ARTICLE DOI: 10.1038/s41467-017-00263-7 OPEN Systems analysis of apoptotic priming in ovarian cancer identifies vulnerabilities and predictors of drug response Ioannis K. Zervantonakis 1, Claudia Iavarone1, Hsing-Yu Chen1, Laura M. Selfors1, Sangeetha Palakurthi2, Joyce F. Liu3, Ronny Drapkin 4, Ursula Matulonis3, Joel D. Leverson5, Deepak Sampath6, Gordon B. Mills7 & Joan S. Brugge1 The lack of effective chemotherapies for high-grade serous ovarian cancers (HGS-OvCa) has motivated a search for alternative treatment strategies. Here, we present an unbiased systems-approach to interrogate a panel of 14 well-annotated HGS-OvCa patient-derived xenografts for sensitivity to PI3K and PI3K/mTOR inhibitors and uncover cell death vulner- abilities. Proteomic analysis reveals that PI3K/mTOR inhibition in HGS-OvCa patient-derived xenografts induces both pro-apoptotic and anti-apoptotic signaling responses that limit cell killing, but also primes cells for inhibitors of anti-apoptotic proteins. In-depth quantitative analysis of BCL-2 family proteins and other apoptotic regulators, together with computational modeling and selective anti-apoptotic protein inhibitors, uncovers new mechanistic details about apoptotic regulators that are predictive of drug sensitivity (BIM, caspase-3, BCL-XL) and resistance (MCL-1, XIAP). Our systems-approach presents a strategy for systematic analysis of the mechanisms that limit effective tumor cell killing and the identification of apoptotic vulnerabilities to overcome drug resistance in ovarian and other cancers. 1 Department of Cell Biology, Ludwig Center at Harvard, Harvard Medical School, Boston, MA 02115, USA. 2 Belfer Center for Applied Cancer Research, Dana Farber Cancer Institute, Boston, MA 02115, USA. 3 Department of Medical Oncology, Dana Farber Cancer Institute, Boston, MA 02115, USA. 4 Penn Ovarian Cancer Research Center, Department of Obstetrics and Gynecology, University of Pennsylvania School of Medicine, Philadelphia, PA 19104, USA. 5 Oncology Development, AbbVie, Inc, North Chicago, IL 60064, USA. 6 Translational Oncology, Genentech, South San Francisco, CA 94080, USA. 7 Department of Systems Biology, MD Anderson Cancer Center, Houston, TX 77030, USA. Correspondence and requests for materials should be addressed to J.S.B. (email: [email protected]) NATURE COMMUNICATIONS | 8: 365 | DOI: 10.1038/s41467-017-00263-7 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00263-7 igh-grade serous ovarian cancer (HGS-OvCa) accounts multiple drugs targeting different nodes of the PI3K/AKT/mTOR for 70–80% of ovarian cancer deaths and, despite opti- pathway has revealed limited efficacy as single-agents13 and H 14, 15 mized surgery and chemotherapy protocols, treatment multiple resistance mechanisms have been identified . The resistance ultimately emerges in most cases1. Therefore, there is identification of drugs that optimally synergize with PI3K/AKT/ an urgent need to develop new therapies to improve patient mTOR inhibition is critical for effective targeting of HGS-OvCa outcomes2. Although therapeutically actionable recurrent point tumors with PI3K/AKT/mTOR pathway activation15, 16. mutations are uncommon in HGS-OvCa, genomic and proteomic Preclinical studies using established ovarian cancer cell lines characterization of primary tumors have uncovered commonly have described combinations of PI3K inhibitors with che- deregulated signaling pathways that represent attractive targets motherapy17 and various agents targeting the RAS/ERK path- – for therapeutic intervention3 5. In particular, multiple compo- way18, EGFR19, mTOR20, and BCL-2-family proteins21, 22. nents of the phosphoinositide 3-kinase/AKT/mammalian target However, genomic23 and tumor xenograft studies24, 25 have called of rapamycin (PI3K/AKT/mTOR) pathway are genetically altered into question the suitability of many commonly used ovarian in HGS-OvCa tumors (PTEN3 copy number loss, AKT3 and cancer cell lines as models of HGS-OvCa. Patient-derived xeno- PIK3CA3, 6 copy number amplification) and there is evidence for graft (PDX) models, on the other hand, represent a more clini- pathway activation based on increased phosphorylation of key cally relevant tool for studying drug treatment efficacy, as they nodes (phospho-AKT7, phospho-GSK37, phospho-PRAS408, have been shown to mirror clinical responses and recapitulate phospho-p70RSK9, an phospho-S68). High PI3K/AKT/mTOR resistance mechanisms seen in patients26, 27 and retain the genetic pathway activity in HGS-OvCa tumors has been associated with heterogeneity of human tumors more faithfully than established – – decreased patient survival10 12 and therefore represents an cell lines28 30. Given their genomic heterogeneity, PDX models important therapeutic target. To date, clinical evaluation of may also be more relevant for biomarker discovery31 to enable a Analysis of drug Identification of drug Short-term culture in vitro Computational modeling sensitivity resistance mechanisms & and analysis: & & Orthotopic PDX in vivo Predictive biomarkers Proteomic profiling Validation of biomarkers bc2.0 15 M) 1.5 µ ( tment) 50 trea 5 1.0 Cell Death (%) 0.5 (GNE-493 GNE- 493 IC 493 GNE- –5 0.0 DF68 DF14DF20 DF59 DF09 DF83DF86 DF68 DF14DF20 DF59 DF09 DF83DF86 DF101 DF172 DF216 DF118DF149 DF106DF181 DF101 DF172 DF216 DF118DF149 DF106DF181 de–1.00 –0.67 –0.33 0.00 DF101-D DF68-D DF172-D DF14-D DF216-D DF59-D DF181-D DF83-D DF20-D DF118-D DF149-D DF09-D DF106-D DF86-D 0.33 2.0 0.67 AKT_pS473 1.00 AKT_pT308 GSK-3A-B_pS21_S9 DF83 PRAS40_pT246 1.5 DF181 M) p27_pT157 µ p27_pT198 ( DF86 50 INPP4b DF106 PTEN 1.0 4E-BP1_pS65 DF09 DF216 DF59 DF14 4E-BP1_pT37_T46 r = 0.68 DF149 mTOR_pS2448 p = 0.01 GNE- 493 IC 493 GNE- Protein expression p70-S6K_pT389 0.5 DF118 DF172 DF101 S6_pS235_S236 DF20 DF68 S6_pS240_S244 RICTOR_pT1135 0.0 PI3K/Akt score 123 TSC/mTOR score PI3K / Akt pathway score (a.u.) Fig. 1 Analysis of PI3K/AKT/mTOR pathway activation at the protein level and sensitivity to PI3K/mTOR inhibition. a A systems approach strategy using a panel of 14 HGS-OvCa PDX and integrated drug profiling/proteomics to identify drug resistance mechanisms and validate response biomarkers. b PI3K/ S473 mTOR inhibitor (GNE-493) IC50 values ranked according to PI3K/AKT activation status (phospho-AKT protein levels). Data is derived from three independent experiments and error bars denote SEM. c cell death after PI3K/mTOR treatment (GNE-493 0.3 μM, 96 h) using laser scanning cytometry. Data is representative of three independent experiments. Error bars represent SEM for n = 6 replicate wells. d Heatmap showing protein levels (median- centered) of PI3K/AKT/mTOR pathway targets for PDX samples under short-term in vitro culture conditions. Samples ranked according to PI3K/AKT activation status (data is representative of two independent experiments). e Correlation analysis of GNE-493 IC50 values with PI3K/AKT pathway score (Pearson r = 0.68, p = 0.01) 2 NATURE COMMUNICATIONS | 8: 365 | DOI: 10.1038/s41467-017-00263-7 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/s41467-017-00263-7 ARTICLE a b –1.00 1.5 –0.67 –0.33 0.00 0.33 0.67 1.00 DF68-G/D DF172-G/D DF14-G/D DF20-G/D DF216-G/D DF59-G/D DF118-G/D DF149-G/D DF181-G/D DF83-G/D DF86-G/D DF101-G/D DF09-G/D DF106-G/D 1.0 S6_pS235_S236 S6_pS240_S244 4E-BP1_pS65 NDRG1_pT346 CDK1 0.5 p70-S6K_pT389 PLK1 (GNE-493/DMSO) CYCLIN-B1 4E-BP1_pT37_T46 Rb_pS807_S811 phospho-S6 protein level ratio 0.0 ACC_pS79 PAR FOXM1 DF68 DF14DF20 DF59 DF09 DF83DF86 DF101 DF172 DF216 DF118DF149 DF106DF181 ACC1 CHK1 PRAS40_pT246 c 2.0 mTOR_pS2448 HSP27_pS82 FASN MAPK_pT202_Y204 p90RSK_pT573 1.5 IGFBP2 PKC-b-II_pS660 eEF2 MDM2_pS166 Rictor_pT1135 1.0 MEK1 EMA Pathway activation Pathway B-Raf_pS445 eIF4G by PI3K/mTOR inhibition IRS1 0.5 SMAC GLUTAMATE-D1-2 PUMA RTKs VHL BIM ApoptosisCell Cycle Hormone HormoneA Ras/MAPK B HISTONE-H3 DNA Damage H2AX_pS140 HER3_pY1289 FIBRONECTIN d –0.50 CLAUDIN-7 –0.33 –0.17 AKT_pS473 0.00 BECLIN 0.17 0.33 ANNEXIN-I 0.50 DF101-G/D DF68-G/D DF172-G/D DF14-G/D DF20-G/D DF216-G/D DF59-G/D DF118-G/D DF149-G/D DF09-G/D DF106-G/D DF181-G/D DF83-G/D DF86-G/D PORIN BCL-2 PDGFR-b ATP5A BCL-XL survival COMPLEX-II-SUBUNIT pro - MCL-1 HER3 BCL2-A1 IGFRb PAI-1 XIAP PDCD4 BAD_pS112 AXL BAK IGF1R_pY1135_Y1136 SDHA BAX apoptotic MITOCHONDRIA BID pro - AKT_pT308 PUMA SOD2 SMAC TFAM CASPACE-7-CLEAVED CASPASE-3 BIM Fig. 2 Rewiring of multiple pathways after PI3K/mTOR inhibition and upregulation of pro-apoptotic and anti-apoptotic proteins in all PDX samples. a Analysis of phospho-S6 protein levels after treatment with 0.5 μM GNE-493 (48 h). Mean values for ratio of GNE-493 over DMSO protein levels for each PDX model across three replicate samples. b Heatmap showing proteins exhibiting the largest fold increases (red) or decreases (blue) relative to baseline after GNE-493 treatment across all 14 PDX samples. Protein lysates from PDX cells treated with 0.5 μM GNE-493 for 48 h were analyzed by RPPA and all signals from all samples were normalized to DMSO control. Data is representative of two independent experiments and represents the mean log2 transformed value for n = 3 replicates. c Analysis of pathway scores (mean value for the ratio of GNE-493 over DMSO score values, data is derived from b). d Heatmap of anti-apoptotic and pro-apoptotic relative protein levels after GNE-493 treatment (data represents the ratio of protein levels of GNE-493 treatment to DMSO (G/D) and is derived from b) appropriate patient selection, an important consideration given level. We show that, despite diverse signaling responses in the that PI3K/AKT/mTOR-therapies in combination with other tar- PDX models, PI3K/mTOR inhibition results in elevated apoptotic geted agents are currently under clinical evaluation16.