SLIDES & TRANSCRIPTS
Tuesday, March 6

Resistance to Therapy: Molecular Markers in GI Oncology
Heinz-Josef Lenz, MD

I actually will talk about two different areas. One is the characterization of genetic profiling of tumors as predictors of outcome to chemotherapy.
I actually will not go into protein expression. P53 will be actually discussed with Dr. Ford. I will show you some new data on genomic polymorphism as a new way to characterize patients who will respond to chemotherapy, and maybe actually characterized by toxicity to chemotherapy.

TOP

You all are familiar with the 5FU metabolism, and I use this as a paradigm to see and explain how molecular markers can be identified and characterized. You know, in the 5FU pathways there are three critical enzymes. The most critical is the thymidine synthase, which is the target gene of 5FU. You know, 5 FU gets metabolized to F-dump. F-dump binds into a suicidal complex with TS and inhibits DNA synthesis. So, it was not far fetched to test a hypothesis, that if you have a high expression of TS, that it would predict resistance to 5 FU, because you are not able to deliver enough 5 FU to inhibit the enzyme.

One other enzyme that is very important is DPD, dipyrimidine dehydrogenase, which is a detoxifying enzyme. High levels would detoxify at 5 FU, unable to inhibit TS. There is a third enzyme which has kind of different areas of action. You heard yesterday from Dr. Radinsky that thymidine phosphorylase is a very important factor in angiogenesis and actually a poor prognostic marker in patients with different cancers, including gastric cancer. In the 5 FU pathway, it is an activating enzyme, making 5 FU more active metabolized by F-dump, provides more F-dump to inhibit TS.

TOP

When we started out testing the hypothesis, this is more or less the summary that came up, and it was just published in Clinical Cancer Research in May 2000. You see on the left-hand side patients who responded to chemotherapy had actually a profile including that all three enzymes I mentioned -- TS, DPD and TP -- had low expression levels. If one, two or all three enzymes were increased, we did not see any more response to chemotherapy, which was the first time with the molecular profiling you could not only pick out the patients who will not respond to chemotherapy, but also the patients who will respond to chemotherapy.

In this case, these are patients with metastatic colon cancer treated with protected infusions, 5 FU. This would give us the first hint that actually, identifying the molecular target may guide us in what kind of chemotherapy would be beneficial. Based on this data, prospective clinical trials are ongoing at Sloan Kettering and in Germany, that 5FU or CPT-11 chemotherapy will be selected based on the TS gene expression levels prior to treatment beginning.

TOP

When we actually put up the data together, as you can see, when we used three enzymes or two enzymes, over 90 percent of the patients will respond to chemotherapy, just using two or three markers identifying the patients with clinical benefit.

TOP

The big question is, is this only a number jungle, or does it really translate into a survival benefit for patients?
In order to make it very clear, we used two enzymes, TS and DPD. It becomes very clear that patients with low expression levels of these target enzymes actually live significantly longer than patients who have either one or both enzymes elevated.
There is a significant survival benefit. Actually, the median survival in our patient population goes to 18 months with single agent 5 FU.

With markers, we did not only select out the patients who respond, but also who have a survival benefit. In particular, in the area of choosing toxic regimens, this is of significant benefit. Coming back to gastric cancer, would this data reflect and be replicated in gastric cancer patients? This is a trial initiated by Larry Leichman at USC in a neoadjuvant setting. Patients were treated with two series of 5 FU cisplatin treatment and prior surgery. The end point was also a pathological end point. It came up in the discussions how important neoadjuvant approaches could be, because of the access to evaluate the biological activity of chemotherapy at the time of resection.

TOP

Coming back to the TS expression, what we saw was similar data to the data we saw in colon cancer. We had TS gene expressions above or below a certain cut off. We did not see any more response to chemotherapy than TS was elevated to a certain level, which actually are the same data we saw in colon cancer, and it didn't make a lot of sense. However, low expression levels did not predict and select out the patients responding 100 percent, but increased the probability significantly.
Now, because we had a combination chemotherapy with 5 FU and cisplatin, we looked for markers which could predict the efficacy of cisplatin agents.
Now, at this time, in vitro data suggested that a DNA repair enzyme -- ERCC-1 -- was directly associated with cisplatin resistance. ERCC-1 is a DNA repair enzyme which actually repairs intra and inter-strand DNA adducts efficiently, and would therefore make a lot of sense to actually be correlated with cisplatin efficacy.

TOP

When we set up a technology to measure ESCC-1 gene expression levels, again in endoscopic biopsy prior to the start of chemotherapy, we saw the following data.
Patients with a high expression of DNA repair capacity were actually resistant to cisplatin, which made a lot of sense.
Patients with a low expression level were more sensitive to chemotherapy and actually were highly likely to respond to chemotherapy.

TOP

Again, we have two markers for two different agents. When you combine them, did we actually see an additive effect.
Patients with ERCC-1 or TS, we increased the selectivity of the prediction of response significantly. Actually, we were able to predict response in 85 percent of this patient population. Again, in this neoadjuvant setting in locally advanced gastric cancer, these markers actually predict also outcome in regard to survival, the golden standard of efficacy.
It did. Using ERCC-1, there was a significant survival benefit for patients who had low expression levels and associated response to cisplatin.

TOP

This was true also for TS, low expression level of TS was significantly associated with a better clinical outcome.

TOP

These data were published in JCO in 1996. We actually expanded our program to look more aggressively in the GE junction and esophageal cancer, to see if this data would be true for these particular cancers.
As you already mentioned, there are a lot of other agents that are active. So, we were particularly interested in the activity of other medications used in GE junction or esophageal cancer, including mitomycin and taxanes. Here, we actually did a pilot study with adenocarcinoma of the cardia. Again, TS was a significant predictor of survival. I will spare you all the response data, just cutting to the chase, to give you the impression and try to convince you that molecular profiling is a significant marker to select out patients who have a higher benefit from chemotherapy.

TOP

Now, this technology gives us an incredible tool. For most of the drugs, we know what the targets are. All the new medications that we heard yesterday are actually more targeted than the standard classical anti-cancer drugs. The identification of these targets is even easier.
Here we actually did a study with patients treated with 5 FU taxol and cisplatin from the group from the University of Munchen. What we did is actually looked for better tubulin isoform quantitation in the tumor cells. Patients had actually esophago-cardia adenocarcinomas. You know, taxane targets are the better tubulin isoforms, in particular the isoform-3. You are aware of the data presented at the last ASCO about the better tubulin mutations with the results data. Here, we actually went to the expression level of the target of taxol. What you can see is, in the triple combination, is that actually better tubulin-3 isoform quantitation was predictive of survival in patients treated with a triple combination. It opens up again the opportunity to profile tumors and pick actually the more successful chemotherapy based on molecular profiling.

TOP

We looked actually also for patients who have looked if ESCC-1, because there are some data on the in vitro level that maybe DNA repair may be associated with some of the chemotherapy regimens when taxol is involved, and actually, in this patient population, we again saw some tendency that ESCC-1 may be a predictive marker in the triple combination using 5 FU, cisplatin, taxotere and radiation, again, in cardia adenocarcinomas.

TOP

One of the new drugs which actually was entered in the clinical trial programs is oxaliplatin. Since it is, at the moment, only available for colon cancer, we actually asked the question if oxaliplatin would be able to characterize, by certain gene profiling, in the predictive frame.
We had patients treated at our center, with compassionate use, third line chemotherapy, where we actually measured TS again for patients who failed already 5 FU. I tried to convince you that TS is an important predictor of chemotherapy with 5 FU, and we established a particular cut off level.

The idea that 5 FU could work again in combination with oxaliplatin, that TS may be another predictor level at a different threshold was, for us, actually a very interesting idea. The level of increase and up-regulation of TS may translate into, again, efficacy in a 5 FU combination based on data, as you know, shown and published that maybe CP-11 in combination with 5 FU, or oxaliplatin in combination with 5 FU is downregulating the TS gene.

Here, we actually were totally surprised to see a survival benefit for patients in third line treatment in combination with 5 FU oxaliplatin with an established calculated cut-off level in this patient population, which is 72 patients. There was a median survival here of about two to three months for patients below this level, that is 10 months. So, there was a seven to eight month survival benefit in third line chemotherapy using TS.

TOP

The gene that we thought would be the most successful would be ERCC-1. In fact, it was as well as TS, showing a significant predictor of outcome in third line treatment with oxaliplatin for patients who failed 5 FU and leucovorin and CP-11. This is actually a combination. When we did TS and ESCC-1 together, there is a significant survival benefit.

TOP

This opens up the question that ERCC-1 may be actually a significant predictive marker for patients treated with oxaliplatin.

TOP

As I mentioned before, we actually also looked for other genes involved in metabolic pathways. This did not reach statistical significance. This was for data generated with Dr. Isakoff from UCLA, to look if DT diaphorase may be an important predictor to mitomycin-based chemotherapy. As you well know, DT diaphorase is an activating enzyme in the mitomycin pathway.

TOP

One of the newer interesting targets is GST-pi-1 in the action of cisplatin activity. As you know, GST is, at the moment, also the target of the new development of small molecules going into clinical trials in colon and pancreas cancer. We have actually established, as the expression level of glutathione GST-P-1 as a significant predictor in survival for patients with cardia adenocarcinoma with a statistical significance, and patients who have high expression level did significantly better. This opened up the possibility with adding molecular markers in the pathway of these metabolic agents to actually characterize the outcome of these patients.

TOP

This is actually not limited to genes in the tumors. It could be also applied to receptors. This is very preliminary. We have done more data in colon cancer, but I just want to give you an idea. These were patients involved in the national trial with CP-11 C225, patients with metastatic colon cancer. All these patients stained 2-plus for each EGFR protein.
When we actually microdissected the tumors, we actually measured the gene expression level of EGFR. Yesterday we heard that actually the presence is not enough, we just need an active receptor. What we found, this is the expression level of EGF gene expression level. CR is the complete remission, PR is for partial response, SD for stable disease and we had one patient with progressive disease.
I don't want to draw too much conclusion, but it seems like the expression level may be of importance for EGF receptor in regard to the response to C225.

TOP

If you want an example that it really works, these are patients matched with paraffin-embedded tissue sections with fresh frozen tissues, actually done at different time points, because we had to go back and get the paraffin plug. It matched the clinical outcome.

TOP

What was our biggest surprise was the ability of this technology to go back 18 years in paraffin plugs to recover RNA which would be in quality and quantity enough to measure gene expression levels. That was our biggest surprise.

TOP

How powerful this technology is, is just shown as an example in this slide. We had a primary tumor, we had a liver MET, gene of interest. We could microdissect from one paraffin-embedded single section tumor cells, measure the quantitative aspect of genes which we thought would be interest at this aspect, and evaluate then the clinical outcome to it.

TOP

The last couple of minutes I want to spend a little bit of time on genetic polymorphism. You know, usually these polymorphisms are related, and a major focus of research in epidemiology as potential cancer risk factors.
As you can imagine, enzymes which play a role in metabolizing certain anti-cancer drugs should be of clinical significance, particularly for toxicity.

TOP

We were focusing on the significance of certain genes in outcome to chemotherapy. We looked at TS polymorphism and, as you know, we have established TS as a molecular predictor to outcome.
So far, the regulatory relevance of TS is not very well understood. So, we wanted to know if maybe a polymorphism which was described in vitro, and associated with TS activity in vitro, would actually be applicable in a clinical setting, that polymorphism may be an explanation for gene upregulation in the tumor and normal tissue.

TOP

One of the interesting concepts about the genomic polymorphisms, because it is usually tested in blood, is that there are significant differences in different ethnic populations.
For TS, just recently, Howard McCloud from Rush University showed that there are particular genomic polymorphisms in the TS genes in African Americans. We actually identified at USC a particular new polymorphism in the UGG-18 in Latinos, which showed it had more clinical toxicity to CPT-11 treatment.

TOP

When we did the testing, we actually found a very interesting correlation between genotype and intratumor gene expression. LL stands for long, which is a triple repeat of the 27 base sequence in the 5 UTR region of the gene. S stands for short. It is a double repeat, and the SL is the heterozygote.
What we found, we actually here had 42 patients where we had blood samples and biopsies for metastatic liver lesions from colon cancer patients. What we found is that the mean TS expression level in patients with an LL genotype, which in vitro had shown higher TS activity, showed significantly increased gene expression levels with a confidence interval between five and 16. Now, if you have looked very carefully in our first slides, our cut off in response to 5 FU is around 4 to 4.5. This patient should not respond to chemotherapy.

This is also the first time, to my knowledge, that a genomic polymorphism is linked and associated with intra-tumoral gene expression. The SS genotype actually had a mean of 2.6 with a confidence interval of 1.3 to 4.8. This indicated that this patient should have a high likelihood of responding to chemotherapy. Almost like it is made up, the heterozygotes are exactly in the middle. In the meantime, because of the suggestive nature of this, we actually now identified a transcription factor binding to that area, and we are in the process of characterizing this transcription factor. Now, another interesting component of this was that we had also access to normal liver. It is not as good control as a normal colon because it is metastatic colon cancer, but we wanted to see if there is any relationship between polymorphism and normal tissue expression. What we found is that there is actually a significant difference between the LL and the SS genotype in the TS mean expression level. What this could mean is that this actually could have the potential to predict toxicity. This should have the potential to predict response, and that is exactly what we found.

TOP

The TS genotype predicted response to 5 FU chemotherapy. Patients with an LL genotype, only two out of 23 responded. The patients with the SS genotype, four out of seven responded. The patient numbers are small, but it reached statistical significance.

TOP

When we looked for clinical toxicities, that is what we also found, a relationship between genotypes and toxicities. Patients with the LL types should have normally, based on our data, very little clinical toxicity. That is what most of the patients had.
On the other hand, all SS genotypes had either moderate or severe clinical toxicities, indicating that TS polymorphism, on a genomic level, actually can predict response to and toxicity of chemotherapy.

TOP

We actually expanded our genotype program into other areas, particularly the cisplatins. For patients who were treated with 5 FU oxaliplatin, we identified one other polymorphism in the XRCC-1, which is more or less the cousin of the ERCC-1, another DNA repair enzyme.
What we found in our limited population was that we reached statistical significance in predicting response to 5 FU oxaliplatin, using another polymorphism in the DNA repair enzymes. We are at the moment in the process to actually include other genotyping and other enzymes, polymorphisms to characterize other agencies, including EGFR. As you know, there are polymorphisms in the receptor.

TOP

From my point of view, I would like to see these genetic profiling markers included in clinical trials. There was a big ASCO plenary session to ask, are we ready for prime time. How can we use this data we have now generated that showed statistical significance when we are doing clinical design. I think that some of this data should be included, even at a level where we start out treating patients like our colleagues in lung cancer did, using ERCC-1, using TS, and even better tubulin mutations to design prospective randomized trials, looking for the predictive value of these markers, and actually learn more about the mechanisms of resistance in this patient population. Thank you.

TOP