- Email: [email protected]
USING MOLECULAR MARKERS TO PREDICT OUTCOME
0022-5347/04/1725-0018/0 THE JOURNAL OF UROLOGY® Copyright © 2004 by AMERICAN UROLOGICAL ASSOCIATION
Vol. 172, S18 –S22, November 2004 Printed in U.S.A.
DOI: 10.1097/01.ju.0000142448.58831.d9
USING MOLECULAR MARKERS TO PREDICT OUTCOME MARK A. RUBIN* From the Department of Pathology, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts
ABSTRACT
Purpose: Developing molecular tests to predict prostate cancer progression requires first defining meaningful clinical end points and defining strategies to take advantage of emerging technology. Materials and Methods: The select relevant literature was reviewed concerning clinical trials, clinical prostate cancer nomograms, molecular biomarker development and molecular prostate cancer imaging. Results: There is controversy regarding the use of prostate specific antigen or biochemical failure following prostatectomy or radiation therapy for clinically localized prostate cancer as a marker of progression. As a consequence, advances in prostate cancer biomarker development may require using population based cohorts or cases from clinical trials to identify meaningful associations. Whereas the discovery of novel candidate biomarkers was slow 5 to 10 years ago and often resulted from serendipity, advances in high throughput technologies have led to the identification of a large number of candidate genes. Strategies to identify candidate genes include the use of expression array analysis, single nucleotide polymorphism arrays (single nucleotide polymorphism chips), proteomics and bioinformatics. Monitoring the progression of prostate cancer has been limited to standard approaches such as computerized tomography or magnetic resonance imaging, which in general do not delineate the extent of disease. By carefully selecting novel prostate cancer biomarkers future work should allow in vivo monitoring of prostate cancer. This will represent a revolutionary advance in our ability to monitor prostate cancer progression and ultimately it may be one of the most important applications of cancer biomarkers. Conclusions: Emerging technology should allow us to analyze clinical prostate cancer trials with sufficient followup to help develop meaningful markers of prostate cancer progression. KEY WORDS: prostate; prostatic neoplasms; gene expression; tumor markers, biological; polymorphism, single nucleotide
Indolent prostate carcinoma exists but we are unable to diagnose it in a practical manner. Some colonic adenomas progress with time into cancers and even at the earliest phases they share molecular alterations identical to those of cancer, although they are not classified as carcinoma. One important difference between colonic adenoma and indolent prostate cancer is that, whereas adenoma can be resected from the colonic mucosa and followed by endoscopy, the prostate tumor is not accessible to continuous monitoring. The terms expectant therapy or watchful waiting have been coined for the process of monitoring men with biopsy proven prostate cancer. However, this strategy leaves the patient and doctor living with the uncertainty of how the cancer will progress with time. Are they following an indolent tumor? Are they sitting on a large aggressive tumor that has been only partially sampled? Using a combination of clinical, pathological and molecular modalities, we should be able to distinguish indolent from aggressive prostate cancer. DEFINING INDOLENT PROSTATE CANCER REQUIRES STUDYING THE CORRECT POPULATION
In the United States prostate cancer remains the most common solid tumor malignancy in men with approximately 200,000 men diagnosed yearly. A total of 30,000 men will die yearly in the United States due to prostate cancer. Epidemiological data suggest a tremendous disparity between men
diagnosed with clinically localized prostate cancer and those presenting with advanced metastatic disease with 100% and less than 50% 5-year survival rates, respectively.1, 2 Many investigators believe that the improvement in 5-year survival has less to do with improved treatment and more to do with early detection since the introduction of the serum blood test, prostate specific antigen (PSA), in 1987. Increasing the time from initial cancer diagnosis to clinical presentation due to symptoms by widespread serum PSA screening, as in the United States, definitely improves 5-year survival statistics. As a result, some aggressive cancers are detected earlier but of equal concern are the significantly larger numbers of indolent tumors detected through PSA screening. These indolent tumors may not have posed any risk of clinical morbidity or mortality but, nonetheless, they are treated in a similar manner.3 Therefore, an overall goal is to develop new molecular tests to improve our ability to detect significant prostate cancers. A hurdle to discovery is that we are testing biomarkers against patient populations with insufficient followup. In the United States most surgical series describe low risk patient populations identified by PSA screening. Therefore, given a combination of PSA lead time bias and short followup, most series have few significant events and focus on PSA recurrence (ie biochemical failure) as the study end point. In fact, it has been suggested that PSA recurrence may not be the most relevant end point to determine clinically relevant events such as the development of metastatic disease and disease specific death.4, 5 Therefore, given the limitations of populations identified using PSA screening, 3 examples of cohorts that should be extremely useful in characterizing the molecular progression
Supported by SPORE National Cancer Institute Grants P50CA90381 (MAR) and P50CA69568 (MAR), and National Cancer Institute Grants CA97063 (MAR) and R01AG21404 (MAR). * Correspondence and requests for reprints: Department of Pathology, Brigham and Women’s Hospital/Harvard Medical School, 75 Francis St., Boston, Massachusetts 02115 (telephone: 617-525-6747; FAX: 617-264-5169; e-mail: [email protected]). S18
USING MOLECULAR MARKERS TO PREDICT OUTCOME
of prostate cancer will be briefly described. Two of these trials come from Scandinavia6 –9 and the third is from the Physicians’ Health Study.10 The first study is the randomized Scandinavian trial evaluating the benefit of prostatectomy over watchful waiting.6 Approximately 700 men were randomized into a surgery arm or a watchful waiting arm. Surgery was seen to decrease significantly the incidence of metastatic prostate cancer and cancer specific death by 50% at 8-year followup. Longer followup may demonstrate even more dramatic benefits. However, the study also determined that 17 men needed surgery to prevent 1 prostate cancer related death. The implication that a significant number of men do not benefit from treatment is a clear message that should not change dramatically. A second study with significantly longer followup is the watchful waiting trial performed in the Orebro region of Sweden.7, 8 This population based study follows men diagnosed with localized prostate cancer in the pre-PSA era between 1977 and 1984 in a well-defined catchment area of 190,000 men living in south central Sweden. Study results at 10 and 15 years suggested that there is minimal benefit from localized treatment (ie surgery or external beam radiation). A recent report of Johansson et al on the 20-year follow up of these patients demonstrated some surprising late events.9 From approximately 200 men on this watchful waiting cohort 31 with prostate cancer had metastatic disease, 27 died of prostate cancer, 113 died of other causes and 27 were still alive. When averaged during the first 15 years, the rate of progression to metastatic disease was 18/1,000 person-years and the prostate cancer mortality rate was 15/1,000 personyears. In contrast, an approximately 3-fold higher rate was found for progression and death during followup between years 15 and 20.9 Lastly, the Physicians’ Health Study is a population based study done at the Harvard School of Public Health, and Brigham and Women’s Hospital that has followed more than 20,000 male physicians since 1982.10 Most men were enrolled into the study while in their early 40s. Detailed dietary information was gathered, as were serum samples. To date approximately more than 1,700 participants are estimated to have prostate cancer. Future work will be able to focus on serum and tissue biomarkers that help determine which men had prostate cancer during the course of this study. Therefore, in studies evaluating the natural history of prostate cancer the epidemiological data are complex, suggesting that localized prostate cancer is being dramatically over treated. However, even apparently indolent tumors may progress at a long interval following initial diagnosis. These observations make the characterization of pure indolent tumors more challenging and suggest that the use of outcomes studies using surrogate end points, such as PSA biochemical recurrence, may not be sufficient. LIMITATIONS OF PATHOLOGICAL FINDINGS AND CLINICAL NOMOGRAMS
Urologists are most familiar with the pretreatment nomograms that have been developed and validated to predict prostate cancer recurrence after treatment for localized disease.11–14 These nomograms classically consider serum PSA levels, prostate needle biopsy Gleason score and clinical stage. At best these clinical nomograms have an area under the ROC curve of 0.75. Kattan et al recently noted that the addition of 2 novel serum markers (interleukin-6SR and transforming growth factor-1) could improve the area under the ROC curve to 0.83, suggesting that molecular markers combined with clinical parameters could improve existing predictive nomograms.15 Therefore, to date the best clinical nomograms can predict which men may benefit from surgery. However, these nomograms do not address if we are over treating these men. A reason that current clinical nomo-
S19
grams are of limited value is that we still cannot determine the volume of high grade prostate cancer that is present before treatment.16 In the absence of endoscopy for the prostate we need to develop molecular tools to help see into the prostate gland and determine the extent of aggressive disease. STRATEGIES FOR THE DEVELOPMENT OF MOLECULAR BIOMARKERS
If the goal is to create a molecular definition to help distinguish indolent from aggressive prostate cancer, the development of the appropriate biomarkers is needed. If we define a biomarker as a molecular test that provides additional information over current clinical data (including pathological review), we are currently lacking any bona fide prostate cancer biomarkers to define indolent disease. There is a massive ongoing effort in this post-genomic era to discover novel cancer biomarkers to help solve this problem. Expression array analysis. Recent advances in genomic and proteomic technologies suggest that molecular signatures of disease can be used to diagnose,17, 18 predict survival19, 20 and define novel molecular subtypes of disease.21 Several studies have used expression array analysis to characterize the gene expression profiles of prostate cancer compared with benign prostate disease and normal prostate tissue.22–27 Several interesting candidate genes have been consistently identified, including AMACR, fatty acid synthase, PIM-1 kinase, hepsin and EZH2. Future work in the area of cDNA expression array analysis may allow for its use in the clinical setting. The use of expression array analysis to predict clinical outcome has been attempted in the area of lymphoma,28 breast29 and prostate24 cancer. However, there are no robust clinical models that can be applied today. Several prospective trials have expression array analysis built into their design but beyond these proof of principle studies expression array analysis has been more of a biomarker discovery tool. Single nucleotide polymorphism (SNP) analysis. In a recent study SNPs in prostate cancer have begun to be explored using Affymetrix SNP arrays (Affymetrix, Santa Clara, California).30 A group recently reported the use of these SNP chips to evaluate loss of heterozygosity (LOH) by comparing tumor and normal samples.30 This initial study was performed using SNP chips with 1.5K SNPs. The current chip has 11.5K SNPs and in 2004 SNP chips will be available with more than 120K SNPs. There are several exciting applications of this technology for biomarker development. These high resolution SNP chips can detect LOH, deletions and amplification at the DNA level in 120,000 known SNPs. Therefore, novel tumor classes may become evident, as has previously been seen with cDNA expression array analysis using various unsupervised and supervised clustering methods. As an example, in the quest for characterizing indolent tumors that do not progress our research group is using this technology to test clinical samples from the Orebro Watchful Waiting trial, as described.7, 8 SNP analysis also allows the development of individual biomarkers. For example, gains in the long arm of chromosome 8 (8q) are believed to be associated with poor outcome and the development of hormone refractory prostate cancer. Based on a meta-analysis of gene expression microarray data from multiple prostate cancer studies31 a candidate oncogene, tumor protein D52 (TPD52), was identified in the 8q21 amplicon. TPD52 is a coiled coil, motif bearing protein that is potentially involved in vesicle trafficking. The mRNA and protein levels of TPD52 were highly elevated in prostate cancer tissues using tissue microarrays. Amplification analysis using SNP arrays demonstrated increased DNA copy number in the region encompassing TPD52, which was confirmed by fluorescence in situ hybridization.32 These findings suggest that dysregulation of
S20
USING MOLECULAR MARKERS TO PREDICT OUTCOME
TPD52 by genomic amplification may have a role in prostate cancer progression and it is an example of combining several technologies to characterize molecular biomarkers. Proteomics. Recent work has concentrated on performing proteomics to detect prostate cancer specific fingerprints.33–36 For example, Petricoin et al identified unique profiles seen in serum from patients with prostate cancer but not controls.34 This work promises to identify proteins that may be used for prognosis and diagnosis of prostate cancer. Currently several proteomic approaches are being used, including 2-dimensional electrophoresis and surface enhanced laser desorption ionization-time of flight proteomics.17 A limitation to many of these techniques is that proteins are difficult to characterize from this analysis. Another important criticism is that the majority of proteins identified in the serum will not be from the tumor, making analysis of the numerous peaks extremely difficult. However, it is still feasible that a serum test could be developed without knowing what the specific candidate proteins are and, nonetheless, be clinically helpful in distinguishing indolent from aggressive prostate cancer. For example, serum samples from the Physicians’ Health Study are being evaluated to identify profiles of men who had prostate cancer during the longterm year followup of this study.10 Bioinformatics. Informatics tools have an important role in the discovery process. A rapidly emerging field, bioinformatics, is starting to alter the way research is being performed. Using information from large databases in silico studies can be done to discover and validate new candidate genes and pathways significant in areas such as the development of prostate cancer. For example, Rhodes et al recently identified lists of significant prostate cancer-related genes by performing a meta-analysis on publicly available cDNA expression array datasets.31 This study was also able to extrapolate prostate cancer related pathways by piecing together data from multiple studies. This approach has now become available on an Internet based website called ONCOMINE (www.oncomine.org), which allows user to perform a meta-analysis of genes of interest and contains links to other websites that provide information regarding their genes of interest.37 In addition to currently existing bioinformatics tools for microarray data, such as cluster, TreeView (http://taxonomy.zoology.gla.ac.uk/rod/treeview/treeview_ manual.html) and dCHIP (http://biosun1.harvard.edu/complab/ dchip/install.htm), new tools are constantly being developed (http://chip.dfci.harvard.edu/). For example, Lin et al are developing tools to perform hierarchical clustering of SNP data, amplification analysis and LOH analysis.38 MOLECULAR IMAGING
Even if elegant molecular profiles for prostate cancer progression are developed, could we use this information to impact prostate cancer treatment? The answer will depend on whether the expression of these biomarkers is readily detected in serum or in prostate biopsy samples. For this to be reliable the development of markers expressed in a homogenous manner would be preferable for the development of an indolent molecular profile. Perhaps the greatest impact on defining indolent prostate cancer will come in the area of molecular imaging. High resolution imaging will continually improve our ability to view anatomical structures but correlations with molecular imaging may provide a way of monitoring prostate cancer progression. For example, a group recently reported that highly lymphotropic superparamagnetic nanoparticles can gain access to lymph nodes and be visualized at extremely high resolution, sufficient to detect displacement by metastatic prostate cancer.38 A limitation of this technology is that it targets host macrophages and is not cancer specific. Other ongoing work targets prostate cancer specific markers, which are located on the tumor cell surface. An example is hepsin, a serine protease, which is known to be
expressed early in prostate cancer development.24 Hepsin, which is expressed by other organs including the liver, would be a poor serum marker but it may turn out to be a powerful molecular beacon. Therefore, with the correct tools molecular imaging could have the greatest impact in following prostate cancers with time, thus, helping stage and classify tumors as indolent and aggressive based on progression and invasiveness. CONCLUSIONS
Briefly, the goal of predicting prostate cancer progression, that is distinguishing indolent from aggressive disease, within the next 5 to 10 years is palpable. Even if this cannot be defined from a pure biological standpoint, molecular imaging or the development of serum tests may make monitoring of men with biopsy proven prostate cancer feasible. The development of these tests will require evaluation of the correct clinical cohorts. REFERENCES
1. Chu, K. C., Tarone, R. E. and Freeman, H. P.: Trends in prostate cancer mortality among black men and white men in the United States. Cancer, 97: 1507, 2003 2. Weir, H. K., Thun, M. J., Hankey, B. F., Ries, L. A., Howe, H. L., Wingo, P. A. et al: Annual report to the nation on the status of cancer, 1975–2000, featuring the uses of surveillance data for cancer prevention and control. J Natl Cancer Inst, 95: 1276, 2003 3. Etzioni, R., Penson, D. F., Legler, J. M., Di Tommaso, D., Boer, R., Gann, P. H. et al: Overdiagnosis due to prostate-specific antigen screening: lessons from u.s. Prostate cancer incidence trends. J Natl Cancer Inst, 94: 981, 2002 4. Pound, C. R., Partin, A. W., Eisenberger, M. A., Chan, D. W., Pearson, J. D. and Walsh, P. C.: Natural history of progression after PSA elevation following radical prostatectomy. JAMA, 281: 1591, 1999 5. D’Amico, A. V., Moul, J. W., Carroll, P. R., Sun, L., Lubeck, D. and Chen, M. H.: Surrogate end point for prostate cancerspecific mortality after radical prostatectomy or radiation therapy. J Natl Cancer Inst, 95: 1376, 2003 6. Holmberg, L., Bill-Axelson, A., Helgesen, F., Salo, J., Folmerz, P., Ha¨gmann, M. et al: A randomized trial comparing radical prostatectomy with watchful waiting in early prostate cancer. N Engl J Med, 347: 781, 2002 7. Johansson, J. E., Holmberg, L., Johansson, S., Bergstrom, R. and Adami, H. O.: Fifteen-year survival in prostate cancer. A prospective, population-based study in Sweden. JAMA, 277: 467, 1997 8. Johansson, J. E., Adami, H. O., Andersson, S. O., Bergstrom, R., Holmberg, L. and Krusemo, U. B.: High 10-year survival rate in patients with early, untreated prostatic cancer. JAMA, 267: 2191, 1992 9. Johansson, J. E., Andren, O., Andersson, S. O., Dickman, P. W., Holmberg, L., Magnuson, A. et al: Natural history of early, localized prostate cancer. JAMA, 291: 2713, 2004 10. Manson, J. E., Buring, J. E., Satterfield, S. and Hennekens, C. H.: Baseline characteristics of participants in the Physicians’ Health Study: a randomized trial of aspirin and betacarotene in U.S. physicians. Am J Prev Med, 7: 150, 1991 11. Partin, A. W., Kattan, M. W., Subong, E. N., Walsh, P. C., Wojno, K. J., Oesterling, J. E. et al: Combination of prostate-specific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update. JAMA, 277: 1445, 1997 12. Kattan, M. W., Eastham, J. A., Stapleton, A. M., Wheeler, T. M. and Scardino, P. T.: A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst, 90: 766, 1998 13. Kattan, M. W., Zelefsky, M. J., Kupelian, P. A., Scardino, P. T., Fuks, Z. and Leibel, S. A.: Pretreatment nomogram for predicting the outcome of three-dimensional conformal radiotherapy in prostate cancer. J Clin Oncol, 18: 3352, 2000 14. Kattan, M. W., Potters, L., Blasko, J. C., Beyer, D. C., Fearn, P., Cavanagh, W. et al: Pretreatment nomogram for predicting freedom from recurrence after permanent prostate brachytherapy in prostate cancer. Urology, 58: 393, 2001
USING MOLECULAR MARKERS TO PREDICT OUTCOME 15. Kattan, M. W., Shariat, S. F., Andrews, B., Zhu, K., Canto, E., Matsumoto, K. et al: The addition of interleukin-6 soluble receptor and transforming growth factor beta1 improves a preoperative nomogram for predicting biochemical progression in patients with clinically localized prostate cancer. J Clin Oncol, 21: 3573, 2003 16. Rubin, M. A., Mucci, N. R., Manley, S., Sanda, M., Cushenberry, E., Strawderman, M. et al: Predictors of Gleason pattern 4/5 prostate cancer on prostatectomy specimens: can high grade tumor be predicted preoperatively? J Urol, 165: 114, 2001 17. Adam, B. L., Qu, Y., Davis, J. W., Ward, M. D., Clements, M. A., Cazares, L. H. et al: Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men. Cancer Res, 62: 3609, 2002 18. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P. et al: Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science, 286: 531, 1999 19. Rosenwald, A., Wright, G., Chan, W. C., Connors, J. M., Campo, E., Fisher, R. I. et al: The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med, 346: 1937, 2002 20. Takahashi, M., Rhodes, D. R., Furge, K. A., Kanayama, H., Kagawa, S., Haab, B. B. et al: Gene expression profiling of clear cell renal cell carcinoma: gene identification and prognostic classification. Proc Natl Acad Sci USA, 98: 9754, 2001 21. Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A. et al: Molecular portraits of human breast tumours. Nature, 406: 747, 2000 22. Luo, J., Duggan, D. J., Chen, Y., Sauvageot, J., Ewing, C. M., Bittner, M. L. et al: Human prostate cancer and benign prostatic hyperplasia: molecular dissection by gene expression profiling. Cancer Res, 61: 4683, 2001 23. Magee, J. A., Araki, T., Patil, S., Ehrig, T., True, L., Humphrey, P. A. et al: Expression profiling reveals hepsin overexpression in prostate cancer. Cancer Res, 61: 5692, 2001 24. Dhanasekaran, S. M., Barrette, T. R., Ghosh, D., Shah, R., Varambally, S., Kurachi, K. et al: Delineation of prognostic biomarkers in prostate cancer. Nature, 412: 822, 2001 25. Singh, D., Febbo, P. G., Ross, K., Jackson, D. G., Manola, J., Ladd, C. et al: Gene expression correlates of clinical prostate cancer behavior. Cancer Cell, 1: 203, 2002 26. Welsh, J. B., Sapinoso, L. M., Su, A. I., Kern, S. G., Wang-Rodriguez, J., Moskaluk, C. A. et al: Analysis of gene expression identifies candidate markers and pharmacological targets in prostate cancer. Cancer Res, 61: 5974, 2001 27. Luo, J., Dunn, T., Ewing, C., Sauvageot, J., Chen, Y., Trent, J. et al: Gene expression signature of benign prostatic hyperplasia revealed by cDNA microarray analysis. Prostate, 51: 189, 2002
S21
28. Shipp, M.-A., Ross, K.-N., Tamayo, P., Weng, A.-P., Kutok, J.-L., Aguiar, R.-C. T. et al: Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med, 8: 68, 2002 29. van ’t Veer, L. J., Dai, H., van de Vijver, M. J., He, Y. D., Hart, A. A., Mao, M. et al: Gene expression profiling predicts clinical outcome of breast cancer. Nature, 415: 530, 2002 30. Lieberfarb, M. E., Lin, M., Lechpammer, M., Li, C., Tanenbaum, D. M., Febbo, P. G. et al: Genome-wide loss of heterozygosity analysis from laser capture microdissected prostate cancer using single nucleotide polymorphic allele (SNP) arrays and a novel bioinformatics platform dChipSNP. Cancer Res, 63: 4781, 2003 31. Rhodes, D. R., Barrette, T. R., Rubin, M. A., Ghosh, D. and Chinnaiyan, A. M.: Meta-analysis of microarrays: interstudy validation of gene expression profiles reveals pathway dysregulation in prostate cancer. Cancer Res, 62: 4427, 2002 32. Rubin, M. A., Varambally, S., Beroukhim, R., Tomlins, S. A., Rhodes, D. R., Paris, P. L. et al: Overexpression, amplification, and androgen regulation of TPD52 in prostate cancer. Cancer Res, 64: 3814, 2004 33. Paweletz, C. P., Charboneau, L., Bichsel, V. E., Simone, N. L., Chen, T., Gillespie, J. W. et al: Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene, 20: 1981, 2001 34. Petricoin, E. F., 3rd, Ornstein, D. K., Paweletz, C. P., Ardekani, A., Hackett, P. S., Hitt, B. A. et al: Serum proteomic patterns for detection of prostate cancer. J Natl Cancer Inst, 94: 1576, 2002 35. Yasui, Y., Pepe, M., Thompson, M. L., Adam, B. L., Wright, G. L., Jr., Qu, Y. et al: A data-analytic strategy for protein biomarker discovery: profiling of high-dimensional proteomic data for cancer detection. Biostatistics, 4: 449, 2003 36. Simone, N. L., Remaley, A. T., Charboneau, L., Petricoin, E. F., 3rd, Glickman, J. W., Emmert-Buck, M. R. et al: Sensitive immunoassay of tissue cell proteins procured by laser capture microdissection. Am J Pathol, 156: 445, 2000 37. Rhodes, D. R., Yu, J., Shanker, K., Deshpande, N., Varambally, R., Ghosh, D. et al: ONCOMINE: A Cancer Microarray Database and Integrated Data-Mining Platform. Neoplasia, 6: 1, 2004 38. Lin, M., Wei, L. J., Sellers, W. R., Lieberfarb, M., Wong, W. H. and Li, C.: dChipSNP: significance curve and clustering of SNP-array-based loss-of-heterozygosity data. Bioinformatics, 2004 39. Harisinghani, M. G., Barentsz, J., Hahn, P. F., Deserno, W. M., Tabatabaei, S., van de Kaa, C. H. et al: Noninvasive detection of clinically occult lymph-node metastases in prostate cancer. N Engl J Med, 348: 2491, 2003
DISCUSSION Dr. Philip W. Kantoff. What is the heterogeneity in terms of the pathology interpretation of something like HER-2/neu, which is clinically used in breast cancer? Dr. Mark A. Rubin. From my perspective it is a mistake to think that you can have a Food and Drug Administration approved test that can be evaluated by pathologists in a uniform way to determine whether it is positive or negative. HER-2/neu may be able to distinguish among borderline cases at academic centers that specialize in it but at many centers there is a lack of reproducibility. Dr. Neal Rosen. Most true positives with enormous levels of over expression do not respond for biological reasons that we do not understand. Most of the time we get false-positives. Doctor Kantoff. There are some differences in classification among institutions. Doctor Rosen. I agree. The use of molecular biomarkers can be done pretty well at power centers but not at smaller places. I understand the biological interest but what is the clinical utility to being able to find tiny occult metastatic disease with imaging techniques? Doctor Rubin. I think that the important point would be to develop the ability to image actual tumor size, growth and change with time. Dr. Anthony V. D’Amico. The clinical utility is to assess response to therapy. In fact, you are going to have micrometastatic disease in patients that you never would have found with standard radiological measures. When you start to find micrometastatic disease with more sophisticated measures, you do not know if you should
S22
USING MOLECULAR MARKERS TO PREDICT OUTCOME
change management based on it because you do not have prospective data to assess whether or not changing management in that setting based on this more sophisticated scan actually matters. However, to assess response to treatment would actually be useful because you could then see if the particular treatment you are using is eradicating micrometastatic disease. Doctor Rosen. Even that is a research question. If it is not eradicating micrometastatic disease, you would have to understand how the pattern of response, as seen on image, is correlated with clinical outcome. It is possible that stasis or even growth to a certain degree of micrometastatic disease will be associated with response in the long run. Once we can image micrometastatic disease, we do not know which of those tumors would progress to gross metastases with time. Also, the stem cell story would suggest that gross response, that is the tumor cells that you see going away, are not the ones that kill you. Doctor Rubin. If you knew that a patient had a tiny cancer and you could follow it at 6-month intervals with and without treatment, and see that it does not progress, that would be a tremendously useful tool. Dr. Joel B. Nelson. How many of the small indolent primary tumors would you need to have before you could identify genotypes, that is say that 1 type of tumor was genetically different from another? Doctor Rubin. That is why we start by looking at populations and seeing how they progress. Some small tumors might do worse than a large tumor. You have to study the right cohorts, which we are just now learning to do. We have selected our local tumors and we can certainly say things about PSA failure, but we do not have the important, hard end points to do the supervised-type analysis that you are describing. Doctor Nelson. We have always been surprised at how few cases you actually need to identify some of the important differences between phenotypes. If you had the right cases, it could be done with very few numbers. Doctor Rubin. You want to look for the cases that are actually going to differentiate the Gleason 6 case that does not progress from the Gleason 6 that does progress. Doctor Nelson. What we do not have are identical cases in which some tumors progress and the patients die and others do not. Since the difference is not Gleason score, stage or PSA, let’s look at the molecular profile. We do not have that cohort to study, and so we can only conjecture at this point. Doctor Kantoff. Dr. William Sellers took a group of prostate tumors from patients with early disease and performed SNP profiling on them. Using a nonsupervised learning model I believe that 4 subsets of genetic prostate cancer were found. Doctor Nelson. I am concerned that when you are looking at 100,000 genes and there are maybe 50 different types of prostate cancer, we are not sophisticated enough as clinicians to be able to judge which patient should receive hormones and radiation vs some other therapy. Doctor Rubin. There are different approaches. Using a global approach you can look at a thousand genes. However, if you look at individual genes, you might understand the biology and that is an important approach as well.
Vol. 172, S18 –S22, November 2004 Printed in U.S.A.
DOI: 10.1097/01.ju.0000142448.58831.d9
USING MOLECULAR MARKERS TO PREDICT OUTCOME MARK A. RUBIN* From the Department of Pathology, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts
ABSTRACT
Purpose: Developing molecular tests to predict prostate cancer progression requires first defining meaningful clinical end points and defining strategies to take advantage of emerging technology. Materials and Methods: The select relevant literature was reviewed concerning clinical trials, clinical prostate cancer nomograms, molecular biomarker development and molecular prostate cancer imaging. Results: There is controversy regarding the use of prostate specific antigen or biochemical failure following prostatectomy or radiation therapy for clinically localized prostate cancer as a marker of progression. As a consequence, advances in prostate cancer biomarker development may require using population based cohorts or cases from clinical trials to identify meaningful associations. Whereas the discovery of novel candidate biomarkers was slow 5 to 10 years ago and often resulted from serendipity, advances in high throughput technologies have led to the identification of a large number of candidate genes. Strategies to identify candidate genes include the use of expression array analysis, single nucleotide polymorphism arrays (single nucleotide polymorphism chips), proteomics and bioinformatics. Monitoring the progression of prostate cancer has been limited to standard approaches such as computerized tomography or magnetic resonance imaging, which in general do not delineate the extent of disease. By carefully selecting novel prostate cancer biomarkers future work should allow in vivo monitoring of prostate cancer. This will represent a revolutionary advance in our ability to monitor prostate cancer progression and ultimately it may be one of the most important applications of cancer biomarkers. Conclusions: Emerging technology should allow us to analyze clinical prostate cancer trials with sufficient followup to help develop meaningful markers of prostate cancer progression. KEY WORDS: prostate; prostatic neoplasms; gene expression; tumor markers, biological; polymorphism, single nucleotide
Indolent prostate carcinoma exists but we are unable to diagnose it in a practical manner. Some colonic adenomas progress with time into cancers and even at the earliest phases they share molecular alterations identical to those of cancer, although they are not classified as carcinoma. One important difference between colonic adenoma and indolent prostate cancer is that, whereas adenoma can be resected from the colonic mucosa and followed by endoscopy, the prostate tumor is not accessible to continuous monitoring. The terms expectant therapy or watchful waiting have been coined for the process of monitoring men with biopsy proven prostate cancer. However, this strategy leaves the patient and doctor living with the uncertainty of how the cancer will progress with time. Are they following an indolent tumor? Are they sitting on a large aggressive tumor that has been only partially sampled? Using a combination of clinical, pathological and molecular modalities, we should be able to distinguish indolent from aggressive prostate cancer. DEFINING INDOLENT PROSTATE CANCER REQUIRES STUDYING THE CORRECT POPULATION
In the United States prostate cancer remains the most common solid tumor malignancy in men with approximately 200,000 men diagnosed yearly. A total of 30,000 men will die yearly in the United States due to prostate cancer. Epidemiological data suggest a tremendous disparity between men
diagnosed with clinically localized prostate cancer and those presenting with advanced metastatic disease with 100% and less than 50% 5-year survival rates, respectively.1, 2 Many investigators believe that the improvement in 5-year survival has less to do with improved treatment and more to do with early detection since the introduction of the serum blood test, prostate specific antigen (PSA), in 1987. Increasing the time from initial cancer diagnosis to clinical presentation due to symptoms by widespread serum PSA screening, as in the United States, definitely improves 5-year survival statistics. As a result, some aggressive cancers are detected earlier but of equal concern are the significantly larger numbers of indolent tumors detected through PSA screening. These indolent tumors may not have posed any risk of clinical morbidity or mortality but, nonetheless, they are treated in a similar manner.3 Therefore, an overall goal is to develop new molecular tests to improve our ability to detect significant prostate cancers. A hurdle to discovery is that we are testing biomarkers against patient populations with insufficient followup. In the United States most surgical series describe low risk patient populations identified by PSA screening. Therefore, given a combination of PSA lead time bias and short followup, most series have few significant events and focus on PSA recurrence (ie biochemical failure) as the study end point. In fact, it has been suggested that PSA recurrence may not be the most relevant end point to determine clinically relevant events such as the development of metastatic disease and disease specific death.4, 5 Therefore, given the limitations of populations identified using PSA screening, 3 examples of cohorts that should be extremely useful in characterizing the molecular progression
Supported by SPORE National Cancer Institute Grants P50CA90381 (MAR) and P50CA69568 (MAR), and National Cancer Institute Grants CA97063 (MAR) and R01AG21404 (MAR). * Correspondence and requests for reprints: Department of Pathology, Brigham and Women’s Hospital/Harvard Medical School, 75 Francis St., Boston, Massachusetts 02115 (telephone: 617-525-6747; FAX: 617-264-5169; e-mail: [email protected]). S18
USING MOLECULAR MARKERS TO PREDICT OUTCOME
of prostate cancer will be briefly described. Two of these trials come from Scandinavia6 –9 and the third is from the Physicians’ Health Study.10 The first study is the randomized Scandinavian trial evaluating the benefit of prostatectomy over watchful waiting.6 Approximately 700 men were randomized into a surgery arm or a watchful waiting arm. Surgery was seen to decrease significantly the incidence of metastatic prostate cancer and cancer specific death by 50% at 8-year followup. Longer followup may demonstrate even more dramatic benefits. However, the study also determined that 17 men needed surgery to prevent 1 prostate cancer related death. The implication that a significant number of men do not benefit from treatment is a clear message that should not change dramatically. A second study with significantly longer followup is the watchful waiting trial performed in the Orebro region of Sweden.7, 8 This population based study follows men diagnosed with localized prostate cancer in the pre-PSA era between 1977 and 1984 in a well-defined catchment area of 190,000 men living in south central Sweden. Study results at 10 and 15 years suggested that there is minimal benefit from localized treatment (ie surgery or external beam radiation). A recent report of Johansson et al on the 20-year follow up of these patients demonstrated some surprising late events.9 From approximately 200 men on this watchful waiting cohort 31 with prostate cancer had metastatic disease, 27 died of prostate cancer, 113 died of other causes and 27 were still alive. When averaged during the first 15 years, the rate of progression to metastatic disease was 18/1,000 person-years and the prostate cancer mortality rate was 15/1,000 personyears. In contrast, an approximately 3-fold higher rate was found for progression and death during followup between years 15 and 20.9 Lastly, the Physicians’ Health Study is a population based study done at the Harvard School of Public Health, and Brigham and Women’s Hospital that has followed more than 20,000 male physicians since 1982.10 Most men were enrolled into the study while in their early 40s. Detailed dietary information was gathered, as were serum samples. To date approximately more than 1,700 participants are estimated to have prostate cancer. Future work will be able to focus on serum and tissue biomarkers that help determine which men had prostate cancer during the course of this study. Therefore, in studies evaluating the natural history of prostate cancer the epidemiological data are complex, suggesting that localized prostate cancer is being dramatically over treated. However, even apparently indolent tumors may progress at a long interval following initial diagnosis. These observations make the characterization of pure indolent tumors more challenging and suggest that the use of outcomes studies using surrogate end points, such as PSA biochemical recurrence, may not be sufficient. LIMITATIONS OF PATHOLOGICAL FINDINGS AND CLINICAL NOMOGRAMS
Urologists are most familiar with the pretreatment nomograms that have been developed and validated to predict prostate cancer recurrence after treatment for localized disease.11–14 These nomograms classically consider serum PSA levels, prostate needle biopsy Gleason score and clinical stage. At best these clinical nomograms have an area under the ROC curve of 0.75. Kattan et al recently noted that the addition of 2 novel serum markers (interleukin-6SR and transforming growth factor-1) could improve the area under the ROC curve to 0.83, suggesting that molecular markers combined with clinical parameters could improve existing predictive nomograms.15 Therefore, to date the best clinical nomograms can predict which men may benefit from surgery. However, these nomograms do not address if we are over treating these men. A reason that current clinical nomo-
S19
grams are of limited value is that we still cannot determine the volume of high grade prostate cancer that is present before treatment.16 In the absence of endoscopy for the prostate we need to develop molecular tools to help see into the prostate gland and determine the extent of aggressive disease. STRATEGIES FOR THE DEVELOPMENT OF MOLECULAR BIOMARKERS
If the goal is to create a molecular definition to help distinguish indolent from aggressive prostate cancer, the development of the appropriate biomarkers is needed. If we define a biomarker as a molecular test that provides additional information over current clinical data (including pathological review), we are currently lacking any bona fide prostate cancer biomarkers to define indolent disease. There is a massive ongoing effort in this post-genomic era to discover novel cancer biomarkers to help solve this problem. Expression array analysis. Recent advances in genomic and proteomic technologies suggest that molecular signatures of disease can be used to diagnose,17, 18 predict survival19, 20 and define novel molecular subtypes of disease.21 Several studies have used expression array analysis to characterize the gene expression profiles of prostate cancer compared with benign prostate disease and normal prostate tissue.22–27 Several interesting candidate genes have been consistently identified, including AMACR, fatty acid synthase, PIM-1 kinase, hepsin and EZH2. Future work in the area of cDNA expression array analysis may allow for its use in the clinical setting. The use of expression array analysis to predict clinical outcome has been attempted in the area of lymphoma,28 breast29 and prostate24 cancer. However, there are no robust clinical models that can be applied today. Several prospective trials have expression array analysis built into their design but beyond these proof of principle studies expression array analysis has been more of a biomarker discovery tool. Single nucleotide polymorphism (SNP) analysis. In a recent study SNPs in prostate cancer have begun to be explored using Affymetrix SNP arrays (Affymetrix, Santa Clara, California).30 A group recently reported the use of these SNP chips to evaluate loss of heterozygosity (LOH) by comparing tumor and normal samples.30 This initial study was performed using SNP chips with 1.5K SNPs. The current chip has 11.5K SNPs and in 2004 SNP chips will be available with more than 120K SNPs. There are several exciting applications of this technology for biomarker development. These high resolution SNP chips can detect LOH, deletions and amplification at the DNA level in 120,000 known SNPs. Therefore, novel tumor classes may become evident, as has previously been seen with cDNA expression array analysis using various unsupervised and supervised clustering methods. As an example, in the quest for characterizing indolent tumors that do not progress our research group is using this technology to test clinical samples from the Orebro Watchful Waiting trial, as described.7, 8 SNP analysis also allows the development of individual biomarkers. For example, gains in the long arm of chromosome 8 (8q) are believed to be associated with poor outcome and the development of hormone refractory prostate cancer. Based on a meta-analysis of gene expression microarray data from multiple prostate cancer studies31 a candidate oncogene, tumor protein D52 (TPD52), was identified in the 8q21 amplicon. TPD52 is a coiled coil, motif bearing protein that is potentially involved in vesicle trafficking. The mRNA and protein levels of TPD52 were highly elevated in prostate cancer tissues using tissue microarrays. Amplification analysis using SNP arrays demonstrated increased DNA copy number in the region encompassing TPD52, which was confirmed by fluorescence in situ hybridization.32 These findings suggest that dysregulation of
S20
USING MOLECULAR MARKERS TO PREDICT OUTCOME
TPD52 by genomic amplification may have a role in prostate cancer progression and it is an example of combining several technologies to characterize molecular biomarkers. Proteomics. Recent work has concentrated on performing proteomics to detect prostate cancer specific fingerprints.33–36 For example, Petricoin et al identified unique profiles seen in serum from patients with prostate cancer but not controls.34 This work promises to identify proteins that may be used for prognosis and diagnosis of prostate cancer. Currently several proteomic approaches are being used, including 2-dimensional electrophoresis and surface enhanced laser desorption ionization-time of flight proteomics.17 A limitation to many of these techniques is that proteins are difficult to characterize from this analysis. Another important criticism is that the majority of proteins identified in the serum will not be from the tumor, making analysis of the numerous peaks extremely difficult. However, it is still feasible that a serum test could be developed without knowing what the specific candidate proteins are and, nonetheless, be clinically helpful in distinguishing indolent from aggressive prostate cancer. For example, serum samples from the Physicians’ Health Study are being evaluated to identify profiles of men who had prostate cancer during the longterm year followup of this study.10 Bioinformatics. Informatics tools have an important role in the discovery process. A rapidly emerging field, bioinformatics, is starting to alter the way research is being performed. Using information from large databases in silico studies can be done to discover and validate new candidate genes and pathways significant in areas such as the development of prostate cancer. For example, Rhodes et al recently identified lists of significant prostate cancer-related genes by performing a meta-analysis on publicly available cDNA expression array datasets.31 This study was also able to extrapolate prostate cancer related pathways by piecing together data from multiple studies. This approach has now become available on an Internet based website called ONCOMINE (www.oncomine.org), which allows user to perform a meta-analysis of genes of interest and contains links to other websites that provide information regarding their genes of interest.37 In addition to currently existing bioinformatics tools for microarray data, such as cluster, TreeView (http://taxonomy.zoology.gla.ac.uk/rod/treeview/treeview_ manual.html) and dCHIP (http://biosun1.harvard.edu/complab/ dchip/install.htm), new tools are constantly being developed (http://chip.dfci.harvard.edu/). For example, Lin et al are developing tools to perform hierarchical clustering of SNP data, amplification analysis and LOH analysis.38 MOLECULAR IMAGING
Even if elegant molecular profiles for prostate cancer progression are developed, could we use this information to impact prostate cancer treatment? The answer will depend on whether the expression of these biomarkers is readily detected in serum or in prostate biopsy samples. For this to be reliable the development of markers expressed in a homogenous manner would be preferable for the development of an indolent molecular profile. Perhaps the greatest impact on defining indolent prostate cancer will come in the area of molecular imaging. High resolution imaging will continually improve our ability to view anatomical structures but correlations with molecular imaging may provide a way of monitoring prostate cancer progression. For example, a group recently reported that highly lymphotropic superparamagnetic nanoparticles can gain access to lymph nodes and be visualized at extremely high resolution, sufficient to detect displacement by metastatic prostate cancer.38 A limitation of this technology is that it targets host macrophages and is not cancer specific. Other ongoing work targets prostate cancer specific markers, which are located on the tumor cell surface. An example is hepsin, a serine protease, which is known to be
expressed early in prostate cancer development.24 Hepsin, which is expressed by other organs including the liver, would be a poor serum marker but it may turn out to be a powerful molecular beacon. Therefore, with the correct tools molecular imaging could have the greatest impact in following prostate cancers with time, thus, helping stage and classify tumors as indolent and aggressive based on progression and invasiveness. CONCLUSIONS
Briefly, the goal of predicting prostate cancer progression, that is distinguishing indolent from aggressive disease, within the next 5 to 10 years is palpable. Even if this cannot be defined from a pure biological standpoint, molecular imaging or the development of serum tests may make monitoring of men with biopsy proven prostate cancer feasible. The development of these tests will require evaluation of the correct clinical cohorts. REFERENCES
1. Chu, K. C., Tarone, R. E. and Freeman, H. P.: Trends in prostate cancer mortality among black men and white men in the United States. Cancer, 97: 1507, 2003 2. Weir, H. K., Thun, M. J., Hankey, B. F., Ries, L. A., Howe, H. L., Wingo, P. A. et al: Annual report to the nation on the status of cancer, 1975–2000, featuring the uses of surveillance data for cancer prevention and control. J Natl Cancer Inst, 95: 1276, 2003 3. Etzioni, R., Penson, D. F., Legler, J. M., Di Tommaso, D., Boer, R., Gann, P. H. et al: Overdiagnosis due to prostate-specific antigen screening: lessons from u.s. Prostate cancer incidence trends. J Natl Cancer Inst, 94: 981, 2002 4. Pound, C. R., Partin, A. W., Eisenberger, M. A., Chan, D. W., Pearson, J. D. and Walsh, P. C.: Natural history of progression after PSA elevation following radical prostatectomy. JAMA, 281: 1591, 1999 5. D’Amico, A. V., Moul, J. W., Carroll, P. R., Sun, L., Lubeck, D. and Chen, M. H.: Surrogate end point for prostate cancerspecific mortality after radical prostatectomy or radiation therapy. J Natl Cancer Inst, 95: 1376, 2003 6. Holmberg, L., Bill-Axelson, A., Helgesen, F., Salo, J., Folmerz, P., Ha¨gmann, M. et al: A randomized trial comparing radical prostatectomy with watchful waiting in early prostate cancer. N Engl J Med, 347: 781, 2002 7. Johansson, J. E., Holmberg, L., Johansson, S., Bergstrom, R. and Adami, H. O.: Fifteen-year survival in prostate cancer. A prospective, population-based study in Sweden. JAMA, 277: 467, 1997 8. Johansson, J. E., Adami, H. O., Andersson, S. O., Bergstrom, R., Holmberg, L. and Krusemo, U. B.: High 10-year survival rate in patients with early, untreated prostatic cancer. JAMA, 267: 2191, 1992 9. Johansson, J. E., Andren, O., Andersson, S. O., Dickman, P. W., Holmberg, L., Magnuson, A. et al: Natural history of early, localized prostate cancer. JAMA, 291: 2713, 2004 10. Manson, J. E., Buring, J. E., Satterfield, S. and Hennekens, C. H.: Baseline characteristics of participants in the Physicians’ Health Study: a randomized trial of aspirin and betacarotene in U.S. physicians. Am J Prev Med, 7: 150, 1991 11. Partin, A. W., Kattan, M. W., Subong, E. N., Walsh, P. C., Wojno, K. J., Oesterling, J. E. et al: Combination of prostate-specific antigen, clinical stage, and Gleason score to predict pathological stage of localized prostate cancer. A multi-institutional update. JAMA, 277: 1445, 1997 12. Kattan, M. W., Eastham, J. A., Stapleton, A. M., Wheeler, T. M. and Scardino, P. T.: A preoperative nomogram for disease recurrence following radical prostatectomy for prostate cancer. J Natl Cancer Inst, 90: 766, 1998 13. Kattan, M. W., Zelefsky, M. J., Kupelian, P. A., Scardino, P. T., Fuks, Z. and Leibel, S. A.: Pretreatment nomogram for predicting the outcome of three-dimensional conformal radiotherapy in prostate cancer. J Clin Oncol, 18: 3352, 2000 14. Kattan, M. W., Potters, L., Blasko, J. C., Beyer, D. C., Fearn, P., Cavanagh, W. et al: Pretreatment nomogram for predicting freedom from recurrence after permanent prostate brachytherapy in prostate cancer. Urology, 58: 393, 2001
USING MOLECULAR MARKERS TO PREDICT OUTCOME 15. Kattan, M. W., Shariat, S. F., Andrews, B., Zhu, K., Canto, E., Matsumoto, K. et al: The addition of interleukin-6 soluble receptor and transforming growth factor beta1 improves a preoperative nomogram for predicting biochemical progression in patients with clinically localized prostate cancer. J Clin Oncol, 21: 3573, 2003 16. Rubin, M. A., Mucci, N. R., Manley, S., Sanda, M., Cushenberry, E., Strawderman, M. et al: Predictors of Gleason pattern 4/5 prostate cancer on prostatectomy specimens: can high grade tumor be predicted preoperatively? J Urol, 165: 114, 2001 17. Adam, B. L., Qu, Y., Davis, J. W., Ward, M. D., Clements, M. A., Cazares, L. H. et al: Serum protein fingerprinting coupled with a pattern-matching algorithm distinguishes prostate cancer from benign prostate hyperplasia and healthy men. Cancer Res, 62: 3609, 2002 18. Golub, T. R., Slonim, D. K., Tamayo, P., Huard, C., Gaasenbeek, M., Mesirov, J. P. et al: Molecular classification of cancer: class discovery and class prediction by gene expression monitoring. Science, 286: 531, 1999 19. Rosenwald, A., Wright, G., Chan, W. C., Connors, J. M., Campo, E., Fisher, R. I. et al: The use of molecular profiling to predict survival after chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med, 346: 1937, 2002 20. Takahashi, M., Rhodes, D. R., Furge, K. A., Kanayama, H., Kagawa, S., Haab, B. B. et al: Gene expression profiling of clear cell renal cell carcinoma: gene identification and prognostic classification. Proc Natl Acad Sci USA, 98: 9754, 2001 21. Perou, C. M., Sorlie, T., Eisen, M. B., van de Rijn, M., Jeffrey, S. S., Rees, C. A. et al: Molecular portraits of human breast tumours. Nature, 406: 747, 2000 22. Luo, J., Duggan, D. J., Chen, Y., Sauvageot, J., Ewing, C. M., Bittner, M. L. et al: Human prostate cancer and benign prostatic hyperplasia: molecular dissection by gene expression profiling. Cancer Res, 61: 4683, 2001 23. Magee, J. A., Araki, T., Patil, S., Ehrig, T., True, L., Humphrey, P. A. et al: Expression profiling reveals hepsin overexpression in prostate cancer. Cancer Res, 61: 5692, 2001 24. Dhanasekaran, S. M., Barrette, T. R., Ghosh, D., Shah, R., Varambally, S., Kurachi, K. et al: Delineation of prognostic biomarkers in prostate cancer. Nature, 412: 822, 2001 25. Singh, D., Febbo, P. G., Ross, K., Jackson, D. G., Manola, J., Ladd, C. et al: Gene expression correlates of clinical prostate cancer behavior. Cancer Cell, 1: 203, 2002 26. Welsh, J. B., Sapinoso, L. M., Su, A. I., Kern, S. G., Wang-Rodriguez, J., Moskaluk, C. A. et al: Analysis of gene expression identifies candidate markers and pharmacological targets in prostate cancer. Cancer Res, 61: 5974, 2001 27. Luo, J., Dunn, T., Ewing, C., Sauvageot, J., Chen, Y., Trent, J. et al: Gene expression signature of benign prostatic hyperplasia revealed by cDNA microarray analysis. Prostate, 51: 189, 2002
S21
28. Shipp, M.-A., Ross, K.-N., Tamayo, P., Weng, A.-P., Kutok, J.-L., Aguiar, R.-C. T. et al: Diffuse large B-cell lymphoma outcome prediction by gene-expression profiling and supervised machine learning. Nat Med, 8: 68, 2002 29. van ’t Veer, L. J., Dai, H., van de Vijver, M. J., He, Y. D., Hart, A. A., Mao, M. et al: Gene expression profiling predicts clinical outcome of breast cancer. Nature, 415: 530, 2002 30. Lieberfarb, M. E., Lin, M., Lechpammer, M., Li, C., Tanenbaum, D. M., Febbo, P. G. et al: Genome-wide loss of heterozygosity analysis from laser capture microdissected prostate cancer using single nucleotide polymorphic allele (SNP) arrays and a novel bioinformatics platform dChipSNP. Cancer Res, 63: 4781, 2003 31. Rhodes, D. R., Barrette, T. R., Rubin, M. A., Ghosh, D. and Chinnaiyan, A. M.: Meta-analysis of microarrays: interstudy validation of gene expression profiles reveals pathway dysregulation in prostate cancer. Cancer Res, 62: 4427, 2002 32. Rubin, M. A., Varambally, S., Beroukhim, R., Tomlins, S. A., Rhodes, D. R., Paris, P. L. et al: Overexpression, amplification, and androgen regulation of TPD52 in prostate cancer. Cancer Res, 64: 3814, 2004 33. Paweletz, C. P., Charboneau, L., Bichsel, V. E., Simone, N. L., Chen, T., Gillespie, J. W. et al: Reverse phase protein microarrays which capture disease progression show activation of pro-survival pathways at the cancer invasion front. Oncogene, 20: 1981, 2001 34. Petricoin, E. F., 3rd, Ornstein, D. K., Paweletz, C. P., Ardekani, A., Hackett, P. S., Hitt, B. A. et al: Serum proteomic patterns for detection of prostate cancer. J Natl Cancer Inst, 94: 1576, 2002 35. Yasui, Y., Pepe, M., Thompson, M. L., Adam, B. L., Wright, G. L., Jr., Qu, Y. et al: A data-analytic strategy for protein biomarker discovery: profiling of high-dimensional proteomic data for cancer detection. Biostatistics, 4: 449, 2003 36. Simone, N. L., Remaley, A. T., Charboneau, L., Petricoin, E. F., 3rd, Glickman, J. W., Emmert-Buck, M. R. et al: Sensitive immunoassay of tissue cell proteins procured by laser capture microdissection. Am J Pathol, 156: 445, 2000 37. Rhodes, D. R., Yu, J., Shanker, K., Deshpande, N., Varambally, R., Ghosh, D. et al: ONCOMINE: A Cancer Microarray Database and Integrated Data-Mining Platform. Neoplasia, 6: 1, 2004 38. Lin, M., Wei, L. J., Sellers, W. R., Lieberfarb, M., Wong, W. H. and Li, C.: dChipSNP: significance curve and clustering of SNP-array-based loss-of-heterozygosity data. Bioinformatics, 2004 39. Harisinghani, M. G., Barentsz, J., Hahn, P. F., Deserno, W. M., Tabatabaei, S., van de Kaa, C. H. et al: Noninvasive detection of clinically occult lymph-node metastases in prostate cancer. N Engl J Med, 348: 2491, 2003
DISCUSSION Dr. Philip W. Kantoff. What is the heterogeneity in terms of the pathology interpretation of something like HER-2/neu, which is clinically used in breast cancer? Dr. Mark A. Rubin. From my perspective it is a mistake to think that you can have a Food and Drug Administration approved test that can be evaluated by pathologists in a uniform way to determine whether it is positive or negative. HER-2/neu may be able to distinguish among borderline cases at academic centers that specialize in it but at many centers there is a lack of reproducibility. Dr. Neal Rosen. Most true positives with enormous levels of over expression do not respond for biological reasons that we do not understand. Most of the time we get false-positives. Doctor Kantoff. There are some differences in classification among institutions. Doctor Rosen. I agree. The use of molecular biomarkers can be done pretty well at power centers but not at smaller places. I understand the biological interest but what is the clinical utility to being able to find tiny occult metastatic disease with imaging techniques? Doctor Rubin. I think that the important point would be to develop the ability to image actual tumor size, growth and change with time. Dr. Anthony V. D’Amico. The clinical utility is to assess response to therapy. In fact, you are going to have micrometastatic disease in patients that you never would have found with standard radiological measures. When you start to find micrometastatic disease with more sophisticated measures, you do not know if you should
S22
USING MOLECULAR MARKERS TO PREDICT OUTCOME
change management based on it because you do not have prospective data to assess whether or not changing management in that setting based on this more sophisticated scan actually matters. However, to assess response to treatment would actually be useful because you could then see if the particular treatment you are using is eradicating micrometastatic disease. Doctor Rosen. Even that is a research question. If it is not eradicating micrometastatic disease, you would have to understand how the pattern of response, as seen on image, is correlated with clinical outcome. It is possible that stasis or even growth to a certain degree of micrometastatic disease will be associated with response in the long run. Once we can image micrometastatic disease, we do not know which of those tumors would progress to gross metastases with time. Also, the stem cell story would suggest that gross response, that is the tumor cells that you see going away, are not the ones that kill you. Doctor Rubin. If you knew that a patient had a tiny cancer and you could follow it at 6-month intervals with and without treatment, and see that it does not progress, that would be a tremendously useful tool. Dr. Joel B. Nelson. How many of the small indolent primary tumors would you need to have before you could identify genotypes, that is say that 1 type of tumor was genetically different from another? Doctor Rubin. That is why we start by looking at populations and seeing how they progress. Some small tumors might do worse than a large tumor. You have to study the right cohorts, which we are just now learning to do. We have selected our local tumors and we can certainly say things about PSA failure, but we do not have the important, hard end points to do the supervised-type analysis that you are describing. Doctor Nelson. We have always been surprised at how few cases you actually need to identify some of the important differences between phenotypes. If you had the right cases, it could be done with very few numbers. Doctor Rubin. You want to look for the cases that are actually going to differentiate the Gleason 6 case that does not progress from the Gleason 6 that does progress. Doctor Nelson. What we do not have are identical cases in which some tumors progress and the patients die and others do not. Since the difference is not Gleason score, stage or PSA, let’s look at the molecular profile. We do not have that cohort to study, and so we can only conjecture at this point. Doctor Kantoff. Dr. William Sellers took a group of prostate tumors from patients with early disease and performed SNP profiling on them. Using a nonsupervised learning model I believe that 4 subsets of genetic prostate cancer were found. Doctor Nelson. I am concerned that when you are looking at 100,000 genes and there are maybe 50 different types of prostate cancer, we are not sophisticated enough as clinicians to be able to judge which patient should receive hormones and radiation vs some other therapy. Doctor Rubin. There are different approaches. Using a global approach you can look at a thousand genes. However, if you look at individual genes, you might understand the biology and that is an important approach as well.