say even to this point in time there still remain substantially loud dissenting voices in this area.

TOP

Nonetheless there appears to be clear evidence that microsatellite instability also relates to outcome. The initial data came from studies of populations such as the HNPCC registries in Scandinavia. Here is a study published a few years ago comparing the HNPCC patients with localized disease or with metastases to lymph nodes or disseminated disease in comparison to the native population in Scandinavia where the autopsy rate is essentially 100 percent, so that death from disease is well documented on a population basis.

What this study clearly shows is that the patients in the registry with HNPCC stage for stage had improved survival compared to patients who did not have that condition.

TOP

As Eric has pointed out, the hallmark of HNPCC is microsatellite instability and in a very important paper, again from the New England Journal of Medicine published January 19, comes a population-based study from Toronto looking at about 600 patients under the age of 50, identifying those with microsatellite instability and then expanding this to look at outcome relative to the molecular markers and stage. What this shows is that for all patients in the study as well as by the individual patients with Stage A, B, C and disseminated disease, Stage I, II, III and disseminated disease, patients with microsatellite instability have improved outcomes compared to those patients who have microsatellite stable tumors.

TOP

Now we can move then into a second area, and that is the question of are there any markers that are indicators of response to therapy. Obviously radiation, chemotherapy, 5-fluorouracil and recent agents like CPT-11 or oxaliplatin are clearly the mainstays in the therapy of patients with colorectal cancer, and so as a consequence there has now been increasing interest in trying to look at some of these molecular pathways as indicators for responsiveness or resistance to these various markers.

This is, again, from a very important study published in Cancer Research in 1997, from the NCI group looking at the p53 gene pathway that Eric discussed relative to responsiveness in vitro of the NCI cell lines that have been used for screening for sensitivity to chemotherapeutic agents, and what this shows is that at two levels of radiation in vitro this is the percent G1 arrest as an intermediate biomarker for responsiveness to radiation therapy. What one clearly sees here, as indicated by various cut points within the data set, is that all of the tumor cell lines that are characterized by high levels of G1 arrest and then expected sensitivity to radiation therapy are those that have wild type p53, whereas the lines that are characterized by mutations in p53 are generally poorly responsive or unresponsive to radiation therapy.

TOP

Another target that was looked at in this study that is going to be relevant to what I am going to present to you in just a moment is the p21 WAF1 protein. This is a cyclin-dependent kinase inhibitor that is regulated by p53. This was, again, looked at in this in vitro study and shows that the response with mRNA induction of the p21 protein in these various cell lines again was very strongly associated with the presence of wild type p53 in those individual cell lines, suggesting that this may be a very important marker pathway for radiation therapy.

TOP

In addition, this group looked at chemotherapy responses to a wide variety of agents, and the one of obvious relevance for colon cancer is 5-fluorouracil. This plots the negative log of the inhibitory concentration of 5-fluorouracil in vitro and again shows that the wild type p53 cell lines require in general more concentrations of 5-FU than those that are characterized by mutated p53 although there is obviously extensive overlap in these groups.

TOP

Can this be translated into trying to look at clinical applications? Again, this is a major question because clearly as we begin to look at some of the clinical markers there turn out to be differences from what occurs in vitro.

These are studies done with microsatellite instability lines done at the University of San Diego looking at response of lines that are characterized by microsatellite instability and microsatellite stability to oxaliplatin, and what this shows is that after a single dose of oxaliplatin the microsatellite stable lines show a decrease in tumor volume in the animals followed by a regrowth.

In contrast to that, the lines that are characterized by microsatellite instability are unresponsive to that particular agent, and a number of other studies have shown similar resistance in vitro to 5-fluorouracil.

TOP

We have now tried to address in the Eastern Cooperative Oncology Group translating some of these markers from the laboratory setting and understanding their basic biology into the utilization as clinical markers.

These are from patients who were enrolled in E2284 and E2288, the ECOG component of the intergroup trials 35 and 89. What we did in this was to accumulate the blocks from these patients and then to analyze them for some of the markers that we have been talking about and in particular 18q allelic loss, 17P allelic loss and p53 overexpression of the type that has seen mutation as surrogates for trying to do mutation analysis. We really did not want to take on trying to sequence the p53 gene in 500 cases for formalin fixed paraffin-embedded tissue. So we elected this surrogate.

Looking at p21 WAF1 that I talked about a few minutes ago, which is a p53 regulated gene, and finally at microsatellite instability including some of the genes like TGF beta R2 and BAX that Eric told you about a few minutes ago that are often mutated because of the presence of microsatellites in the coding region of these genes.

These are the frequencies of the abnormalities across this spectrum of cases ranging from about 300 for some of the markers up to almost 500 for the patients who had immunohistochemical markers looked at,

TOP

and from this data analysis done by Paul Catalano, who is in the audience, we identified that there are two sets of markers that appear to be related to response in Stage III patients treated with 5-FU based chemotherapy regimens, and these are 18q allelic loss and microsatellite instability in a particular mutation of the TGF beta R2 gene.

These data have not yet been published, are going to be submitted for publication very shortly, but I will go ahead and share them with you.

In the upper left hand corner are all patients stratified by 18q status including tumors in the chromosomal instability pathway as well as those with microsatellite instability where allelic loss is uncommon. What this shows is that in the subset of patients who have retained 18q allele the 5-year survival rates after 5-FU therapy in Stage III patients is about 75 percent, whereas in patients who have lost 18q alleles the 5-year survival rate is about 50 percent, essentially what would be predicted in untreated patients.

If we subdivide this group in a subset analysis and look at the patients who have microsatellite stable tumors, excluding those that have microsatellite instability, the split in the survival curve becomes a bit wider, and the marker becomes more predictive, again, with 18q allelic loss absent within these tumors 5-year survival rates of about 75 percent. With 18q loss present in the microsatellite stable subset about 45 percent 5-year survival with a hazard ratio that approaches three.

Looking at microsatellite instability alone in these groups there was somewhat of a surprise. We expected to see improved outcome based upon what existed in the literature, but there were no statistically significant differences, although there was a suggestion that the microsatellite instability high tumors did better, but what was of interest is the fact that when we looked at the specific genes that are mutated in this subset of tumors, those that are characterized by TGF beta R2 mutation had improved survival rate within the subset with high levels of microsatellite instability compared to those that had not mutated the TGF beta R2 gene in the polyadenine tract, and again the survival rates here are about 75 percent versus 50 percent at 5 years in these subsets, again, with hazard ratios approaching three.

TOP

What this has done then is to allow us to potentially put forth a stratification scheme based on these molecular markers in tumors. In Stage III patients who are treated with 5-FU adjuvant chemotherapy, we can evaluate now the long arm of chromosome 18q and the microsatellite instability status, first asking the question is the cancer characterized by high levels of microsatellite instability. If the answer to that is yes, the next question is whether the TGF beta R2 gene is mutated. In those cases that have that gene mutated the 5-year survival rate is about 75 percent compared to about 50 percent in patients who don't have that mutation.

In the chromosomal instability microsatellite stability pathway on the other branch of this, the next question to be asked is there 18q allelic present. If 18q alleles are lost, again, this puts the patients into a high-risk category of about 50 percent 5-year survival rates after adjuvant therapy, whereas in the patients who have retained 18q alleles the 5-year survival rate is about 75 percent, and again, of considerable interest to us was the fact that our analysis of the p53 pathway at both the DNA and protein level was unrelated to outcome in this series of patients.

TOP

What about the problem of markers in general? Obviously as I said at the beginning of the talk there are a tremendous number of markers that are potentially available and can be applied in the clinical setting, and the problem that we really have is in trying to sort these out as time goes on.

Obviously the process of metastatic disease is an extremely complicated one involving large numbers of biochemical and other steps that have to be undertaken by cancer cells in establishing satisfactory, or in the case of the patient, unsatisfactory metastatic foci.

This offers large numbers of opportunities to look at markers, but obviously the question is how we are going to begin to use these in the clinical setting and try to reduce this very large number of potential markers down to a usable number.

TOP

The other major issue that has to be addressed is that of how these are going to be evaluated and what the problems are in trying to bring these to clinical utilization.

You are going to hear from Peter Dannenberg tomorrow about another approach of trying to look at drugs that are involved in metabolism of 5-FU as another potential set of markers. So there are lots of ways to skin the cat in this area.

What is important is to recognize, however, that there is really a subtle difference between what we talk about as prognostic or predictive markers and the process of getting to the development of these individual markers.

What we do is a sort of study that I have shown you this morning where we take a large series of patients and we analyze the tumor that has been removed for them looking for these various markers. We come up with statistically significant differences with P values and hazard ratios.

The problem in the area of clinical application and the evaluation of markers that will really make a difference in the way individual patients are managed is in the area of prognosis, because what you are dealing with there is the outcome in one patient.

You now have to move your thinking into looking at the topic of predictive values and sensitivity and specificity, not of hazard ratio.

Furthermore, the problem here is that in patients who have their colon cancer removed surgically and then the tumor studied, the outcome in these patients is not determined by what the pathologist is looking at under the microscope but rather by the micrometastatic disease, the tumor that has not been removed from the patient or that is not amenable to the postoperative adjuvant therapy, and finally, it is worthwhile to point out that this whole process requires basically fortune telling. We are having to try to predict what is going to happen in the future based on some understanding of what has happened already in trying to characterize the alterations that are present in tumors, and obviously, from Eric's talk, the complexity of this area is well apparent.

Thank you very much for your attention. I will be glad to answer questions.

(Applause.)

TOP