Molecular Pathogenesis of Small Cell Lung Cancer: Potential Clinical Implications by Adi Gazdar, MD

SLIDES & TRANSCRIPTS
Tuesday, September 14, 2000

Molecular Pathogenesis of Small Cell Lung Cancer: Potential Clinical Implications
Adi F. Gazdar, MD

Slide 1:

DR. SAXMAN: Again, we will have an opportunity to ask more questions after the presentations this morning and certainly in the breakout sessions. I think many more of these things need to be discussed and elaborated.

The next speaker is Dr. Adi Gazdar who is professor of pathology and Deputy Director of the Hamon Center for Therapeutic Oncology Research at the University of Texas, Southwestern Medical Center.

Dr. Gazdar is going to speak with us about molecular pathogenesis of small cell lung cancer, particularly as it relates to the clinical implications.

DR. GAZDAR: I appreciate the opportunity to be here and am thankful for the invitation. As we all know, lung cancer for clinical purposes, at least and for a number of other reasons, is divided up into non-small cell and small cell. In this country small cell is approximately 20 to 25 percent of lung cancer.

One thing that has puzzled me and is seldom addressed is why the incidence of small cell is so low in certain developing countries such as India where it is well under 5 percent, perhaps 3 or 4 percent, and why smoking-related cancers in different countries show these different histologic distributions is something that really hasn't been well addressed.

Slide 2:

As you all heard and know, in fact, that small cell is different for a number of reasons. It is widely metastatic at diagnostic. It is initially responsive to cytotoxic therapy. It is resistant to therapy on relapse, and it is a neuroendocrine tumor. If we understood the mechanisms of what caused these different things, I think we would have the key to curing small cell lung cancer, but in fact, our present state of knowledge is somewhat rudimentary in many of these areas.

One of the other problems, of course, as you heard being such a widely metastatic tumor, it is not usually treated by surgery, curative surgery, unless the diagnosis is in doubt. It is a peripheral tumor. You don't know what it is, and it is removed and therefore, you still seldom have an opportunity to get it because the diagnosis is made post-surgery, and one of the problems is what do you use to study this disease. Cell lines that John Minna in his infinite wisdom forced me to establish at the NCI VA Branch, which later became the NCI Navy Branch, we had about 70-odd lines from both treated and untreated cases, but limited disease and extensive disease cases, and then I estimate about 50 other lines worldwide. So we have a good pool of cell lines.

Tissue is the big problem, the major problem that remains, and one reason why I think our knowledge of small cell biology exploded in the eighties and has somewhat slowed down and non-small cell biology has emerged as the dominant studies in the nineties. I think that the lack of tissue is a major concern which has to be addressed, and how do we obtain this? Bruce Johnson has obtained autopsy tissues, but these, of course, are poor quality DNA. They cannot be used for RNA, and they are from heavily treated patients. So I don't think autopsy tissues are suitable for most types of studies.

We have tried to get a bank of paraffin tissues just like Bill Travis has, and I think one of the things that may come out of conferences like this is cooperative groups that pool their resources and establish tissue banks, even if there is some paraffin that we can all share.

I should say another word about cell lines. Cell lines are often maligned. You heard already some negative comments about cell lines. We have studied, compared the cell lines of breast and non-small cell with the original tumors they came from, and they are very good models. They represent what we started off with, but the tumors are, also, atypical. They are bad prognostic tumors, tumors with bad prognostic features. They are widely metastatic, but those may not hold true in small cell because all the tumors are widely metastatic and bad prognosis. So, it could well be that small cell lines, particularly from untreated patients, are pretty good models.

We don't have good animal models to study lung cancer and small cell in particular, and so we are left with these cell lines.

Slide 3:

There are many changes as you know in cancer cells, inappropriate response to external signals, such as ras and HER2/neu, loss of cell cycle control and a variety of different genes involved there, immortality associated with telemerase and other factors, loss of the apoptosis pathway, again P53 and BCL-2, loss of contact inhibition, the ability to metastasize, angiogenesis and autocrine growth loops. Many of these features have been studied in small cell.

Many of these features have been well described years ago. Even though many of the people who have done most of the seminal work are in this room today, you will forgive me if I pass over your work in a bullet point or two, partly because many of these things are well described in the literature for many years.

Slide 4:

Now, there are major differences with the small cell and non-small cell incidence. We talked about the neuroendocrine properties that we will go into a little bit later on. As pointed out, they are non-small cell, the neuroendocrine properties. Approximately 15 percent have the entire range of neuroendocrine properties, and we showed back in the early nineties that cell lines from non-small cells with neuroendocrine properties were explicitly sensitive to chemotherapy agents in vitro at least, the same as small cell lines were, and this has not been well confirmed in very small poorly carried out trials, I think partly because most people have used immunohistochemistry to define neuroendocrine properties. I think immunohistochemistry is a poor way to define the neuroendocrine spectrum, and we should use some of the new molecular tools to identify neuroendocrine properties and perhaps rethink some of these clinical trials for non-small cell tumors with neuroendocrine properties.

Now, ras mutations, we will come back to this point later on, C-myc amplification work and in fact myc genes done by Bruce Johnson and others, these are amplified. A certain percentage are small cells, particularly those that we treated with certain cytotoxic agents. We see them less commonly nowadays. I don't know if Bruce agrees with that, but I haven't seen myc amplification tumors, particularly those so-called "tumors with large cell features" of late.

Her2/neu, again very rare in small cell, relatively common in certain non-small cell, particularly adenocarcinomas, BCL-2 expressions and differences in incidence. P53 gene inactivation of mutations in a very high percentage of small cell and somewhat less in non-small cell. RB gene inactivation, as you heard, in most if not all small cells and much rarer in non-small cell, while the opposite P16 gene inactivation is common in non-small cell and rare in small cell.

Slide 5:

Just again to put the cell cycle perturbations then are somewhat different, being a small cell and non-small cell where the retinoblastoma gene is the major mechanism of disturbance in small cell and the P16 and non-small cell.

The end result is the same. So both these mutations end up with the same end result.

Slide 6:

Now, ras gene mutations are somewhat interesting in that they are commonly found in all human cancers. However, I don't think I have ever seen a bona fide report of a ras mutation in a small cell. Here is this highly metastatic, highly malignant tumor without any ras mutation described, a remarkable finding or a remarkable non-finding. I think it suggests, also, that perhaps there are downstream disturbances of the ras signaling pathway which John Minna and others have looked at somewhat and haven't really come to any deep conclusions, but I suspect that this is something that will emerge in time.

Also, as you have heard, farnesyl transferase inhibitors inhibit small cell and when I began to work with farnesyl transferase inhibitors from Genentech, to my amazement, I found that they caused rapid apoptotic death in small cell and they caused growth inhibition without apoptosis in non-small cell. So despite the fact that these tumors didn't have ras mutations, the farnesyl transferase inhibitors had a much more dramatic effect on small cell than they did in non-small cell.

Slide 7:

One of the advantages of having these cell lines is that we also have paired cell lines, paired B lymphoblastoid and tumor cell lines, and if there is a need for more cell lines in small cell perhaps it is to get more pairs. There are about 15 or 17 or so that I know about, most of them from the NCI group.

We also have a larger number of non-small cell so we can do detailed allelotyping. These are done by hand using selected targets, targets that we know from the literature are frequently deleted in one or another types of cancers. John Minna is in the process of doing a complete non-specific genome-wide allelotyping. So we are going to have some more hot spots identified, but these are hot spots in more than 30 percent of cell lines, and most of these have been confirmed by looking at tumors. As you can see the remarkable thing is how many there are.

There are over 50 different hot spots now identified, and many of them are not on this map, but there are differences between small cell in red, non-small cell in blue and common ones in green.

Slide 8:

One way to look at this loss is by looking at the overall loss using an index which reflects the number of regions lost per specimen.

Slide 9:

These are not cell lines now, these are tumors. We notice that small cell has the highest fractional loss if we select these markers to target lung cancers. We see that 85 percent of the regions we looked at were lost in small cell, some losing 100 percent incidence, squamous somewhat less, and these two differences are not significant, but certainly adenocarcinoma is significantly lower than either of these other two cancers.

Slide 10:

If you look at different hot spots that we know are frequently deleted in lung cancers, we also see differences. Adenocarcinoma, as you might predict from the previous slide, is usually at the bottom of the three. Squamous and small cell you might think look very similar, but there are differences. Not all the differences are identified on the slide. Certainly with chromosome 5q we see a relatively high incidence of loss in small cell and somewhat lower in squamous cell and virtually none in adenocarcinoma.

By contrast, 8p has more frequent loss in squamous cell than in the other types,

Slide 11:

and 3p, John Minna's favorite target for so many years and both a joy and frustration, I call 3p the Hope diamond of molecular biology. Anyone who works on it is cursed, but in any case what you see here the bars represent the overall loss using these multiple markers up and down the whole arm, and the bars represent the areas of loss, and several things you see. One is that in small cell and squamous cell, frequently the entire or most of the arm is lost, while in adenocarcinoma the losses are both less frequent and often much more discreet.

Slide 12:

Despite intense study, I think the nature of the genes, multiple genes on 3p, remain somewhat controversial, and I will let John Minna wax on that point perhaps later on.

One of the reasons we have neglected the literature and we have turned our attention to partly was the mesotheliomas; we knew that they were Chromosome 4 rearrangements in mesotheliomas. So we look at mesothelioma and compared them with small cell and later with breast cancer and now with colon cancer, which I don't show here.

Interestingly, loss at four discrete points is frequent in many types of cancer, including all these four —not so much in non-small cell as small cell and colon cancer, which I don't show here. We found this very high incidence, perhaps the highest incidence of loss of any finding in mesothelioma, at what we call the R1 region at the telemeric end of 4q. The same is found in small cell. After 3p losses, the most common event in small cell that we have investigated, a 90 percent incidence of loss, both in cell lines and tumors in this region here. In other regions, you see fairly high loss, particularly in this R2 and R3 region in both mesotheliomas and small cell.

In breast cancer, again a high incidence of loss, a slightly different pattern of loss and colon cancer the same thing. The other interesting thing about Chromosome 4 is that these losses, unlike Chromosome 3, are very discrete and marked by closely linked markers, so that finding the relevant oncogenes may be somewhat easier than with the 3p story.

One other thing of interest about 4p is that it is among the earliest charges that are seen. 3p is the earliest change we found, and 4p appears to be as early, and in some cases perhaps even earlier than, 3p loss in both lung, perhaps breast and colon cancer as well. So it is a very early event and important event in many tumor types.

Slide 13:

Now 8p, as I mentioned before, a very high incidence of loss in squamous cell, somewhat less in small cell and adenocarcinoma and the same borne out by the cell lines.

I should mention non-small cell lines I think are much poorer models for lung cancer than small cell lines, partly because most non-small cell lines come from adenocarcinomas and advanced adenocarcinomas with bad prognostic features and bad survival rates. So I think small cell lines are good models. Non-small cell, you have got to take with a grain of salt.

Slide 14:

To come to another point, which has been touched on by Bill Travis, is of course that small cell is a neuroendocrine tumor with the entire spectrum of neuroendocrine properties. There is not one neuroendocrine I am aware of that hasn't been described in small cell often at relatively low amounts but nevertheless present.

Carcinoids are much better differentiated neuroendocrine tumors. Low grade tumors as you heard can be divided up into typical and atypical forms, but both of these have relatively low metastatic potential and have a relatively good prognosis.

The carcinomas which, of course, the major one is small cell lung cancer, but this large cell neuroendocrine carcinoma put on the map by Bill Travis and increasingly more and more diagnosed nowadays by pathologists who are becoming aware of this entity. These tumors in the past were either called large cell tumors or funny looking small cells or perhaps even atypical carcinoids. So we don't really have a good knowledge of the natural history of these tumors. They are highly aggressive and do kill the patient, although survival curves seem to be better than for small cell.

How they respond to therapy is not fully understood. There is some anecdotal data that they do respond to therapy, although whether they respond like small cell to initial therapy we don't know. Perhaps some multi-institutional trials should be started to look at this tumor and its response to therapy.

Slide 15:

The neuroendocrine products of small cell lung cancer are numerous. There are over 30 different products being described, often multiple ones in the same tumor, and I often think of them as both eutopic, meaning products of the normal precursor cell, the endocrine cells of the lung as well as ectopic meaning that they are neuroendocrine products not made by the endocrine cells present in the normal lung. And as I said, production of one or more of these is almost universal in small cell and carcinoids.

Slide 16:

One way of looking at these neuroendocrine tumors by histology alone is that they represent a gradation from the low-grade carcinoids to the high-grade carcinomas,

Slide 17:

but when you actually look at their properties there are many differences that suggest that they are in fact two different tumor types. The carcinoids have very high amounts of these neuroendocrine properties while in the carcinomas there are much lower levels.

The carcinoids have a slight female preponderance while the carcinoma are male preponderance. Carcinoids are not really smoking associated while carcinomas are. The metastatic potential of carcinoids is low, while that of carcinomas is high. P53 mutations are rare or relatively rare in carcinoids, frequent in carcinomas.

The pattern of allelic losses, there are relatively few described in carcinoids and frequent in carcinomas. Gene mutations are frequent in carcinoids and rare in carcinomas.

Slide 18:

I think a better way of looking at these neuroendocrine tumors is to divide them into two different categories, the low-grade carcinoids and the high-grade carcinomas, acknowledging that they probably have different pathogenesis and different etiologic features and

Slide 19:

different response to therapy, the carcinoids being highly resistant to cytotoxic agents.

The autocrine pathways, again, Paul Bunn talked about them. You are going to hear a lot about them. So, perhaps I will skip that slide in the interest of time.

Slide 20:

Again, we are going to hear about angiogenesis. I would like to point out one feature. This is a histologic change not described by Wilbur Franklin but put on the map by Wilbur Franklin at Colorado. What it is is the formation of these micropapillary-like lesions best shown in the middle panel here where the stroma come into the epithelium, often carrying blood vessels which you will see better in the next picture, and this is a change we frequently see in dysplastic squamous epithelium of smokers.

I think in my experience about 40 percent of patients with dysplasia have one or more of these lesions in their bronchial tract.

Slide 21:

This again comes from Wilbur Franklin. When you do a stain for blood vessels, and I have forgotten which one this is, factor 8 or VEGF, but in any case you can see a very high density of blood vessels actually penetrating into the epithelium and these micropapillary-like fingers. The point I am trying to make is that disturbances of angiogenesis occur very early in the course of lung tumors in smokers with dysplasia, and so, therefore, prevention methods may be applicable as well as therapy methods.

Slide 22:

Now, I would like to address one last thing that we have been interested in of late, and that is the sequential development of lung cancers. As you all know, you have read in your biology books that these things follow a certain course. This isn't strictly true, but the standard line is that normal epithelium goes to hyperplasia, then squamous metaplasia; dysplastic changes appear of various grades, then carcinoma in situ which is still limited by the basement membrane, and finally, an invasive cancer.

This is all very well, but this holds true for squamous cell cancers and not for other types of lung cancers. Also, I don't think hyperplasia, squamous metaplasia are true premalignant lesions. They regress after smoking cessation, while most of the advanced dysplasias do not regress even 15 or more years after smoking cessation. I think these advanced changes are there for life once they develop, and

Slide 23:

so for squamous cell we can see that the sequence of events is from normal epithelium through these mild changes which I have not listed here, to dysplasia, to CIS and finally, invasive cancer.

Slide 24:

For adenocarcinomas a somewhat different sequence via the peripherally occurring tumors that are much harder to study sequentially. There is this change described in the periphery called atypical adenose hyperplasia, which is a hyperplasia of the type 2 pneumocytes. Perhaps these may go on to develop invasive cancer, although the association is more by inference, rather than by solid proof.

Slide 25:

Now, the problem is in small cell line cancer where we don't really have good precursor lesions. Occasionally you might see neuroendocrine cell hyperplasia. I think in about 5,000 cases I have looked at I recognize six or seven examples of carcinoma in situ. So, virtually, it seems to be a movement from normal epithelium that may go through some intermediate stages when we usually cannot recognize them to invasive cancer, and why is this?

Slide 26:

So, just to sum up what I just showed you: squamous cancers and adenocarcinomas seem to go through these intermediate stages to invasive cancer, we cannot recognize these intermediate stages in small cell lung cancer,

Slide 27:

and so we looked at these different pathways. I show the slide mainly to recognize the work of Evonne Wistuba who did most of the work.

Slide 28:

We looked at both the allelotype of these different tumor types, which I have already spoken about, and then using this data we looked at the pattern of loss in the accompanying bronchial epithelium of respective lung cancers.

Slide 29:

We looked at about 68 resected tumors, including 22 small cells. From these we identified bronchial epithelium that was mainly hyperplastic or normal. We ignored the dysplasias because we seldom found them in small cell lung cancers. So we focused on histologically normal epithelium and hyperplastic epithelium, reactive changes.

Slide 30:

We dissected out these by laser microsection, and we looked at these markers that you heard about that are frequently deleted in lung cancer. We found major differences between the major types.

Slide 31:

In the tumors we already heard the data. Small cell and squamous cell carcinomas have a high frequency of loss, adenocarcinoma somewhat less. These are the changes in accompanying bronchial epithelial foci, and one thing you notice that they are very low again in adenocarcinoma. This is a peripherally occurring tumor, and we are looking at bronchus, but it is squamous cancer which is arising from bronchial epithelium for the most part. Again, a modest number of changes, much more so than adenocarcinoma, but now look at small cell. Here it is 30 times more than adenocarcinoma.

Also, not only is it more frequent, the relative ratio of changes in normal epithelium or mildly abnormal—hyperplastic epithelium to the corresponding tumor, the ratio is one to three. In other words an enormous number of changes in histologically normal epithelium adjacent to small cell lung cancer, while in adenocarcinoma, even though they start off with fewer changes the ratio is much, much higher and intermediate in squamous cell. So this suggests to me that there is an incredible amount of damage in the bronchial epithelium of patients who are going to get small cell lung cancer.

Slide 32:

Again, if you look at the patterns of these different genes, we can almost always see differences. One thing that really stood out was losses at the P53 gene, which we very seldom see except at the dysplastic stage in non-small cell cancer and in most smokers. They had a very high incidence of loss at the P53 gene in the normal and hyperplastic epithelium of patients with small cell lung cancer, suggesting again that not only were there lots of changes but that they seemed to specifically target the P53 locus.

Slide 33:

If we map out the size of the 3p deletions, not only their frequency but their size and numbers—this is for squamous cell cancer and I am showing this for a reason—we see that the lesions are much smaller and discrete. John Minna has more detailed data to show you. In the hyperplasia and metaplasia it becomes slightly more frequent. In dysplasia it is not any more frequent. They are much larger, and for carcinoma in situ the pattern closely resembles that of invasive cancer.

If you look at the normal epithelium of small cell lung cancers, what we see is a pattern more closely resembling the dysplasias of squamous cell cancers, again proving that these lesions are much more advanced at a very early stage in small cell lung cancer.

Slide 34:

These are microsatellite alterations, changes in allele size and again we don't see a difference between normal epithelium and the corresponding tumors, but what we do see again is a much higher incidence in small cell both in the tumors and in the accompanying epithelium than in the other tumor types.

Slide 35:

So to summarize these data regarding the overall frequency of allelic loss in lung cancer, small cell is much higher, is about equal to squamous cell and much greater than adenocarcinoma. The overall frequency of allelic loss in the accompanying bronchial epithelium is much higher in small cell and squamous cell and 30-fold higher than in adenocarcinoma.

The overall frequency of microsatellite alterations in both cancers and in the accompanying bronchial epithelium is higher in small cell than squamous, which is about equivalent to adenocarcinoma.

Slide 36:

So, each of these major histologic types of lung cancer has its characteristic pattern of allelic loss. In the bronchial epithelium we see extensive molecular damage in small cell lung cancer, minimal damage in adenocarcinoma, and intermediate levels in squamous, suggesting that each of these major histological pathways arises by a distinct molecular pathway.

Slide 37:

So now we can say that small cell appears to go directly from normal epithelium through very short, if any intermediate stages towards invasive cancer, and this is reflected by this enormous amount of damage in the normal epithelium of small cell lung cancer.

Slide 38:

I was asked to point out how these things impact on therapy. I think this is the same slide that Paul Bunn showed more or less, which means we went to the same Aspen meeting and learned something from that meeting. I think we will have other opportunities to discuss the therapeutic implications of these findings in our breakout sessions and our reports tomorrow.

Thank you for your attention.