Individualized Systems Medicine Strategy to Tailor Treatments for Patients with Chemorefractory Acute Myeloid Leukemia

Home , Dactolisib, PDCD10, Pipobroman, Tivozanib, Vatalanib, Veliparib

Published OnlineFirst September 20, 2013; DOI: 10.1158/2159-8290.CD-13-0350

RESEARCH ARTICLE

Tea Pemovska 1 , Mika Kontro 2 , Bhagwan Yadav 1 , Henrik Edgren 1 , Samuli Eldfors 1 , Agnieszka Szwajda 1 , Henrikki Almusa 1 , Maxim M. Bespalov 1 , Pekka Ellonen 1 , Erkki Elonen 2 , Bjørn T. Gjertsen 5 , 6, Riikka Karjalainen 1 , Evgeny Kulesskiy 1 , Sonja Lagström 1 , Anna Lehto 1 , Maija Lepistö 1 , Tuija Lundán 3 , Muntasir Mamun Majumder 1 , Jesus M. Lopez Marti 1 , Pirkko Mattila 1, Astrid Murumägi 1, Satu Mustjoki 2, Aino Palva 1 , Alun Parsons 1 , Tero Pirttinen 4 , Maria E. Rämet 4 , Minna Suvela 1 , Laura Turunen 1, Imre Västrik 1 , Maija Wolf 1 , Jonathan Knowles 1 , Tero Aittokallio 1 , Caroline A. Heckman 1 , Kimmo Porkka 2 , Olli Kallioniemi 1 , and Krister Wennerberg 1

ABSTRACT We present an individualized systems medicine (ISM) approach to optimize cancer drug therapies one patient at a time. ISM is based on (i) molecular profi ling and ex vivo drug sensitivity and resistance testing (DSRT) of patients’ cancer cells to 187 oncology drugs, (ii) clinical implementation of therapies predicted to be effective, and (iii) studying consecutive samples from the treated patients to understand the basis of resistance. Here, application of ISM to 28 samples from patients with acute myeloid leukemia (AML) uncovered fi ve major taxonomic drug-response subtypes based on DSRT profi les, some with distinct genomic features (e.g., MLL gene fusions in subgroup IV and FLT3 -ITD mutations in subgroup V). Therapy based on DSRT resulted in several clinical responses. After progression under DSRT-guided therapies, AML cells displayed signifi cant clonal evolution and novel genomic changes potentially explaining resistance, whereas ex vivo DSRT data showed resistance to the clinically applied drugs and new vulnerabilities to previously ineffective drugs.

SIGNIFICANCE: Here, we demonstrate an ISM strategy to optimize safe and effective personalized cancer therapies for individual patients as well as to understand and predict disease evolution and the next line of therapy. This approach could facilitate systematic drug repositioning of approved targeted drugs as well as help to prioritize and de-risk emerging drugs for clinical testing. Cancer Discov; 3(12); 1416–29. ©2013 AACR.

See related commentary by Hourigan and Karp, p. 1336.

Authors’ Affi liations: 1 Institute for Molecular Medicine Finland, FIMM; T. Aittokallio, C.A. Heckman, K. Porkka, O. Kallioniemi, and K. Wennerberg 2 Hematology Research Unit Helsinki, Helsinki University Central Hospital, shared senior authorship of this article. University of Helsinki, Helsinki; 3 Department of Clinical Chemistry and Current address for M.M. Bespalov: Stem cells and Neurogenesis Unit, TYKSLAB, Turku University Central Hospital, University of Turku, Turku; Division of Neuroscience, San Raffaele Scientifi c Institute, Milan, Italy. 4Department of Internal Medicine, Tampere University Hospital, Tampere, Finland; 5 Department of Clinical Science, Hematology Section, University Corresponding Author: Krister Wennerberg, Institute for Molecular Medi- of Bergen; and 6 Department of Internal Medicine, Hematology Section, cine Finland, University of Helsinki, Tukholmankatu 8, 00290 Helsinki, Haukeland University Hospital, Bergen, Norway Finland. Phone: 358-9-191-25764; Fax: 358-9-191-25737; E-mail: krister .wennerberg@fi mm.fi Note: Supplementary data for this article are available at Cancer Discovery Online (http://cancerdiscovery.aacrjournals.org/). doi: 10.1158/2159-8290.CD-13-0350 T. Pemovska and M. Kontro shared fi rst authorship of this article. ©2013 American Association for Cancer Research.

1416 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

INTRODUCTION therapy ( 10 ). In light of the genomic and molecular diversity of AML, and its continuous evolution in response to chemother- Adult acute myeloid leukemia (AML) is a prototype exam- apy, it is important to better understand the potential utility of ple of the challenges of modern cancer drug discovery, devel- all targeted cancer drugs that are already available in the clinic. opment, and patient therapy. With the exception of the These drugs could be systematically repurposed as off-label retinoic acid–sensitive acute promyelocytic leukemia subtype, indications to responding subgroups of AML. Furthermore, molecularly targeted therapeutic approaches for AML are yet comparative information on the effi cacy of the hundreds of to be translated into clinical advances. The disease has tra- emerging targeted anticancer agents, as well as their potential ditionally been subdivided into different subtypes (M0–M7) combinations, in patient-derived ex vivo samples could dramati- based on cellular lineage and biomarkers ( 1 ). Current World cally help prioritize clinical development of such agents. Health Organization classifi cation refl ects the fact that a grow- To facilitate testing of already clinically available drugs as ing number of AML cases can be categorized on the basis of well as emerging targeted inhibitors for patients with AML, their underlying genetic abnormalities that defi ne distinct clin- we undertook a comprehensive functional strategy to directly icopathologic entities (2 ). Genomic changes in AML are now determine the drug dependency of cancer cells based on ex vivo relatively well understood, with each AML sample containing drug sensitivity and resistance testing (DSRT). First, we applied roughly 400 genomic variants, of which an average of 13 reside a systematic large panel of drugs covering both cancer chemo- in coding regions (3, 4). Recurrent changes have highlighted therapeutics and many clinically available and emerging molec- potential driver genes, including NPM1, CEBPA , DNMT3A , ularly targeted drugs. Second, we developed a new way to score TET2 , RUNX1 , ASXL1 , IDH2 , and MLL , with mutations in FLT3 , for differential drug response in AML cells as compared with IDH1 , KIT , and RAS genes modifying the disease phenotype ( 5 ). the effi cacy of these drugs in normal bone marrow cells. Third, Although several of the recurrent genetic alterations link to trac- we verifi ed DSRT predictions in vivo by treating patients with table drug targets, genetic testing of patients with AML has yet AML with off-label drugs. Fourth, we assessed the molecular to result in effective personalized or stratifi ed therapies. mechanisms underlying development of cancer progression Patients with AML have a poor outcome, with a 5-year and drug resistance by repeat sampling from relapsed patients, survival of 30% to 40% (6, 7). The standard therapy for most followed by genomic and transcriptomic profi ling as well as adult patients with AML is conventional chemotherapy con- a new DSRT analysis to understand both coresistance as well sisting of the nucleoside analog cytarabine combined with as new vulnerabilities. Taken together, we term this approach a topoisomerase II inhibitor (8, 9). A number of second-line individualized systems medicine (ISM; Fig. 1 ). treatment options have been applied in patients with AML after Here, we demonstrate that the ISM strategy made it possible to relapse, but the response rates have remained low. Furthermore, (i) create a taxonomy of comprehensive drug responses in AML, patients with AML at relapse exhibit an increased number of (ii) identify clinically actionable AML-selective targeted drugs, genetic alterations, which can be attributed to disease progres- (iii) clinically apply such therapies for individual chemorefrac- sion and/or DNA-damaging agents used for routine chemo- tory AML patients predicted to be sensitive to targeted drugs,

DECEMBER 2013CANCER DISCOVERY | 1417

RESEARCH ARTICLE Pemovska et al.

Individualized drug treatment selection

DSRT

100 Effective drugs 80 60 40

% Survival 20 Selective DSS Selective Resistant drugs 0 –9 –8 –7 –6 –5 Log conc (mol/L) Understanding biology of disease Diagnosis Understanding Relapse 1 Molecular profiling Result drug sensitivity and resistance Relapse 2 Genome Transcriptome Signalome Rapid introduction of therapies

Figure 1. Functional ISM platform for improved AML therapy. The platform involves (i) comprehensive direct DSRT of 187 approved and investigational oncology compounds in ex vivo primary cells from serial AML samples; (ii) clinical implementation of testing results in individual patients with relapsed and refractory disease; (iii) deep molecular and genomic profi ling of the patients with AML from consecutive samples before and after relapse and drug resistance for monitoring disease progression and clonal evolution; and (iv) integrating drug sensitivity, next-generation sequencing, and clinical follow-up data to understand the biology of disease, drug sensitivity, and resistance that can lead to rapid introduction of novel therapies to the clinic. DSS, Drug Sensitivity Score. and (iv) follow individually optimized therapies in patients by out a pilot test for the implementation of optimized therapies analysis of the clonal evolution of leukemic cells and molecular for 8 patients with chemorefractory AML. Patient treatment profi ling to understand mechanisms of drug resistance. outcome was assessed by clinical criteria, but also by genomic profi ling to understand the clonal architecture underlying drug RESULTS response and emerging resistance. Data on recurrent, paired samples identifi ed drugs that showed effi cacy after the develop- ISM Strategy for Personalized AML Therapy ment of drug resistance and highlighted drugs that could be To uncover the mechanisms of drug response and resistance applied in combination. This approach provided the framework as well as to monitor therapy response at the level of individual for a real-time continuous cycle of learning and optimization of AML subclones, we combined DSRT and deep molecular profi l- therapies, one patient at a time, thereby creating an individual- ing data. DSRT was implemented to AML blast cells ex vivo using ized systems medicine process for improving cancer care. a comprehensive set of 187 drugs, consisting of conventional chemotherapeutics and a broad range of targeted oncology DSRT Identifi ed Signal Transduction Inhibitors as compounds (Table 1 and Supplementary Table S1). Each drug Putative AML-Selective Drugs was tested over a 10,000-fold concentration range, allowing for Ex vivo DSRT was performed on 28 samples obtained from the establishment of accurate dose–response curves for each 18 patients with AML (Supplementary Table S3). Eighteen sam- drug in each patient and control sample with identifi cation of ples were collected at relapse and 10 at diagnosis, mainly from half-maximal effective concentrations and maximal responses patients with adverse or intermediate cytogenetic risk (according (Supplementary Table S2). Although these responses refl ect the to European LeukemiaNet; ref. 9 ). Out of the 187 drugs tested, drug–sample interaction, the detailed interpretation of curve sDSS indicated the most selective ex vivo effective drugs for each parameters is diffi cult. To overcome this issue, we found that the individual patient, with a focus on leukemia-specifi c effi cacies by most informative way to assess quantitative drug sensitivities comparing responses with normal bone marrow mononuclear was to convert the information to a Drug Sensitivity Score (DSS), cells. The results were expressed as a patient-specifi c water- a metric used to determine the area under the dose–response fall plot ( Fig. 2A ). Several targeted drugs exhibited a selective curves. Importantly, we developed a new scoring system for response in a subset of the AML samples, with only a minimal assessing leukemia-selective effects by comparing the DSS in the response in the control samples (Fig. 2B and C and Supplemen- AML cells with those of healthy donors (selective DSS: sDSS). tary Fig. S2), suggesting that these AML samples were addicted Analysis of the drug-response data revealed that DSRT is highly to the signaling pathways inhibited by the drugs. In contrast, reproducible ( r = 0.98; Supplementary Fig. S1) and that all con- the average sensitivity to conventional chemotherapeutics did trol samples exhibited similar response profi les (Supplementary not signifi cantly differ between the patient samples and controls Fig. S1). The analysis of DSRT results was completed in 4 days, (Fig. 2B and D). This refl ects the known limited therapeutic rapid enough for clinical implementation. Indeed, we carried window for these drugs and the diffi culty in predicting their

1418 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

Table 1. Drug classes and drugs represented in the DSRT screening platform

Classes of drugs Drugs Alkylating agents Altretamine, azacitidine, busulfan, carboplatin, carmustine, chlorambucil, cyclophosphamide, dacarbazine, ifosfamide, lomustine, pipobroman, procarbazine, streptozocin, temozolomide, thioTEPA, uracil mustard Antimetabolites Allopurinol, capecitabine, cladribine, clofarabine, cytarabine, decitabine, fl oxuridine, fl udarabine, fl uorouracil, gemcitabine, mercaptopurine, methotrexate, nelarabine, pentostatin, thioguanine Antimitotics ABT-751, docetaxel, indibulin, ixabepilone, paclitaxel, patupilone, S-trityl-L-cysteine, vinblastine, vincristine, vinorelbine Antitumor antibiotics Bleomycin, dactinomycin, mitomycin C, plicamycin BCL-2 inhibitors Navitoclax, obatoclax HDAC inhibitors Belinostat, CUDC-101, entinostat, panobinostat, tacedinaline, vorinostat Hormone inhibitors Abiraterone, aminoglutethimide, anastrozole, exemestane, fi nasteride, fl utamide, fulvestrant, goserelin, letrozole, megestrol acetate, nilutamide, raloxifene, tamoxifen HSP90 inhibitors Alvespimycin, BIIB021, NVP-AUY922, tanespimycin Immunomodulators Celecoxib, dexamethasone, fi ngolimod, imiquimod, lenalidomide, levamisole, methylprednisolone, plerixafor, prednisolone, prednisone, tacrolimus, thalidomide Kinase inhibitors, AGC Alisertib, AT9283, AZD1152-HQPA, BI2536, bryostatin 1, danusertib, enzastaurin, fasudil, midostaurin, MK-2206, ruboxistaurin, sotrastaurin, UCN-01 Kinase inhibitors, CAMK AZD7762, PF-00477736 Kinase inhibitors, CK1 MK-1775 Kinase inhibitors, CMGC Alvocidib, AZ 3146, doramapimod, palbociclib, seliciclib, SNS-032 Kinase inhibitors, PIKL AZD8055, dactolisib, idelalisib, OSI-027, PF-04691502, pictilisib, XL147, XL765 Kinase inhibitors, STE Pimasertib, refametinib, selumetinib, trametinib Kinase inhibitors, TK Afatinib, axitinib, BMS-754807, canertinib, cediranib, crizotinib, dasatinib, dovitinib, EMD1214063, erlotinib, foretinib, gandotinib, gefi tinib, imatinib, lapatinib, lestaurtinib, linsitinib, masitinib, MGCD-265, motesanib, nilotinib, pazopanib, ponatinib, quizartinib, regorafenib, ruxolitinib, saracatinib, sorafenib, sunitinib, tandutinib, tivozanib, tofacitinib, vandetanib, vatalanib Kinase inhibitors, TKL Vemurafenib PARP inhibitors Iniparib, olaparib, rucaparib, veliparib Proteasome inhibitors Bortezomib, carfi lzomib Rapalogs Everolimus, sirolimus, temsirolimus Smoothened (Hh) inhibitors Erismodegib, vismodegib Topoisomerase I/II inhibitors Amonafi de, camptothecin, daunorubicin, doxorubicin, etoposide, idarubicin, irinotecan, mitoxantrone, teniposide, topotecan, valrubicin Miscellaneous antineoplastics Bexarotene, hydroxyurea, mitotane, tretinoin Other 2-Methoxyestradiol, anagrelide, bimatoprost, pilocarpine, Prima-1 Met, serdemetan, tarenfl urbil, tipifarnib, XAV-939, YM155

Abbreviations : AGC: PKA, PKG, and PKC kinase group; CAMK: calcium/calmodulin–dependent protein kinase group; CK1: casein kinase group; CMGC: CDK, MAPK, GSK3, and CLK kinase group; PIKL: phosphoinositide 3-kinase (PI3K)-like (PI3K inhibitors and inhibitors of related atypical kinases: mTOR, DNA-PK, ATM, ATR); STE: sterile kinase group; TK: tyrosine kinase group; TKL: tyrosine kinase-like group.

clinical effi cacy based on ex vivo testing. Interestingly, cytarabine, such as temsirolimus in 9 (32%), foretinib in 9 (32%), ponatinib in an established and effective AML drug, showed higher selective 7 (25%), ruxolitinib in 7 (25%), dactolisib in 7 (25%), MK-2206 in effi cacy against AML cells than other cytotoxic agents ( Fig. 2B 6 (21%), sorafenib in 6 (21%), and quizartinib in 5 (18%; Fig. 2E ). and Supplementary Fig. S2), suggesting the feasibility of com- Thus, we identifi ed several highly (ex vivo) AML-selective drugs bining cytarabine with promising molecularly targeted drugs in that are not currently approved for AML but that are approved the future clinical testing of relapsed or primary AML. for other cancer indications, and would therefore be available Signifi cant anticancer-selective effects were observed for the in the clinic. Furthermore, a number of effective investiga- tyrosine kinase inhibitors (TKI) dasatinib in 10 of the AML tional drug classes were seen, such as AKT inhibitors and ATP- patient samples (36%) and sunitinib in 10 (36%), MAP–ERK competitive mTOR inhibitors, which could be prioritized for kinase (MEK) inhibitors such as trametinib in 10 (36%), rapalogs future clinical studies of chemorefractory AML.

DECEMBER 2013CANCER DISCOVERY | 1419

RESEARCH ARTICLE Pemovska et al.

Controls AB393_3 25 AML samples 20 35 15 30 10 25 5 20

sDSS 0 DSS 15 − 5 10 − 10 5 − 15 0

YM155 abine Alisertib BIIB021 Ponatinib

Paclitaxel ametinib Sunitinib Idarubicin Docetaxel Entinostat Cladribine Vorinostat ytar Dasatinib Ruxolitinib Navitoclax AZD8055 Trametinib Pimasertib Topotecan Idarubicin Danusertib Azacitidine CUDC-101 Fingolimod Vinblastine Teniposide VincristineTr Vinorelbine Saracatinib

Clofarabine C Lestaurtinib Selumetinib

Refametinib Clofarabine Prima-1 Met Serdemetan Gemcitabine Mitoxantrone Alvespimycin Panobinostat Lenalidomide Dactinomycin PF-00477736 NVP-AUY922 Tanespimycin Camptotechin

Dexamethasone Chemotherapy Targeted agents

Trametinib Idarubicin C 10 D 10

5 5

0 0 Z -score Z -score

−5 −5

−10 −10 C1 C4 C5 C2 C3 C7 C6 C2 C7 C5 C1 C4 C6 C3 718 250 800 800 250 718 1690 1064 1993 2095 1280 1497 1867 2080 1064 1993 1280 1497 1867 2095 1690 2080 252_3 600_3 600_1 252_5 560_1 600_2 252_2 393_1 252_1 560_9 393_3 784_2 784_1 393_3 600_3 393_1 252_3 252_2 784_1 252_1 560_1 252_5 784_2 600_2 600_1 560_9 1145_1 1145_6 1145_3 1145_3 1145_1 1145_6 4701001 4701001

E 40 Figure 2. Targeted compounds exhibit cancer-selective ex vivo drug responses in AML. A, waterfall plot that highlights the most potent 30 cancer-selective drugs for each individual patient as well as drugs that are most likely to exhibit resistance/no sensitivity. B, comparison of cancer selectivity of ex vivo drug responses (DSS) of four clinically 20 approved drugs, conventional chemotherapeutic agents (left), and four signal transduction inhibitors (right). The DSSs are compared between = 10 healthy bone marrow samples (controls; n 7) and AML patient samples (n = 28) with the range and median DSS value depicted. C and D, the distribution of the sensitivity to trametinib (MEK inhibitor) and

Significant responders (%) 0 idarubicin (topoisomerase II inhibitor) in 28 AML patient and 7 control b samples expressed as Z -score (SD from the average control DSS drug response). E, percentages of AML patient samples selectively respond- metinib onatinib Dasatinib Sunitinib ForetinibP MK-2206Sorafenib ing ex vivo to selected signal transduction inhibitors as assessed by Tra RuxolitinibDactolisi Quizartinib Temsirolimus sDSS, represented as a bar graph.

Taxonomy of AML Based on the Comprehensive signal-driven phenotype as seen by selective responses to a Drug-Response Profi les group of immunosuppressive drugs (e.g., dexamethasone or Overall drug-response patterns of the patient samples prednisolone), Janus-activated kinase (JAK) family kinase were visualized with unsupervised hierarchical clustering. inhibitors, and MEK inhibitors. Group III, IV, and V AMLs Although each individual sample showed a unique leukemia- were selectively sensitive to a broad range of TKIs, indicating selective drug-response profi le, the overall drug-response that they were driven or addicted to receptor tyrosine kinase profi les segregated the AML patient samples into fi ve robust signaling pathways. Group III AMLs displayed a similar functional subgroups (Fig. 3 ; Supplementary Figs. S3 and sensitivity pattern to HSP90, HDAC, and phosphoinositide S4). Thus, despite the underlying genomic and phenotypic 3-kinase (PI3K)/mTOR inhibitors as group IV and V, albeit variability in AML, similar drug sensitivity patterns were with lower selectivity. Group IV AMLs were especially sensi- observed among the AML patient samples for certain drug tive to MEK and PI3K/mTOR inhibitors, whereas group V classes. Compared with controls, all fi ve groups showed AMLs showed selective responses to receptor TKIs target- increased sensitivity to navitoclax, a BCL-2/BCL-XL inhibitor, ing ABL, VEGF receptor (VEGFR), platelet derived growth HSP90 inhibitors, and histone deacetylase (HDAC) inhibi- factor receptor (PDGFR), FLT3, KIT, PI3K/mTOR as well tors. Group I exhibited a strong selective response to navito- as topoisomerase II inhibitors (Fig. 3 ). Overall, 19 of clax and lack of sensitivity to the remainder of the tested 28 samples were sensitive to tyrosine kinase inhibition, compounds. Group II AMLs were largely nonresponsive to correlating well with the fi ndings in the recent study by receptor TKIs, but instead showed a potential infl ammatory Tyner and colleagues (11 ).

1420 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

Figure 3. Functional taxonomy of AML based on comprehensive drug response and mutation profi les. A dendrogram of the DSRT responses shows clustering of the AML patient samples in fi ve functional groups (group I, II, III, IV, and I II III IV V V). Samples were clustered with the complete linkage method using Euclidean distance measures. This approach provides a data-driven way to classify samples based on drug effi cacies 560_9 600_2 1690 4701001 560_1 252_1 252_5 600_1 1064 393_3 252_2 1993 1145_1 252_3 1145_6 1497 784_1 250 2095 800 393_1 1280 600_3 1867 1145_3 718 784_2 and drugs based on their differential bioac- Disease stage DD D*RRD*DRD*RRDRRRD RRRRRRRRRDR tivity across patient samples. Sensitivity or nonsensitivity to either navitoclax, ruxolitinib, Navitoclax MEK inhibitors, dasatinib, quizartinib, sunitinib, Ruxolitinib PI3K/mTOR inhibitors, and topoisomerase II Dexamethasone inhibitors drives the sample groupings. The MEKi molecular profi les (signifi cant AML mutations PI3K/mTORi and recurrent gene fusions), disease stage, and Quizartinib adverse karyotype status of the patient sam- Sunitinib ples are also shown to illustrate the correlation Dasatinib between functional drug sensitivity and somatic Topoisomerase IIi mutation and cytogenetics data. D, diagnosis; D*, secondary AML diagnosis; R, relapsed and/or TP53 refractory. RUNX1 DNMT3A NRAS/KRAS MLL-X fusions WT1 ETV6–NTRK3 KIT NUP98–NSD1 FLT3-ITD NPM1 IDH1/IDH2 PTPN11

Adverse karyotype FAB subtypes M1M2 M2 M2 M2 M2 M5M5M5 M5 M5 M5 M5 M5 M5 M5 M5 M1M1

Taxonomy of Cancer Drugs Based on the possibilities that could not have easily been discovered with- Comprehensive Drug-Response Proﬁ les in AML out the unbiased DSRT data.

The hierarchical clustering also stratifi ed the drugs based Genomic and Molecular Findings Underlying the on the variability of responses among the patients with Drug-Response Profi les AML (Supplementary Fig. S3). In this unsupervised analysis, To test whether the molecular profi les of the patient sam- drugs with similar modes of action clustered together, such ples correlated with the overall drug responses, we compared as the PI3K/mTOR inhibitors, MEK inhibitors, HSP90 and the distribution of signifi cant AML mutations and recurrent HDAC inhibitors, VEGFR-type TKIs, PDGFR inhibitors, and gene fusions ( 4 ) with the DSRT-driven clustering ( Fig. 3 ). ABL-like kinase inhibitors, antimitotics, and topoisomerase FLT3-mutated samples appeared in several different functional II inhibitors. Thus, the unsupervised clustering of drugs groups, but all patient samples in group V, the most tyrosine into subgroups defi ned by their intended targets strongly kinase–dependent group, carried these mutations. Hence, the supports the consistency and reproducibility of the DSRT group V drug-response pattern is a strong indicator of driver analysis as well as its ability to acquire biologically and medi- FLT3 mutations, and, as expected, an FLT3 mutation is an cally relevant information. However, there were also notable indicator that the cells are likely to respond to TKI treatment. deviations from the expected patterns. Importantly, the FLT3 Several FLT3 inhibitors, such as quizartinib, sunitinib, and inhibitor quizartinib clustered with the topoisomerase II foretinib, were among the most selective drugs for group V. inhibitors but not with other TKIs. Furthermore, the recently Importantly, TKIs without FLT3 inhibitory activity, such as approved BCR–ABL1 and FLT3 inhibitor ponatinib clustered dasatinib, were highly effective in this group (P = 0.03), suggest- with cytarabine, HSP90, and HDAC inhibitors and not with ing that FLT3-driven AMLs are also dependent on other tyrosine other TKIs. These unexpected links may represent underlying kinase signals. Furthermore, mutations in RAS genes correlated key molecular mechanisms of these drugs in the AML con- signifi cantly with ex vivo sensitivity to MEK inhibitors ( P = text, including unexpected off-target effects. Furthermore, 0.001). Samples from two patients with MLL fusions clustered these drug-clustering patterns from human AML patient together in group IV, possibly linking MLL fusions with sensitiv- specimens ex vivo may in the future be critically important in ity to MEK inhibitor sensitivity. In addition, an enrichment of designing novel therapeutic combination strategies for clini- TP53 mutations in groups I and II was observed (3 of 4 cases), cal trials in AML. Therefore, this provides new combinatorial and all patient samples in taxonomic group II were associated

DECEMBER 2013CANCER DISCOVERY | 1421

RESEARCH ARTICLE Pemovska et al. with adverse karyotypes. Beyond these examples, the majority (q35;p15.5) was detected by RNA sequencing in sample 600_2. of the clustering of drug sensitivities could not be attributed to This oncogenic fusion ( 12 ) is relatively common in cytogeneti- obvious alterations in the AML tumor genomes. cally normal pediatric AML (13, 14), but relatively rare in adult AML ( 15 ). The NUP98 – NSD1 fusion was also detected in the DSRT Was Predictive of Clinical Responses and diagnostic sample (600_0) and all follow-up samples, suggest- Recapitulates Acquisition of Resistance In Vivo ing that this fusion was the initiating event in the development The results of DSRT were considered to be therapeutically of the patient’s disease. Exome sequencing revealed a diverse actionable if (i) a distinct leukemia-selective response pattern subclonal architecture highlighted by two FLT3 -ITD (Supple- was seen, (ii) drugs showing sensitivity were available for commentary Fig. S5A and S5B) and four different WT1 mutations passionate or off-label use without signifi cantly delaying the (Supplementary Fig. S6A and S6B; Supplementary Table S5; treatment, and (iii) no standard therapy was available. Accord- and Supplementary Methods). After induction chemotherapy, ing to European LeukemiaNet ( 9 ) response criteria, 3 of 8 the predominant FLT3 -ITD harboring subclone was no longer evaluable patients had a response to DSRT-guided therapy detectable. Instead, subclones containing WT1 mutations and [Supplementary Table S4; patient 600 (dasatinib–sunitinib– a second FLT3 -ITD emerged (Fig. 4D and E). The sensitivity temsirolimus): complete remission with incomplete platelet to quizartinib, sunitinib, and other FLT3 inhibitors indicated recovery (CRi); patients 718 (sorafenib–clofarabine) and 800 FLT3 as a disease driver in the 600_2 sample. In the DSRT- (dasatinib–clofarabine–vinblastine): morphologic leukemia- relapsed 600_3 sample, the dasatinib, sunitinib, and tem- free state]. Four other patients had responses that did not sirolimus therapy had further selected a specifi c subclone still meet the European LeukemiaNet criteria. Patient 560 showed containing the second FLT3 -ITD even though the response a rapid clearance of blasts in peripheral blood after 5 days to FLT3 inhibitors and other TKIs was lost. The broad loss of treatment (dasatinib–sunitinib), after which therapy was in drug response in the 600_3 sample was also accompanied discontinued because of gastrointestinal toxicity. Patient 252 with decreased phosphorylation of AKT (S473), CHK2 (T68), (AML with three prior relapses) had an 8-week progression-free CREB (S133), ERK1/2 (T202/Y204, T185/Y187), FAK (Y397), period during dasatinib monotherapy (bone marrow blasts p38α (T180/Y182), and STAT1 (Y701; data not shown). 65%–40%–70%). Patient 784 achieved a transient response with dasatinib–sunitinib–temsirolimus therapy: bone marrow blasts DSRT and Molecular Profi ling Defi ned Key Oncogenic decreased from 70% to 35%, but the treatment response was lost Signals and Mechanisms of Drug Resistance due to selection of a resistant clone. Patient 1145 had hemato- A second clinical case was a 37-year-old patient (784_1) who logic improvement during ruxolitinib–dexamethasone therapy. was diagnosed with a recurrent t(11;19)(q23;p13.1) transloca- Even patients with partial or transitional clinical responses had tion and corresponding MLL – ELL fusion gene. The patient had a profound effect on the clonal composition of the tumors, relapsed from three previous rounds of conventional therapy. including selection of potential drug resistance–associated Initial DSRT results (784_1) showed selective responses to mutations. Therefore, detailed genomic analysis of such cases is MEK inhibitors, rapalogs, and several TKIs, including dasat- important to measure the impact of therapy and to understand inib (Fig. 5A ). This patient was also treated with dasatinib, the potential mechanisms of response and resistance. sunitinib, and temsirolimus, which led to a rapid decrease in Here, we present in detail two clinical examples on the both peripheral leukocytosis and bone marrow blast counts, implementation of DSRT results in patients with AML, but the effect was short-lived. The ex vivo drug sensitivity of including the consecutive sampling and serial monitoring of the resistant sample (784_2) revealed that the cells had lost drug-sensitivity profi les and clonal evolution in vivo. In the their response to dasatinib and rapalogs, but preserved their fi rst case, the bone marrow cells from a relapsed and refractory response to ATP-competitive mTOR inhibitors (such as dac- 54-year-old patient (sample 600_2) with a normal-karyotype tolisib and AZD8055). Interestingly, the resistant sample gained AML FAB M5 were subjected to DSRT and deep molecular sensitivity to a TKI that was previously ineffective in the DSRT, profi ling. The patient had previously failed three consecutive BMS-754807 [insulin-like growth factor-I receptor (IGF-IR)/ induction therapies ( Fig. 4A ). The DSRT results highlighted Trk inhibitor], as well as crizotinib (TKI) and tipifarnib (farnesyl- dasatinib, sunitinib, and temsirolimus among the top fi ve transferase inhibitor), several topoisomerase II inhibitors, and most selective approved drugs. In an off-label compassion- immunomodulatory and differentiating compounds ( Fig. 5B ). ate use setting, the patient received a combination of these Using the DSRT data on kinase inhibitors with published targeted drugs, resulting in rapid reduction of the bone comprehensive biochemical profi ling data (16 ), we identifi ed marrow blast count and marked improvement in the poor putative kinases to which the cells were addicted. Impor- performance status. Concomitantly, the blood counts rapidly tantly, a comparison between the fi rst and the relapsed sam- normalized resulting in CRi. However, 30 days after achieving ple identifi ed a major switch in kinase addiction with a loss the CRi response, resistance emerged. A new DSRT analysis of addiction to Src family kinases, PI3K and p38 mitogen- from the relapsed sample (600_3) showed that the drugs used activated protein kinases (MAPK), and a gain of addiction to in patient treatment exhibited remarkably reduced anticancer ALK and Trk family receptor tyrosine kinases ( Fig. 5C ). activity as compared with the pretreatment sample (Fig. 4B ), The resistance to the dasatinib, sunitinib, and temsirolimus demonstrating a match between ex vivo and in vivo responses. treatment in this patient was not associated with any novel In this patient, the ex vivo responses to many other drugs were detected mutations or other genetic alterations, but coincided also strongly reduced in the relapsed sample ( Fig. 4C ). with more than 1,000-fold enrichment of two fusion tran- A fusion transcript joining NUP98 exon 12 and NSD1 exon scripts, ETV6–NTRK3 and STRN-ALK (Supplementary Fig. S7), 6 resulting from a cryptic chromosome translocation t(5;11) suggesting that resistance emerged from the selection of

1422 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

A 100 5.0 B Dasatinib Sunitinib Bone marrow blasts (%) 15 20 Neutrophils (x10E9) 15 80 10 10 sDSS 60 sDSS 5 5 % 2.5 40 10E9 0 0 600_2 600_3 600_2 600_3 Temsirolimus 20 15 10 0 0.0 0 50 100 150 180 200 220 5

Chemotherapy Das–Sun–Tem Das–Tem sDSS 0 –5 600_0 600_1 600_2 600_3 600_2 600_3 CDD 20 600_0 600_2 600_3

Clone1 74 % 15 FLT3-ITD #1

33% <10% Clone 3 BIIB021 WT1 #2 6% Plicamycin NUP98–NSD1 10 NVP-AUY922 20% Tivozanib FLT3-ITD #2 15% <10% Navitoclax Tanespimycin 600_3 sDSS Lestaurtinib WT1 #1 WT1 #4 Alvespimycin Ponatinib Clone 2 15% Clone 5 5 Cytarabine Entinostat Dasatinib 37% Axitinib WT1 #3 Etoposide Sorafenib Sunitinib Clone 4 90% PF-04691502 Quizartinib Temsirolimus Sirolimus AZD7762 0 Mitoxantrone Chemotherapy Das–Sun–Tem 0 5 10 15 20 600_2 sDSS E

NUP98–NSD1IST1(MP243-) fussion FLT3-ITDFLT3-ITD #1 WT1(-379?) #2 H2AFZ(V110G) (#1) SDR42E1(R303Q)WT1(-376R?)WT1(-370?) (#2)WT1(-380?) (#3) (#4) Clone 1 Clone 2 Clone 3 Clone 4 Clone 5

Figure 4. Clinical implementation of DSRT predictions and clonal evolution analysis in a heavily refractory AML patient. A, clinical follow-up of patient 600 from diagnosis through relapse, depicting percentage of bone marrow blasts and number of neutrophils. B, initial DSRT results from relapsed and refractory patient 600 showed that the patient cells ex vivo exhibited sensitivity to several kinase inhibitors (including dasatinib and sunitinib) and rapalogs, and as a result, the patient was treated with a combination of dasatinib, sunitinib, and temsirolimus. The patient achieved complete remission under this drug regimen, but relapsed 5 weeks later. Bar graphs show the before and after relapse sDSS for sensitivity to dasatinib, sunitinib, and temsirolimus. C, comparison of DSRT responses of the initial (600_2) and relapsed sample (600_3), revealing the loss of sensitivity to the majority of tested compounds. Drugs for which the DSS decreased greater than 10 from 600_2 to 600_3 are marked in blue, other drugs with an sDSS greater than 10 in 600_2 are marked in pale blue, and drugs with an sDSS greater than 10 in both 600_2 and 600_3 are marked in green. D, clonal progression of the disease in the patient from diagnosis to relapse; further information is included in the Supplementary Data and Supplementary Table S5. E, summary heatmap illustrating the key putative oncogenic genetic alterations in the clones depicted in D.

preexisting small subclones. The ETV6–NTRK3 encodes for 784 initially also had FLT3 -ITD mutations (Supplementary the oncogenic fusion protein TEL–TrkC, whereas the STRN- Fig. S9A and S9B) that were lost after induction chemo- ALK fusion was out of frame and therefore did not result therapy, and a mutation to WT1 augmented by LOH that in a functional fusion protein (Fig. 6A and Supplementary persisted throughout the course of the disease (Fig. 6C and Fig. S8). These genetic events were accompanied by increases D and Supplementary Table S6). The gained sensitivity to in p70S6 kinase (T389) and CREB (S133) phosphorylation the dual IGF-IR/TrkC inhibitor BMS-754807 ﬁ ts with the (Fig. 6B ), suggesting that mTORC1 was hyperactivated, a model of the TEL–TrkC fusion protein as a new driver, known mechanism of resistance toward rapalogs ( 17 ). The because the oncogenic potential of this fusion has been shown fusion gene MLL – ELL was detected from the diagnostic and to be dependent on the activity of IGF-IR ( 18, 19 ) and lead subsequent samples, suggesting that this was an initiating to hyperactivation of mTORC1 (20, 21). Thus, we predict driver event in this leukemia. Similar to patient 600, patient that in this patient, the mechanism of resistance involved

DECEMBER 2013CANCER DISCOVERY | 1423

RESEARCH ARTICLE Pemovska et al.

ABTop selective drugs in the 784_1 sample 20 Refametinib - MEKi Pimasertib - MEKi Sirolimus - Rapalog Temsirolimus - Rapalog* Pimasertib AZD8055 Refametinib Everolimus - Rapalog 15 BMS-754807 Tipifarnib Lestaurtinib PF-04691502 - mTOR/PI3Ki Dactinomycin Plicamycin Trametinib - MEKi Docetaxel Danusertib Paclitaxel Dactolisib AZD8055 - mTORi Topotecan BIIB021 Plicamycin - RNA synth.i Mitoxantrone Vincristine Trametinib Selumetinib - MEKi Navitoclax Selumetinib BIIB021 - HSP90i 10 Vinblastine Dexamethasone Dasatinib - Broad TKI* PF-04691502 Lestaurtinib - Broad TKI Tivozanib - TKI (VEGFR) 784_2 sDSS Tanespimycin - HSP90i Doramapimod - p38i 5 Dactolisib - mTOR/PI3Ki Idelalisib - PI3Ki Everolimus Pictilisib - PI3Ki Temsirolimus CUDC-101 - HDACi Sirolimus 0 Dasatinib 0 5 10 15 20 0 5 10 15 20 sDSS 784_1 sDSS C 784_1 784_2

GSK3A CDK11A GSK3A CDK11A

CDK11B CDK11B CHK2 CDK13 CHK2 CDK13 CDK9 CDK18 CDK9 CDK18 CDK4 CDK4 CDK2 CDK2 γ γ PhK 2 PhKγ1 PhK 2 PhKγ1 CDK16 CDK16 CDK17 CDK17 BIKE DAPK1 p110α BIKE DAPK1 p110α AAK1 ULK2 AAK1 ULK2 β β TrkC p110 TrkC p110 p110γ p110 γ TrkA TrkA ITK ITK ALK δ ALK δ LTK p110 LTK p110

IRAK3 MER MEKK2 MEK2 IRAK3 MER MEKK2 MEK2 MEK1 MEK1

CK1ε CK1ε YES1 TXK YES1 TXK

BTK FYN BTK FYN CK1δ BMX CK1δ BMX

EPHB4 CSK EPHB4 CSK JNK2 HCK JNK2 HCK EPHA4 EPHA4

p38γ p38α p38γ p38α EPHB2 p38β EPHB2 p38β

Figure 5. Functional DSRT and kinase inhibitor sensitivity deﬁ ned key oncologic signals and mechanisms of resistance. A, initial DSRT results from refractory patient 784, highlighting selective sensitivity to MEK inhibitors, rapalogs, and several TKIs. On the basis of these results, the patient was treated with dasatinib, sunitinib, and temsirolimus (marked with asterisks). B, correlation of the DSRT results of the initial (784_1) and resistant (784_2) sample, illustrating that the relapsed cells had lost sensitivity to dasatinib and rapalogs but retained sensitivity to ATP-competitive mTOR inhibitors and gained sensitivity to BMS-754807, crizotinib, danusertib, and tipifarnib. Drugs for which the DSS decreased greater than 10 from 784_1 to 784_2 are marked in blue, other drugs with an sDSS greater than 10 in 784_1 are marked in pale blue, drugs for which the DSS increased greater than 10 from 784_1 to 784_2 are marked in red, other drugs with an sDSS greater than 10 in 784_2 are marked in pink, and drugs with an sDSS greater than 10 in both 600_2 and 600_3 are marked in green. C, kinase inhibitor sDSS responses matched with target proﬁ les described by Davis and colleagues (16 ) yield putative kinase addiction subnetworks in the two patient samples.

TEL–TrkC-mediated activation of IGF-IR signaling, which pro- DISCUSSION moted hyperactivation of mTORC1. Supporting this hypothesis, we observed synergistic activities between the BMS-754807 We present here an ISM strategy based on the systematic func- IGF-IR/TrkC inhibitor and dactolisib, an ATP-competitive tional testing of patient-derived primary cancer cell sensitivity to mTOR inhibitor (Fig. 6E ). The combination of BMS-754807 targeted anticancer agents coupled with genomic and molecu- and the MEK inhibitor trametinib, on the other hand, did not lar proﬁ ling. Importantly, the intent is to guide treatment result in synergism, indicating that the TEL–TrkC/IGF-IR/ decisions for individual cancer patients coupled with monitor- mTORC1 dependency represents a separate signal than the ing of subsequent responses in patients to measure and under- one leading to addiction to MEK signaling (Fig. 6F ). Taken stand the efﬁ cacy and mechanism of action of the drugs. The together, these results indicate how ISM strategy helps to iden- ISM strategy allows for an iterative adjustment of therapies for tify not only the mechanisms of resistance, but also potential patients with cancer, one patient at a time, with repeated sam- ways to counteract it with combinatorial therapies. pling playing a major role in understanding and learning from

1424 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

A B Zinc finger, CXXC-type 5 784_1 (IPR002857) MLL ELL 4 784_2 1 140646 622 RNA pol II elongation factor ELL (IPR019464) 3 ELL MLL Occludin/RNA pol II elongation 2 factor ELL domain (IPR010844) Fold change 1 Pointed doman TEL TrkC 0 (IPR003118) 1 336 466 825 /2 Tyr kinase catalytic pp38α pCREB domain (IPR020635) pMEK1 pp70 S6K NTRK3 ETV6 pGSK-3α/β C D 784_0 784_1 784_2

Clone 1 FLT3-ITD #1 2T) 3 H

LO Clone 2 Clone 4 ETV6–NTRK3 FLT3-ITD #2 MLL–ELL fusWT1ion PDCD10(R157H)TP63(I4 FLT3-ITD #1FLT3-ITD #2 WT1 ETV6–NTRK3STRN–ALK fusion fusion MLL–ELL Clone 5 Clone 1 WT1 STRN–ALK Clone 2 Clone 3 Clone 3 WT1 LOH Clone 4 Chemotherapy Das–Sun–Tem Clone 5

E Dactolisib Trametinib F mTOR/PI3Ki MEKi 0 0.1 1 10 100 1,000 0 0.025 0.25 2.5 25 250 784_1 784_2 0 Dasatinib TEL-TrkC/ target IGF-IR 1 Das–Sun–Tem 10 therapy

100 TORC1 Resistance Hyperactivated

IGF-IR/TrKCi TORC1 BMS-754807 1,000

10,000 Synergistic Nonsynergistic

Figure 6. IGF-IR and mTOR inhibition has a synergistic effect in ETV6 – NTRK3 -driven AMLs. A, validation sequencing and resulting predicted protein structure of two fusions identiﬁ ed in this patient with RNA sequencing. The MLL – ELL fusion was present throughout the disease, whereas the ETV6– NTRK3 (TEL–TrkC) fusion was detected in the dasatinib–temsirolimus resistant sample. B, phosphoproteomic proﬁ ling of the initial and resistant samples displayed a signaling switch in the leukemic cells. C, clonal evolution of the AML from diagnosis to relapse. D, summary heatmap illustrating the key putative oncogenic genetic alterations in the clones depicted in C. E, combinatorial treatment of the patient cells with the IGF-IR/TrkC inhibitor BMS- 754807 and either dactolisib (ATP-competitive mTOR inhibitor) or trametinib (MEK inhibitor) revealed that there is a synergistic effect between mTOR and IGF-IR/TrkC inhibition. F, a model of kinase driver switch and drug resistance in this patient.

each success and failure. Furthermore, ISM facilitates learning setting and need to be verifi ed. However, the clinical implemen- by discovering molecular or functional patterns from past tation of ISM in individual patients is a powerful way to create patient cases to help therapeutic assessment of new patients. hypotheses to be tested in systematic clinical studies, both for Application of the DSRT technology to 28 AML patient existing and emerging drugs. Indeed, we also identifi ed several samples identifi ed effective molecularly targeted compounds investigational oncology drug classes, such as ATP-competitive inducing selective toxic or inhibitory responses in AML cells PI3K/AKT/mTOR pathway inhibitors, MEK inhibitors, and over normal bone marrow mononuclear cells. Our data suggest JAK inhibitors, which deserve attention as drugs to priori- the intriguing possibility that anticancer agents already in clini- tize for future clinical trials in AML. Furthermore, the ISM cal use for other diseases, such as dasatinib (current approved approach could also help to identify effective drug combina- indications are chronic myeloid leukemia and Ph+ acute lym- tions based on associating drug sensitivities. phoblastic leukemia), sunitinib (renal cell cancer and gastroin- Unsupervised clustering of the patient samples identifi ed testinal stromal tumors), and temsirolimus (renal cell cancer), fi ve functional taxonomic groups in AML based on ex vivo could be repositioned for subsets of AML patients. Although drug responses. FLT3 mutations were highly enriched among we did not achieve long-term cures in patients with AML receiv- the most TKI-sensitive functional group (group V), with FLT3 ing DSRT-guided therapies, the clinical responses seen are inhibitors being the most selective class of drugs for this group. encouraging, given the fact that most of the patients with AML However, these samples were also selectively sensitive to other studied here had complex chemorefractory, end-stage disease. kinase inhibitors, such as dasatinib, that lack FLT3 activity, sug- Obviously, the clinical results arise from a nonrandomized gesting that oncogenic FLT3 signaling may be dependent on the

DECEMBER 2013CANCER DISCOVERY | 1425

RESEARCH ARTICLE Pemovska et al. signaling of other tyrosine kinases whose inhibition may syner- This hypothesis is supported by the acquisition of sensitivity, gize with the therapeutic effects of FLT3 inhibitors. Given that based on ex vivo testing, to the dual IGF-IR/TrkC inhibitor the clinical implementation of FLT3 inhibitors has proven very BMS-754807 exclusively in the relapsed sample. challenging, the identifi cation of druggable synergistic kinase In conclusion, we present an individual-centric, functional signals or drugs could be extremely important. Furthermore, we systems medicine strategy to systematically identify drugs to observed a signifi cant association between activating mutations which individual patients with AML are sensitive and resistant, in RAS genes and sensitivity to MEK inhibitors, in line with pre- implement such strategies in the clinic, and learn from the inte- vious results showing that trametinib exhibited favorable clini- grated genomic, molecular, and functional analysis of drug sen- cal responses in patients with RAS -mutated refractory AML ( 22 ). sitivity and resistance in paired samples. ISM strategy provides Finally, we identifi ed a tendency of clustering of TP53 mutations a powerful way to create hypotheses to be tested in formal clini- in the two least-responsive functional groups (groups I and II) cal trials, for existing drugs, emerging compounds, and their and mutations and fusions linking to epigenetic modulation in combinations. In the future, ISM may pave a path for routine groups III and IV. However, most of the drug-response classifi ca- individualized optimization of patient therapies in the clinic. tions and variabilities could not yet be attributed to the main mutations detected in the patient samples. Such drug responses may arise from nongenomic causes and complex combinatorial METHODS molecular pathways, and these fi ndings therefore highlight the Study Patients and Material value of DSRT in (i) functionally validating suggestions aris- Twenty-eight bone marrow aspirates and peripheral blood samples ing from the genomic profi les and (ii) discovering other drug (leukemic cells) and skin biopsies (noncancerous cells for germline dependencies that have yet to be deduced from genomic or genomic information) from 18 AML and high-risk [according to the transcriptomic data. The relationship between genomic changes WHO classifi cation-based Prognostic Scoring System (WPSS); ref. 39 ] and drug response may also be more complicated in the chem- patients with myelodysplastic syndromes (MDS) and 7 samples from orefractory patient samples studied here. different healthy donors (controls) were obtained after informed con- Despite multiple efforts over several decades, the direct present with approval (No. 239/13/03/00/2010, 303/13/03/01/2011). Patient characteristics are summarized in Supplementary Table S3. diction of cancer cell chemosensitivity in the clinical setting has Mononuclear cells were isolated by Ficoll density gradient (Ficoll- remained an elusive goal. Compared with previously published Paque PREMIUM; GE Healthcare), washed, counted, and suspended approaches, the ISM approach presented here has distinct differ- in Mononuclear Cell Medium (MCM; PromoCell) supplemented with ences. First, our studies focused on targeted drugs, whereas most 0.5 μg/mL gentamicin and 2.5 μg/mL amphotericin B. One sample of the previous efforts of ex vivo drug testing have focused on from patient 393, a secondary AML after MDS with 20% myeloblasts, conventional chemotherapeutics (23–27 ). The ex vivo responses was enriched for the CD34+ cell population (sample 393_3, cor- to these agents are often nonselective and more diffi cult to inter- responding to the blast cell population) using paramagnetic beads pret and translate to clinical patient care. Second, we focused according to the manufacturer’s instructions (Miltenyi Biotech). on leukemias, whereas many previous studies have focused on solid tumors (28–33 ), where representative samples are diffi cult Development of the Compound Collection to acquire and consecutive samples from recurrent disease are The oncology compound collection covers the active substances from typically not available. Third, we measured selective anticancer the majority of U.S. Food and Drug Administration/European Medi- effects as compared with normal bone marrow mononuclear cines Agency ( FDA/EMA)–approved anticancer drugs (n = 123), as well = cells, which makes it possible to identify cancer-selective drugs as emerging investigational and preclinical compounds (n 64) covering with potential for less systemic toxicity. Fourth, we performed a a wide range of molecular targets (Supplementary Table S1). The compounds were obtained from the National Cancer Institute Drug Testing rapid analysis of the in vivo effects of the drugs from consecutive Program (NCI DTP) and commercial chemical vendors: Active Biochem, samples. Our novel endpoint for therapy effi cacy in patients was Axon Medchem, Cayman Chemical Company, ChemieTek, Enzo Life an impact on the clonal composition of the cancer sample. Sciences, LC Laboratories, Santa Cruz Biotechnology, Selleck, Sequoia The different types of responses in our clinically translated Research Products, Sigma-Aldrich, and Tocris Biosciences. patient cases highlight the diffi culty in predicting mechanisms of resistance and support the importance of repeated DSRT functional testing such as DSRT during disease progression Ex vivo DSRT was performed on freshly isolated primary AML cells to identify changes in drug sensitivities. In patient 600, the derived from patient samples as well as mononuclear cells derived leukemic cells carried a FLT3 -ITD mutation and showed from healthy donors. The compounds were dissolved in 100% dimethyl addiction to this oncogene based on highly selective responses sulfoxide (DMSO) and dispensed on tissue culture–treated 384-well to FLT3 inhibitors, such as quizartinib and sunitinib. Inter- plates (Corning) using an acoustic liquid handling device, Echo 550 estingly, the same FLT3 -ITD variant remained in the resistant (Labcyte Inc.). The compounds were plated in fi ve different concentra- cells, but the response to quizartinib, sunitinib, and other tions in 10-fold dilutions covering a 10,000-fold concentration range FLT3 inhibitors was lost, and the cells acquired pan-resistance (e.g., 1–10,000 nmol/L). The predrugged plates were kept in pressu- to almost all agents tested. In contrast, in patient 784, resist- rized StoragePods (Roylan Developments Ltd.) under inert nitrogen gas until needed. The compounds were dissolved with 5 μL of MCM ance to dasatinib, sunitinib, and temsirolimus treatment was while shaking for 30 minutes. Twenty microliters of single-cell sus- linked to enrichment of clonal populations carrying a tyro- pension (10,000 cells) was transferred to each well using a Multi Drop sine kinase fusion gene that mediated resistance. The ETV6– Combi (Thermo Scientifi c) peristaltic dispenser. The plates were incu- ° NTRK3 fusion and the resulting TEL–TrkC fusion protein is bated in a humidifi ed environment at 37 C and 5% CO2 , and after a known oncogenic driver in AML and other cancers ( 34–36 ) 72 hours cell viability was measured using CellTiter-Glo luminescent and is dependent on IGF-IR kinase activity (18–20 , 37, 38). assay (Promega) according to the manufacturer’s instructions with a

1426 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

Molecular Devices Paradigm plate reader. The data were normalized to sample from patient 1497 was sequenced using the Illumina TruSeq negative control (DMSO only) and positive control wells (containing Amplicon Cancer Panel and the MiSeq sequencer (Illumina). 100 μmol/L benzethonium chloride, effectively killing all cells). Somatic Mutation Calling and Annotation Generation of Dose–Response Curves and Analysis of Data from Exome-Sequencing Data The plate reader data were uploaded to Dotmatics Browser/Studies Sequence reads were processed and aligned to the reference genome software (Dotmatics Ltd.) for a normalized calculation of percentage as described previously ( 41, 42 ). Somatic mutation calls were made survival for each data point and generation of dose–response curves for exome capture target regions of the NimbleGen SeqCap EZ v2 for each of the drugs tested. The dose–response curves were fi tted on Capture Kit (Roche NimbleGen) and the fl anking 500 bps. High con- the basis of a four-parameter logistic fi t function defi ned by the top fi dence somatic mutations were called for each tumor sample using

and bottom asymptote, the slope, and the infl ection point (EC50 ). the VarScan2 somatic algorithm ( 43 ) with the following parameters: In the curve fi tting, the top asymptote of the curve was fi xed to 100% – strand-fi lter 1 viability, whereas the bottom asymptote was allowed to fl oat between 0% – min-coverage-normal 8 and 75% (i.e., drugs causing < 25% inhibition were considered inactive). – min-coverage-tumor 1 – somatic- P value 0.01 Scoring and Clustering of DSRT Data – normal-purity 0.95 To quantitatively profi le individual patient samples in terms – min-var-freq 0 of their DSRT-wide drug responses, as well as to compare drug responses across various AML patient samples, a single measure was Mutations were annotated with SnpEff (44 ) using the Ensembl v68 developed: DSS. The curve fi tting parameters were used to calculate annotation database. To fi lter out false-positive calls due to genomic the area under the dose–response curve (AUC) relative to the total repeats, somatic mutation calls in regions defi ned as repeats in the area (TA) between 10% threshold and 100% inhibition . Furthermore, RepeatMasker track obtained from the University of California, Santa the integrated response was divided by a logarithm of the top asymp- Cruz (UCSC) Genome Browser were removed from the analysis. To fi l- tote (a ). Formally, the DSS was calculated by: ter out misclassifi ed germline variants, population variants included in dbSNP (Single Nucleotide Polymorphism database) version 130 100 × AUC were removed. Remaining mutations were visually validated using the DSS = TA× log a Integrated Genomics Viewer (Broad Institute).

We scored for differential activity of the drugs in AML blast cells Analysis of Mutation Frequencies in Serial Samples in comparison with control cells, sDSS. Clustering of the drug sen- To examine frequencies of the identifi ed mutations in samples sitivity profi les across the AML patient and control samples was per- where the mutations did not pass the criteria for high confi dence formed using unsupervised hierarchical complete-linkage clustering mutations, variant frequencies and read counts for each mutation using Spearman and Euclidean distance measures of the drug and were retrieved from a set of unfi ltered variant calls generated by sample profi les, respectively. Reproducibility of the clustering and VarScan2 with the following parameters: the resulting drug-response subtypes detected was evaluated using – strand-fi lter 0 the bootstrap resampling method with the Pvclust R-package ( 40 ). – min-coverage-normal 8 Prediction of Kinase Addictions – min-coverage-tumor 1 – somatic- P value 1 sDSSs of kinase inhibitors were further used to predict sample- – normal-purity 1 specifi c kinase addictions. Sample-specifi c sDSS responses were – min-var-freq 0 compared with target profi les for 35 kinase inhibitors overlapping between our compound panel and the panel profi led by Davies In addition, we used variant allele frequencies from control–leuke- and colleagues ( 16 ). More specifi cally, for each kinase target, we mia pairs to identify regions of LOH. calculated a Kinase Inhibition Sensitivity Score (KISS) by averaging the sDSS values among those compounds that selectively target FLT3-ITD Detection by Capillary Sequencing and qPCR the kinase. These putative selective kinases were compared with For determination of patients’ FLT3 -ITD status, genomic DNA was gene expression to exclude nonexpressed targets, and the remaining extracted from the bone marrow mononuclear cell fraction. Qualita- kinases defi ned a putative “kinaddictome” for each patient sample. tive PCR was performed as described by Kottaridis and colleagues ( 45 ) For displaying purposes, the resulting kinases were depicted in a tar- by using a 6-carboxyfl uorescein (FAM)-labeled forward primer. The get similarity network, in which edges connect kinases with similar PCR products were separated on an agarose gel and in capillary elec- inhibitor specifi city profi les (ref. 16 ; Spearman rank-based correla- trophoresis with an ABI3500D× Genetic Analyzer and sequenced using tion > 0.5; Szwajda and colleagues, unpublished data). M13-tailed direct sequencing. Assessment of minimal residual disease (MRD) level was performed with real-time quantitative PCR (qPCR). A DNA Sequencing patient ITD-specifi c ASO (allele-specifi c oligonucleotide) primer was Genomic DNA was isolated using the DNeasy Blood & Tissue Kit designed at the ITD junction region and used together with a down- (Qiagen). Exome sequencing was performed on the patient samples stream TaqMan probe and reverse primer (primer sequences available highlighted in Fig. 3 . In addition, whole-genome sequencing was upon request). Albumin gene qPCR was additionally performed to performed using DNA from the skin and AML cells from sample normalize the variability in DNA quality in the follow-up samples. 784_2. DNA (3 μg) was fragmented and processed according to the NEBNext DNA Sample Prep Master Mix protocol. Exome capture Amplicon Sequencing was performed using the Nimblegen SeqCap EZ v2 capture Kit Amplicons were amplifi ed using locus-specifi c PCR primers carrying (Roche NimbleGen). Sequencing of exomes and genomes was done Illumina compatible adapter sequences, grafting sequences (P5 and using HiSeq1500, 2000, or 2500 instruments (Illumina). For germ- P7), and an amplicon-specifi c 6-bp index sequence (Supplementary line control samples 4 × 10 7 and for tumor samples 10 × 107 2 × 100 Table S7). The PCR reaction contained 10 ng of sample DNA, 10 μL bp paired-end reads were sequenced per sample. The leukemia DNA of 2× Phusion High-Fidelity PCR Master Mix (Thermo Scientifi c),

DECEMBER 2013CANCER DISCOVERY | 1427

RESEARCH ARTICLE Pemovska et al. and 0.5 μmol/L of each primer. Following PCR amplifi cation, samples Acquisition of data (provided animals, acquired and managed were purifi ed using Performa V3 96-Well Short Plate and QuickStep2 patients, provided facilities, etc.): M. Kontro, M.M. Bespalov, P. Ellonen, SOPE Resin (EdgeBio). Sequencing of PCR amplicons was performed E. Elonen, B.T. Gjertsen, R. Karjalainen, E. Kulesskiy, A. Lehto, M. Lepistö, using the Illumina HiSeq2000 instrument (Illumina). Samples were T. Lundán, P. Mattila, A. Murumägi, S. Mustjoki, A. Parsons, M.E. Rämet, sequenced as 101-bp paired-end reads and one 7-bp index read. M. Suvela, L. Turunen, M. Wolf, C.A. Heckman, K. Porkka Analysis and interpretation of data (e.g., statistical analysis, Library Preparation, Sequencing, and Data Analysis biostatistics, computational analysis): T. Pemovska, B. Yadav, of Transcriptomes H. Edgren, S. Eldfors, A. Szwajda, H. Almusa, E. Elonen, T. Lundán, Total RNA (2.5 to 5 μg) was used for depletion of rRNA (Ribo-Zero M.M. Majumder, J.M.L. Marti, S. Mustjoki, I. Västrik, T. Aittokallio, rRNA Removal Kit; Epicentre), purifi ed (RNeasy clean-up-kit; Qia- C.A. Heckman, K. Porkka, O. Kallioniemi, K. Wennerberg gen), and reverse transcribed to double-stranded cDNA (SuperScript Writing, review, and/or revision of the manuscript: T. Pemovska, Double-Stranded cDNA Synthesis Kit; Life Technologies). Random M. Kontro, H. Edgren, S. Eldfors, E. Elonen, B.T. Gjertsen, T. Lundán, hexamers (New England BioLabs) were used for priming the fi rst- A. Murumägi, S. Mustjoki, J. Knowles, T. Aittokallio, C.A. Heckman, strand synthesis reaction. K. Porkka, O. Kallioniemi, K. Wennerberg Illumina-compatible Nextera Technology (Epicentre) was used for Administrative, technical, or material support (i.e., report- preparation of RNAseq Libraries. High Molecular Weight (HMW) ing or organizing data, constructing databases): T. Pemovska, buffer and 50 ng of cDNA were used for tagmentation as recommended R. Karjalainen, A. Lehto, A. Palva, A. Parsons, T. Pirttinen, M. Suvela, by the manufacturer. After the tagmentation reaction, the fragmented L. Turunen, I. Västrik, C.A. Heckman, K. Porkka cDNA was purifi ed with SPRI beads (Agencourt AMPure XP, Beckman Study supervision: E. Elonen, J. Knowles, T. Aittokallio, C.A. Heck- Coulter). The RNAseq libraries were size selected (350–700 bp frag- man, K. Porkka, O. Kallioniemi, K. Wennerberg ments) in 2% agarose gel followed by purifi cation with QIAquick gel extraction kit (Qiagen). Acknowledgments Each transcriptome was loaded to occupy one third of the lane The authors thank the patients for donating their samples for capacity in the fl ow cell. C-Bot (TruSeq PE Cluster Kit v3, Illumina) research, Jani Saarela and Ida Lindenschmidt (FIMM Technology was used for cluster generation, and an Illumina HiSeq2000 platform Centre, High Throughput Biomedicine Unit) for technical assist- (TruSeq SBS Kit v3-HS reagent kit) was used for paired-end sequenc- ance, Biocenter Finland research infrastructures (Drug Discovery ing with 100-bp read length. Nextera Read Primers 1 and 2 as well and Chemical Biology platform as well as the Genome-Wide Methods as Nextera Index Read Primer were used for paired-end sequencing network), EATRIS and BBMRI ESFRI infrastructures for technical and index read sequencing, respectively. RNA-seq data analysis, such and infrastructural support, and Labcyte Inc. for technical assistance as fusion gene identifi cation, mutation calling, and gene expression in developing the DSRT pipeline as part of an ongoing collaboration quantitation (Tophat and Cuffl inks), was done as described previ- with FIMM. ously ( 46 ). Primers used to validate as well as quantify the fusion genes detected are listed in Supplementary Tables S8 and S9. Grant Support This study was fi nancially supported by the Academy of Finland (to Proteomic Analysis T. Aittokallio and O. Kallioniemi); Biocenter Finland (to O. Kallioniemi Phosphoproteomic analysis of the AML patient samples was per- and K. Wennerberg); ERDF: the European Regional Development formed using Proteome Profi ler antibody arrays (R&D Systems) Fund (to I. Västrik); EU FP7 BioMedBridges (to O. Kallioniemi); according to the manufacturer’s instructions. Lysates containing 300 Finnish Cancer Societies (to O. Kallioniemi and K. Porkka); the Jane μg of protein were applied to the arrays, and fl uorescently labeled and Aatos Erkko Foundation (to K. Wennerberg); the Sigrid Jusélius streptavidin (IRDye 800 CW streptavidin; LI-COR) and an Odyssey Foundation (to O. Kallioniemi); and Tekes: Finnish Funding Agency imaging system (LI-COR) were used for detection. for Technology and Innovation (to O. Kallioniemi and J. Knowles).

Other Statistical Analyses Received July 8, 2013; revised September 12, 2013; accepted Sep- Statistical analysis was performed using GraphPad Prism 5. A tember 13, 2013; published OnlineFirst September 20, 2013. Pearson correlation test was used to determine the correlations between drug sensitivity profi les of healthy donor samples. A two-tailed Student t test was used to assess the correlation between RAS or FLT3 mutations REFERENCES and MEK inhibitor or dasatinib sensitivity, respectively. A correlation 1. Marcucci G , Haferlach T , Dohner H . Molecular genetics of adult < in sensitivity was considered statistically signifi cant when P 0.05. acute myeloid leukemia: prognostic and therapeutic implications. J Clin Oncol 2011 ; 29 : 475 – 86 . Disclosure of Potential Confl icts of Interest 2. Swerdlow SH , Campo E , Harris NL , Jaffe ES , Pileri SA , Stein H , et al. S. Mustjoki has honoraria from the Speakers Bureaus of Novartis WHO classifi cation of tumours of haematopoietic and lymphoid tis- and Bristol-Myers Squibb and is a consul tant/advisory board member sues. 4th ed. Lyon, France: IARC Press; 2008 . of the same. O. Kallioniemi has received commercial research support 3. Welch JS , Ley TJ , Link DC , Miller CA , Larson DE , Koboldt DC , et al. from IMI-Project Predect (a consortium of pharmaceutical compa- The origin and evolution of mutations in acute myeloid leukemia . nies) and is a consultant/advisory board member of Medisapiens. No Cell 2012 ; 150 : 264 – 78 . potential confl icts of interest were disclosed by the other authors. 4. Network CGAR. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med 2013 ; 368 : 2059 – 74 . Authors’ Contributions 5. Dohner H , Gaidzik VI . Impact of genetic features on treatment decisions in AML. Hematology Am Soc Hematol Educ Program 2011 ; 2011 : 36 – 42 . Conception and design: T. Pemovska, M. Kontro, E. Elonen, I. Västrik, 6. Rowe JM , Kim HT , Tallman MS . Reply to induction therapy and out- J. Knowles, C.A. Heckman, K. Porkka, O. Kallioniemi, K. Wennerberg come in acute myeloid leukemia. Cancer 2011 ; 117 : 2237 . Development of methodology: T. Pemovska, M. Kontro, B. Yadav, 7. Rowe JM , Tallman MS . How I treat acute myeloid leukemia. Blood H. Edgren, S. Eldfors, M.M. Bespalov, R. Karjalainen, E. Kulesskiy, 2010 ; 116 : 3147 – 56 . S. Lagström, M. Lepistö, M.M. Majumder, P. Mattila, L. Turunen, M. Wolf, 8. Buchner T , Berdel WE , Haferlach C , Haferlach T , Schnittger S , T. Aittokallio, C.A. Heckman, K. Porkka, O. Kallioniemi, K. Wennerberg Muller-Tidow C , et al. Age-related risk profi le and chemotherapy dose

1428 | CANCER DISCOVERYDECEMBER 2013 www.aacrjournals.org

ISM Approach to Therapy Selection RESEARCH ARTICLE

response in acute myeloid leukemia: a study by the German Acute 27. Yamada S , Hongo T , Okada S , Watanabe C , Fujii Y , Ohzeki T. Clini- Myeloid Leukemia Cooperative Group. J Clin Oncol 2009 ; 27 : 61 – 9 . cal relevance of in vitro chemoresistance in childhood acute myeloid 9. Dohner H , Estey EH , Amadori S , Appelbaum FR , Buchner T , Burnett leukemia. Leukemia 2001 ; 15 : 1892 – 7 . AK , et al. Diagnosis and management of acute myeloid leukemia in 28. Iwadate Y , Fujimoto S , Namba H , Yamaura A . Promising survival for adults: recommendations from an international expert panel, on patients with glioblastoma multiforme treated with individualised behalf of the European LeukemiaNet. Blood 2010 ; 115 : 453 – 74 . chemotherapy based on in vitro drug sensitivity testing. Br J Cancer 10. Ding L , Ley TJ , Larson DE , Miller CA , Koboldt DC , Welch JS , et al. 2003 ; 89 : 1896 – 900 . Clonal evolution in relapsed acute myeloid leukaemia revealed by 29. Bosanquet AG , Bell PB . Ex vivo therapeutic index by drug sensitivity whole-genome sequencing. Nature 2012 ; 481 : 506 – 10 . assay using fresh human normal and tumor cells. J Exp Ther Oncol 11. Tyner JW , Yang WF , Bankhead A III, Fan G , Fletcher LB , Bryant J , 2004 ; 4 : 145 – 54 . et al. Kinase pathway dependence in primary human leukemias 30. Villman K , Blomqvist C , Larsson R , Nygren P . Predictive value of in determined by rapid inhibitor screening. Cancer Res 2013 ; 73 : 285 – 96 . vitro assessment of cytotoxic drug activity in advanced breast cancer . 12. Wang GG , Cai L , Pasillas MP , Kamps MP . NUP98–NSD1 links H3K36 Anticancer Drugs 2005 ; 16 : 609 – 15 . methylation to Hox-A gene activation and leukaemogenesis . Nat Cell 31. Ehemann V , Kern MA , Breinig M , Schnabel PA , Gunawan B , Schulten Biol 2007 ; 9 : 804 – 12 . HJ , et al. Establishment, characterization and drug sensitivity testing 13. Cerveira N , Correia C , Doria S , Bizarro S , Rocha P , Gomes P, et al. in primary cultures of human thymoma and thymic carcinoma. Int J Frequency of NUP98–NSD1 fusion transcript in childhood acute Cancer 2008 ; 122 : 2719 – 25 . myeloid leukaemia . Leukemia 2003 ; 17 : 2244 – 7 . 32. Lehnhardt M , Muehlberger T , Kuhnen C , Brett D , Steinau HU , Jafari 14. Hollink IH , van den Heuvel-Eibrink MM , Arentsen-Peters ST , Prat- HJ , et al. Feasibility of chemosensitivity testing in soft tissue sarco- corona M , Abbas S , Kuipers JE , et al. NUP98/NSD1 characterizes a mas. World J Surg Oncol 2005 ; 3 : 20 . novel poor prognostic group in acute myeloid leukemia with a dis- 33. Brigulova K , Cervinka M , Tosner J , Sedlakova I . Chemoresistance tinct HOX gene expression pattern. Blood 2011 ; 118 : 3645 – 56 . testing of human ovarian cancer cells and its in vitro model. Toxicol 15. Fasan A , Haferlach C , Alpermann T , Kern W , Haferlach T , Schnittger S . A In Vitro 2010 ; 24 : 2108 – 15 . rare but specifi c subset of adult AML patients can be defi ned by the cytoge- 34. Knezevich SR , McFadden DE , Tao W , Lim JF , Sorensen PH . A novel netically cryptic NUP98–NSD1 fusion gene. Leukemia 2013 ; 27 : 245 – 8 . ETV6–NTRK3 gene fusion in congenital fi brosarcoma. Nat Genet 16. Davis MI , Hunt JP , Herrgard S , Ciceri P , Wodicka LM , Pallares G , et al. 1998 ; 18 : 184 – 7 . Comprehensive analysis of kinase inhibitor selectivity . Nat Biotech- 35. Liu Q , Schwaller J , Kutok J , Cain D , Aster JC , Williams I R, et al. Signal nol 2011 ; 29 : 1046 – 51 . transduction and transforming properties of the TEL-TRKC fusions 17. Dowling RJ , Topisirovic I , Alain T , Bidinosti M , Fonseca BD , Petrou- associated with t(12;15)(p13;q25) in congenital fi brosarcoma and lakis E , et al. mTORC1-mediated cell proliferation, but not cell acute myelogenous leukemia. EMBO J 2000 ; 19 : 1827 – 38 . growth, controlled by the 4E-BPs. Science 2010 ; 328 : 1172 – 6 . 36. Tognon C , Knezevich SR , Huntsman D , Roskelley CD , Melnyk N , M ath- 18. Lannon CL , Martin MJ , Tognon CE , Jin W , Kim SJ , Sorensen PH. A ers JA , et al. Expression of the ETV6–NTRK3 gene fusion as a primary highly conserved NTRK3 C-terminal sequence in the ETV6–NTRK3 event in human secretory breast carcinoma. Cancer Cell 2002 ; 2 : 367 – 76 . oncoprotein binds the phosphotyrosine binding domain of insulin 37. Perl AE , Kasner MT , Tsai DE , Vogl DT , Loren AW , Schuster SJ, receptor substrate-1: an essential interaction for transformation . J et al. A phase I study of the mammalian target of rapamycin inhibitor Biol Chem 2004 ; 279 : 6225 – 34 . sirolimus and MEC chemotherapy in relapsed and refractory acute 19. Morrison KB , Tognon CE , Garnett MJ , Deal C , Sorensen PH . ETV6– myelogenous leukemia. Clin Cancer Res 2009 ; 15 : 6732 – 9 . NTRK3 transformation requires insulin-like growth factor 1 receptor 38. Tognon CE , Somasiri AM , Evdokimova VE , Trigo G , Uy EE , Melnyk signaling and is associated with constitutive IRS-1 tyrosine phospho- N , et al. ETV6–NTRK3–mediated breast epithelial cell transforma- rylation. Oncogene 2002 ; 21 : 5684 – 95 . tion is blocked by targeting the IGF1R signaling pathway. Cancer Res 20. Tognon CE , Martin MJ , Moradian A , Trigo G , Rotblat B , Cheng S W, 2011 ; 71 : 1060 – 70 . et al. A tripartite complex composed of ETV6–NTRK3, IRS1 and 39. Malcovati L , Germing U , Kuendgen A , Della Porta MG , Pascutto C , IGF1R is required for ETV6–NTRK3-mediated membrane localiza- Invernizzi R , et al. Time-dependent prognostic scoring system for tion and transformation. Oncogene 2012 ; 31 : 1334 – 40 . predicting survival and leukemic evolution in myelodysplastic syn- 21. Tamburini J , Chapuis N , Bardet V , Park S , Sujobert P , Willems L , et al. dromes. J Clin Oncol 2007 ; 25 : 3503 – 10 . Mammalian target of rapamycin (mTOR) inhibition activates phos- 40. Suzuki R , Shimodaira H . Pvclust: an R package for assessing the phatidylinositol 3-kinase/Akt by up-regulating insulin-like growth uncertainty in hierarchical clustering . Bioinformatics 2006 ; 22 : 1540 – 2 . factor-1 receptor signaling in acute myeloid leukemia: rationale for 41. Koskela HL , Eldfors S , Ellonen P , van Adrichem AJ , Kuusanmaki H , therapeutic inhibition of both pathways. Blood 2008 ; 111 : 379 – 82 . Andersson EI , et al. Somatic STAT3 mutations in large granular lym- 22. Borthakur G , Popplewell L , Boyiadzis M , Foran JM , Platzbecker U , phocytic leukemia. N Engl J Med 2012 ; 366 : 1905 – 13 . Vey N , et al. Phase I/II trial of the MEK1/2 inhibitor trametinib 42. Sulonen AM , Ellonen P , Almusa H , Lepisto M , Eldfors S , Hannula S , (GSK1120212) in relapsed/refractory myeloid malignancies: evidence et al. Comparison of solution-based exome capture methods for next of activity in patients with RAS mutation-positive disease . In: Proceed- generation sequencing. Genome Biol 2011 ; 12 : R94 . ings of the 54th ASH Annual Meeting and Exposition ; 2012 Dec 8–11 ; 43. Koboldt DC , Zhang Q , Larson DE , Shen D , McLellan MD , Lin L, et al. Atlanta, GA: Georgia World Congress Center. p. 120. Abstract nr 677. VarScan 2: somatic mutation and copy number alteration discovery 23. Gustavsson A , Olofsson T . Prediction of response to chemotherapy in cancer by exome sequencing. Genome Res 2012 ; 22 : 568 – 76 . in acute leukemia by in vitro drug sensitivity testing on leukemic stem 44. Cingolani P , Platts A , Wang le L , Coon M , Nguyen T , Wang L , et al. cells. Cancer Res 1984 ; 44 : 4648 – 52 . A program for annotating and predicting the effects of single nucle- 24. Larsson R , Fridborg H , Kristensen J , Sundstrom C , Nygren P . In vitro otide polymorphisms, SnpEff: SNPs in the genome of Drosophila testing of chemotherapeutic drug combinations in acute myelocytic melanogaster strain w1118; iso-2; iso-3. Fly 2012 ; 6 : 80 – 92 . leukaemia using the fl uorometric microculture cytotoxicity assay 45. Kottaridis PD , Gale RE , Frew ME , Harrison G , Langabeer SE , Belton (FMCA). Br J Cancer 1993 ; 67 : 969 – 74 . AA , et al. The presence of a FLT3 internal tandem duplication in 25. Larsson R , Nygren P . Laboratory prediction of clinical chemothera- patients with acute myeloid leukemia (AML) adds important prognos- peutic drug resistance: a working model exemplifi ed by acute leukae- tic information to cytogenetic risk group and response to the fi rst cycle mia. Eur J Cancer 1993 ; 29A : 1208 – 12 . of chemotherapy: analysis of 854 patients from the United Kingdom 26. Pieters R , Loonen AH , Huismans DR , Broekema GJ , Dirven MW , He y- Medical Research Council AML 10 and 12 trials. Blood 2001 ; 98 : 1752 – 9 . enbrok MW , et al. In vitro drug sensitivity of cells from children with 46. Edgren H , Murumagi A , Kangaspeska S , Nicorici D , Hongisto V , leukemia using the MTT assay with improved culture conditions . Kleivi K , et al. Identifi cation of fusion genes in breast cancer by Blood 1990 ; 76 : 2327 – 36 . paired-end RNA-sequencing. Genome Biol 2011 ; 12 : R6 .

DECEMBER 2013CANCER DISCOVERY | 1429

Individualized Systems Medicine Strategy to Tailor Treatments for Patients with Chemorefractory Acute Myeloid Leukemia

Tea Pemovska, Mika Kontro, Bhagwan Yadav, et al.

Cancer Discovery 2013;3:1416-1429. Published OnlineFirst September 20, 2013.

Updated version Access the most recent version of this article at: doi:10.1158/2159-8290.CD-13-0350

Supplementary Access the most recent supplemental material at: Material http://cancerdiscovery.aacrjournals.org/content/suppl/2013/09/18/2159-8290.CD-13-0350.DC1 http://cancerdiscovery.aacrjournals.org/content/suppl/2021/03/02/2159-8290.CD-13-0350.DC2

Cited articles This article cites 44 articles, 18 of which you can access for free at: http://cancerdiscovery.aacrjournals.org/content/3/12/1416.full#ref-list-1

Citing articles This article has been cited by 65 HighWire-hosted articles. Access the articles at: http://cancerdiscovery.aacrjournals.org/content/3/12/1416.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerdiscovery.aacrjournals.org/content/3/12/1416. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.