CDK 4 phosphorylation status and a linked gene expression profile predict sensitivity to palbociclib

Abstract

Cyclin D‐ CDK 4/6 are the first CDK complexes to be activated in the G1 phase in response to oncogenic pathways. The specific CDK 4/6 inhibitor PD 0332991 (palbociclib) was recently approved by the FDA and EMA for treatment of advanced ER ‐positive breast tumors. Unfortunately, no reliable predictive tools are available for identifying potentially responsive or insensitive tumors. We had shown that the activating T172 phosphorylation of CDK 4 is the central rate‐limiting event that initiates the cell cycle decision and signals the presence of active CDK 4. Here, we report that the profile of post‐translational modification including T172 phosphorylation of CDK 4 differs among breast tumors and associates with their subtypes and risk. A gene expression signature faithfully predicted CDK 4 modification profiles in tumors and cell lines. Moreover, in breast cancer cell lines, the CDK 4 T172 phosphorylation best correlated with sensitivity to PD 0332991. This gene expression signature identifies tumors that are unlikely to respond to CDK 4/6 inhibitors and could help to select a subset of patients with HER 2‐positive and basal‐like tumors for clinical studies on this class of drugs. image Biomarkers able to predict breast cancer sensitivity to the newly FDA / EMA ‐approved CDK 4/6 inhibitor palbociclib are still lacking. Based on the T172‐phosphorylation of CDK 4 as critical to the kinase activation, an 11‐gene classifier helps to segregate cancer subtypes and predict palbociclib sensitivity. In frozen breast cancers samples ( n  = 56), three distinct CDK 4 modification profiles are observed using 2D‐gel electrophoresis: profile H, preponderant presence of T172‐phosphorylated CDK 4 (most luminal B and HER 2‐positive cancers); profile L, minor presence of phosphorylated CDK 4 (90% of lower risk luminal A tumors); profile A, no phosphorylated CDK 4 despite high proliferation index (60% of triple‐negative cancers). An 11‐gene classifier that correctly predicts the three CDK 4 modification profiles was generated and optimized using these breast tumors. In 4,000 breast cancer samples, the CDK 4 modification profiles as predicted by the 11‐gene classifier are associated with subtype, grade, relapse status and risk profile. The 11‐gene classifier correctly predicts the observed presence or absence of CDK 4 phosphorylation in 24 out of 25 breast cancer cell lines and the observed or reported sensitivity to palbociclib in 50 out of 52 breast cancer cell lines. Once transposed to a qPCR assay compatible with formalin‐fixed breast tumor samples, this classifier should predict the potential sensitivity or resistance to CDK 4 inhibitors and the risk and grade of tumors.