Articles, Blog

J. Michael Bishop (UCSF) Part 3: The cancer genome and therapeutics

August 27, 2019


Hello. I am Mike Bishop from the University of California, San Francisco. And I am now going to tell the third and final chapter of my story about cancer. And this chapter begins with the German medical scientist Paul Ehrlich, who towards the end of the 19th century became interested in using chemical dyes to stain human tissue. He serendipitously discovered that some of these dyes preferentially stained microbes like bacteria and parasites, as opposed to human tissue. And this inspired a vision of what he called “magic bullets”. If chemical dyes could preferentially stain a bacteria as opposed to a human cell, perhaps drugs could be developed that would preferentially kill bacteria, as opposed to human cells. He initiated what was probably the first drug screen in the history of medical science. His first major discovery was a drug that was effective against trypanosomiasis. But he became an international celebrity with his 606th drug that proved to be an effective therapy for syphilis. He called it Salvarsan, as in salvation. Most people do not know that Paul Ehrlich was not finished at that point because his ultimate objective was cancer cells. He had a vision of a magic bullet that would kill cancer cells rather than normal cells. Between 1904 and 1909, he put many hundreds of chemicals onto normal and cancer cells in the laboratory without ever finding his magic bullet. He received the Nobel Prize actually for another discovery, independent of his work with dyes and therapeutics, but he died a disillusioned man telling his wife on his death bed that his life had been wasted. This apparently was traced to his failure to find a magic bullet for cancer. A century or more later, we now are on the verge of having Paul Ehrlich’s magic bullets for cancer. And it is based on the genetic paradigm of cancer that I have discussed in the first two chapters of my story. The fundamental tenet is that all cancer arises from the malfunction of genes. This tenet suggests that we can create magic bullets. The targets would be cancer genes and their protein products. These would create distinctive therapeutic targets, distinctive to the cancer cell as opposed to the normal cell. In other words, we would hope to kill the outlaw cell without harming the innocent bystander. There are two culprits involved: malfunctioning proto-oncogenes which suffer gain of function, and malfunctioning tumor suppressor genes that suffer loss of function, and as I have explained in my first two chapters, these combine to give rise to the malignant phenotype. How would we approach targeting these ailments? Well, we would inhibit a gain of function. We would want to replace a loss of function. We know how to inhibit gain of function and it’s an active exercise with some early successes. But we do not yet know how to replace a loss of function. There is a third newly emerging and still relatively experimental approach in which neither the cancer gene or its protein product is a direct target. Instead, we attack from the flank. I am going to use experimental data to illustrate inhibition of gain of function and two examples of attacking from the flank. One directed to a refractory gain of function, and a second addressed to loss of function. My first example is actually the only truly curative targeted therapy yet developed. And it is to treat a disease known as Acute Promyelocytic Leukemia, and the leukemic cells are illustrated in this picture. This was a serendipitous targeting, but now that we understand it, it is exquisitely specific. This was an untreatable disease, an incurable disease, until 1986 when Chinese scientists in Shanghai discovered that a relative of vitamin A, all-trans retinoic acid, could induce remissions, not cures, but remissions of this disease. Then they combined all-trans retinoic acid (ATRA) with conventional poisons, conventional chemotherapy, and developed a cure for about 80% of the patients. This became standard procedure around the world and that was where things stood for quite some time. Meanwhile, it was a mystery as to why the all-trans retinoic acid was working until about a decade or more later after its discovery. The leukemic cell of acute promyelocytic leukemia contains a chromosomal translocation. And this translocation fuses major portions of two genes: one the gene that encodes a major receptor for retinoic acid, so called retinoic acid receptor alpha, and the second a heretofore unknown gene known as promyelocytic leukemia gene now, PML. This fusion protein is apparently a driver in the development and maintenance of the leukemia, and the therapy with retinoic acid was presumably due to binding to the retinoic acid receptor portion of this mongrel protein. Now what about the 20% failure rate? It was of two sorts. Some people never responded. Some people relapsed after the conventional therapy and died. And no one did much about this for quite some time, but meanwhile out in the provinces of China a solution was in the making. A group of physicians and oncologists in Harbin province were exploring the use of traditional Chinese medicines, and one of them is known as Ai-Lin 1. And in 1992 they reported in a Chinese journal that this folk medicine elicited remissions of acute promyelocytic leukemia. That report inspired the scientists in Shanghai who had developed retinoic acid therapy to take a look. First of all, they became convinced that this was actually happening, that this folk medicine was eliciting remissions. And then they decided to identify the active ingredient. Ai-Lin 1 is composed of ground toad bladder, a rock that is rich in mercury, and a rock that is rich in arsenic. The Chinese scientists decided to ignore the toad bladder. It was too complicated. Mercury is a deadly poison. Arsenic when used in reasonable doses was already a therapeutic for some infections and for some leukemia, so they thought, “Let’s check out the arsenic.” The active ingredient of Ai-Lin 1 in the treatment of acute promyelocytic leukemia proved to be arsenic trioxide. In 1997 a persuasive clinical trial was published by the scientists in Shanghai demonstrating that arsenic trioxide elicits remissions following relapse from conventional therapy. This was the way the drug was tested and used for quite some years. Not as a primary therapy, but as a last resort after the conventional therapy had failed. Meanwhile Diane Brown and Scott Kogan in my laboratory had developed a mouse model for acute promyelocytic leukemia. They had linked the mongrel gene PML-RAR alpha taken from a human leukemia cell to a control element called MRP8 which targets the expression of the mongrel protein to the promyelocyte, the cell in which the leukemia allegedly originates. To our great satisfaction, these mice developed acute promyelocytic leukemia. In fact we could see a pre-leukemic state in their bone marrow long before the disease itself occurred. They also developed benign epidermal polyps, but that was of no consequence to our work. About the time that we had this model up and running, we learned from Hugues de The in France of the Chinese experiments with all-trans retinoic acid… arsenic trioxide. So we resolved to collaborate with de The and his colleagues, particularly Lallemand-Breitenbach, and test the utility of arsenic trioxide in our mice as a single therapy and as a combination with all-trans retinoic acid. The results were stunning. This is the survival of the untreated mice. They all die quickly and almost synchronously. There is a modest extension of lifespan here the dotted line is treated with all-trans retinoic acid alone. Treatment with arsenic trioxide also extended lifespan modestly, but when we used the two together, the mice were cured. In fact they live normal lifespans, and all evidence of leukemic cells or translocation disappeared from their bloodstream or their bone marrow. This inspired us to make two proposals. First of all, combination of all-trans retinoic acid and arsenic trioxide might alone be curative. We might be able to eliminate the highly toxic chemotherapy part of the standard therapy, anf secondly why wait for a failure? Why not use this combination as frontline therapy? It took ten years, as the development and analysis of clinical trials usually do, but eventually the Chinese in Shanghai again came through demonstrating that combination therapy with all-trans retinoic acid and arsenic trioxide as frontline therapy, as first treatment, led to a 95% five-year survival. That was of 2009, most of these patients are truly cured by now. Now they were still using the combination with chemotherapy as well, but in more recent years it has become apparent that many patients will respond to the combination of all-trans retinoic acid and arsenic trioxide without the other chemotherapeutics just as our mice did. And it is now possible to classify patients as to those who don’t need the additional toxic chemotherapy and those who do. Now why this efficacy? Well there are three explanations. One of them is a touch hypothetical, and that is there may be only a single driver in this leukemia, which would make this, that is to say, an exceptionally vulnerable pathogenesis. We do know that the leukemia stem cell, this is one of the tumors in which stem cells have been authoritatively demonstrated. I spoke on this in chapter two of my talk. We know that this therapy kills the stem cell, not just the full-blown leukemic cell. And we now know that it also represents a bimodal attack on a single target. Now what do I mean by that? Well, here is a cartoon taken from Hugues de The’s writings showing the mongrel protein bound to DNA. That is not the point. The point is that the all-trans retinoic acid attacks the retinoic acid receptor part of the protein. The arsenic, it’s now been shown, attacks the PML part of the protein. The mechanisms of action are entirely different. The likelihood that any single cellular lineage will develop resistance to both of these in the same molecule is infinitesimally small. Hence the remarkable efficacy, and the failure of resistance to emerge and the consequent ability to cure. So we’ve learned some lessons from this experience. First, our ability to predict the outcome of therapy as it might occur in humans, and did occur in humans presaged a new era for preclinical models. The classical mouse model for testing cancer therapies was essentially discredited by the time we began this work. But many scientists have now been working to develop the kind of model that I described, which is based on the genetic lesion found in a human tumor. And it is my belief that this type of model will set a new standard for preclinical testing of cancer therapy, and I am not alone in that. Secondly, it is important to attack the tumor initiating cell, which means that we are going to have to devise ways to identify that and distinguish it from the bulk population of the tumor and evaluate its therapeutic response. And third, we are not going to escape the need for combination therapy. It is, if nothing else, the way to avoid the emergence of resistance, and the combination of the bimodal attack on the mongrel protein in acute promyelocytic leukemia is a dramatic example of that. Which brings me to the use of flank attack. This is an approach that exploits a phenomenon known as synthetic lethality. It represents a cooperation between therapeutic agent and the oncoprotein, the tumor driver. Synthetic lethality has been known for many years. It was first discovered in bacteria and yeast. And it is a simple phenomenon, but not always fully understood. Two strains of a microbe with two different gene mutations, A and B, each of these separately are viable, but when combined together they are lethal. Now in our thinking about this, my colleagues and I and others simply substituted the cancer gene as mutation A, and the therapeutic as mutation B. Both of these might well… certainly the cancer gene is consistent with viability. And the therapeutic might be consistent with viability as well, but put the two together and perhaps we can get a synthetic lethal interaction. We resolved to test this with the proto-oncogene MYC. This encodes a highly pleiotypic transcription factor. Some reports allege that it is involved in the control of as much as half of the human genome. 10,000 human genes. The physiological expression of MYC is involved in a number of normal functions: cell proliferation; cell growth, increase in size; differentiation. And if the gene is overexpressed, it can lead to suicide, apoptosis, destabilize the genome leading to chromosomal abnormalities and other genomic maladies. If you install MYC in a transgenic animal and drive its expression to a particular lineage as we did with the mongrel protein of acute promyelocytic leukemia, it can be tumorigenic in various organ systems. And most importantly, overexpression of MYC is among the most common ailments in human cancer, and it is found in a wide variety of human cancers including some very important, such as breast cancer. Now, so this sounds like a good target for therapy. The problem is that it has got two liabilities. First of all, MYC is essential for normal cells. And we might well wreak havoc by inhibiting it directly. And secondly, as a transcription factor, it is not a popular target for pharmaceutical chemists. This may change soon. There certainly is a lot of work being done on it. But at the time that we were thinking about it, it was certainly considered, quote, “undruggable.” So why not attack from the flank? I am going to show you two examples of how we have done this. The first example involves inhibition of the G2 to M transition in the cell cycle. And the second example involves cytokinesis. The first is the work of Andrei Goga with some help from Aaron Tward and David Morgan and myself. And it derived from Andrei’s interest in using inhibitors of cell cycle kinases. The two that he had in mind were CDK2 and CDK1. We set CDK2 aside because there were reports of failures with this. CDK1, however, had never been addressed. Andrei utilized an inhibitor that had been developed in Peter Schultz’s lab called Purvalanol. This drug inhibits CDK1 preferentially, although not exclusively. And it creates as these data will show, I am not going into detail, it creates a block at the G2-M transition. If you create that block in a normal cell, it is reversible. Take the drug away after a day or two and the cells recover and grow. What Andrei discovered was that any cell, be it an otherwise normal cell or a human tumor cell, that was overexpressing MYC would die when he used Purvalanol to inhibit CDK1. This is a partial example of his results. So there’s no killing of human fibroblasts, epithelial cells, rat cells, rat cells expressing an oncogene, the RAS oncogene, and no killing of this cell line. But, any cell line that he tested that was overexpressing MYC died promptly. This led them to do some preclinical tests in mouse models. The first model is a liver tumor which arises from targeting the expression of MYC to the hepatocyte. This model had been developed in my lab, and Andrei used it to do a preclinical test. And this was a more sophisticated model than the one I mentioned before. We used a now familiar trick of placing the expression of the transgene under the control of doxycycline. So in this instance doxycycline keeps the transgene off, so Andrei was able to keep the animals, keep the transgene turned off until they were weaned, and then turn it on. And if he treated the animals with a placebo, if you will, the livers looked like this after a few weeks. But if he treated the animals during the same time period, the livers looked like this. These animals are not cured but they have obviously undergone a dramatic resistance or regression of the tumor, which eventually recurs and we have not explored the reason for that. We also tested a B-cell lymphoma elicited by overexpression of MYC, and again, this time he was able to examine survival. Notice that these data were collected after a single brief treatment of Purvalanol. And he got a significant extension of lifespan just by inhibiting CDK1, not by addressing MYC directly. So, we have here a classical synthetic lethal interaction between a therapeutic drug and an anomalous expression of a cancer gene. The pre-clinical results certainly suggest potential therapeutic efficacy, and indeed, Andrei is now as an independent member of the faculty at UCSF in the process of mounting a clinical trial to test this idea. And it is also, we would suggest, that overexpression of MYC is going to prove to be a biomarker for sensitivity to treatment of any kind of tumor by inhibition of CDK1. Time will tell. The other example of attacking MYC from flank involves a chromosomal passenger protein complex, and this is the work of Dun Yang. The chromosomal passenger protein complex has three major components. There is at least one and perhaps more, others, but the one that we care about is called the Aurora B kinase. Now this complex is involved in a number of stages in cell division. And one that is important to us is cytokinesis. Dun used a drug called VX680, which was developed by Merck and is readily available. This drug inhibits Aurora kinases, not only B, but others. It blocks cytokinesis, and it disables the spindle checkpoint, the cellular response to mishaps in chromosomal segregation. It has a reversible effect on normal cells. A crucial feature to developing any synthetic lethal therapeutic. Dun first tested human cells. These RPE cells are epithelial cells, normal human epithelial cells. Either as controls or as cells that he manipulated to overexpress MYC. And he discovered that the normal cells did not die, and if he withdrew the drug here at the white arrow, the normal cells recouped and grew and they had doubled in time within a few days. The cells that were overexpressing MYC died. The remarkable finding was that you could withdraw the drug after say three days, as was done here, and the cells overexpressing MYC continued to die to extinction. So we had what we called an early phase death and a later phase death. The death by memory as opposed to the death by direct exposure, immediate exposure to the drug. Now the early phase death is due to apoptosis. Here we, Dun, monitored apoptosis, and this is rise to about a third, a little less than a third of the population of cells. And then the apoptosis disappears. That leaves the delayed death that I defined in the previous diagram unexplained. When Dun examined these cells closely he discovered that by electron microscopy, they were rife with autophagy or “autophagy”, take your choice. He validated this with all the standard tests for autophagy. I won’t go into them in detail, but cogniscenti if any are watching or listening would recognize these. They’re all satisfied. These criteria are all satisfied by the cells in question. So we came to the conclusion that more than likely the delayed death was due to the autophagy. Dun authenticated that by inhibiting ATG genes. The machinery for autophagy is encoded by a battery of genes known as ATG, and Dun used three different tricks. I’ll just refer to the two most persuasive. He used interfering RNA to block the activity of either the fifth or the seventh ATG gene or cells that were genetically deficient in the atg5 gene. Blocking any of these genes in this manner spared the cells from the delayed death. So we would argue that the autophagy is integrally involved in the killing. So, incidentally we then here have a bimodal form of killing reminiscent of the bimodal form of therapeutic that I talked about before. Dun did a preclinical test in three different models. First, a B cell lymphoma driven by a MYC transgene. And the results represent about a threefold extension in lifespan. He tested a T-cell lymphoma driven by a transgenic MYC. Again, a substantial increase in lifespan. Finally he tested that very aggressive liver tumor that Andrei Goga had also used and got a significant extension of lifespan with a brief intermittent therapy. He also has looked at a battery of human cancer cell lines, 75 all totaled, representing quite a variety of human tumors. And most of them respond as they should; if MYC is overexpressed, they die, and if it is not, they do not die. There are exceptions, but let me show you two sets of data of successes. Brain tumors and lung cancer cells. And the take home message from these data is that in either instance, if the tumor cells were expressing MYC, they died. That’s the black bars. And if they were not expressing MYC, they survived the treatment with the inhibitor, with VX680. Synthetic lethality combining VX680 with overexpression of MYC. So we have here a synthetic lethal interaction that has potential therapeutic value. We have a biomarker once again for sensitivity to this class of drug. VX680 is no longer in use, and there are second and third generation drugs that are superior for a variety of reasons. We simply use VX680 for a matter of convenience. This bimodal killing may impede the development of drug resistance, just as the bimodal attack by therapeutics impedes it in the case of acute promyelocytic leukemia. And the fact that normal cells can recover raises the idea that pulse therapy may be a particularly innocuous approach to using this form of synthetic lethality and treating cancers. And in fact Dun used pulse therapy in those pre-clinical trials in mouse models that I showed you. All told now between our work and work in the literature, there are five known perturbations of the cell that give a synthetic lethal interaction with overexpression of MYC. They are listed here. I won’t go into any more detail. The three from our lab include the inhibition of CDK1 and the inhibition of cytokinesis that I told you about, and work that we have not yet published on glutamine deprivation. Other labs have reported that if you stimulate particular receptors on the surface of the cell, you get a synthetic lethal interaction although this is very toxic. And there has been a report of a synthetic lethal interaction between the inhibition of CDK2 and overexpression of MYC’s cousin, MYC-N, that you heard about when I was talking about prognostic markers with cancer genes. Now this works with other genes. And the first example to come to light involved the mutant RAS gene. This is from Mariano Barbacid and his colleagues in Spain. Now this is important because RAS is another, to this day, undruggable but very widespread cancer gene. So just as MYC had its disadvantages as a therapeutic direct target, so does RAS. But Barbacid and his colleagues have found that indeed you can get a synthetic lethal interaction between mutant K-RAS, one of the forms of the RAS family, and inhibition of another kinase, CDK4. Well, this is another gain of function lesion, another therapeutic combining in a synthetic lethal interaction in non-small cell lung cancer, again a very refractory cancer for which any new therapeutic would be welcome. Now you can broaden the reach of synthetic lethal therapeutics to just screening the entire genome, and this has been done by a number of laboratories. And what’s involved here is that you simply take a target gene, like RAS, so you take cells expressing a mutant version of RAS, and then you use a genome wide screen with interfering RNA. You systematically knockdown expression of every gene, and this sort of screen is now quite common. And you identify those genes which when inhibited have a synthetic lethal interaction with mutant RAS. These would be potential therapeutic targets for flanking attack on RAS. The first report of this identified something over 70, over 70, potential targets that might have a druggable synthetic lethal interaction with RAS. Another great virtue of, or at least potential virtue, of synthetic lethality is attacking loss of function because we have no other recourse at the moment. And I am going to illustrate that with the breast cancer genes, BRCA1 and 2. These genes were discovered by studying families in which breast cancer was inherited in a very strong manner. And this is a pedigree of a family and all of the circles, the red circles, are women with breast cancer. The square is a male carrier. So you can see the strong inheritance of this tumor, of this cancer, due to deficiency in a tumor suppressor gene, In this instance it is BRCA1. Now what do we know about these genes? Well, they are essential for DNA repair. As a result a deficiency in these genes leads to an increase in spontaneous DNA damage, unrepaired damage of the sort that is occurring in all of our cells throughout our lives. This inherited deficiency and the subsequent increase in DNA damage creates a high risk of breast, ovarian, and prostrate cancer. In July of 2009 this report in the New England Journal of Medicine electrified the oncology community. An inhibitor of an enzyme known as poly(ADP-Ribose) polymerase showed remarkable efficacy against tumors carrying BRCA mutations. We know how this works. It is a synthetic lethal interaction. In normal cells, the BRCA genes are part of one form of DNA repair, so called homologous recombination. And the PARP enzyme is involved in another form of DNA repair known as base excision repair. In the tumors that are deficient in BRCA1, the tumor cell still survives. But if you then bring in a drug that inhibits the other, another major form of repair, that is just too much for the cell to bear. That combination leads to death of a cancer cell. So, these are our principles of targeted therapy as we understand them now. Therapeutic inhibitors for gain of function. Synthetic lethality for either gain or loss of function. And combination therapy almost always essential. What about combination therapy? Well, we can imagine three different forms, at least. First of all you might attack two different signaling pathways. An example that I’ve given here is the MEK kinase and the PI3 kinase that will be familiar to those of you who know something about signaling. These are parallel pathways, and so you are ganging up on the tumor cell by inhibiting two of its supporting elements. But you could also attack sequential targets in the same pathway. For example in the RAS signaling pathway downstream of it is a RAF kinase and then the MEK kinase. And by attacking two in the same pathway, you might also reduce the likelihood of resistance emerging because for the same statistical arguments that have been used for the bimodal therapy before. And then of course there is the bimodal attack, which is a reality with PML-RAR for acute promyelocytic leukemia, which is under consideration for the Philadelphia chromosome, BCR-ABL, as a way to avoid the inevitable resistance to Gleevec that emerges eventually. And in those various forms of synthetic lethality for MYC, you could imagine combination therapy by utilizing several of those together. So here is a sampler, looking at it from the standpoint now of the therapeutic. Combination of all-trans retinoic acid and arsenic trioxide is curative for acute promyelocytic leukemia. Herceptin for breast cancer extends lifespan, sometimes extraordinarily so. Gleevec offers a remarkable extension of lifespan for chronic myeloid leukemia and some rare tumors such as gastrointestinal stromal tumor or GIST. I told you about two drugs that attack a switch on the surface of lung cancer cells that give a remission in select cases that can be identified by genomics. A recently described drug that is effective at least short term against melanoma, which has received immense press coverage because this is a deadly disease in its advanced stages, and this drug gives remissions in the advanced stage of the disease. And the PARP inhibitors, which should be useful at the least against breast, ovarian, and prostate cancer, that have got a deficiency in one or the other of the BRCA genes. You will notice that only the PARP inhibitors so far represent a therapy that’s actually been utilized in the clinic and proven to have effect against loss of function. The rest of these are all therapeutics for gain of function. Now there is much more out there. For example, this is a compilation of the targets and the drugs presently… so these are the targets in lung cancer, and the drugs that have been developed to attack them that are currently at one stage of study or another. And beyond this is much more. At last count there were well over two hundred clinical trials of various targeted therapeutics underway in the United States alone. So we have much to look forward to in all likelihood. So, we have made Paul Ehrlich’s vision a reality. Whether it will work as well as he hoped and we hope remains to be seen, but there is certainly reason for hope, and both the public and even the press have got that message. Thank you very much for listening.

1 Comment

  • Reply vobgla August 19, 2014 at 1:54 pm

    GreatĀ talks but I'm not convinced on the explanation for synthetic lethality for PARP inhibitors against a background of BRCA1/2 mutations.

  • Leave a Reply