Researchers Identify Five Molecular Prostate Cancer Subtypes That Could Change Risk Assessment

August 3, 2015

Treatment decisions in prostate cancer are still largely guided by the classical diagnostic parameters, including Gleason scoring, TNM classification, and prostate specific antigen (PSA) levels, despite growing evidence that molecular signatures may be useful for assessing risk.

1-6Identifying patients at risk of progression and untimely death (and who therefore would benefit most from aggressive treatment) remains problematic.1A recent report details a new molecular profiling strategy that divides patients with prostate cancer into five distinct groups and may have important ramifications for developing more personalized treatment plans.1

In the study by Ross-Adams et al published inEBioMedicine, molecular analyses were performed using a discovery cohort (called Cambridge) composed of some 356 fresh-frozen samples from 156 men in the UK, and a previously described validation cohort (Stockholm), composed of some 206 samples from 103 primary prostate cancers with matching germline DNA, and 99 samples of mRNA.1,4Relevant diagnostic parameters, including preoperative and 6-monthly PSA, TNM staging, and Gleason score, were collected for each cohort.

Unlike earlier studies, which used genome-wide copy number (CN) alterations or mRNA transcript profiling (transcriptome analysis) methods, individually, to stratify patients with prostate cancer,2,3investigators in this study used an integrative analysis combining both copy number and gene expression data in an effort to identify so-called genome-wide expression quantitative trait loci (eQTL).1

Using the eQTL features in a joint latent variable framework for the integrated analysis, the researchers identified five distinct molecular subtypes (termed iClusters), each with characteristic patterns of CN alterations and gene expression profiles.1Further, they were able to identify a core set of 100 genes, which had both alteration in CN, and changes in mRNA expression, which could account for these differences in the clustering.

Using survival analysis, it was shown that the five patient subgroups (driven by the set of 100 genes) showed distinct patterns of better or worse prognosis, based on biochemical relapse (BCR) data (for the Cambridge and Stockholm cohorts,P= .0017 andP= .016, respectively, for the two highest versus the two lowest risk groups).

Interestingly, tumors with a poor prognosis (those with Gleason grade 4 or higher +3) were distributed evenly across the clusters (ie, no Gleason score predominated in any one subtype), suggesting that these distinctions were not driven solely by well-established parameters like tumor grade. Although some differences between clusters could be explained by BCR and extracapsular extension, the molecular subtypes could not be distinguished by other known prostate cancer risk factors, thus suggesting that additional biologic detail of potential prognostic significance may be described by these molecular profiles.

Notably, when considering the analysis based on CN alone in the Cambridge discovery cohort, the findings were consistent with previously observed loss of chromosome 8p and gain of chromosome 8q. However, in a survival analysis considering BCR, clustering patients using CN data alone was not significant (logrankP= .063). Similarly, although high levels of intertumor variability were found in multiple established prostate cancer risk genes and candidate genes, when considering mRNA expression data, clustering based on expression data alone was also not predictive of prognosis in survival analysis (logrankP= .11).

Using the integrative analysis of the two data sets in the study (Cambridge and Stockholm) showed that iClusters 2 and 4 had the best outcomes and showed relatively little change in CN or gene expression profiles. By comparison, iClusters 1 and 3 had the worst outcomes and were much more genomically unstable, with extensive CN gain or loss and a large number of genes differentially expressed.

iClusters 1 and 3 were more effective at identifying men at risk for relapse than well-established risk factors such as elevated Gleason score, high PSA level, extra-capsular extension, and positive surgical margins. The 100-gene signature that was identified from the analysis identified patients with poor prognoses at high risk for early relapse, with greater efficiency than other gene signatures.

Overall, the study shows that the integration of CN and transcriptome data could effectively stratify patients with prostate cancer into risk cohorts in two distinct datasets, totaling 259 men. The refined 100-gene set also appeared to be more informative in distinguishing positive versus negative outcomes in patients with prostate cancer compared with previously identified signatures.

Using this gene signature, distinct treatment approaches can be designed for men who were previously stratified into the low, intermediate, or high- risk profiles according to conventional criteria. Early adjuvant therapy immediately following prostatectomy, for example, could be used in a patient with a molecular signature associated with the highest risk. The authors illustrate this potential by describing (in supplemental material) a case of a man falling into the iCluster3 group, with low/intermediate risk disease by conventional criteria (ie, Gleason, PSA, stage), who nonetheless showed early relapse.

References:

1. Ross-Adams H, Lamb AD, Dunning MJ, et al. Integration of copy number and transcriptomics provides risk stratification in prostate cancer: A discovery and validation cohort study. EBioMedicine.2015.http://www.ebiomedicine.com/article/S2352-3964(15)30071-2/pdfAccessed July 31, 2015

2. Sun J, Liu W, Adams TS, et al. DNA copy number alterations in prostate cancers: a combined analysis of published CGH studies.Prostate.2007; 67 (7), 692-700.

3. Gorlov IP, Yang JY, Byun J, et al. How to get the most from microarray data: advice from reverse genomics.BMC Genomics.2014; 15, 223.

4. Liu W, Xie CC, Thomas CY, et al. Genetic markers associated with early cancer-specific mortality following prostatectomy.Cancer.2013; 119 (13): 2405-2412.

5. Taylor BS, Schultz N, Hieronymus H, et al. Integrative genomic profiling of human prostate cancer.Cancer Cell.2010; 18 (1), 11-22.

6. Varambally S, Yu J, Laxman B, et al. 2005 Integrative genomic and proteomic analysis of prostate cancer reveals signatures of metastatic progression.Cancer Cell.2005; 8 (5), 393-406.