Predicting Carfilzomib Resistance Mechanisms and Therapeutics Using Computational Modelling of Genomics and Proteomics

Background: Multiple myeloma (MM) is a malignancy of plasma cells accounting for around 10% of all hematologic cancers. MM is an incurable heterogeneous malignancy which impacts the response rate due to complex nature of the disease. Although with standard of care treatment, including proteasome inhibitors (PI), a significant response and remission is achieved, the majority of patients still develops resistance. There is no precise method for determining the pathways which govern the acquired resistance to PI. Computational biology modelling (CBM), which uses genomics for creating the MM specific disease characteristics, can be used for deciphering the dominance of signaling changes in acquired resistance and accordingly select right treatment combination. Predicting a priori treatment response based on disease characteristics would enable optimal treatment selection and potentially reduce the trial and error approach that impacts outcome and health care costs.

Aim: The objective of the study was to predict response to proteasome inhibitor Carfilzomib (CFZ) in AMO1 MM disease model and its resistant counterparts using CBM, to determine mechanism of resistance to CFZ and identify customized therapy options for CFZ resistant AMO1 (AMO-CFZ) cells.

Methods: AMO-CFZ resistant cells were derived by long-term exposure of the cells to increasing doses of CFZ. AMO1 PI-sensitive and AMO-CFZ were analyzed using cytogenetics and mutations (profiled by next-generation sequencing; NGS). At the same time quantitative proteomic analysis was done to find proteins which were upregulated as well as downregulated in the resistant cells. Genomic and proteomic data was input to generate disease-specific protein network maps, biomarkers and pathway characteristics unique to AMO1 and AMO-CFZ cells respectively. Digital drug models of CFZ and other targeted-FDA approved drugs are simulated on the disease models individually and in combination at varying doses. Drug impact was assessed by quantitatively measuring the effect on a cell growth score, which is a composite of cell proliferation, viability and apoptosis. Unique therapy combinations identified in AMO-CFZ cells will be prospectively experimentally validated; at the same time key signaling changes were deciphered in AMO-CFZ resistance with respect to AMO1 PI-sensitive MM model.

Results: Genomic analysis using CBM showed that AMO1 MM model harbored amplification of MYC, POU2AF1, POU2F1, TXNIP and knock down (KD) of proteasome subunit PSMA6 which lead to Carfilzomib sensitivity. The predictions in AMO-CFZ identified loss of PSMA6 and upregulation of proteasomal subunits PSMB5/7/2/1 which reduced ER stress and ATF4. Predictions in AMO-CFZ cells identified higher glycolysis and pentose phosphate pathway resulting in higher glutathione levels compared to AMO1 PI-sensitive cells. The anti-oxidant proteins NQO1, TXN, PRDX1/4/5 were higher in AMO-CFZ-resistant cells, in addition to PRDX6 overexpression (OE), which was common with PI-sensitive cells that reduced oxidative stress. Carfilzomib drug exporter ABCB1 was overexpressed in CFZ-resistant cells and Nelfinavir increased sensitivity of the cells towards Carfilzomib. CBM identified novel therapeutic targets CDK5 and NFE2L2 that can be used to overcome CFZ-resistance. Moreover, Venetoclax was more cytotoxic in AMO-CFZ cells than in AMO-1 PI-sensitive, because BCL2 was higher in CFZ-resistant version than the sensitive one due to its overexpression and TP53 knockdown.

CBM predicted novel therapeutics for investigation which include: GEMCITABINE plus MELPHALAN; GEMCITABINE plus OLAPARIB; MELPHALAN plus OLAPARIB. DNA damage is high in AMO-CFZ due to knockdown of CHEK1, ERCC3, FANCB/G, LIG4, XRCC6 and loss of function of FANCD2, ATM.

Conclusions: Using genomics and proteomics, CBM identified the resistance mechanisms of AMO-CFZ and new treatment strategies. These were validated showcasing a methodology for a clinically relevant workflow using NGS information.