Acquisition of proteomic changes following molecularly targeted drug treatments in OVSAHO cells.

(a) Dose-response curves of the single small-molecule inhibitors, used for the primary perturbation followed by MS protein response profiling in OVSAHO cells, including AKT inhibitor MK-2206 (AKTi), BCL2 inhibitor Venetoclax (BCL2i), GSK3β inhibitor CHIR-99021 (GSK3βi), MEK inhibitor PD-0325901 (MEKi), PKC inhibitor Bisindolylmaleimide VIII (BIM VIII, PKCi), and SRC inhibitor Bosutinib (SRCi). Data is aggregated from three biological replicates each with three technical replicates and error bars represent standard deviations of the data. (b) Label-free mass spectrometry (MS)-based proteomics workflow used to measure protein changes in response to drug treatment at IC50. Data-driven network analysis was later applied to the acquired MS data to identify potential resistance mechanisms for the nomination of drug combination candidates. (c) Number of proteins quantified upon each drug treatment of OVSAHO cells after 72h at inhibitor IC50 concentration. Error bars: standard deviation of three biological replicates for each condition. (d) Multidimensional scaling (MDS, Euclidean distance) 2D projection of protein levels measured after treatment. Colors indicate different drug perturbation conditions with weak dotted lines around each group of biological replicates to indicate the closeness of the replicates.

Identification of frequently responsive proteins and processes induced by pharmacological inhibitors.

(a) Proportion of proteins per treatment condition that changed compared to the control treatment. Strongly responsive proteins are defined as proteins whose absolute log2 expression change is at least 0.5 (p-value < 0.05 and BH-based FDR < 0.2 in t-test). Constant proteins are the remaining proteins; protein expression values are averaged over three biological replicates. (b) Cumulative numbers of proteins that strongly respond to a certain number of perturbation conditions. Proteins that strongly respond in 4 or more treatment conditions are defined as frequently responsive proteins (red). The set of 2587 drug-specific proteins (black in the pie chart) is the union of the strongly responsive proteins for the six drugs (black rectangles in (a)). (c) Protein expression change of identified frequently responsive proteins; “frequently” refers to both positively and negatively responding proteins in at least 4 conditions (red in (b)). Most proteins respond with consistent trends (e.g., increasing expression) across several or all perturbation conditions, consistent with the notion of being part of a general stress response. (d) Gene set enrichment analysis (GSEA) identified both activated (increased protein expression) and suppressed (decreased protein expression) biological processes. Suppressed processes likely reflect the cytotoxic effects of drugs while activated processes might suggest a general resistance mechanism.

Drug-specific protein response networks identified using NetBox.

(a) Drug-specific protein responses were grouped into functional modules using the NetBox algorithm and the modules were characterized by functional labels based on enrichment analysis (Material and Methods). Subsets of protein modules for each drug perturbation were chosen based on their relevance to the known functions of the inhibitors and their implications for potential resistance mechanisms. Nodes are proteins colored by protein expression change, log2(expression ratio perturbed/unperturbed). Edges are undirected protein interactions from the background network used by the NetBox algorithm (Reactome FI Network or INDRA network). (b) Individual protein expression change ratios that are justification for the nominated drug combinations.

Proposed combination drug candidates based on the analysis of the proteomic response profiles.

These combinations were tested experimentally. Potential resistance mechanisms were identified based on the analysis of proteomic profiling after drug treatment using three complementary approaches, (i.e., NetBox analysis, GSEA analysis, and individual protein expression analysis). Small molecule inhibitors were selected to target the corresponding pathways indicative of resistance as combination partners of the originally profiled drugs. Drugs are prioritized if they are more relevant in clinical settings, i.e. FDA-approved or in late-stage clinical trials.

Drug combinations experimentally tested for effect on proliferation.

The most interesting drug combinations were experimentally tested across several ovarian cell lines and the inferred combination indices (CI) were calculated by the Chou-Talalay method. For drug combinations with 3 technical replicates, cell response was averaged across the replicates to perform the CI computation. CI < 1.00 indicates synergy (red) and CI > 1.00 the opposite. IGROV-1 etc (columns) are cell lines. NA: not available for technical reasons.

OVCA HGSOC cell line measures of similarity to patient samples and drug response to rucaparib.

Cell line similarity to TCGA HGSOC patient samples as characterized by Domcke et al. and TumorComparer (TC) (Sinha et al. 2021; Domcke et al. 2013). Additionally, a select set of genes including ones commonly altered in HGSOC (i.e., BRCA1, BRCA2, RB1) (Cancer Genome Atlas Research Network 2011). Rucaparib IC50 (µM, half-maximal inhibitory concentration) single drug screening of the cell lines as reported in Genomics of Drug Sensitivity in Cancer (GDSC) (Iorio et al. 2016). Mutations are protein-coding mutations as measured in particular cell line studies; similarly, CNAs are copy number alterations. KRAS etc. are proteins, altered (green check) or not (blank) by mutation or CNA.

Inhibition of proliferation by various drug combinations in a representative ovarian cancer cell line (OVCAR-4).

The combinations were chosen based on pathway analysis of protein response (Fig. 3). TVB-2640 (‘TVB’, FASNi), rucaparib (‘Ruc’, PARPi), bortezomib (‘Bort’, proteasome inhibitor), Apcin (APCi), GC7 (DNA synthase inhibitor), novobiocin (‘Novo’, DNA polymerase inhibitor), infigratinib (‘Inf’, FGFR3i), venetoclax (‘Ven’, BCL2i), everolimus (‘Eve’, MTORi), Palbociclib (‘Pal’, CDK4/6i), PD-0325901 (‘PD’, MEKi), idelalisib (‘Ide’, PI3Ki), bosutinib (‘Bos’, SRCi), auranofin (‘Aur’, PRDXi), simvastatin (‘Sim’, statin), GPX4-IN-3 (‘GPX’, GPX4i), RSL3 (GPX4i), DM4 (tubulin inhibitor). Each drug was applied to OVCAR-4 cells with a serial dilution factor of 3 or 4 to establish a dose-response curve for the single drug. Each drug combination was applied at the same concentrations as the single drugs. The combination index for each drug pair is calculated using the Chou-Talalay method.