40O - Identification of gene expression signatures of palbociclib (PD) response in breast cancer (BC)

Date 07 May 2015
Event IMPAKT 2015
Session Best abstracts session
Topics Anti-Cancer Agents & Biologic Therapy
Breast Cancer
Translational Research
Presenter Ilenia Migliaccio
Citation Annals of Oncology (2015) 26 (suppl_3): 15-24. 10.1093/annonc/mdv117
Authors I. Migliaccio1, S. Piazza2, R. Verardo2, C. Guarducci1, M. Bonechi1, Y. Ciani2, E. Moretti3, O. Siclari3, A. Di Leo3, L. Malorni3
  • 1Translational Research Unit, Hospital of Prato, Istituto Toscano Tumori, 59100 - Prato/IT
  • 2Functional Genomics & Bioinformatics Units, Proxenia S.r.l, 34149 - Trieste/IT
  • 3Medical Oncology Unit, Hospital of Prato, Istituto Toscano Tumori, 59100 - Prato/IT



Introduction: The CDK4/6 inhibitor PD has shown promising activity in patients (pts) with ER + /HER2- BC, but resistance is commonly seen. Retinoblastoma (RB) genetic loss is a known marker of resistance to PD, but only a proportion of tumors retaining RB responds to PD. We hypothesize that a signature of functional RB loss would predict resistance to PD and aim to identify a predictive gene expression signature of PD response.

Materials and methods: We established two different signatures, one of functional RB loss from BC in the TCGA dataset (RBsig) and one of PD sensitivity identified by comparing 13 PD sensitive and 13 PD resistant cell lines in the Cancer Cell Line Encyclopedia (CCLE) dataset (PDSENSsig). We then tested these signatures for their prognostic and predictive value in a broad BC gene expression meta-dataset (N = 3458). Finally, we tested the ability of these signatures to discriminate PD sensitive vs resistant cell lines, according to Finn et al. (PMID:19874578), in a validation dataset of BC cell lines (GSE48213).

Results: The RBsig (87 genes) was predictive of RB status in the TCGA dataset across all molecular subtypes. Untreated or endocrine treated pts with ER+ tumors expressing high RBsig had significantly worse recurrence free survival (RFS) compared to those with low RBsig (HR = 2.34 [1.75–3.13]; P = 3.4 e − 09 and HR = 2.6 [1.92–3.53]; P = 1.9 e − 10 respectively). A robust performance of the RBsig predictor in evaluating sensitivity/resistance to PD was obtained in the validation dataset (ROC area under curve (AUC) = 0.93). PDSENSsig was composed of 20 genes upregulated in the PD sensitive cell lines. Untreated or endocrine treated pts with ER+ tumors expressing high PDSENSsig had significantly better RFS compared to those with low PDSENSsig (HR = 0.63 [0.47–0.84] P = 0.0017 and HR = 0.64 [0.47–0.87] P = 0.0038, respectively). A good performance of the PDSENSsig predictor was obtained in the validation dataset (ROC AUC = 0.76).

Conclusions: Here we derived two signatures that might be helpful in selecting pts more likely to benefit from PD treatment. Further validation in cohorts of BC pts treated with PD is warranted.

Disclosure: A. Di Leo: Honoraria from AstraZeneca and Novartis. Research grant from Pfizer.

L. Malorni: Research grant from Pfizer.

All other authors have declared no conflicts of interest.