Identification of anti-tumour biologics using primary tumour models, 3-D phenotypic screening and image-based multi-parametric profiling

Abstract

BackgroundMonolayer cultures of immortalised cell lines are a popular screening tool for novel anti-cancer therapeutics, but these methods can be a poor surrogate for disease states, and there is a need for drug screening platforms which are more predictive of clinical outcome. In this study, we describe a phenotypic antibody screen using three-dimensional cultures of primary cells, and image-based multi-parametric profiling in PC-3 cells, to identify anti-cancer biologics against new therapeutic targets.MethodsScFv Antibodies and designed ankyrin repeat proteins (DARPins) were isolated using phage display selections against primary non-small cell lung carcinoma cells. The selected molecules were screened for anti-proliferative and pro-apoptotic activity against primary cells grown in three-dimensional culture, and in an ultra-high content screen on a 3-D cultured cell line using multi-parametric profiling to detect treatment-induced phenotypic changes. The targets of molecules of interest were identified using a cell-surface membrane protein array. An anti-CUB domain containing protein 1 (CDCP1) antibody was tested for tumour growth inhibition in a patient-derived xenograft model, generated from a stage-IV non-small cell lung carcinoma, with and without cisplatin.ResultsTwo primary non-small cell lung carcinoma cell models were established for antibody isolation and primary screening in anti-proliferative and apoptosis assays. These assays identified multiple antibodies demonstrating activity in specific culture formats. A subset of the DARPins was profiled in an ultra-high content multi-parametric screen, where 300 morphological features were measured per sample. Machine learning was used to select features to classify treatment responses, then antibodies were characterised based on the phenotypes that they induced. This method co-classified several DARPins that targeted CDCP1 into two sets with different phenotypes. Finally, an anti-CDCP1 antibody significantly enhanced the efficacy of cisplatin in a patient-derived NSCLC xenograft model.ConclusionsPhenotypic profiling using complex 3-D cell cultures steers hit selection towards more relevant in vivo phenotypes, and may shed light on subtle mechanistic variations in drug candidates, enabling data-driven decisions for oncology target validation. CDCP1 was identified as a potential target for cisplatin combination therapy.Electronic supplementary materialThe online version of this article (doi:10.1186/s12943-015-0415-0) contains supplementary material, which is available to authorized users.

CDCP1 cleavage is necessary for homodimerization-induced migration of triple-negative breast cancer

“…41,42 The induction of apoptosis in the presence of extracellular matrix components, collagen I and Matrigel (collagen IV, laminin, enactin/nidogen-1 and proteoglycans), may indicate that ECC inhibits stellate structure formation, which is indicative of cell binding and anchoring to these substrates during invasion. This explanation is supported by a study showing preferential and direct binding of β 1 integrin to cCDCP1 over flCDCP1.…”

Drug combination therapy increases successful drug repositioning

“…Executing this approach requires a characteristic phenotype associated with the disease that is known. Cell-based phenotypic assays usually use primary cells [11], isolated pathogens [12], engineered cell lines [13] or the recently emerged iPSC-derived cells including neuronal cells, cardiomyocytes, hepatocytes and epithelial cells [14–16]. As an example of this, Eggan, Woolf and co-workers discovered hyperexcitability as a result of a reduced delayed-rectifier potassium channel as a disease phenotype in iPSC-derived motor neurons from amyotrophic lateral sclerosis (ALS) patients [16].…”

Phenotypic screening reveals TNFR2 as a promising target for cancer immunotherapy

“…We have also demonstrated that phage display may have advantages over hybridoma-based screening, due to the ability to rationally deselect antibodies to known or abundant antigens, therefore enabling access to wider pools of targets during the screening process [17, 18]. …”