A large-scale chemical screen identifies YK-135 as a selective cytotoxic agent against EMT-subtype gastric cancer cell lines. (A) Classification of the 37 gastric cancer cell lines according to their EMT signatures. Columns represent EMT signature genes (149 upregulated genes on the left, 161 downregulated genes on the right). (B) Screening workflow. All 48,467 chemicals in the library were screened once at a concentration of 2.5 μM in three EMT (Hs746T, SNU484 and SNU668) and one non-EMT (SNU719) cell lines. Four hundred fifty-seven molecules with selective response profiles were re-screened at 2.5 μM in six cell lines (3 EMT (Hs746T, SNU484 and SNU668) vs. 3 non-EMT (SNU719, MKN45 and NCC59)). One hundred seventy-five chemicals with selective viability were selected and re-assayed in the same conditions, followed by filtering by chemical structures or properties (e.g., availability of resupply and chemical structural similarity) (Supplementary Data 1). Seventy-five chemicals were re-screened in a multi-dose format (12-point dose responses) against a panel of 37 cell lines in singlicate. Nineteen chemicals with EMT-selective toxicity were selected and re-assayed across the 37 cell lines using 12 doses (half-log dilution series). (C) Dot plot of KS test distances of AUC in EMT cell lines compared to non-EMT cell lines against 18 selected chemicals (left). Red dots indicate selected 7 hit candidates (ΔAUC KS distance ≥ 0.6). Cumulative distribution function plots of 7 hit candidates demonstrating enrichment for AUC of EMT (red) and non-EMT (blue) cell lines (right). (D) Structures of the 7 hit candidate chemicals. (E) Dot plot of GSEA results illustrating GO gene sets decreased in EMT-type gastric tumor of TCGA stomach adenocarcinoma cohort. The figure shows the top 16 enriched GO terms with NES < −2.0 and adjusted P-value < 0.03. Red text indicates mitochondria-related gene sets. (F) Basal respiration measurement using a Seahorse bioanalyzer. All values are mean ± SD of triplicated experiments. P-values were determined by one-way ANOVA, followed by a Tukey multiple comparison test (**P < 0.01, ***P < 0.0001). (G) YK-135 dose-response in 4 EMT, 3 non-EMT-subtype gastric cancer cell lines and 3 normal cell lines.
