|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
Journal of Clinical Oncology, Vol 23, No 29 (October 10), 2005: pp. 7253-7256 © 2005 American Society of Clinical Oncology. DOI: 10.1200/JCO.2005.03.9792
Genomic Profiling of Cancer: What Next?
Cancer Research United Kingdom Centre for Cancer Therapeutics, The Institute of Cancer Research, Sutton, Surrey, United Kingdom During the last decade, the development of a number of high-throughput genomic technologies has begun to have significant implications for both clinical and basic research in cancer. The ability to obtain gene expression profiles or genetic fingerprints at a genome-wide level has begun to have an impact on the diagnosis and prognostic classification of tumors, as well as on the prediction of response of individual patients to specific chemotherapeutic regimes. In addition, the identification of novel tumor targets that can be exploited therapeutically has also become a reality. Advances in genomic technologies such as gene expression microarrays, high-throughput sequencing, and molecular and functional imaging approaches such as magnetic resonance spectroscopy and imaging and positron emission tomography have facilitated the integration of genomic and functional data to allow better understanding of the molecular biology of a given disease and to facilitate clinical decision making. These powerful technologies are being used to study many biologic and clinical processes and have provided important insights in areas such as cancer pathogenesis, cancer diagnosis, pharmacogenetics and prediction of clinical outcome, as well as the development of novel targeted therapeutics. This potential is exemplified by the studies presented in this issue of the Journal of Clinical Oncology, which examine genomic expression signatures and specific patterns of loss of heterozygosity (LOH) and polymorphisms in growth factor receptors as important clues to the pathogenesis of disease or as potential genomic signatures for prognosis or prediction of outcome to therapy. They highlight the potential to exploit our understanding of tumor biology at the genetic and molecular level to improve cancer treatment and illustrate the biggest challenge of modern biology and medicine: to understand the "software and hardware" of the cancer cell and the cancer patient in comparison with the nondiseased state. For the clinician, the appropriate diagnosis and classification of cancer is of key importance for the selection of the most appropriate treatment or for defining patient prognosis. During the last 50 years, we have depended on the anatomic and morphologic classification of tumors and have treated patients accordingly. However, the application of modern genomic technology is revealing a new genetic taxonomy that has begun to facilitate the reclassification of cancers based on a molecular signature. Many tumors that until now appeared indistinguishable at the morphologic level are molecularly very distinct, and such molecular distinction may be helpful when applied to patient treatment. We now have more and better drugs, together with improved knowledge of cancer as a disease, and this trend will continue. However, human cancers are a diverse group of diseases that quickly display their heterogeneity, in particular when exposed to all forms of chemotherapy and, more recently, to novel targeted therapies. Therefore, the challenge for the next decade will be the development of clinical trials that combine targeted drugs and cytotoxics along with genomic profiles to identify subsets of patients who are most likely to benefit from certain treatments, thereby avoiding the needless cost and toxicity of ineffective treatment. Indeed this approach has the potential to confer enormous advantages in clinical trial design, allow for more efficient and cost-effective drug development, and ultimately reduce the burgeoning cost of this new technology. Therefore, the utilization of these genomic technologies and the application of pharmacogenomics must become a central platform in patient treatment and for cancer drug development for the future Despite the enormous advances that have been made, there remain significant hurdles to be overcome to harness their full potential. These include a lack of a focused and disciplined stepwise approach in their implementation, poorly defined protocols and inadequately sized clinical study groups, as well as unreliable bioassays and a lack of suitable clinical material. Unless these issues are urgently addressed in a comprehensive manner, the incorporation of these new technologies into the management of patients will continue to be problematic and slow. Pharmacogenomic approaches can be broadly separated into two categories: those based on candidate genes, or those based on a genome-wide approach. The review by Walgren et al1 summarizes the advantages and disadvantages to each approach. The candidate gene approach assumes knowledge of the pathology and pharmacology and allows the selection of candidate genes for which expression may impact therapeutic response and offers the advantage of a potential cost savings. In addition, the smaller number of starting genes reduces the risk of false-positive findings that may occur in a genome-wide approach, but at the risk of excluding other genes that may be important. Whole-genome approaches identify genes that play a significant role in a given phenotype and have the ability to reveal genes that might be anticipated to be involved, but also genes not known to play a significant role that may therefore add important new insights into the pathophysiology or pharmacology in a specific setting. A more optimal selection of candidate genes for pharmacogenomics and predictive oncology may be achieved by combining in vitro discovery using cell lines, including Centre d'Etude Polymorphisme Humain pedigree lines, comparative studies using ex vivo samples from patients, tissues, and appropriate studies using mouse models. However, despite these genome-wide and candidate gene approaches, no multigene expression signature has yet been widely adopted in clinical practice to date. Walgren et al point out that linkage of genetics to individual variations in drug response is not new. Differences in pharmacodynamic responses to irinotecan due to polymorphic variation in the glucuronidation of the metabolite SN38 by UGT1A1 is well established. The effects of the polymorphic cytochrome P450 genes on the response to drugs is also well known and new drug development programs seek to avoid this liability. Polymorphisms in the thymidylate synthase gene promoter affect responsiveness to fluorouracil.
More recently, it has been established that somatic mutations in the EGFR gene, which encodes the drug target, can account for a large amount, but not all, of the variation in sensitivity to gefitinib and erlotinib. The latter discovery was made by a combination of sequencing of a candidate gene (ie, the drug target) and high-throughput sequencing of larger numbers of kinases.2-4 Subsequent mutations in EGFR can lead to resistance.5,6 The power of high-throughput mutation analysis was exemplified by the discovery of B-RAF as an oncogene in melanoma and other cancers.7 High-throughput kinome sequencing identified putative activating mutations in several kinases,8 including the PIK3CA gene that encodes the p110 In addition to genome sequencing, gene expression microarray analysis is also a powerful new approach. The development of genomic-based biomarker classifiers useful for improving treatment decisions and sufficiently validated for broad clinical application is problematic. Simon's14 excellent review attempts to clarify some of the misconceptions regarding the development and validation of multigene signature classifiers and highlights the steps needed to move genomic signatures into clinical application as therapeutically relevant diagnostics. He points to the fact that prognostic marker studies are generally performed with no written protocol, no eligibility criteria, no primary end point or hypothesis, and no defined analysis plan. In order to correct these deficiencies, he suggests that a focused development pathway must be developed involving several key decision steps. These include developing the genomic classifier for a specific therapeutic decision problem and using material from clinical cases relevant to that decision context; consideration must be given to the treatment options and costs of misclassification, such that any classifier that is derived is likely to be used. In addition, Simon suggests that an internal validation of the classifier is performed to assess whether it appears sufficiently accurate relative to standard prognostic factors to justify further development for broad clinical application. Finally, the classifier must be robust, reproducible and independently validated in a prospectively planned study. Simon pays particular attention to the critical issue of validation of genes of therapeutic relevance. These issues are exemplified by the Caldas group, who examine the current status of the molecular reclassification and the state of predictive oncology within breast cancer.15 Many expression microarray studies have addressed broad prognostic questions in breast cancer but none satisfy the criteria for good prognostic studies or provide a sufficient level of evidence for the widespread clinical implementation of candidate markers.16 The two largest studies completed to date have sought to identify genomic markers for recurrence or survival after surgical treatment of node-negative disease. Van't Veer et al17 identified a 70-gene classifier set that predicted early relapse in a cohort of 98 node-negative breast cancers with onset before the age of 55. In a similar-sized study, Wang et al18 derived a 76-gene set that predicted early relapse from 115 node-negative cancers. Of note, despite similar clinical and statistical designs, the two independent genomic classifiers shared only three genes in common. This may have been due to the use of different microarray platforms, which can lead to differences in data repeatability and gene discovery and a marked variability in the derived prognostic signatures. This also arises because the difference between gene-outcome correlations is small, making it difficult to identify the most predictive genes, and the process of choosing the genes is highly dependent upon the subset of patients used to develop the prediction model. Taken together, these results highlight the difficulties of using limited numbers as a study sample together with a technology that may be poorly reproducible. More importantly, they underscore the importance of adopting a standardized approach as suggested by Simon to translate correlations between gene expression and prognosis into robust diagnostics that are clinically relevant, in a timely manner. The critical review from Brenton et al15 is complemented by two research reports on gene expression profiling in breast cancer. Gianni et al19 set out to identify gene-expression markers that predict the likelihood of chemotherapy response and to test whether the response correlated with a 21-gene Recurrence Score.20 Pathologic complete response was associated with a high expression of proliferation- and immune-related genes and a lower expression of estrogen receptor–related genes. The Recurrence Score was positively associated with likelihood of pathologic complete response. Espinosa et al21 sought to reproduce using quantitative reverse transcriptase polymerase chain reaction (qRT-PCR) results obtained with the previously described 70 gene signature.22 The gene profile was found to be significantly associated with relapse-free survival and overall survival, and the authors conclude that qRT-PCR was able to reproduce the prognostic gene expression profile in early stage breast cancer. Sieben et al23 have studied ovarian serous borderline tumors that behave in an indolent manner and that have a high incidence of mutations in B-RAF and K-RAS, in contrast to serous carcinomas. Gene expression profiling by microarray showed that although the mitogenic pathway is activated in these tumors, activation of genes involved in the degradation of extracellular matrix is absent. Furthermore, two genes involved in regulating the uncoupling of these events, namely those encoding the ERK inhibitor Dusp 4 and the uPA inhibitor Serpina 5, are downregulated in the serous carcinomas, with activation of the metalloprotease MMP-9, in contrast to serous borderline tumors. The results suggest that Dusp 4 and Serpina 5 are putative tumor suppressor proteins that may explain the indolent nature of the borderline tumors. The Gly338Arg polymorphism in the fibroblast growth factor 4 (FGFR4) receptor tyrosine kinase is associated with poor prognosis in various cancers. Spinola et al24 performed a case-control study of 274 patients with adenocarcinoma of the lung and 401 healthy control subjects; mRNA analysis was carried out in healthy lung of cancer patients. The results suggested the FGFR4 Gly388Arg polymorphism may be predictive of prognosis in lung adenocarcinoma. Bilke et al25 present a model of tumor progression for neuroblastoma based on DNA copy number. By comparing the least and most aggressive stages of neuroblastoma, the authors' progression model was found to be compatible with current models based on ploidy changes. However, the model is not a simple linear one and reflects the heterogeneity of the clinical behavior. Grundy et al26 investigated whether LOH for chromosomes 16q or 1p was associated with a poorer prognosis for children with favorable-histology Wilms tumor. LOH was rare in clear-cell sarcoma of the kidney or malignant rhabdoid tumor of the kidney. Tumor-specific LOH for both 16q and 1p identified a subset of favorable-histology Wilms tumor patients with a significantly increased risk of relapse and death. The authors suggest that LOH for these regions can be used as an independent prognostic factor, so that treatment intensity can be targeted to high-risk patients. Chen et al27 investigated whether gene expression profile predicts for survival in surgically curable gastric cancer patients following surgical resection. A survival prediction model was developed on the basis of three genes, namely those encoding CD36 antigen, signaling lymphocytic activation molecule and the PIM-1 oncoprotein. The authors conclude that RT-PCR expression profiling of these three genes, derived initially from microarray data, can be used to predict outcome in this setting. Deregulation of the cyclin D genes CCND1, CCND2 and CCND3 is common in multiple myeloma.28 The so-called TC2 group showed extra copies of the CCND1 locus and no immunoglobulin heavy-chain locus translocations or 13q deletions, and was characterized by overexpression of genes involved in protein biosynthesis at the translational level. The authors conclude that identification of the particular gene expression pattern in TC2 patients may improve risk stratification. They also speculate that further analysis of this profile may suggest new therapeutic targets. As new genomic technologies and increasing therapeutic options emerge for cancer treatment it will be essential to use our increasing knowledge of tumor biology to optimize and individualize therapy. However, to date, no genomic markers have been developed to the point of allowing reliable and widespread use in clinical practice. The development of any potential genomic classifier set for clinical use must be explored initially by large retrospective studies such as those presented in this issue of the Journal. When evaluated alongside complete data on other prognostic variables, a more convincing and complete picture may be presented. However, only when a genomic classifier has been independently associated with outcome after adjusting for such variables can it be considered a promising prognosticator and one that should be taken forward for further study of its clinical utility in a prospective clinical study setting. In order to achieve this, the next generation of clinical studies must be designed in a manner that allows the incorporation of genomic technologies alongside clinical drug development. Authors' Disclosures of Potential Conflicts of Interest The authors indicated no potential conflicts of interest. REFERENCES
1. Walgren RA, Meucci MA, McLeod HL: Pharmacogenomic discovery approaches: Will the real genes please stand up? J Clin Oncol 23:7342-7349, 2005
2. Lynch TJ, Bell DW, Sordella R, et al: Activating mutations in the epidermal growth factor receptor underlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med 350:2129-2139, 2004
3. Paez JG, Janne PA, Lee JC, et al: EGFR mutations in lung cancer: Correlation with clinical response to gefitinib therapy. Science 304:1497-1500, 2004
4. Pao W, Miller V, Zakowski M, et al: EGF receptor gene mutations are common in lung cancers from "never smokers" and are associated with sensitivity of tumors to gefitinib and erlotinib. Proc Natl Acad Sci U S A 101:13306-13311, 2004 5. Pao W, Miller VA, Politi KA, et al: Acquired resistance of lung adenocarcinomas to gefitinib or erlotinib is associated with a second mutation in the EGFR kinase domain. PLoS Med 2:e73, 2005[CrossRef][Medline]
6. Kobayashi S, Boggon TJ, Dayaram T, et al: EGFR mutation and resistance of non-small-cell lung cancer to gefitinib. N Engl J Med 352:786-792, 2005 7. Davies H, Bignell GR, Cox C, et al: Mutations of the BRAF gene in human cancer. Nature 417:949-954, 2002[CrossRef][Medline]
8. Bardelli A, Parsons DW, Silliman N, et al: Mutational analysis of the tyrosine kinome in colorectal cancers. Science 300:949, 2003
9. Samuels Y, Wang Z Bardelli N, et al: High frequency of mutations of the PIK3CA gene in the human cancer. Science 304:554, 2004 10. Workman P: Genomics and the second golden era of cancer research. Mol BioSystems. 1:17-27, 2005 11. International Human Genome Sequencing Consortium: Finishing the euchromatic sequence of the human genome. Nature 431:931-945, 2004[CrossRef][Medline] 12. Futreal PA, Coin L, Marshall M, et al: A census of human cancer genes. Nat Rev Cancer 4:177-183, 2004[CrossRef][Medline] 13. Vogelstein B, Kinzler KW: Cancer genes and the pathways they control. Nat Med 10:789-799, 2004[CrossRef][Medline]
14. Simon R: A roadmap for developing and validating therapeutically relevant genomic classifiers. J Clin Oncol 23:7332-7341, 2005 15. Brenton J, Carey LA, Ahmed AA, et al: Molecular classification and molecular forecasting of breast cancer: Ready for clinical application? J Clin Oncol, in press 16. Perou CM, Sorlie T, Eisen MB, et al: Molecular portraits of human breast tumors. Nature 406:747-752, 2000[CrossRef][Medline] 17. van 't Veer LJ, Dai H, van de Vijver MJ, et al: Gene expression profiling predicts clinical outcome of breast cancer. Nature 415:530-536, 2002[CrossRef][Medline] 18. Wang Y, Klijn JG, Zhang Y, et al: Gene expression profiles to predict distant metastasis of lymph node negative primary breast cancer. Lancet 365:671-679, 2005[Medline]
19. Gianni L, Zambetti M, Clark K, et al: Gene expression profiles in paraffin-embedded core biopsy tissue predict response to chemotherapy in women with locally advanced breast cancer. J Clin Oncol 23:7265-7277, 2005
20. Paik S, Shak S, Tang G, et al: A multigene assay to predict recurrence of tamoxifen-treated, node-negative breast cancer. N Engl J Med 351:2817-2826, 2004
21. Espinosa E, Fresno Vara JA, Redondo A, et al: Breast cancer prognosis determined by gene expression profiling: A quantitative RT-PCR study. J Clin Oncol 23:7278-7285, 2005
22. van de Vijver MJ, He YD, van't Veer LJ, et al: A gene expression signature as a predictor of survival in breast cancer. N Engl J Med 347:1999-2009, 2002
23. Sieben N, Oosting J, Flanagan A, et al: Differential gene expression in ovarian tumors reveals Dusp 4 and Serpina 5 as key regulators for benign behavior of serous borderline tumors. J Clin Oncol 23:7257-7264, 2005
24. Spinola M, Vera L, Pignatiello C, et al: The functional FGFR4 Gly388Arg polymorphism predicts prognosis in lung adenocarcinoma patients. J Clin Oncol 23:7307-7311, 2005
25. Bilke S, Chen QR, Westerman F, et al: Inferring a tumor progression model for neuroblastoma from genomic data. J Clin Oncol 23:7322-7331, 2005
26. Grundy PE, Breslow NE, Li S, et al: Loss of heterozygosity for chromosomes 1p and 16q is an adverse prognostic factor in favorable histology Wilms tumor: A report from the National Wilms Tumor Study Group. J Clin Oncol 23:7312-7321, 2005
27. Chen C-N, Lin J-J, Chen JJW, et al: Gene expression profile predicts patient survival of gastric cancer after surgical resection. J Clin Oncol 23:7286-7295, 2005
28. Agnelli L, Bicciato S, Mattioli M, et al: Molecular classification of multiple myeloma: A distinct transcriptional profile characterizes patients expressing CCND1 and negative for 14q32 translocations. J Clin Oncol 23:7296-7306, 2005
This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2005 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
catalytic subunit of phosphatidylinositol-3 kinase (PI3K).
