An analysis of the genomic landscape of patients with prostate cancer shows that tumors separated into 3 clusters that aligned with the luminal A, luminal B, and basal subtypes found in breast cancer tumors, according to a presentation during the 2017 American Urological Association Annual Meeting.
Matthew R. Cooperberg, MD, MPH
An analysis of the genomic landscape of patients with prostate cancer shows that tumors separated into 3 clusters that aligned with the luminal A, luminal B, and basal subtypes found in breast cancer tumors, according to a presentation during the 2017 American Urological Association Annual Meeting.1
A comparison of the genomic risk profiles of patients with low-risk prosate cancer who are candidates for active surveillance against higher-risk patients, which was completed in order to characterize the genomics of clinically low-risk prostate cancer, discovered that substantial genomic heterogeneity exists among patients with prostate cancer. "Most low-risk prostate cancers have a clear guideline statement that they should be managed with active surveillance. Yet we know that some low-risk tumors are undersampled. At diagnosis, that there’s a small proportion of clinically low-risk tumors that will progress quickly, and that there are a larger proportion of these tumors that are entirely indolent and never show any signs of growth over the patient’s lifetime,” said Matthew R. Cooperberg, MD, MPH, associate professor of urology and Helen Diller Family Chair in Urology at the University of California, San Francisco (UCSF), School of Medicine.
Although there are several emerging biomarkers that help classify and determine the prognosis of patients with prostate cancer, some of which are already in clinical practice, they typically resect only 1 or a few aspects of the overall tumor biology. This study aimed to generate a more comprehensive description of what genomic expression looks like in clinically low-risk prostate cancer to determine which patients are good candidates for active surveillance.
The study considered 408 biopsies from a cohort of patients in the UCSF urologic oncology database who were considered potentially suitable for active surveillance. These patients had Gleason 3+3, low-volume 3+4, stage ≤T2N0M0 disease, or a prostate-specific antigen (PSA) score ≤10 ng/mL. The biopsy samples were profiled by GenomeDX Biosciences using the Affymetrix human exon microarray to generate RNA expression data. These patients had also undergone surgery, so their pathologic outcomes were made available to investigators.
The biopsies were compared with nearly 1300 cases in the GenomeDX database, which had been previously profiled for clinical decision making; most of the patients had higher-risk features. Their UCSF Cancer of the Prostate Risk Assessment (CAPRA) risk profile was assessed using a point system, looking at the patients’ age at diagnosis (younger or older than 50), PSA at diagnosis, Gleason score of the biopsy (primary/secondary), clinical stage, and percent of the biopsy cores with cancer (less or more than 34%).2Based on the CAPRA score, 75.5% of patients were classified as low risk (score 0-2) and 22.5% were intermediate risk (score 3-5); the reference specimens were mostly higher risk.
Based on 21 prognostic signature risk models, patients were divided into quartiles. Fifty-four percent of patients were in the first quartile, 26% in the second, 15% in the third, and 5% in the fourth. Cooperberg noted that several of the low-risk patients were included in the third and fourth quartiles.
Compared with previous research of other genomic risk scores, the results of the study were consistent with most of the previously published risk scores, according to Cooperberg, with high-risk patients, in particular, lining up with previous research.
“The interesting thing is, if we look at the individual pathways, we see a tremendous amount of heterogeneities, so it is not the case that high-risk [disease] is being consistently driven by only 1 or a few pathways,” he commented. When narrowing the analysis down to only those patients who were Gleason 3+3, Cooperberg noted that there are many patients who have very aggressive genomic expression.
In an unsupervised cluster analysis, allowing for the factoring of the different genomic pathways regardless of any previously published risk algorithm, 3 distinct clusters of genomic expressions emerged, “agnostic to any previously described standard clinical risk assessment,” he said.
Notably, Cooperberg and his colleagues classified these 3 clusters as luminal A, luminal B, and basal subtypes, similar to breast cancer tumors, according to PAM50 testing criteria. “We found patterns that look very much like these breast cancer expression patterns, and we’ve seen similar patterns now in bladder cancer as well,” Cooperberg commented.
When digging further into these subtypes, Cooperberg found that they do not necessarily function as prognostic signifiers, but there is a distinct biology driving the different tumors, he noted. Among Cluster 2, the luminal B-like tumors, these tumors had a slightly higher tumor volume and lower PSAs, whereas Cluster 3, the luminal A-like tumors, had slightly lower tumor volume and higher PSAs (TABLE). According to Cooperberg, these subtypes seem to be driven by androgen response elements in particular. However, he noted that there were no major differences in the clinical risk strati cation between these subtypes.
These classifications in prostate cancer in line with breast cancer subtypes were recently acknowledged in a paper published in JAMA Oncology.3This study found further associations with these subtypes, including increased androgen receptor expression in both luminal subtypes. There was an association with poor outcomes among those with luminal B−like prostate cancer compared with the other subtypes.
“We found substantialand frankly, somewhat surprising—genomic diversity among these cancers, which clinically, histologically, and pathologically have always looked homogenous. I think the prostate cancer cases we are currently seeing across different cohort studies classify in ways that, in many respects, look like other cancer types,” Cooperberg said.
This research, including Cooperberg’s conclusions for this study, may be pushing toward a clinical molecular taxonomy for cancer that goes beyond the organ of origin to a more biologic description of why the cancer originated and how best to treat it. This could help drive further improvements in management decision making, both in terms of who should be treated and how treating urologists can tailor treatment and surveillance protocols.