Systematic Screening of Targeted Chemical Combinations for Cancer Therapy Systematic Screening of Targeted Chemical Combinations for Cancer Therapy Systematic Screening of Targeted Chemical Combinations for Cancer Therapy Systematic Screening of Targeted Chemical Combinations for Cancer Therapy Adrian Heilbut, Joseph Lehár , Glenn F. Short III, Grant R. Zimmermann and Curtis T. Keith Adrian Heilbut, Joseph Lehár , Glenn F. Short III, Grant R. Zimmermann and Curtis T. Keith Adrian Heilbut, Joseph Lehár , Glenn F. Short III, Grant R. Zimmermann and Curtis T. Keith CombinatoRx Inc., Cambridge, MA 02142, USA CombinatoRx Inc., Cambridge, MA 02142, USA #1410 #1410 Abstract Systematic Multi-Target Mechanism Screen Preliminary Results and Analysis Follow-on Studies Abstract Systematic Multi-Target Mechanism Screen Preliminary Results and Analysis Follow-on Studies Systematic Multi-Target Mechanism Screen Preliminary Results and Analysis Follow-on Studies Clinical experience and theoretical analyses suggest that multi-target approaches are required to Synergy Score Statistics Objective Modulators of DNA damage response pathways Clinical experience and theoretical analyses suggest that multi-target approaches are required to overcome redundant and adaptive oncogenic mechanisms, and cotherapies are indeed the standard of Synergy Score Statistics Experimental error estimate from self combinations Objective Screen diverse inhibitors of cellular processes to find 9 Histogram of Additivity Volumes Combos Combos + 0.1mM Topotecan Modulators of DNA damage response pathways DNA damage drugs remain central to cancer therapy care for many cancers. Identification of optimal and selective combinations of the many targeted agents Experimental error estimate from self combinations Scores > 1 are significant at p ~99% Screen diverse inhibitors of cellular processes to find selective antiproliferative synergies. Probe Set Cells 8 HCT116 A549 MRC9 + 0.1mM Topotecan DNA damage drugs remain central to cancer therapy becoming available presents a critical challenge. CombinatoRx has developed a platform for combination high throughput screening (cHTS TM ) that we have deployed to screen combinations of Scores > 1 are significant at p ~99% ~35% of Score>1 synergies are artifacts selective antiproliferative synergies. Cytoskeleton Pleiotropic Viral replication undefined Cytoskel. Probe Set Cells 7 MRC9 self/self Can combinations selectively modulate responses? combination high throughput screening (cHTS TM ) that we have deployed to screen combinations of approved drugs for the discovery of therapeutically relevant synergies in cell based models. In addition ~35% of Score>1 synergies are artifacts Score < 1 tails dominated by artifacts Receptor, neural Receptor, hormone Cytoskeleton Fungal cell wall DNA damage DNA Receptor Cytoskel. HCT116 carcinoma 6 approved drugs for the discovery of therapeutically relevant synergies in cell based models. In addition to providing effective treatments, chemical synergies can provide information on interactions between Score < 1 tails dominated by artifacts 1. Probe Library DNA metabolism DNA synthesis Transcription, activation Receptor, adenosine Receptor, adrenergic Receptor, growth factor DNA Receptor carcinoma (colon) 5 (Count) Selected 24 probes of survival and death pathways to providing effective treatments, chemical synergies can provide information on interactions between targeted pathways, elucidating previously unappreciated connections between disease mechanisms. 430 compounds, ~250 diverse targets Transcription, chromatin Signaling, kinase, PKC Signaling, kinase, MAPK Transcript. Kinase 3 4 log( Selected 24 probes of survival and death pathways Screening combinations in DNA damage background targeted pathways, elucidating previously unappreciated connections between disease mechanisms. Scores are correlated across cell types Probes not biased to therapeutic targets Transcription, machinery Translation, ribosome Signaling, kinase, PKA Signaling, kinase, PKB Kinase A549 2 3 Screening combinations in DNA damage background Here we extend systematic combination screening to probe perturbations of diverse cellular Scores are correlated across cell types Cell line commonalities dominate over artifacts Even coverage of sampled mechanisms Protein processing Signaling, kinase, lipid Signaling, kinase, tyrosine Protein A549 carcinoma (lung) 1 2 Synergies persist on top of DNA damage background Here we extend systematic combination screening to probe perturbations of diverse cellular mechanisms to discover novel pathway interactions with therapeutic potential, and to evaluate the utility Cell line commonalities dominate over artifacts Even coverage of sampled mechanisms Protein modification Protein degradation Signaling, phosphatase Signaling, phosphodiesterase Signaling, kinase modif. Lipid (lung) -4 -3 -2 -1 0 1 2 3 4 0 True 3 way synergies are rare (< ~0.25%) mechanisms to discover novel pathway interactions with therapeutic potential, and to evaluate the utility of combination effect measures for predicting mechanisms of action for novel compounds. A set of 180 Significant synergies occur at ~1% rate Proteasome Metabolism Metabolism, metals Signaling, lipid Signaling, phosphatase Lipid signal MRC9 T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 p53-/- T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 T116 -4 -3 -2 -1 0 1 2 3 4 Additivity Volume True 3 way synergies are rare (< ~0.25%) chemical probes were selected that modulate molecular targets involved in diverse cellular functions. All Significant synergies occur at ~1% rate 2. Screen Design Metabolism, energy Metabolism, redox Metabolism, lipid Signaling, apoptosis Metab. MRC9 fibroblast HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT HCT116 HCT116 p53 -/- Correlations Between Cell Lines 16,110 pair wise combinations of these probes were tested at multiple concentrations and ratios using cHTS in a proliferation assay with HCT116 human colon cancer cells. Interesting combinations were Synergies are rare but more common than 2. Screen Design Metabolism, sphingolipid Metabolism, sterol Metabolism, leukotriene Signaling, intracellular Signaling, cell cycle Metab. Ion Signaling (lung) HCT C-10331 C-10331 C-10677 C-10739 C-10896 C-11035 C-11343 C-11465 C-12220 C-12240 C-12312 C-12316 C-12317 C-12318 C-12353 C-12390 C-12395 C-12440 C-12460 C-12463 C-12477 C-12487 C-12488 C-12490 C-12505 C-10331 C-10677 C-10739 C-10896 C-11035 C-11343 C-11465 C-12220 C-12240 C-12312 C-12316 C-12317 C-12318 C-12353 C-12390 C-12395 C-12440 C-12460 C-12463 C-12477 C-12487 C-12488 C-12490 C-12505 HCT116 HCT116 p53 -/- 3 re Most combinations behave similarly cHTS in a proliferation assay with HCT116 human colon cancer cells. Interesting combinations were further evaluated for tumor selectivity using additional cell lines. The screen identified both previously Genotype-Specific Synergies r = 0.27±0.08 genetic interactions (~0.5%, Tong et al. 2004) Selected 180 probes covering ~120 targets Ion transport Signaling, ion Signaling, inflammatory Signaling, neural Ion Signaling C-10331 C-10677 C-10739 C-10896 C-11035 1 2 y scor Most combinations behave similarly across isogenic cell lines further evaluated for tumor selectivity using additional cell lines. The screen identified both previously reported and novel synergies and antagonisms that reflect connections between pathways relevant to Genotype-Specific Synergies Does synergy depend on p53 status? Only 1/3 of probes account for 70% of synergies All pairwise combinations in sparse dose matrix C-11035 C-11343 C-11465 C-12220 C-12240 r = 0.20±0.08 -1 0 ditivity reported and novel synergies and antagonisms that reflect connections between pathways relevant to cancer proliferation. Synergy profiles of compounds with proximal targets were found to be correlated, Does synergy depend on p53 status? Only 1/3 of probes account for 70% of synergies Finding unexpected synergies requires Two cancer and one “normal” cell line C-12240 C-12312 C-12316 C-12317 C-12318 -2 -1 C9 add cancer proliferation. Synergy profiles of compounds with proximal targets were found to be correlated, suggesting that such profiles may be used to infer mechanism of action. Combination screening data Finding unexpected synergies requires very large combination screens identify cancer selective synergies A549 MRC9 C-12318 C-12353 C-12390 C-12395 C-12440 r = 0.19± 0.08 -5 -4 -3 -2 -1 0 1 2 3 -4 -3 MRC r = 0.27 ± 0.08 suggesting that such profiles may be used to infer mechanism of action. Combination screening data using cell lines analyzed in the context of emerging knowledge of cancer genotypes and expression Screened combinations of the same 24 probes very large combination screens C-12460 C-12463 C-12477 C-12487 r = 0.19± 0.08 -5 -4 -3 -2 -1 0 1 2 3 HCT116 synergy score using cell lines analyzed in the context of emerging knowledge of cancer genotypes and expression profiles may lead to the development of more selective, personalized, and effective cancer therapies. 384-well Incubate 72h Add ATP-Lite C-12488 C-12490 C-12505 profiles may lead to the development of more selective, personalized, and effective cancer therapies. preliminary data suggests rare selective 3. Quality Control and Filtering 384-well Assay Plate Add ATP-Lite (Luminescence) Profiles for Two Statins Correlation No single-agent effect; Qualitatively different synerty Single-agent effect References: preliminary data suggests rare selective synergies Synergy profiles 2 2.5 3. Quality Control and Filtering Failed plates and bad wells flagged for exclusion Correlation HCT p53 +/+ HCT p53 -/- 1.5 Qualitatively different synerty from secondary target Single-agent effect and modest synergy orB 7 7 Borisy AA et al. “Systematic Discovery of Multicomponent Therapeutics” PNAS 100 (13):7977 (2003) Keith CT, Borisy AA, Stockwell BR. “Multicomponent Therapeutics for Networked Systems” Nat Rev Drug Discov. 4(1):71 (2005) synergies Profile = vector of all scores involving a probe 1 1.5 Failed plates and bad wells flagged for exclusion 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 A B C 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 A e HCT p53 +/+ HCT p53 -/- 0.5 1 core fo (uM) .83 1.7 (uM) .83 1.7 Keith CT, Borisy AA, Stockwell BR. “Multicomponent Therapeutics for Networked Systems” Nat Rev Drug Discov. 4(1):71 (2005) Zimmermann G, Lehar J & Keith CT. “Multi-target therapeutics: when the whole is greater than the sum of the parts” Drug Disc Tod 12(1): 34 Profile = vector of all scores involving a probe Expect similar mechanism correlated profiles 0 0.5 C D E F B C D E Score nhib -0.5 0 Sc lysin ( 1 .41 lysin ( 1 .41 Zimmermann G, Lehar J & Keith CT. “Multi-target therapeutics: when the whole is greater than the sum of the parts” Drug Disc Tod 12(1): 34 Lehar J et al. “Chemical combination effects predict connectivity in biological systems” Mol Syst Biol 3:80 (2007) Eg: p53 pathway modulator + kinase inhibitor Expect similar mechanism correlated profiles -1.5 -1 -0.5 Intra plate artifacts quantified with control wells and G H I J E F G H S ase in -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -1.5 -1 r = 0.57 ± 0.06 Probe number ascapl .1 .21 ascapl .1 .21 Lehar J et al. “Chemical combination effects predict connectivity in biological systems” Mol Syst Biol 3:80 (2007) potential secondary target of p53 pathway 0 20 40 60 80 100 120 140 160 180 -1.5 Intra plate artifacts quantified with control wells and corrected computationally J K L M N I J K L Score for A Kina -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 -1.5 Fa 0 N=1 Fa 0 N=1 Combination High Throughput Screening potential secondary target of p53 pathway probe whose effect is unmasked in corrected computationally 2005 Q4 Mech Onc MTM Sparse6 / aheilbut-2006-09-06-1 / CT00032194 N O P M N O P Combination High Throughput Screening probe whose effect is unmasked in absence of p53 Cluster analysis HCT116 Compounds with unstable single agent activity Plate Responses 2005 Q4 Mech Onc MTM Sparse6 / aheilbut-2006-09-06-1 / CT00032194 2005 Q4 Mech Onc MTM Sparse6 / aheilbut-2006-09-06-1 / CT00032194 P P53 pathway inhib P53 pathway inhib absence of p53 Clustered profiles by score correlation S D W 3 M HCT116 Score Compounds with unstable single agent activity Plate Responses after Quality Control Sparse Dose Matrix P53 pathway inhib P53 pathway inhib Cell based phenotypic assays Highlights Clustered profiles by score correlation Many probes with similar annotations S 1 S I H P C C G C P P F L U C S M D 2 M S D identified and excluded from analysis Cell based phenotypic assays Highlights Many probes with similar annotations group together T M A 3 P R C M C S P A D T G 5 E H O R S S 1 Statins Preserve biological networks to measure effects in a disease relevant context, and provide group together Profiles contain information on mechanism M T P N N C T C S M V D C E F T I M G B L C T M Dose Response Curve Statins Tubulin inhibitors an opportunity to observe target interactions. Comprehensive, well sampled survey Chemical library Profiles contain information on mechanism A A I A 4 R E J S 8 B F A C S P A I M T C M T 4. Quantifying combination effects I X 1:10? Dose Response Curve Tubulin inhibitors MAPK inhibitors Comprehensive, well sampled survey of antiproliferative chemical combinations Chemical library 2,400+ bioactive agents include: T R L H N S C T C N A M A A U P T N T 2 E A A 4. Quantifying combination effects Different types and patterns of synergy: ect I ug Y 1:1? 1:10? MAPK inhibitors of antiproliferative chemical combinations 2,400+ bioactive agents include: • approved pharmaceutical ingredients J C A R C 2 I R U U R D C O B N E B G 3 T T R Different types and patterns of synergy: Effe Dru 10:1? • approved pharmaceutical ingredients • chemical probes with known targets C H W F P S A G H N 2 F D T A P R W F t T J C potency shift vs. efficacy boost at unknown ratios Concentration X Drug X RTK inhibitors Synergistic antiproliferative combinations are rare • chemical probes with known targets • drugs in development A S A D E T C N L T C S Z D O S G P F T P C Selected Hits M) Dose matrices test many doses and ratios Inhibition (%) Dose Matrix Top view Response Surface RTK inhibitors Synergistic antiproliferative combinations are rare • drugs in development • biologics (antibodies, proteins, siRNA) F R Z S A Probes, by profile similarity Selected Hits Combinations for further investigation: (uM 10 Dose matrices test many doses and ratios Inhibition (%) Dose Matrix Top view Response Surface Synergistic hits reflect both known biology and novel interactions • biologics (antibodies, proteins, siRNA) Combinations for further investigation: d B 3 2.5 1 Response = Inhibition = (untreated-treated)/untreated 5 4 n step Y ion C Y 80 90 100 100 80 80 90 100 100 80 %) Synergistic hits reflect both known biology and novel interactions Dose matrix data . synergistic, selective, novel mechanisms ound 039 .16.63 Response = Inhibition = (untreated-treated)/untreated 3 2 entration rug Y centrati 40 50 60 70 80 60 40 40 50 60 70 80 60 40 ition I (% HCT116 Inhibition (%) DT-1032204 71 70 63 Dose matrix data Inhibition (%) DT-1024585 81 87 89 mpo 0 .01 .0 1 0 Conce Dr Conc 0 00 .00 0 00 0.00 0 10 20 30 5 1.0 .0 20 0 0 00 .00 0 00 0.00 0 10 20 30 5 1.0 .0 20 0 Inhibi MRC9 HCT116 uM) 26 79 43 71 44 70 46 63 46 Subset being followed up with DNA damage and loss of p53 Orthogonally titrated compounds generate 20 81 25 87 65 89 85 Some hits reflect known biology Com Compound A (uM 0 .01 .039.16 .63 2.5 10 Compare response to Loewe additivity model 0 0 1 2 3 4 5 Concentration step Drug X 0.00 0.01 0.02 0.04 0.08 0.16 0.31 0.63 1.25 2.50 5.0 10. 0.00 0.01 0.02 0.04 0.08 0.16 0.31 0.63 1.25 2.50 5.0 10 C o m p o un d B ( u M ) C o m po und A (u M ) 0.0 .03 .13 .25 0.5 1 .03 .13 .25 0.5 1. C o n ce ntra t ion X C o n c e n tra tio n Y .06 .06 0.00 0.01 0.02 0.04 0.08 0.16 0.31 0.63 1.25 2.50 5.0 10. 0.00 0.01 0.02 0.04 0.08 0.16 0.31 0.63 1.25 2.50 5.0 10 C o m p o un d B ( u M ) C o m po und A (u M ) 0.0 .03 .13 .25 0.5 1 .03 .13 .25 0.5 1. C o n ce ntra t ion X C o n c e n tra tio n Y .06 .06 hib hib enin (u 2.9 8.8 -1.6 8.8 19 34 32 36 33 Subset being followed up with DNA damage and loss of p53 a two dimensional dose response matrix. 2 18 4.5 27 17 44 38 Some hits reflect known biology SAMD and ODC inhibs cut off parallel metabolic (expected response for drug-with-itself) Concentration C X C o n ce n Y C o n ce n Y C inh C inh Apige .98 2 =1-2 0 -1.5 -1.6 15 19 20 18 25 32 28 30 32 33 -.4 1.7 4.7 1.7 .1 .4 15 21 33 SAMD and ODC inhibs cut off parallel metabolic pathways in spermine biosynthesis Isobologram Isobologram Additive model Excess (data-model) Observed data ODC ODC 0 Methylglyoxal bis(guanylhydrazone) 0 .41 1.2 3.7 11 33 N= 0 15 20 25 28 32 Synergies and antagonisms can be Methods Methylglyoxal bis(guanylhydrazone) 0 .41 1.2 3.7 11 33 -.4 4.7 1.7 .4 15 33 pathways in spermine biosynthesis Synergy Score = Volume between data and Loewe EC Y 1 d d i t i v i t y 1 d d i t i v i t y O O SAMD inhib SAMD inhib Methylglyoxal bis(guanylhydrazone) dihydrochloride hydrate (uM) identified over many doses and ratios. Methods dihydrochloride hydrate (uM) Synergy Score = Volume between data and Loewe C Y /E . 5 A d 7 5 . 5 A d 7 5 SAMD inhib SAMD inhib Inhibition (%) 2 DT-1032209 -5.8 -12 40 Cell Culture Hits suggest novel biological interactions near 0. on ( %)= 7 0. on ( %)= 7 Inhibition (%) DT-1024587 A549 HCT116 b b M) 17 52 -3.1 -5.8 -10 -12 -.7 40 64 Quantify synergy relative to reference Cell Culture Cell lines were obtained from ATCC and grown in RPMI-1640 media with 10% FBS, 2 mM glutamine, and 1% penicillin/streptomycin. Protein modification inhibitor + Kinase inhibitor Isobologram shows amount of potency shifting Lin 0 0.5 1 0 I n h i b iti o 0 0.5 1 0 I n h i b iti o 0 1.9 -.9 7.2 16 70 76 nhib nhib 490 (uM 1.9 5.8 3.3 8.2 -6.2 -7.7 -7.5 11 55 models. Cell lines were obtained from ATCC and grown in RPMI-1640 media with 10% FBS, 2 mM glutamine, and 1% penicillin/streptomycin. P53-null HCT116 cells were obtained from the Vogelstein Lab at Johns Hopkins. Protein modification inhibitor + Kinase inhibitor Suggests an additional mode of regulation of Linear C X /EC X 0 0.5 1 0 0.5 1 -1.2 -1.5 .4 1.5 4.2 30 54 se in se in AG-4 .64 1 =1-2 -1.2 -.1 3.3 -8.5 -6.2 -4.1 -7.1 -.3 -7.5 3.5 6.3 50 55 Compounds Suggests an additional mode of regulation of a specific kinase signaling pathway X X ∑ 1.6 -1.2 .4 3 4.2 17 54 inas inas 0 Trichostatin A (uM) 0 3e-3 .0091 .027 .081 .24 N= -1.2 -8.5 -4.1 -.3 3.5 50 cHTS TM provides a powerful tool to: Compounds Stock solutions of each compound were prepared in DMSO, and serial dilutions were created in 384 well plates using a MiniTrak (PerkinElmer). Compounds were diluted 1:100 into culture media to create 10X stock on the day of the assay. Combinations were created by a specific kinase signaling pathway 5. Screen Results A549 MRC9 HCT116 Trichostatin A (uM) 0 3e-3 .0091 .027 .081 .24 0 -2.4 4.3 7.1 15 30 Ki Ki Trichostatin A (uM) cHTS provides a powerful tool to: • Discover multi-target mechanisms and (PerkinElmer). Compounds were diluted 1:100 into culture media to create 10X stock on the day of the assay. Combinations were created by diluting (1:10) the plates containing individual compounds into assay plates. 5. Screen Results Full screen performed in HCT116 cells; A549 MRC9 HCT116 Trichostatin A (uM) Prot. Modif inhib Prot. Modif inhib • Discover multi-target mechanisms and therapies diluting (1:10) the plates containing individual compounds into assay plates. Assays Potential therapeutic combination Full screen performed in HCT116 cells; HCT116 MRC9 therapies • Characterize patterns of drug synergy Assays Cells were seeded at 1500 cells per well (A549, HCT116) or 3000 cells per well (MRC9) in 384 well plates and allowed to recover overnight at DNA Damage Agent + Kinase Inhibitor (uM) 3 40 (uM) 3 40 120x120 subset tested in A549 and MRC9 age age • Characterize patterns of drug synergy and antagonism Cells were seeded at 1500 cells per well (A549, HCT116) or 3000 cells per well (MRC9) in 384 well plates and allowed to recover overnight at 37°C. Compounds were added and assay plates incubated at 37°C for proliferation differences to occur. Total time of compound treatment was 72 hours. Effect on cell proliferation was quantified using ATP-Lite 1Step (PerkinElmer), and luminescence was measured using a Wallac Novel interaction between two targets in ipivoxil 4.4 13 ipivoxil 4.4 13 ama dama and antagonism • Guide clinical cotherapy development was 72 hours. Effect on cell proliferation was quantified using ATP-Lite 1Step (PerkinElmer), and luminescence was measured using a Wallac VictorV or Envision plate reader. Percent inhibition of proliferation was calculated compared to vehicle-treated controls. Novel interaction between two targets in clinical development fovir Di .49 1.5 3 fovir Di .49 1.5 2 NA da NA d • Guide clinical cotherapy development decisions VictorV or Envision plate reader. Percent inhibition of proliferation was calculated compared to vehicle-treated controls. Acknowledgements clinical development Selective synergy in HCT116 over MRC9 Adef 0 N=1-3 Adef 0 N=1-2 Synergy Score for the 120x120 probe subset DN DN decisions Acknowledgements We would like to thank John Healy and Yiqun Bai for assistance with screening, and Margaret Lee and Ricky Rickles for valuable advice. Selective synergy in HCT116 over MRC9 Kinase inhib Kinase inhib We would like to thank John Healy and Yiqun Bai for assistance with screening, and Margaret Lee and Ricky Rickles for valuable advice. Kinase inhib AACR Annual Meeting, Los Angeles, April 2007 AACR, Los Angeles, CA, USA, April 2007 AACR Annual Meeting, Los Angeles, April 2007 AACR, Los Angeles, CA, USA, April 2007