Introduction: Clinical trials of palbociclib in combination with endocrine therapy have recently shown unprecedented activity for the treatment of hormone receptor positive/HER2 negative (HR+/HER2neg) advanced breast cancer. However, de novo and acquired resistance to palbociclib limit its clinical utility. Moreover, combining palbociclib with endocrine therapy increases toxicity and costs of the treatment. Identifying patients more likely to benefit from this compound and understanding the mechanisms of resistance to palbociclib is critical. In this study we investigated whether metabolomic profiles of breast cancer cell lines with acquired resistance to palbociclib (PDR) differ from their sensitive counterpart (PDS). In addition we sought to identify metabolic biomarkers of sensitivity to palbociclib by analyzing a breast cancer cell line unable to acquire resistance to palbociclib.

Material and methods: We have established in our lab three PDR HR+/HER2neg breast cancer models (MCF7L, T47D and ZR75-1) by chronically exposing cells to escalating doses of palbociclib. PDR derivatives show IC50 values 6 to 30 times higher than their PDS counterparts. One additional model, CAMA-1, was unable to develop resistance. Whole-cell lysates and conditioned cell culture media from five replicates of each of the PDS and PDR models and from CAMA-1 were analyzed by nuclear magnetic resonance (NMR). Principal component analysis (PCA) was used as first exploratory analysis and as dimension reduction technique. Canonical Analysis (CA) was used to discriminate different groups. Differentially expressed metabolites between PDR and PDS models and between CAMA-1 and PDS cells were analyzed.

Results: Unsupervised PCA analyses of H NMR spectra, in which no information about PDS and PDR was inserted in the statistical model, correctly identified individual cell lines on both whole-cell lysates and conditioned media. However this analysis did not discriminate PDS from PDR within each model. Using a supervised approach, in which the statistical model was trained to discriminate between PDS and PDR, these groups were categorized with accuracy of 80% using whole-cell lysates and of 65% using conditioned media, using a cross-validation analysis by repeatedly testing the model on blind samples. CAMA-1 was correctly identified as a PDS model; however it showed a distinct metabolic profile compared to other PDS models. Over 30 metabolites were identified as differentially expressed between PDS and PDR models in lysates and conditioned media, but only glycerophosphocholine levels in conditioned media remained significantly higher in PDR compared to PDS models after correction for multiple testing.

Conclusions: In this study we show that analysis of metabolic profile of cells lysates discriminates PDR from PDS cell lines with a high accuracy. Analysis of metabolic pathways implicated in resistance/sensitivity to palbociclib is ongoing and might help identifying new targets to overcome resistance. Additionally, metabolites associated with palbociclib resistance may be potentially tested in clinical samples as biomarkers for patients stratification. Further studies are warranted.

Citation Format: Bonechi M, Guarducci C, Meoni G, Tenori L, Biagioni C, Schiff R, Osborne CK, Luchinat C, Di Leo A, Malorni L, Migliaccio I. Metabolomic analysis by nuclear magnetic resonance spectroscopy discriminates hormone receptor positive/HER2 negative breast cancer cell lines resistant to palbociclib [abstract]. In: Proceedings of the 2016 San Antonio Breast Cancer Symposium; 2016 Dec 6-10; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2017;77(4 Suppl):Abstract nr P6-02-07.