PerspectiveMedicine

# Personalized Cancer Diagnostics

See allHide authors and affiliations

Science  02 Dec 2011:
Vol. 334, Issue 6060, pp. 1217-1218
DOI: 10.1126/science.1216427

More than a decade into the age of molecularly targeted cancer therapeutics, most clincal laboratories, which are required to operate under standards established by the U.S. Food and Drug Administration called the Clinical Laboratory Improvement Amendments (CLIA), are still using a one gene–one test approach to molecular diagnostics. For example, such tests are routinely used to screen for mutations in the gene encoding the signaling protein KRAS in colorectal carcinomas, and in the gene encoding the epidermal growth factor receptor in non–small cell carcinomas of the lung. There is a growing need, however, for broader approaches that can identify more rare mutations (e.g., mutations in the ERBB2 and BRAF genes in lung carcinomas) that could have an impact on clinical care. Several CLIA labs have introduced multiplexed screens that cover as many as several hundred mutations across dozens of cancer genes (1, 2). But even these approaches are limited to mutation “hotspots” and, for technical reasons, necessarily favor oncogenes over tumor suppressors. Larger panels of genes based on next-generation sequencing will be introduced by a number of labs in the immediate future; even so, some are asking: Why not sequence the entire genome of each patient's tumor?

Whole-genome sequencing can be used to devise unique tests to detect the recurrence of an individual patient's tumor (3). Sequencing the entire genome of a leukemia uncovered a cryptic fusion gene that prompted a major change in the clinical management of the patient (4). Roychowdhury et al. (5) have now taken the approach one step further, sequencing not only the whole genome, but also the whole exome (the coding regions of the genome) and the whole transcriptome (the transcribed RNAs) of individual tumors in an effort to identify all potentially important anomalies. They show that this “sequence everything” approach can be done in a cost-effective and timely manner, delivering the ultimate in personalized cancer diagnostics and further opening the door to the new era of clinical cancer genomics.

The approach of Roychowdhury et al. focuses on cancer patients with advanced disease and uses a consent process that includes upfront genetic counseling and the option to accept or decline information on incidental genetic findings. Fresh biopsies were collected for whole-genome sequencing of the tumor DNA (5× to 15× coverage), whole-exome sequencing of tumor and matched normal DNA (70× to 100×), and whole-transcriptome sequencing. This combination of approaches allows orthogonal confirmation of the findings. For example, of the four cases presented, one was a metastatic colorectal carcinoma in which both genomic and exomic sequencing data predicted amplification of a region in chromosome 13q that includes the gene encoding cyclin-dependent kinase 8 (CDK8); overexpression of CDK8 was confirmed by the transcriptome sequence. The same tumor harbored a mutation in the NRAS gene that was evident by all three modalities.

The time from biopsy to initial results was streamlined to just 24 days, which is within the time period often required for all of an administered drug to be eliminated from the body in patients transitioning to a clinical trial. The 72 hours of computer processing time needed to assemble and analyze the data were included in this total. With the data being generated so quickly, the issue becomes how best to distill it all for clinical use. Here, Roychowdhury et al. introduce an innovative concept: a multidisciplinary “sequencing tumor board” that includes clinicians, geneticists, pathologists, biologists, bioinformaticians, and bioethicists. The board discusses the findings of each tumor and determines what should be reported (see the figure). Members focus on a list of genes derived from the Sanger Institute's Cancer Gene Census, complemented by the Catalog of Somatic Mutations in Cancer (COSMIC), the genes of the human kinome (the protein kinase genes), and genes known to be targeted in current oncology trials. The results are categorized into three groups: those that may have a direct impact on care of the current cancer, those that may be biologically important but not currently actionable, and those that are of unknown importance. Because the sequencing data are generated in a research lab, mutations deemed important to clinical care require confirmation in a CLIA lab.

Roychowdhury et al. calculated the costs of sequencing and analysis to be $5400 per patient during the study, but indicated that they have since dropped to$3600, which approaches that of some of the multiplexed tests currently offered by CLIA labs and is below the monthly cost for some of the new targeted therapeutics. However, this figure does not include charges for the image-directed biopsy needed to obtain fresh tumor, nor does it reflect the very substantial investment in equipment and personnel needed to establish a high-quality data pipeline suitable for directing patient care.

It was not the goal of the study to prove that massively parallel sequencing of tumors can improve and extend the lives of cancer patients. Nevertheless, it is interesting to look at the cases that were presented. The colorectal carcinoma that was examined had been obtained after the patient was treated with an Aurora kinase (AURKA) inhibitor; interestingly, the tumor showed dual–copy number gain of the gene encoding AURKA, as well as a point mutation. The NRAS mutation found in this tumor would make the patient eligible for ongoing trials of MEK (mitogenactivated protein kinase kinase) inhibitors. Two samples of mouse xenografts derived from metastatic prostate cancers were used in the pilot phase of the project. Both showed the presence of a gene fusion (TMPRSS2-ERG) [making the tumor potentially sensitive to poly(ADP-ribose) polymerase (PARP) inhibitors], deletion of the PTEN gene (making the tumor potentially sensitive to phosphatidylinositol 3-kinase inhibitors), and mutation of TP53, the gene encoding the tumor suppressor p53. One sample also showed amplification of the gene encoding the androgen receptor (AR), for which new antiandrogen compounds might be effective; the other showed elevated expression of PLK1, theoretically targetable with a Polo kinase inhibitor. The fourth tumor described was a melanoma with a HRAS mutation (new in this tumor type) that, again, might be targeted through MEK inhibitors. There was also a rearrangement of CDKN2C for which use of a CDK inhibitor might be invoked.

Treatment outcomes were not provided by the study, and the effectiveness of the “sequence everything” approach remains to be established. One could argue that the actionable alterations identified by the sequencing tumor board in the four presented cases (HRAS and NRAS mutations; loss of PTEN; amplification of AR) could have been detected through assays that are already clinically available, and that the massively parallel sequencing approach is akin to driving a pin with a sledgehammer. But this view goes against the genuine hope that identification of new gene mutations, fusions, and other alterations will lead to more appropriate and effective use of targeted therapeutics. As next-generation sequencing is introduced into the clinical arena, it is important that this be done by groups that have extensive experience in data generation and analysis, and that the results be examined in a multidisciplinary manner so that patients can be advised appropriately. Roychowdhury et al. have shown that it is technically feasible to perform deep sequencing in a clinically relevant time frame. The next critical step is to prove that this approach not only improves patient care, but also makes the most efficient use of available healthcare resources.

View Abstract