Oncogenes: A Genetic Paradigm for Cancer
Transcript of Part 2: The Cancer Genome: Challenge and Promise
00:00:01.27 Hello. I am Mike Bishop from the University of California, San Francisco, 00:00:06.18 and I want to continue with the second chapter in my story about cancer. 00:00:12.21 At the end of the first chapter we had reached the sound conclusion that all cancer arises from the malfunction of genes. 00:00:20.19 And we had identified two culprits, genes we know as proto-oncogenes, which suffer gain of function in cancer cells, 00:00:29.13 that essentially become jammed accelerators, 00:00:33.03 and tumor suppressor genes, which suffer loss of function, as if they were a failed brake. 00:00:39.18 And these combine to give rise to the malignant phenotype. 00:00:43.04 Now in order to apply this insight, we need to have a complete inventory of cancer genes. 00:00:50.19 A complete catalog for each form of human cancer, 00:00:54.19 and then with that catalog we need to distinguish what we call the drivers, genetic damage that is contributing to the malignancy, 00:01:04.25 from passengers, collateral damage if you will. 00:01:07.21 There are rules established for doing this, but it still seems a daunting task. 00:01:15.21 But it is less daunting than it used to be because of dramatic advances in DNA sequencing as illustrated by this plot. 00:01:27.01 The cost has come down many orders and orders of magnitude. 00:01:30.13 It is now possible to sequence all of the protein coding regions of the human genome (we call that the complete exome sequence) 00:01:37.28 for about 1500 dollars. That is the cost of an MRI exam in a hospital. 00:01:45.23 And there has been a similar dramatic increase in the speed of sequencing to the point where first-rate sequencing facilities 00:01:54.04 can now sequence the complete human genome in a week or so. 00:01:56.29 In order to achieve the inventory with this dramatically improved technology, 00:02:03.27 an International Cancer Genome Consortium has been formed, and its objectives are ambitious. 00:02:10.11 Five hundred genomes, full genome sequences, for each of 50 tumor types. 25,000 full genome cancer sequences. 00:02:21.27 What is the state of play? Well, as of the time I am recording this, there have been several hundred complete genomes sequences 00:02:31.15 for human cancer recorded, and the number is increasing by the day very rapidly. 00:02:37.04 And there have been at least 2000 complete exome sequences reported. 00:02:43.04 So it is still early days, but we can draw some provisional conclusions. 00:02:48.26 First of all, there are clearly numerous mutations in most human cancers. 00:02:52.24 The numbers can be as high as 100,000 or more, although some cancer have many fewer. 00:02:58.19 But among all of those mutations, there are a few that create anywhere from 5 to 20 drivers per tumor. 00:03:05.23 There are many more passengers, and all told, there are at least 400 different drivers that have been identified to date. 00:03:13.22 400 different genes which either suffer gain or loss of function in the production of tumors. 00:03:22.14 And experimental work with mice suggests that the repertoire of drivers could exceed 2000. 00:03:28.24 10% or so of the entire human genome. 00:03:33.09 Two important principles are emerging from this provisional data. 00:03:38.22 First of all, there are distinctive but not entirely unique genetic fingerprints for each type of cancer. 00:03:46.17 There is overlap from one type of cancer to another in the lesions, 00:03:51.18 but there are also distinguishing features. 00:03:55.06 And secondly, these numerous drivers, 400 or more, represent a far more limited number of cellular circuits. 00:04:05.13 This is a fact that will become important when we consider the feasibility of using this knowledge in therapeutics. 00:04:14.02 Let me illustrate this principle of circuitry with the example of pancreatic cancer. 00:04:22.04 Genome sequencing has revealed that there are twelve functions or circuits affected in cancer of the pancreas. 00:04:35.24 But when sequences from two different cell lines derived from two different pancreatic cancers 00:04:42.19 were compared, it was discovered that these twelve circuits were affected by mutations in different genes between the two cell lines. 00:04:56.23 There was only one example, the Ras gene, which was mutated in the two cell lines, 00:05:04.14 representing the K-Ras signaling pathway that is inevitably affected in pancreatic cancers. 00:05:10.18 In other words, if you are thinking about therapy, you will not necessarily think about 00:05:16.10 the mutations in individual cancers such as these two here, 00:05:20.09 but rather the pathways that the mutations share. 00:05:24.28 Now as this information accumulates we are beginning to see how it will be useful 00:05:34.14 in almost every aspect of the study of cancer and the clinical management of the disease. 00:05:41.03 It is going to help us identify causes and urgent need, genetic risk. It will help us improve early detection. 00:05:53.01 It is already transforming the classification of cancer, revealing new sub-types in lymphomas and breast cancers and others. 00:06:00.17 It is going to help in the prognosis, the prediction of outcome. 00:06:04.08 It is already inspiring new therapeutics, as I will explore in detail in my third chapter. 00:06:14.04 It is going to allow us at least to attempt a personalization therapy, the individualization of therapy. 00:06:20.15 With the rapid evaluation of individual responses, we may be able to evaluate 00:06:26.11 whether the tumor is responding in a matter of days rather than months. 00:06:30.17 And it is going to simplify and economize clinical trials. 00:06:37.14 I am going to examine a number of these by way of illustration. 00:06:43.09 First of all, what causes cancer? 00:06:46.09 This is a crucial question because we need to know the answer before we can devise prevention. 00:06:54.18 And it represents one of the most, well certainly what is to my eye, one of the most difficult forms of cancer research. 00:07:02.03 Now for some few major cancers, we know with some certainty at least one cause. 00:07:12.21 Cancer of the cervix is caused by infection with a virus, human papilloma virus. 00:07:16.21 Cancer of the liver is caused by infection with either of two viruses, hepatitis B or hepatitis C virus, 00:07:25.11 and/or by toxins such as those found in contaminated foods. 00:07:31.06 Cancer of the lung, much of it is caused by cigarette smoke. 00:07:36.21 Cancer of the skin you heard me say is caused by sunlight, the ultraviolet light in the sunlight. 00:07:44.10 And most cancer of the stomach is caused by a bacterial infection, 00:07:48.19 but many of our other major cancers remain without an established major cause. 00:07:54.26 And they are big killers, breast, prostate, colon, ovary, pancreas, brain. 00:08:00.27 We have nothing but hunches about what might cause these cancers. 00:08:07.06 Now the tools for discovery here are of two sorts. 00:08:12.09 One is guilt by association, also known as epidemiology. 00:08:17.07 And the other is represented by our newfound genomic tools, and they allow us to do two things. 00:08:25.23 They allow us to detect the presence of previously unrecognized microbes, either viral or bacterial. 00:08:33.18 And they can also reveal lesions in DNA that hint at the nature of the cause of the lesion. 00:08:40.23 I'll illustrate each of these briefly. 00:08:42.21 Perhaps the best example of and particularly successful example of guilt by association involves cancer of the liver. 00:08:50.21 This map has been known for many years. It displays the incidence of chronic hepatitis B virus infection around the world. 00:09:03.07 Quite some years ago medical scientists looked at this map and realized that it could be 00:09:09.13 superimposed on the distribution of liver cancer. 00:09:13.12 This led to more sophisticated epidemiological studies that clearly established that infection with hepatitis B virus, 00:09:23.02 chronic infection in particular, was at least one of the causes of liver cancer. 00:09:28.13 We now have an even more profound proof of that because we have a vaccine against hepatitis B virus, 00:09:37.16 which is being widely used and is clearly having an impact on the incidence of cancer 00:09:42.08 in those areas where the disease has been particularly common. 00:09:45.25 Now what do these genomic tools do for us? 00:09:50.23 Well, first of all, they can help us detect previously unrecognized viruses. 00:09:55.26 There are three examples and they represent a progression from an unsophisticated molecular technique to full genome sequencing. 00:10:07.12 The first of these was a discovery that all cervical cancers contain a previously unrecognized strain of human papilloma virus. 00:10:16.28 This virus was detected with very simple techniques of molecular hybridization. 00:10:23.02 Then the Kaposi Sarcoma Virus was identified by a somewhat more sophisticated technique of molecular hybridization. 00:10:35.06 And the most recent and very rare example is so-called Merkel Cell Carcinoma Virus. 00:10:41.00 This was literally detected by full genome sequencing. 00:10:47.09 Then there is the idea that we can deduce, or at least get a hint about the cause of a cancer 00:10:53.26 by the nature of the chemical damage in DNA. 00:10:57.01 And the premier example is skin cancer, 00:10:59.25 which we were already reasonably certain is due mainly to ultraviolet light in the sunlight. 00:11:06.20 But in the skin tumors, a tumor suppressor gene known as TP53 00:11:13.12 characteristically has a particular change in which just two nucleotides, 00:11:21.03 a pair of C's is converted to a pair of T's. 00:11:25.12 This is a hallmark of damage from ultraviolet light. 00:11:30.00 This is a stunning affirmation of our belief that sunlight has a crucial role in the genesis of skin cancer. 00:11:39.15 There are other examples. The genetic damage of lung cancer reflects the nature of the chemicals in cigarette smoke. 00:11:49.11 The genetic damage in liver cancer can reflect the nature of the toxins that might have been involved in the genesis of the cancer. 00:11:58.04 And finally in the tragic cases where secondary cancers arise as a result of vigorous chemotherapy, 00:12:05.06 the DNA damage there is diagnostic of the chemicals that were used in the therapy. 00:12:11.06 The principal point of learning the cause of cancer of course is to devise preventions, and alas we have only a few of those. 00:12:18.25 We can prevent most lung cancer by avoiding tobacco smoke. 00:12:25.01 We can certainly reduce the incidence of skin cancer by avoiding excessive exposure to sunlight. 00:12:30.06 We have a vaccine for hepatitis B virus, which is reducing the incidence of liver cancer, 00:12:36.14 and we now have a vaccine for human papilloma virus which when widely applied will certainly reduce the incidence of cervical cancer 00:12:46.20 particularly in those developing nations where it is particularly common and a common cause of death among women. 00:12:56.03 When you think of the genome and cancer, it is inevitable that you wonder about the possibility 00:13:06.01 of predicting individual risk of cancer. 00:13:08.20 And there are three basic origins of risk. There is the environment, sunlight for example. 00:13:16.07 There is behavior, cigarette smoking for example, and there is inheritance. 00:13:23.01 And with the advent of genome sequencing, with the advent of modern genomics, genes have come front and center. 00:13:32.11 There are two kinds of genetic risk for cancer. There is a strong risk due to single gene changes 00:13:41.06 that is responsible for perhaps five to ten percent of all cancer, 00:13:46.08 and the example of retinoblastoma I gave you is one, and we will talk about a few others momentarily. 00:13:53.00 Then there are multiple genes, each of which may be making a weak contribution to risk, 00:14:00.29 possibly for all cancers, but these are being identified at a considerable pace. There are many of them. 00:14:11.02 The question remains as to whether this will ever be useful information 00:14:15.18 because the risk contributed to any single change is relatively small. 00:14:21.14 Suffice it to say, it is a work in progress. 00:14:23.27 But we do have a few strong cancer genes that are responsible for inherited cancer and for which we will do genetic testing 00:14:35.18 if it is indicated, and it includes the retinoblastoma gene, the two BRCA breast cancer and ovarian cancer genes, 00:14:44.10 and a gene called APC, a deficiency of which is responsible for polyps and cancer of the colon. 00:14:51.01 Let me illustrate the APC problem. This is a normal colon, and this is a colon taken from a patient 00:15:03.14 who has inherited a deficiency in the APC tumor suppressor gene. 00:15:08.05 It is a sheet of polyps, and some of these polyps will inevitably progress to cancer. 00:15:16.28 It is clear that testing families that have this problem for the presence of the APC deficiency 00:15:25.26 is a valuable approach that can help in the prevention of disease in individual members of the family. 00:15:34.12 Unfortunately that prevention usually involves a complete resection of the colon. 00:15:39.23 Early detection is known to be a valuable in improving the outcome of cancer therapy. 00:15:54.00 And genomics is allowing us to improve early detection. Now the established techniques are only four in nature, in number. 00:16:04.19 The renowned Pap test which was extraordinarily effective in reducing the incidence of cervical cancer. 00:16:12.20 It may someday be replaced by the testing for the presence of the causative agent-the DNA of the human papilloma virus. 00:16:20.11 Then there is colonoscopy, which is an effective means of early detection, but also obviously a burdensome technique. 00:16:28.22 Mammography for breast cancer about which there is some controversy is presently swirling. 00:16:34.23 And the PSA test for prostate cancer, which is deeply mired in controversy and under thorough reconsideration. 00:16:43.11 This is hardly where we would like to be with early detection, and it is possible, for example, we don't have 00:16:53.03 validated tests for any of these major killers: lung cancer, ovary, pancreas, or liver. 00:16:59.14 There is no test available to detect these tumors early in their genesis. 00:17:06.24 We may get a help from what can be called molecular cytology. 00:17:11.03 The human body sheds cells into all of its orifices, into the colon, into breast fluid, into the secretions from the cervix and uterus, 00:17:24.05 from the bladder or the kidney, both coming out into the urine, and in the sputum of the lungs. 00:17:31.25 And these cells can be analyzed for the presence of telltale genetic lesions. 00:17:40.08 Lesions that would hint at a future cancer or the presence of an existing cancer. 00:17:45.19 But I want to illustrate this with a story of Hubert Humphrey, an American statesman who died in 1978 00:17:52.17 from bladder cancer. He first knew he had a problem in 1967 when he had a problem with his bladder 00:18:02.16 that his physicians deemed "not malignant". In 1973 they decided, yes, he might have a mild form of bladder cancer, 00:18:10.03 and they used local therapy. In 1976 they realized that they were dealing with a highly malignant disease. 00:18:16.28 Radical surgery was performed. It was too late. Mr. Humphrey died in 1978 from bladder cancer. 00:18:24.14 Some years ago scientists went back and looked at the specimens from Hubert Humphrey 00:18:32.24 that had been preserved and discovered that from the very outset 00:18:37.00 the tumor suppressor gene, TP53, was mutant. 00:18:41.10 In other words, molecular cytology done at this point in Hubert Humphrey's course 00:18:46.18 would have alerted the physicians that they were dealing with a truly dangerous circumstance, 00:18:52.25 and they could have taken aggressive action at that point, eleven years before Hubert Humphrey's ultimate death. 00:19:02.05 This is a dramatic illustration of what molecular cytology might offer us, 00:19:05.11 and there are vigorous efforts underway to make it a reality. 00:19:08.14 How about prognosis? Both patient and physician want urgently to know what they can expect 00:19:16.14 from the disease itself and from the therapeutic that they might receive. 00:19:22.05 The first example of how gene analysis might help with prognosis remains one of the most powerful. 00:19:32.25 It emerged in the early 80s, not too long after the discovery or at least the solidification of the reality of genetic damage in cancer. 00:19:43.18 It involves a gene known as MYCN, which was originally discovered by virtue of the fact that it is amplified. 00:19:52.05 It is overgrown, sometimes a hundred or a thousand fold in neuroblastomas, a tumor of children. 00:20:00.02 A large national study was conducted. The first purpose was simply to ask whether guilt by association applied here, 00:20:11.10 whether that amplification was common enough in neuroblastoma to be part of the mechanism of tumorigenesis. 00:20:18.14 But a very useful outcome occurred when it was realized that the presence or absence 00:20:26.05 of the amplification of this gene was a profound prognostic indicator. 00:20:33.20 And that is dramatized in this plot. Children whose tumors do not contain an amplification of MYCN 00:20:41.17 have a superb outcome after therapy. 00:20:45.06 Children whose tumors have MYCN amplified are going to be refractory to conventional therapies. 00:20:52.24 This test is now used in all major centers where neuroblastoma is handled. 00:20:59.09 And it remains the most powerful predictor based on genes for the moment. 00:21:05.00 But we have more sophisticated techniques and they are going to change the game. 00:21:13.14 This will be recognized by many of you as what we call a gene array analysis. 00:21:21.15 Simply put, it is now possible to test for the expression of every gene in the human genome. 00:21:32.17 And in this array the red squares indicate a gene that is active, 00:21:37.05 and the green squares indicate a gene that is at normal level or even not active. 00:21:44.02 And this can be used to survey or to compare gene expression in cancer cells with gene expression in normal cells. 00:21:53.03 And let me give you one example, which is actually commercially available, and it is known as "Mammaprint". 00:21:59.15 This is a set of seventy genes which comprises a fingerprint, and if that fingerprint is present 00:22:12.14 the likelihood of a poor outcome is increased. 00:22:20.05 Now this is not perfect. The prognostic accuracy is only about 50%. 00:22:28.04 And the "Mammaprint" is found in about 61% of breast cancers. 00:22:32.29 There is another signature known as the MSP Complex. That signature alone has about a 60% prognostic accuracy, 00:22:42.09 but it only occurs in 14% of human breast cancer. 00:22:48.02 If you put the two together, you get a remarkable improvement in accuracy, 82%, 00:22:54.05 but now the number of patients, the number of tumors, that is the fraction of tumors containing the combined signature is down to 9%. 00:23:03.09 So that is a discouraging limitation on the most effective prediction, 00:23:08.10 but it is also a dramatic reflection of how much more complex breast cancer may be than we had previously realized. 00:23:17.23 In other words that 9% represents a distinct biological and genomic subset of the disease that has not previously been recognized. 00:23:26.10 Which brings us to therapy which is always uppermost or usually uppermost in the public mind. 00:23:36.14 In using genetic lesions in cancer to guide our development of therapies, 00:23:45.20 we are designing an intervention in the elaborate circuitry that controls the lives of our cells, 00:23:52.00 and this is a simplification of that circuitry. The real thing is hundreds, thousands, perhaps tens of thousands of fold more complex. 00:24:00.19 And represented by the red dots are nodes in this circuitry where either a proto-oncogene or a tumor suppressor gene resides. 00:24:09.18 And we would like to target our therapeutics to those nodes that are malfunctioning in the cancer cell. 00:24:20.17 The two obvious ways to do this are either to inhibit a gain of function of a proto-oncogene, 00:24:31.04 or to replace the loss of function of a tumor suppressor gene. 00:24:34.07 Inhibiting gain of function is something we know how to do, and you'll hear a lot more about that from me in my third chapter. 00:24:40.26 It is, as I like to say, a growth industry. Replacement of function is not presently practicable. 00:24:48.12 We simply have no means by way to do that at the moment or in my view for the foreseeable future. 00:24:55.24 And then there is a third, newly emerging technique called, I call it, attacking from the flank, 00:25:01.20 in which neither the cancer gene itself nor its protein product is the direct target for therapy. 00:25:07.24 But the cancer gene or its protein product is never the less being exploited in the therapy, 00:25:13.09 and this will be a major subject of my third chapter. 00:25:16.25 What do we target then? Well, we target proteins, not genes. 00:25:26.15 We target proteins with both pharmaceuticals, small molecules, and biologicals, large molecules like antibodies. 00:25:33.27 We do this now in many instances and we will be doing it more in the future. 00:25:39.13 We don't target genes because the molecular surgery on tumor DNA, or for that matter, normal DNA, is not yet practicable. 00:25:48.08 Just we have no way of doing that. 00:25:50.27 Now the poster child for targeted therapy of cancer is a drug with the trade name Gleevec or the formal name imatinib. 00:26:03.22 And it has been developed and is extraordinarily effective against patients with 00:26:09.14 chronic myeloid leukemia containing the Philadelphia chromosome. 00:26:13.06 This is the translocation I told you about in my first lecture. 00:26:19.18 Now what this translocation does is create a sort of mongrel protein with portions from two different proteins being used. 00:26:27.24 The action of this protein is enzymatic in nature. 00:26:33.25 In the fusion, the mongrel version of the protein, this enzymatic activity is incessantly and excessively on. 00:26:41.02 It cannot be controlled, and that is one of drivers, perhaps the principal driver in the production of this tumor. 00:26:51.10 Some years ago a group of scientists, both academic and commercial, 00:26:56.18 teamed up to develop this chemical known as Gleevec. 00:27:01.03 And they developed it by screening for chemicals that could kill or arrest the growth of leukemia cells in the laboratory. 00:27:13.09 The efficacy in patients proved to be dramatic, and this is a story that is told in great detail by Brian Druker in his iBioSeminar, 00:27:22.23 and I refer you to that if you want to learn more about it. 00:27:25.27 So here are the early returns for what we call targeted therapy. 00:27:33.29 That is to say, therapy that actually has a specific gene product as its target. 00:27:40.08 The disease of acute promyeloblastic leukemia which was once incurable can now be cured. 00:27:46.03 And I will tell you about that in my third chapter. 00:27:48.29 Breast cancer, survival can be extended by several means. 00:27:54.08 The best known being Herceptin, a targeted therapy. 00:27:58.03 Chronic myeloid leukemia. I just told you about Gleevec. 00:28:00.27 This prolongs survival to the point that we are now causing this a chronic disease. 00:28:07.06 The term chronic myeloid leukemia is a bit of misnomer because this disease is ultimately lethal as well. 00:28:12.24 Now it can... if patients are treated early on, they have an outstanding prognosis. 00:28:21.20 And again, Brian Druker's iBioSeminar discusses that in detail. 00:28:26.17 And we have drugs that give us a brief remission in lung cancer 00:28:31.21 and a brief remission in melanoma. Both of these are targeted therapies and they simply point the way to a much more promising future. 00:28:40.20 I will discuss all of this in more detail in my third chapter. 00:28:44.22 Recently cancer scientists have realized that they have a different kind of target they may have to worry about. 00:28:52.19 A distinctive cellular target. 00:28:54.27 This target is known either as a cancer stem cell or a tumor initiating cell. 00:29:01.06 It appears that within the massive population of tumor cells, there is a small subset which is responsible for maintaining the larger mass. 00:29:16.08 This subset functions rather like a normal stem cell in that it is constantly regenerating itself, 00:29:24.16 but also spinning off what you could call a differentiated tumor cell. 00:29:30.27 Now these are the cells that we normally treat. 00:29:34.25 These are the cells whose response we normally measure in our therapy. 00:29:38.25 These are new actors in the game. There is controversy about how universal they are. 00:29:46.12 They have certainly been shown to be... their existence has certainly been well affirmed 00:29:52.22 in certain leukemias and there is strong evidence for them in some solid tumors as well. 00:29:58.06 Suffice it to say, this too is a work in progress, but the question arises: 00:30:04.02 "Do these cells differ from the mass of tumor cells in their therapeutic susceptibilities?" 00:30:11.10 One reason they might is that they are known to proliferate very slowly. 00:30:16.09 And the classical chemotherapeutics exploit the rapid proliferation of cancer cells. 00:30:24.01 So this would make these cells relatively resistant to the classical chemotherapeutics. 00:30:29.19 It is possible that they have a distinctive circuitry of the sort I talked about before, 00:30:36.05 which is in no way targeted by a drug that we develop for the circuitry 00:30:41.02 in the mature tumor cells that have been spun off from the stem cell. 00:30:45.29 Thirdly, these tumors.... these stem cells or tumor initiating cells often carry intrinsic drug resistance. 00:30:54.23 And the reason for that is speculative but makes sense. 00:30:59.16 And that is that if these cells were actually derived originally from normal stem cells, 00:31:07.22 normal stem cells are vital to the maintenance of our adult tissues, 00:31:12.19 and it would be only reasonable to think that they have over the eons evolved intrinsic resistance to environmental toxins of any sort, 00:31:25.06 naturally occurring or more recently of human origin. 00:31:29.19 So let's talk about resistance to cancer therapy for a moment. 00:31:34.24 It takes three forms: pumps that extrude chemicals from the cell, 00:31:43.09 mutations in the targets that we are treating, and peculiarities of the signaling circuitry. 00:31:53.22 Now we have known about these pumps for quite some time. 00:31:57.10 They are transporters that use energy to efflux various chemicals and a huge array of chemicals from the cell. 00:32:07.17 And unfortunately, they often efflux the agents that we use for treating cancer, 00:32:15.09 and this is the classical underpinning of much of the drug resistance against conventional therapeutics. 00:32:23.25 Then there's the problem of mutation in the target gene. 00:32:28.25 Now this I have illustrated here with BCR-ABL, the mongrel protein of chronic myeloid leukemia, that is normally targeted by Gleevec. 00:32:38.12 And this gene develops mutations under the pressure of Gleevec treatment that render the protein resistant to Gleevec 00:32:51.19 and require the development of additional drugs which has been done, successfully, and promises more for the future. 00:33:00.06 But arrayed along this cartoon are various points where mutations have been found that make the protein resistant. 00:33:08.22 This is the drug binding to the protein, so some of the mutations simply impede the binding directly. 00:33:16.03 Other of the mutations change the conformation of the protein in a way that makes it resistant to the therapy. 00:33:21.29 This is a common occurrence for various forms of therapy for various cancers. 00:33:28.09 And it is one that we will have to cope with, even with the new elegance of targeted therapy. 00:33:36.11 Then there's the peculiarities of the circuitry. 00:33:42.01 Now imagine a tumor that has a switch here on the surface that is a hyperactive proto-oncogene. 00:33:54.27 Gain of function. And you want to target that. 00:34:00.08 But if there is also a mutation in one of the downstream signaling elements such as the Ras gene, 00:34:04.28 targeting this will be of no avail because the Ras gene will still be running full tilt, and driving the cancer. 00:34:12.25 This has been found in human cancer as a form of resistance to drug therapy, and I'll say more about that momentarily. 00:34:20.05 And then here is another example of how the circuitry can undermine our therapeutics. 00:34:26.23 This involves the target for Herceptin, the breast cancer gene. 00:34:30.00 Herceptin is an antibody that binds to this surface switch, this gain of function, and shuts it down. 00:34:41.23 But, there are other surface receptors that play on the same downstream signaling. 00:34:47.23 And if any of them are also hyperactive in the cancer cell, shutting down HER-2 alone does not suffice 00:34:55.03 to shut down the signaling that is driving the tumor and gives rise to Herceptin resistance. 00:35:01.01 Trastuzumab-this is the fancy term, the chemical term, the formal term for Herceptin. 00:35:07.04 Clearly if we have the complete genome sequence and the full knowledge of the circuitry in a cancer cell, 00:35:16.10 we can predict either this form of resistance or this form of resistance. 00:35:21.16 So genomics is going to be an assist to us for dealing with this problem. 00:35:26.01 Clinical trials are notorious for a number of difficulties. 00:35:37.01 They need to be very large. Enrolling enough patients is a problem. 00:35:42.06 It takes a long time to get the results. Genomics may offer a solution to that. 00:35:48.04 First of all, as we learn what the drivers are, 00:35:52.13 and as we develop therapeutics for those drivers, we'll be able to define the trial population. 00:35:59.18 We will be able to restrict the population to people with the target. 00:36:03.12 This will make it possible to greatly reduce the size of the cohort, 00:36:10.21 which will reduce the cost, and it will also probably allow us to develop what are known as biomarkers. 00:36:18.00 Molecular or chemical changes that would permit us to evaluate 00:36:22.03 within a matter of days whether there is a response to the therapeutic. 00:36:26.29 And if we can get that kind of feedback, we can then make these trials adaptive. 00:36:33.06 We can change in midstream and use the same cohort to explore a revision of the therapeutic. 00:36:40.04 How many cures do we need for cancer? 00:36:45.03 Well, there is public hope for a panacea, a one hit, a stop cure all. 00:36:51.11 It is not going to be in all likelihood. 00:36:53.29 I have told you that there are varied genetic fingerprints from one tumor to another. 00:37:00.06 Hence there will be no single therapeutic regimen that will deal with all cancers. 00:37:04.11 There will be no single cure for cancer. We will have to for maximum efficacy we will have to personalize the therapy. 00:37:13.05 How do we do this? Well first of all, you have to profile the genome and gene expression in the tumor in question. 00:37:22.10 Identify the potential therapeutic targets. 00:37:27.06 And we may also have to deal with the problem that there are distinctive targets in cancer stem cells. 00:37:33.12 This is essentially an unknown for the moment, but under intense scrutiny. 00:37:41.27 From the genomic data, we'll be able to know whether there is a suitable drug metabolism. 00:37:51.14 Some of the drugs we use, such as tamoxifen for breast cancer, 00:37:53.22 have to undergo a chemical conversion in the cell before they are effective. 00:37:58.23 Genomic data will tell us whether that machinery is present. 00:38:04.02 It will also be able to identify nascent drug resistance of the sort I described in the circuitry, for example. 00:38:11.14 And all of this will allow us to tailor the therapy appropriately and in all likelihood, it will always have to be combination therapy. 00:38:23.16 And I will approach that also in my third chapter. 00:38:27.04 But let me dramatize this with a beginning. And that involves the treatment of non-small cell lung cancer, 00:38:34.14 a horrendous ailment and almost inevitably lethal. 00:38:42.14 It was some years ago discovered that there was a switch, a proto-oncogene protein product, 00:38:52.11 on the surface of lung cancer that is hyperactive. A gain of function. The so called EGF receptor. 00:38:58.18 Drugs to attack this receptor were developed. Their trade names are Iressa and Tarceva. 00:39:09.08 And these are chemical inhibitors of that switch. 00:39:11.15 In the first clinical trials of these inhibitors, there was no prolongation of life. 00:39:16.01 They appeared to have failed. However, in occasional patients, there were remarkable responses. 00:39:23.03 This image shows one patient, whose whole entire left lung is just full of cancer. 00:39:32.26 Six weeks after initiation of treatment with Iressa, the tumor had regressed dramatically. 00:39:40.16 It eventually recurred, but this was a very promising response. 00:39:45.00 But it was limited to only a few. Eventually we learned that it is limited to about 10% of all non-small cell carcinomas of the lung. 00:39:54.10 And that we can identify those patients because they have telltale mutations in that switch. 00:40:01.12 Telltale genetic changes in the gene that encodes the EGF receptor switch, 00:40:07.06 and those changes indicate a susceptibility, so this is a way that we can screen 00:40:11.01 all of these patients and find those that are going to respond to this drug. 00:40:15.04 So, in miniature, we are personalizing the treatment of lung cancer, at least with this drug. 00:40:24.13 We have to screen for the susceptible mutations. 00:40:28.27 We have to screen RAS because it is downstream, if you recall, of the EGF receptor, 00:40:36.19 and if it's mutant, there is no point in turning this off because that won't work. 00:40:39.11 In fact, the drug labeling now advises not to use this drug if there is a mutation in RAS. 00:40:49.14 So you then use these data to make your therapeutic decisions. This is in miniature personalized therapy in the making. 00:40:57.27 Now what about this immense genetic diversity that I've talked about before. Many malfunctioning genes. 00:41:08.02 Well some are relatively common and shared among different tumors, 00:41:10.06 such as overexpression of MYC, such as mutation in RAS. And as I emphasized or illustrated with the pancreas cancer, 00:41:21.20 there is a limited number of signaling pathways that are affected by mutations in a larger number of genes. 00:41:27.18 And we can focus on the signaling pathways in our thinking, and thereby reduce the complexity of the problem. 00:41:34.27 In other words, we will be able to expand the utility of individual therapeutics beyond what we might have expected 00:41:41.05 when we first recognized that there are many malfunctions in genes in cancer cells. 00:41:46.12 In 2010, Nature magazine polled 1,500 cancer scientists. Actually, medical scientists in general. 00:42:01.27 And they asked the question, "How soon do you expect personalized medicine 00:42:06.19 based on human genetic information to become commonplace?". 00:42:10.03 And if you accept those.. if you sort of include those who said, "It's already here", 00:42:15.17 with those that said within 5-10 years or 10-20 years, 00:42:19.18 you get a large majority of these medical scientists who think that personalized medicine is going to become common place. 00:42:26.20 There is one set of profound pessimists: those who say, "not in my lifetime". 00:42:31.00 and I strongly suspect that those are people in my age group. 00:42:34.26 On March 7th, 1986 the Nobel Laureate Renato Dulbecco published an essay in Science magazine. 00:42:44.05 It was one of the first formal calls to sequence the human genome, and he used cancer to justify his argument. 00:42:52.06 And in his conclusion, he said in essence that attention to DNA may, and this is a quote, "close the chapter" on cancer. 00:43:02.29 Well, we haven't closed the chapter yet, but I hope that what I have told you so far 00:43:08.26 has convinced you that we are turning the pages very rapidly. 00:43:13.00 Thank you for listening.