The search for CDK4/6 inhibitor biomarkers has been hampered by inappropriate proliferation assays.

CDK4/6 inhibitors are effective at treating advanced HR+ /HER2- breast cancer, however biomarkers that can predict response are urgently needed. We demonstrate here that previous large-scale screens designed to identify which tumour types or genotypes are most sensitive to CDK4/6 inhibitors have misrepresented the responsive cell lines because of a reliance on metabolic proliferation assays. CDK4/6-inhibited cells arrest in G1 but continue to grow in size, thereby producing more mitochondria. We show that this growth obscures the arrest using ATP-based proliferation assays but not if DNA-based assays are used instead. Furthermore, lymphoma lines, previously identified as the most sensitive, simply appear to respond the best using ATP-based assays because they fail to overgrow during the G1 arrest. Similarly, the CDK4/6 inhibitor abemaciclib appears to inhibit proliferation better than palbociclib because it also restricts cellular overgrowth through off-target effects. DepMap analysis of screening data using reliable assay types, demonstrates that palbociclib-sensitive cell types are also sensitive to Cyclin D1, CDK4 and CDK6 knockout/knockdown, whereas the palbociclib-resistant lines are sensitive to Cyclin E1, CDK2 and SKP2 knockout/knockdown. Potential biomarkers of palbociclib-sensitive cells are increased expression of CCND1 and RB1, and reduced expression of CCNE1 and CDKN2A. Probing DepMap with similar data from metabolic assays fails to reveal these associations. Together, this demonstrates why CDK4/6 inhibitors, and any other anti-cancer drugs that arrest the cell cycle but permit continued cell growth, must now be re-screened against a wide-range of cell types using an appropriate proliferation assay. This would help to better inform clinical trials and to identify much needed biomarkers of response.