A CDK2 activity signature predicts outcome in CDK2-low cancers

• Published in: Oncogene 2017-05, Vol.36 (18), p.2491-2502
• Main Authors: McCurdy, S R, Pacal, M, Ahmad, M, Bremner, R
• Format: Article
• Language: English
• Subjects: 38; 38/61; 38/91; 631/337/2019; 631/67/69; Animals; Apoptosis; Brain tumors; Cancer; Cell Biology; Cell cycle; Cell Cycle - genetics; Colon cancer; Colonic Neoplasms - genetics; Colonic Neoplasms - pathology; Colorectal cancer; Cyclin-dependent kinase 2; Cyclin-Dependent Kinase 2 - genetics; Cyclin-dependent kinases; E2F protein; E2F Transcription Factors - genetics; Gene Expression Regulation, Neoplastic - genetics; Genetic aspects; Glioma; Human Genetics; Humans; Internal Medicine; Kidney - pathology; Kidneys; Male; Medicine; Medicine & Public Health; Mice; Mitosis - genetics; Multivariate analysis; Oncology; original-article; Phosphorylation; Physiological aspects; Prostate cancer; Prostatic Neoplasms - genetics; Prostatic Neoplasms - pathology; Protein kinases; Retinoblastoma Protein - genetics; Thyroid cancer; Thyroid Gland - pathology; Transcription; Transcription factors; Transcriptome - genetics; Transcriptomes; Tumors
• ISSN: 0950-9232; 1476-5594
• DOI: 10.1038/onc.2016.409

The role of cyclin-dependent kinase 2 (CDK2) in cancer is controversial. A major hurdle is the availability of tools to easily assess its activity across many samples. Here, we introduce a transcriptional signature to specifically track CDK2 activity. It responds to genetic and chemical perturbations in the CDK-RB-E2F axis, correlates with mitotic rate in vitro and in vivo and reacts rapidly to changes in CDK2 activity during cell cycle progression. We find that CDK2 activity is specifically elevated in human testes, mirroring its critical function in mice, and report very distinct profiles across human cancers. Increased CDK2 activity decreases risk in colon cancer, but elevates poor outcome two- to fivefold in specific tumors, including low grade glioma, kidney, thyroid, adrenocortical and prostate cancer. These are typically ‘CDK2-low’ cancers, suggesting that above a certain threshold CDK2 promotes progression, but further increases do not influence outcome. Multivariate analysis revealed that the CDK2 signature is the most important predictive feature in these cancers versus dozens of other clinical parameters, such as tumor grade or mitotic index. Thus, transcriptome data provides a novel, straightforward method to monitor CDK2 activity, implicates key roles for the kinase in a subset of human tissues and tumors and enhances cancer risk prediction. The strategy used here for CDK2 could be applied to other kinases that influence transcription.