FRONTIERS IN CANCER RESEARCH PERSPECTIVE The New Era in Cancer Research Harold Varmus For many years, discoveries about the genetic determinants of cancer appeared to be having only minor effects on efforts to control the disease in the clinic. Following advances made over the past decade, however, a description of cancer in molecular terms seems increasingly likely to improve the ways in which human cancers are detected, classified, monitored, and (especially) treated. Achieving the medical promise of this new era in cancer research will require a deeper understanding of the biology of cancer and imaginative application of new knowledge in the clinic, as well as political, social, and cultural changes. he conquest of cancer continues to pose great challenges to medical science. The T disease is notably coniplex, affecting nearly every tissue lineage in our bcdies and arising fiom nonnal cells as a consequence of diverse mutations affecting inany genes. It is also widespread and lethal; currently the second most common cause of death in the United States, it is likely to become the most common in the near hime. Despite large federal and industrial investments in cancer research and a wealth of discoveries abut the genetic, biochemical, and functional changes in cancer cells, cancer is commonly viewed as, at best, minimally controlled by modem medicine, especially when compared with other major dwases. Indeed, the age-adjusted mortality rate for cancer is about the same in the 21st century as it was 50 years ago, whereas the death rates for cardiac, cerebrovascular, and infectious diseases have declined by about hvo-thirds (I). A Perspective on the History of Cancer Research The recent death of Joseph Burchenal(2), one of the pioneers in the use of chemotherapy, provides a vantage point for thinlung about the hstory and the future of cancer research and its implications for control of the disease. Just over 50 years ago, Burchenal (Fig. 1) and his colleagues used analogs of folic acid, methotrexate, and of a nucleoside, 6- mercaptopurine, to induce profound and sustained remissions in children with aggressive leukemias (3). This event-viewed in combination with re- lated, contemporaneous work by Sidney Farber and by Emil Frei and Emil Frekickwas revolution- ary: For the fm time, drugs of known chemical composition that interfered with enzymes engaged in a specific biological process, DNA replication, were used to treat cancers successfully in a rational manner. The sevtral successfid cases Tired the design of clinical trials, pem~tting the mea- protocols. The resulting progress against childhood leukemias, despite the toxicity of the drugs and the lethality of the diseases built confidence in the surement of gractual improvements in treatment Memorial Sloan-Kettering Cancer Center, New York, NY 10021, USA E-mail varmus@mskcc.org notion that biology and chemistry could be harnessed to benefit patients with cancer (4). At the time that Burchenal began treating leukemias, little was known abut the causes of cancers or abut the genetic and molecular mech- anisms by which they arise from nomal cells. His therapeutic strategy was based largely on the pmnise that cancer cells replicate their DNA and divide more freguently than most nonnal cells and hence would be more sensitive to DNA damage. Although this concept has proven to be an overly simplistic explanation, an emphasis on damage to DNA and the mitotic apparatus has guided the development of the many chemothempeutic regi- mens and radiotherapies that have been used for nearly all typ of cancers over the past 50 years. The results have ranged fmm modest at best (in the advanced stages of some of the most common carcinomas of adults), to partially protective am subsequent metastasis (when used as an adjuvant to surgery in the early stages of such diseases), to highly effective (in the treatment of even advanced stages of testicular cancers, some lymphonm. and a few other tumor types). hing most of those 50 years, pharmaceuti- cal chemistry continued to serve cancer patients much more effectively than did cancer biology. Laboratory-based investigations into the nature of cancer cells and clinical efforts to control cancer often seemed to inhabit separate worlds. In the world of labomtov research, the characterization of cancer viruses of aninmls in the 1960s and 70s, the discovery of the first protooncogenes and tu- mor suppressor genes in the 1970s and 8Os, the integration of the p&cts of those genes into cell sipding pathways m the 199Os, and even the repeated unveilings of mutant genes implicated in human cancm beginning in the early 198Os4 seemed to have little or no impact on the methods used by clinicians to diagnose and treat cancers. The Rise of Molecular Oncology During the past decade, perceptions about this sit- uation have been changmg rapidly. Understanding the genetic and biochemical mechanisms by which cancers arise and behave is now widely believed to paend improvements in the way we detect, classify, monitor, and treat these diseases. This message has been driven home, gradually but effectively, by a variety of new and less toxic agents for treating cancers-hormones, antihdies, and enzynie-inhibitory drugs-and especially, by the hiatic arid of a near-mimulous drug, imatinib (Gleevec), a "molecule-specific" agent that induces nearly complete and sustained mnissions in nearly all patients in the early stages of chronic myeloid leukemia (CML), by blocking a protein-tyrosine kinase activated by a well- studied chromosomal translocation (f). These new therapies are ofien called "targeted." Eht in a sense they are not any more targeted than the conventional chemothempies that inta-fere with components of the DNA i~plicatioi~ DNA E@, or mitotic machineries or than radiotherapies that damage DNA in a focused field. The new breeds of treatments usually have specificity for individ- ual cancers, reflecting the particular mutations responsible for that tumor or variations in gene expression-distinctive molecular attributes that are increasingly used to subdivide cancers as- signed to the same standard histopathological subtype (6, 7). These attributes include the pres- ence or absence of receptors that bind to lionnones or to derivative antagonists; the amplification or efficient expression of genes encoding cell surface proteins that are recognized by antibodies that may inhiiit cancer cells (directly or through damaging toxins or isotopes); or the activation of intracellular signaling pathways by mutant proteins that are sensitive to niolecule-specific drugs. Thempatic successes, however limrteed in some situations, have prompted optimism about other uses of genetic and biochemical information- to classify tumors, to detect them early and Fig. 1. Ioseph H. Burchenal. [Photo: courtesy of Memorial Sloan-Kettering Cancer Center] 1162 26 MAY 2006 VOL 312 SCIENCE www.sciencemag.org SPECIALS ECTlO N monitor their growth, and to devise more in- genious ways to inhibit or reverse their growth, Two broad areas of knowledge about cancer in general-and about individual cancers arising in different cell lineages-have been especially significant in ths transformation of thnking about cancer: 1) The genetic basis of cancer. Mutations are now recognized to be the hbental lesions driving neoplasia (8). The mutations are largely somatic, but sometimes heredimy; they affect proto-oncogenes, producing a dominant gain- of-function, and tumor suppressor genes, result- ing in a loss of function. The Cancer Gene Census maintained by the Sanger Center of the Wellcome Trust (9) now lists over 350 genes, situated on every chromosome (except Y), that have been causally implicated in human cancer because they have been repeatedly encountered in mutant form--amplified, deleted, translocated, or damaged by missense, nonsense, or frameshift mutationsin one or more cancer types. The mu- tations are supplemented by epigenetic variations (methylation of DNA or nidfications of Manes or transcription factors) that affect gene expression (10). The mutations and the secondary changes in gene expression provide new tools for classifLrng tumors, for predicting their behavior, for antic- ipating means to detect them early, for designing new tools for imagng, and for developing the rapeutic strategies. In additioi~ gem1 line muta- tions associated with cancers have been observed in 66 genes (9), n&ing them canddates for as- sessment of genetic risks of certain cancers (11). 2) The physiology of cancer. The biological behavior of cancer cells ha$ increasingly been llnked to underlying niutations through an un- derstanding of the signaling pathways that govern the cell cycle and cell growth, programmed cell death (apoptosis), longevity, motility, metabolisni, and genome integrity. Furthermore. in addition to the physiological cha~acteristics of cancer cells themselves, components of a cancer cell's envi- ronment are now recognized to be important for unmding cancer and considering new means to attack it. The swalled hallmarks of cancer (12) include the acquisition of self-sufficient signals for growth, the capacity for extended proliferation, resistance to pwth-inhibiting signals, the ability to evade cell death signals, the potential for tissue invasion and nietastasis, and the power to induce blood-vessel formation (angiogenesis). Some of these traits are the properties of the cancer cells themselves, but others depend on com- munication between the cancer cells and their cellular and macromolecular environments. Each property constitutes a vulnerability in a tumor, to be exploited by new therapies, especially when the underlying mutations and signaling aberrations are known. Still, despite all this new knowledge and de- spite the startling success of imatinib in the treatment of CML, most of the effects of the new era in cancer research are promised, not acheved. Classification of tumors based on analysis of DNA and RNA is stlll an uncertain art and practiced only in a few academic centers, largely on an experimental basis. Development of reliable new biomarkeix for detection of tumors and of novel, hgh-aflinity ligands for imaging, based on evi- dence of changes in the structure or production of ceaain proteins in specific cancers, has yet to occur. The impact of the new generation of mo- lecularly targeted therapies on overall cancer mor- tality rates remains neghgible, because imatinii is effective only in CML and a few other relatively uncommon cancers; because other tyrosine kinase inhibitors dramatically Shrink only those lung cancers with mutations in the epidermal growth factor receptor (13). and the impact on survival in this group of patients has yet to be established in prospective studies; and because antibcdies a- cell surface proteins that are effective as adjuvant therapies, such as anti-HER2 in early breast cancer (14, Is), have not yet been used long and wide- ly enough to affect public health data. Oncogene Dependence So why is there so much excitement about new cancer thmpies? One reason is based on an unexpected consequence of interfering with acti- vated oncogenes. The remarkable reduction in the number of cancer cells observed after treatment with in~tinib and some other tyrosine kinase inhibitors implies that such drugs do not simply mst tumor cell prohfaation when they block oncogene activity; they eliminate tumor cells, most likely by programmed cell death. The idea that cancer cells are dependent on mutant oncogenes for viability, not just growth+often called ``onco- gene dependence" (I@ or "oncogene addiction" (I vis also supported by studies of cancer cell lines and anjmals. In mice carrying oncogenes as transgenes that can be regulated by transcrip tional control, a wide variety of tumor types swiftly regress, mainly by apoptosis, when the oncogenic proteins are de-induced (16, 18). The concept of oncogene dependence encour- ages efforts to destroy cancer cells with new therapeutics directed specifically against the products of mutant oncogenes, but it is still a poorly und- phenomenon At its heart is a vexing question: How did a cell that was originally content without an oncogene become ready to die if deprived of it? Answers to this question could guide shtegies for exploiting a cancer cell's dependence on some of the most ftequently encountered oncogenes, such as members of the RAS and MYC gene families, for which therapeu- tic agents are c~tly lacking. This will entail learning more about the vulnerabilities of cells dependent on oncogenic proteins that do not function as enzymes (e.g., Myc and other oncogenic transcripton factors) or those that have lost a catalytic activity (e.g., mutant Ras proteins lacking guanosine triphosphatase activity). Several other issues require attention be- fore oncogene dependence can be adequately exploited for diagnostic and therapeutic purposes: 1) The mutational repeitok. Most obviously, the catalog of oncogenic mutations associated with the many forms of human cancer is far from complete. The Cancer Genome Atlas (TCGA) initiative, recently announced by the National Institutes of Health (NIH) (19j, should substantial- ly improve this situation over the next decade. The high-throughput technologies that make this initia- tive possible can, in principle, be used to survey sets of hundreds of tumors, each set representing one of the common malignancies, for determi- nation of gene copy number, gene expression pattein and sequences of the exons of 10oO to 2000 genes (20). Development of new methods for DNA sequencing (21 j could appreciably drive down costs of TCGA, and faster methods for karyotyping could extend the project to detect chromosomal reanangements, which are proving to be very common mechanisms of oncogenic mutation (9). TCGA is intended to assist the development of therapeutic strategies, but the portraits of molecular changes in many cancer types should also offer new ideas about diagnos- ing and classlfylng cancers, detecting them earlier with biornarkers, and monitoring them during therapy with novel imaging methods. 2) Mutational hiemchies. Most if not all tu- mors have multiple mutations affecting hown cancer genes, but the relative importance of such mutant genes in maintaining the oncogenicity and viability of a cancer cell is not known. The loss of responsiveness to anti-HER2 antibody after a tumor suppressor gene (PEN) is mutated in hu- man breast cancers (22), and loss of dependence on the c4Qc oncogene in mouse breast tumors when a mutation occurs in another oncogene (ku) (23), imply that thaapies addressing mul- tiple genetic changes will be required. On the other hand, in some genetically enpeered mice, onco- gene dependence is not affected by the coexistence of an oncogenic mutation in another gene (24). Experiments that explore the himhy of mu- tations in Merent types of tumors could guide the selection of the most appropriate molecular targets and the design of multi-agent therapies. 3) Secondary resistance. AU targeted therapies are limited by the appearance of resistance to drugs or antibodies. In some highly instructive cases, resistance can be amibuted to a limited repertoire of secondary mutations in targets such as onco- genic tyrosine kinases (25, 29, providing a basis for screening for drug resistance and for seeking new agents that can prevent or over- come it. Deciphering mechanisms of resistance and developing multi-agent treatment protocols, resembling the anti-HIV combination therapies that reduce the ldcelihood that drug resistance will emerge, will be essential to achieve Iong- term control of cancers. www.sciencernag.org SCIENCE VOL 312 26 MAY 2006 1163 FRONTIERS IN CANCER RESEARCH 4) Heterogeneity and stem cells. The use of differentiation markers reveals heterogeneity among neoplastic cells in a single tumor (-77, 28). Some if not all tumors arc thought to contain a minor population of cells (so-called cancer stem cells) that are responsible for the tumor's continued expansion and for its regeneration when once-effective therapies fail (29,30). Better characterization of cancer stem cells and the means to isolate them may help to monitor this subset of cells during treatment and to design treatments that selectively kill them, thereby eliminating a tumor's potential for regrowth. Ths list is, of course, incomplete. It re- mains to be established, for example, whether all cancers show oncogene dependence; wheth- er there is a relationship between oncogene dependence and metastatic potential; and whether components of signaling mechanisms "downstream" of mutant oncogenic proteins (31) can commonly serve as targets for ther- apeutic intervention. Attacking Cancer Cells Indirectly An enlarged understanding of the tissue en- vironment in which cancers grow is providing new opportunities to develop therapies that are not targeted at the tumor cells themselves. Best known among these novel approaches is the anti-angiogenic strategy, for which drugs and antibodies have already been approved by the Food and Drug Administration (FDA) (32). the^ are grounds for optinism about other approaches that address the tumor's milieu: (i) by interFering with growth-promoting signals supplied by non-neoplastic "stromal cells" that mund a tumor (33); (ii) by inhibiting specific proteases that mold a tumor's environs to promote the dan- gerous escap of tumor cells into the circulation (34, 35) or by using those proteases to activate molecules useful for imaging tumors (36); and (iii) by promoting an immune response against tunior cells-for example, by inactivating factors. such as the T cell surface protein CTLA4 (33, that restrict the immune response to cancer cells. Placing Selective Therapies in Perspective Despite these encouraging ideas, enthusiasm for harnessing new knowledge to combat cancer clinically can seem naively ovqronising; sim- plistic about the medical aid social amibutes of cancer; unperceptive about the history of in- corporating complex technical changes into the gend practice of medicine; and neglectfd of the many other ways cancer can be conhdled. Surgery, chemotherapy, radiation, histopathology, and con- ventional imaging a~ likely to remain the staples of cancer ca~ for many years. And they too are be- coming more effective, even without any molecular advances, through image-guided and minimally in- vasive surgery, positron emission tomography- computed tomography scanning, dose-moctulated radiothempy, and other technologies. Other means to control cancer have also been developed, improved, or more widely used in recent years. These include strategies for prevention [such as smoking cessation pro- grams, vaccines against cancer-promoting viruses (hepatitis B and papilloma viruses), and methods for detection of premalignant le- sions and early cancers (e.g., colonoscopies, mammography, and PAP smears)]; neurotropic medications to control the ancillary symptoms of cancer, most obviously pain and nausea; hematopoietic growth factors to blunt the side- effects of cytotoxic treatments, such as anemia and leukopenia; and psychosocial methods for managing the response of patients and families to the diagnosis and treatment of cancers. Furthermore, as a recent inventory of US. cancer rates and trends makes evident (38), successful control of cancer will require more than just new technologies, whether molecular- ly based or not. It also calls for elimination of disparities in care-and in access to ciue-that are based on racial and economic factors. Gauging the Future It is &cult to appraise the progress that has been made against cancer over the past hdf-century, but even more so to predict the progress that should be anticipated over the next 10 or 50 years, because cancer is such a complex problem, with hundreds of fomq diverse means of controlling it, and daunting social barriem to reducing its burdens. To argue that the fight against cancer has been dis- appointing, one can simply recall that age-adjusted mortality mtes now are about the same as they were 50 years ago. But it is also legitinlate to supprt a more optinnstic view by noting the recent annual 1% declines in mortality rates after seveid decades of steady incms (38); the enormous improvements in treatments of a few adult and several pediatric cancers; the large increases in 5-year patient survival rates for many cancers (39); the recent development and FDA approval of several narrowly targeted therapies with mild side-effects: and the several ways in which living with advanced cancer has been made better by controlling the symptoms of even resilient underlying disease. Regardless of how the current situation is viexved, the United States and many other countries are faced with a daunting demographic reality: With the continued aging of the population, the absolute number of cancer diagnoses will very likely rise substantially in the coming decades. So, for the foreseeable future, we will need better ways to detect and treat cancers, especially the solid tumors of the lung, breast, prostate, colon, pan- creas, ovary, and other organs that are common in older age pups. Aaicles in this issue provide grounds for optimism about the prospects for better means to control such cancers if new research opportunities are fully exploited. But science operates in a cultural context that affects the deployment of the limited financial resources and human talent devoted to cancer. From that perspective, there is a great deal to wony about. The major public support for cancer research in the United States comes froni the National Cancer Ins$itute (NCI) an& to lesser degrees, from several other components of the NIH. Despite a much welcomed doubling of the NIH budget koni 1998 to 2003, appropriations to the NCI specif~cally, and to the NIH generally, have not kept pace with in- flation since then (40). As a result, the buying pow- er of the NIH has been substantially eroded, and the success rates for grant applications have fallen to discouraging levels. In this atniosphere, it is diflicult to take on new and expensive projects and to ath-dct the best young talent even to this exciting and important area of research. Furthennore, the leadership of the nation's cancer efforts has been poorly defined in recent months and will remain so until a new NCI director is appointed (41,42). Traditionally, the public has looked to the pharmaceutical and biotechnology industries for new tools to detect and treat a wide spectrum of diseases, based largely on the results of publicly hded basic science. But a nuniber of factors raise questions about how, in oncology, this tradition may be challenged by a future increasingly influenced by a molecular view of cancer. Will industry lose incentives to develop targeted therapies that address small, precisely &fined classes of tumors? Or will commonalities among tumors, such as the high frequency of mutations in RAS genes (43), sustain n~et sizes? Will the high prices of some recently approved cancer therapies (44) be sustainable, given increasing pres- sures on health care financing? Will govemment agencies and private insm continue to provide adequate reimbment for molecular methcds for detecting diagnosing, and monitoring tumors as thc use of these cimntly expensive technologies expands? Will rebylatory agencies and industry find common ground to allow affordable and interpret- able clinical trials for drugs for uncommon cancers, perhaps by using early indicatm of tlimpeutic success, such as biomarkers in m'? And will companies collabomte to test multi-agent thempies directed at multiple targets? Finally, the new em in cancer research calls for changes in the culture of oncology. These include swonger working relationdups between bench scientists and their clinical colleabwes, between oncologists in academia and those in community hospitals, and between oncologists and other physicians, new iraining programs that provide graduate students in the basic sciences with an opportunity to understand the dilemmas posed by cancer as a human disease; grant mehsms and criteria for advancement in academia that support the kind of teamwork traditionally associated with inchshy; and guarantees of access to the molecular data sets genwdted with public funding, to enhance their usefulncss for investigators, practitioners, and patients and their advocates. 1164 26 MAY 2006 VOL 312 SCIENCE www.sciencemag.org SPECIALS ECTlO N In sum, concerted national efforts to ensure the vitality of all of the components of modern oncology-academic research, industrial devel- opment, and the delivery ofnew methods through- out the health care amx-are essential to an optimistic view of the prospects for transforming an understandmg of oncogenic mechanismis into therapeutic benefits for our entire society. References Statistics-2006- Presentation.asp. 1. www.cancer.org/docrooVPRO/content/PRO-l-l-Cancer- 2. I. Peaice, New York Times, 16 March 2006, p. A25. 3. 1. H. Burchenal, M. L. Murphy, C. T. Tan, Pediatrics 18, 643 (1956). 4. www.beyonddiscovety.org/conten~view.article.asp?a=285. 5. 6. I. Druker et a/., N. Engl. I. Med. 344, 1031 (2001). 6. L. M. Staudt, 5. Dave, Adv. Immunol. 87, 163 (2005). 1. A. H. 8ild et al., Nature 439, 353 (2006). 8. 1. M. Bishop, Cell 64, 235 (1991). 9. www.sanger.ac.ukigenetics/CGP/Census/. 10. C. 6. Yoo, P. A. ]ones, Not Rev. Drug Discov. 5, 37 (2006). 11. 1. E. Garber, K. Offit, 1. Clin. Oncol. 23, 276 (2005). 12. 0. Hanahan, R. A. Weinberg, Cell 100. 57 (2000). 13. W. Pao, V. A. Miller,]. Clin. Oncal 10, 2556 (2005). 14. E. H. Romond et al., N. Engl.1. Med. 353, 1673 (2005). 15. M. I. Piccart-Gebhart et al., N. Engl. I. Med. 353, 1659 (2005). 16. H. Varrnus et al., Cold Spring Harbor Symp. Quont. Bial. 70, l(2006). 17. I. 6. Weinrtein, Science 297, 63 (2002). 18. D. W. Felsher, Curr. Opin. Genet. Dev. 14, 37 (2004). 19. http://cancergenome.nih.gov/index.asp. 20. w.genorne.gov/Pages/AbouVNACHGR/ May2005NACHGRAgendalReportoft heWorkingGroupon BiomedicalTechnology.pdf. 21. R. F. Service, Science 311, 1544 (2006). 22. Y. Nagata et ol., Cancer Cell 6, 117 (2004). 23. C. M. D'Cruz et a/., Nat. Med. 7, 235 (2001). 24. G. H. Fisher et al., Genes Dev. 15, 3249 (2001). 25. M. E. Gorre et al., Science 293, 876 (2001). 26. W. Pao et ol., PLoS Med. 2, e17 (2005). 27. M. AL-Hajj et a/., Proc. Natl. Acad. Sci. USA 100. 3983 28. Y. Li et a/., Proc. Natl Acad. Sci. U.S.A. 100. 15853 29. M. AI-Hajj et al., Curr. Opin. Genet. Dev. 14, 43 (2004). (2003). (2003). 30. T. Reya, 5. 1. Morrison, M. F. CLarke, I. L. Weissman, Nature 414, 105 (2001). 31. D. 6. Solit et a/., Nature 439, 358 (2006). 32. N. Ferrara, R. 5. Kerbel, Nature 438, 967 (2005). 33. N. A. Bhowmick, H. L. Moses, Cur[ Opin. Genet Dev. 15, 97 (2005). 34. C. M. Overall, 0. Kleifeld, Not. Rev. Cancer 6, 227 (2006). 35. C. Iedeszko, B. F. Sloane, BiOl. Chem. 385, 1017 (2004). 36. 1. Grimm et al., Proc. Natl. Acad. Sci. US.A 102, 14404 (2005). 37. 1. G. Egen, M. 5. Kuhns, 1. P. Allison, Not. knmunal. 3, 611 (2002). 38. B. K. Edwards et al., 1. NatL Cancer Inst. 97, 1407 (2005). 39. A. lemal et al., C4 Cancer]. Clin. 55, 10 (2005). 40. www.aaas.org/spp/rdlnih07p.htm. 41. 1. Couzin, ScienceNOW Daily News, 15 March 2006; http://sciencenow.sciencemag.org/cgi/conlen~fulU2006/ 315/1. 42. R. Weirs, The Washington Post. 1 October 2005, p. A03. 43. 1. L. BOS, Cancer Res. 49, 4682 (1989). 44. A. Berenson, The New Yok Times, 15 February 2006, p. Al. 10.1126/science.1126758 www.sciencernag.org SCIENCE VOL 312 26 MAY 2006 1165