Identification of Novel Biomarkers for Prediction of Response to Tyrosine Kinase Inhibitors (TKIs) by Proteomic Profiling of Imatinib as Well as 2^nd and 3^rd Generation TKIs in Vitro

The tyrosine kinase inhibitor (TKI) Imatinib (IM) represents the gold standard firstline treatment for patients with newly diagnosed chronic myeloid leukemia (CML). For patients developing resistance or intolerance to Imatinib, the 2^nd generation TKIs Dasatinib (DASA) and Nilotinib (NILO) which are approved, and Bosutinib (BOSU) which is in clinical development, possess activity against almost all mutant forms of BCR-ABL which confer Imatinib-resistance, except the gatekeeper mutation, T315I. This latter mutation occurs in approximately 15% of clinically observed mutations in chronic phase CML and confers resistance to all currently approved agents. Therefore compounds have been evaluated for activity against T315I mutant BCR-ABL, among which combined Aurora kinase and Abl inhibitors such as PHA-739358 (PHA) have been identified as showing some promise.In the current study, we used a classical proteomics approach to generate drug profiles of the TKIs (IM, DASA, NILO and PHA), in order to identify biomarkers that are either compound or drug group (i.e. 1^st, 2^nd or 3^rd generation TKI) specific and could potentially be used as biomarkers for response prediction in vivo. For in vitro screening, we used murine Ba/F3 cells expressing wild-type (p210, wt) BCRABL or mutants, which are either low grade (M351T) or absolutely resistant (T315I) to Imatinib. Using 2D-gel electropheresis and mass spectrometry, we could identify a total of 68 individual protein spots which were differentially regulated in cells when treated with equieffective concentrations (IC₅₀) of TKIs. Using in silico overlay of the different 2D-gels (Delta2D, Decodon GmbH Greifswald), 42, 38, 41 and 15 spots were found to be specifically differentially regulated in Ba/F3 cells expressing wt BCR-ABL under either IM, NILO, DASA and PHA, respectively. Interestingly, hierarchical cluster analysis based on these candidate proteins identified similar protein expression patterns for IM, NILO and DASA in comparison to PHA. Using genontology analysis (Panther software), the majority of the proteins belonged to the group of nucleic acid binding proteins (25%), cytoskeletal proteins (13%) and chaperones (12%). In contrast to the broad response of the different TKIs on wt Bcr-Abl cells, changes in protein expression patterns induced in cells carrying the M315T BCR-Abl mutation were substantially less pronounced (IM: 9, NILO: 12, DASA: 28, PHA: 17) with the strongest response seen in Dasatinibtreated cells, consistant with the compound being a powerful inhibitor of both wt and M351T BCR-ABL signaling. With the exception of the combined BCR-ABL and Aurorakinase inhibitor PHA (7 proteins), cells expressing T315I BCR-ABL exhibited no altered protein expression in response to treatment with either IM or the other 2^nd generation TKIs. The protein expression patterns identified were used for systems biology network analysis using Metacore software (GeneGo), which enabled the elucidation of signaling pathways and identification of transcription factors involved in TKI response. Besides known regulators of BCR-ABL signaling, such as c-Myc and p53, we were able to identify novel TKI-dependent candidate proteins (e.g. eIF5a) and post-translational modifications (PTM) that, pending validation in primary patient material might effectively be used as biomarkers for response prediction in the near future.