Discussion

Phenotypic screening for novel therapeutics is a well-established discovery platform for small molecules, but in the large-molecule/biologic space most discovery programmes are target-focussed, with relatively few examples of target-agnostic screens. We have established a platform for isolating antibodies/antibody-mimetics against the cell-surface proteomes of intact, disease-relevant cells by phage-display selections. These antibody panels can be expressed in mammalian cells in 96-well format for phenotypic screening.
In collaboration with O-cell-O, we performed a high-content screen on PC-3 cells in a complex 3-D matrix, treated with a mixed panel of antibodies isolated against several different primary tumour cell types. The effects of the antibodies on cell growth morphology were measured, using machine-learning methods to identify a minimal set of phenotypic features that were able to classify the antibody-induced responses. Antibodies that inhibited the invasive phenotype of PC-3 cells were thereby identified for further analysis.
The antibody panel contained a subset of molecules targeting the cell-surface protein CDCP1 (CUB-domain containing protein 1), which had been isolated in phage-selections performed on primary non-small cell lung cancer (NSCLC) cells. CDCP1 has been linked to cancer invasiveness, metastasis and anoikis resistance, and shows promise as a therapeutic antibody target. In our high-content screen, the CDCP1-targetting antibodies did not all behave similarly; instead these molecules divided into two subsets with opposite phenotypic effects on the cells.
This study demonstrates how high-content screens can potentially be used both as a “lead identification” platform and a means to investigate target biology and therapeutic mode-of-action.