(2013) 27, 305–314 & 2013 Macmillan Publishers Limited All rights reserved 0887-6924/13 www.nature.com/leu

ORIGINAL ARTICLE kinase profiles identify a rationale for multitarget kinase inhibition in immature T-ALL

L Lhermitte1, R Ben Abdelali1, P Villare`se1, N Bedjaoui1, V Guillemot2, A Trinquand1, M Libura1, AS Bedin1, A Petit3, H Dombret4, G Leverger3, N Ifrah5, O Hermine1, E Macintyre1 and V Asnafi1

Constitutively activated FLT3 signaling is common in , and is currently under evaluation for , whereas little data is available in T-cell acute lymphoblastic leukemia (T-ALL). We analyzed 357 T-ALL cases for FLT3 and transcript expression. FLT3 mutations (3% overall) and overexpression (FLT3 high expresser (FLT3High)) were restricted to immature/TCRgd T-ALLs. In vitro FLT3 inhibition induced apoptosis in only 30% of FLT3High T-ALLs and did not correlate with mutational status. In order to investigate the mechanisms of primary resistance to FLT3 inhibition, a broad quantitative screen for receptor kinome transcript deregulation was performed by Taqman Low Density Array. FLT3 deregulation was associated with overexpression of a network of receptor kinases (RKs), potentially responsible for redundancies and sporadic response to specific FLT3 inhibition. In keeping with this resistance to FLT3 inhibition could be reversed by dual inhibition of FLT3 and KIT with a synergistic effect. We conclude that immature T-ALL may benefit from multitargeted RK inhibition and that exploration of the receptor kinome defines a rational strategy for testing multitarget kinase inhibition in malignant diseases.

Leukemia (2013) 27, 305–314; doi:10.1038/leu.2012.177 Keywords: T-ALL; multitarget therapy; receptor kinase; FLT3

INTRODUCTION FLT3 autocrine activation has also been reported in over 70% of T-cell acute lymphoblastic (T-ALLs) represent expan- AML cases and represents an alternative mechanism for FLT3 15–17 sions of cells arrested at specific stages of thymic development, oncogenic signaling, regardless of FLT3 mutational status. In with the underlying genetic abnormality often determining the AML, FLT3 overexpression associated with autophosphorylation stage of maturation arrest.1 Several recent studies used different and sensitivity to FLT3 inhibition of the receptor and identified 18 approaches but all suggested the existence of an immature an unfavorable prognostic subgroup. Little data is available, subtype (IM) of T-ALL.2–5 IM T-ALLs lack expression of T-cell however, regarding FLT3 expression and receptor receptor (TCR) alpha-beta or gamma-delta, cytoplasmic TCR beta profiles in either normal human T-cell maturation or in T-ALL. (cTCRb), and frequently CD4/8 and CD1a markers. In contrast, We assessed oncogenic FLT3 /deregulation in a large expression of CD13 and CD33 myeloid antigens markers, as well as series of T-ALL and identified a concerted activated receptor kinase ‘myeloid’ oncogenic lesions, such as MLL and CALM-AF10 (RK) signature as a mechanism for primary resistance to single-agent rearrangements, are common and relatively specific to this FLT3 inhibition, but sensitivity to dual agent inhibition, thus providing subset of T-ALLs. IM T-ALL, early T-cell precursor leukemia or an original rationale for multi-kinase therapeutic strategies. T/myeloid biphenotypic acute leukemia have been reported to respond poorly to chemotherapy and may therefore benefit from specific targeted therapy.6–9 MATERIALS AND METHODS Tyrosine kinase inhibitors may represent such agents, but Patients efficient evaluation of their potential therapeutic impact requires The study population included 357 unselected patients, consisting of 195 their use in subsets of patients that are most likely to benefit. adults (X15 years) and 162 children (o15 years) with newly diagnosed Constitutive activation of the FLT3 T-ALL. Samples were collected centrally for molecular and cellular analysis, pathway by internal tandem duplication (ITD), tyrosine kinase with the support of the adult (GRAALL) and children (FRALLE) protocols. Diagnosis of T-ALL was based on the World Health Organization consensus domain point mutation (D835Y) or an autocrine-loop has been 19 reported in acute myeloid leukemia (AML)10 and MLL rearranged criteria, defined by expression of cytoplasmic CD3, with neither expression of MPO nor strong expression of cytoplasmic CD79a. As such, or hyperdiploid B-cell precursor ALL.11 Activating FLT3 mutations 12,13 mixed phenotype acute leukemias were excluded from the study, in favor have been described in rare cases of T-ALL when they have of pure T-ALL. Maturation stage of arrest was specified further, based been reported to be restricted to the CD117 þ subset in adults or on expression of cTCRbF1 and surface TCR, as previously described.2 associated with a CD4 þ /CD8 À phenotype in pediatric T-ALL, but T-ALL cases were classified as immature (TCR À and cTCRb À ), ab-lineage these data remain controversial.14 (TCRab þ or TCR À /cTCRb þ ) or TCRgd-expressing cases (TCRgd þ ).

1Department of Hematology, Hoˆpital Necker-Enfants-Malades APHP, CNRS UMR 8147, Universite´ Paris-5 Descartes, Paris, France; 2Laboratoire d’Exploration fonctionnelle des Ge´nomes, CEA, Ge´nopole, Evry, France; 3Department of Hematology, Hoˆpital Armand Trousseau, Paris, France; 4Department of Hematology, Hoˆpital St-Louis, Paris, France and 5PRES LUNAM, CHU Angers service des Maladies du Sang et INSERM U 892, Angers, France. Correspondence: Dr V Asnafi, Laboratoire d’He´matologie, Tour Pasteur, Hoˆ pital Necker, 149–161, rue de Se`vres, 75743 Paris cedex 15, France. E-mail: vahid.asnafi@nck.aphp.fr Received 17 September 2011; revised 13 June 2012; accepted 26 June 2012; accepted article preview online 3 July 2012; advance online publication, 27 July 2012 A rationale for multitarget kinase inhibition L Lhermitte et al 306 Normal thymus samples were obtained from children undergoing CD3 expression. Stained samples were analyzed with a FACS Canto II cardiac surgery at the Necker-Enfants-Malades Hospital, with informed (BD Biosciences). Thymic cells were sorted with a combination of CD3, CD4, consent from the parents. CD8, CD34, CD1a and CD45 fluorochrome-labeled using a FACS Aria II cell-sorter (BD Biosciences) (Supplementary Figure S3). Western Molecular analysis (real-time quantitative PCR, FLT3 mutational blotting of FLT3 was performed using anti-FLT3 (clone S18; Santa status and Taqman Low Density Array (TLDA)) Cruz Biotechnology, Santa Cruz, CA, USA) and anti-phosphotyrosine (clone 4G10) as previously described.24 Phosphorylation of KIT was analyzed An unselected series of 357 T-ALLs with available cDNA were screened for 20 using anti-phospho-c-Kit (Tyr719) (Cell Signaling Technology, Berverly, MA, FLT3 ITD and D835Y mutations, as previously described. FLT3 transcript USA) as previously reported.25 expression was quantified by real-time quantitative PCR for 301 cases. Flow-cytometry monitoring of phosphoepitopes of FLT3 and down- Molecular oncogenetic analyses (TLX1, TLX3, SIL-TAL and CALM-AF10) 2,3 stream signaling molecules was performed using a polyclonal anti- were performed centrally, as previously reported. Results of phospho-FLT3 (Y591) Alexa Fluor 488 (Cell Signaling Technology), transcriptional analysis were expressed as target/reference transcript monoclonal anti-phospho-STAT5 (Y694) Alexa Fluor 647 (Clone 47; BD ratios. ABL and glyceraldehyde 3-phosphate dehydrogenase (GAPDH) Biosciences) and monoclonal anti-phospho-ERK1/2 (T202/Y204) Alexa Fluor were shown to be highly correlated, including for cases harboring the 647 (Clone 20a, BD Biosciences). Leukemic samples were incubated in NUP-ABL abnormality, and as such were used as reference culture medium for 3 h and then secondarily subjected to different (Supplementary Figure S1). Quantification of RK expression was performed combinations of activators and/or inhibitors; FL, CEP701 and AC220. on quality-controlled cDNA using predesigned TaqMan Low Density Monitoring of phospho- state was performed after 30 min of microfluidic cards (TLDA Human Array; Applied Biosystems). exposure for FLT3, and 1 h for ERK and STAT5 molecules. Cells were Real-time measurements were performed using the ABI Prism 7900HT immediately fixed by addition of BD CytoFix fixation buffer (BD Biosciences) (Applied Biosystems, Foster City, CA, USA). Samples with GAPDH cycle into the culture well and then incubated for 15 min at 37 1C. After washing, thresholds superior to 21 were excluded from analysis. Each probe value cells were permeabilized using BD PhosFlow Perm Buffer III (BD Biosciences) was normalized with respect to GAPDH copy number and analyzed with 13 according to the manufacturer’s instructions. Samples were then stained Dcycle threshold threshold to discard background noise. Reproducibility of using the above-mentioned reagents added with a backbone combination measurements and correlation with classical real-time quantitative PCR of markers including CD45 V500 (HI30) and CD3 PE (UCHT1) for gating of experiments were validated (Supplementary Figure S2). blast cells. Samples were analyzed using a FACS Canto II (BD Biosciences). Data were processed using Cytobank software.26 Changes in phosphory- Statistical analysis lation of signaling following activator and/or inhibitor exposure All monoparametric measurement comparisons were determined using were calculated using the log2 ratio of the median fluorescence intensity of the Mann–Whitney test or the Student’s test when the number of cells stained with anti-phosphoepitope antibody versus isotype controls, individuals was superior to 30 (PRISM software, Graphpad, La Jolla, CA, and expressed in a color scale in the corresponding figure. USA). Hierarchical classifications were performed using the Pearson’s correlation coefficient as a similarity metric with an average linkage algorithm as described.21 Unsupervised selection of transcripts used to RESULTS determine the RK signature was performed by removing transcripts FLT3 mutations in T-ALL that had s.d. of observed values over the whole series below o1.5. FLT3 ITD (nine cases) and D835Y mutation (one case) were detected in 10 (3%) out of 357 T-ALL cases. Characteristics of T-ALL Cell cultures patients carrying FLT3 mutations according to their phenotypic Cell culture was performed by maintaining plated thawed primary T-ALL and oncogenetic features are reported in Table 1. FLT3 mutations cells in RPMI medium with 10% fetal bovine serum supplemented with human recombinant stem cell factor (SCF) (2 ng/ml), FLT3 (FL) (5 ng/ml) and interleukin 7 (5 ng/ml). Dimethyl sulfoxide or increasing Table 1. Characteristics of T-ALL patients according to FLT3 concentrations of FLT3-specific inhibitor (CEP701 or AC220) were added to mutational status cell cultures 3 h after platting cells. T-cell blasts were analyzed after 48 h of culture by flow cytometry using annexin V APC (Bender MedSystems, FLT3 ITD/ FLT3 WT P-value Vienna, Austria) and propidium iodide (BD-Pharmingen, San Jose, CA, USA) D835Y immunostaining. To assess apoptosis attributable to CEP701, results were plotted for each sample as percentage of untreated control (dimethyl Patient characteristics sulfoxide). Immature T cells were considered sensitive as regards to CEP701 N 10/357 (3%) 347/357 (97%) NA a exposure when demonstrating 70% of cell viability as compared with Age 24 (6–78) 21 (1–66) ns o a untreated cells. The EOL cell line was used as an external control for Sex ratio (M/F) 10/0 269/78 ns a CEP701 efficiency. Children/adults 6/4 156/191 ns

Immunogenetic stage of arrest Analysis of combined drug effects Immature 8/10 (80%) 112/347 (32%) 0.032a Individual drugs were tested to determine the dose required to induce TCRgd 2/10 (20%) 55/347 (16%) nsa apoptosis of cells to 50% of untreated controls (ED50) in an annexin V/ ab lineage 0/10 (0%) 180/347 (52%) 0.0008a propidium iodide assay. Cells were treated with incremental concentra- tions of drugs but fixed ratios between individual drugs. Results were Immunophenotypic features 22,23 analyzed using the median-effect method of Chou and Talalay to CD34 10/10 (100%) 137/345 (40%) 0.0001a produce isobolograms and calculate combination indices (Calcusyn CD117 4/10 (40%) 44/332 (13%) 0.03a Software; Biosoft, Cambridge, UK). The combination index was calculated CD4 þ /8 À 3/10 (30%) 49/339 (15%) nsa based on the conservation assumption of mutually nonexclusive drug Myeloid antigen 9/10 (90%) 110/332 (33%) 0.0004a interactions. The combination index provides a numerical description of expression the effects of a drug combination: a combination index value ¼ 1, o1or 41 predicts an additive, synergic or antagonist effect of drug combination, FLT3 transcript expression respectively, at the particular ratio and concentration examined. Median 840% 25% o0.0001b Range 249–2425% 0.4–1970% Protein expression analysis and cell sorting Abbreviations: NA, not applicable; ns, not significant; WT, wild type. The FLT3 and KIT protein expression was analyzed by flow cytometry with a data reported here refer to the whole series of T-ALLs including 357 T-ALLs. surface-staining protocol using FLT3/CD135 (clone 4G8) and KIT/CD117 All immunophetnoypic markers herewith reported were assessed on blast (clone 104D2) phycoerythrin or PE-Cy7 (BD Biosciences, San Jose, CA, USA) cells recognized based on CyCD3 and CD45 pattern of expression. aFisher’s antibody conjugates. All immunophenotypic analyses were performed by exact test. bStudent’s t-test. gating live blast T cells based on their pattern of CD45 and cytoplasmic

Leukemia (2013) 305 – 314 & 2013 Macmillan Publishers Limited A rationale for multitarget kinase inhibition L Lhermitte et al 307

Figure 1. FLT3 transcript expression in normal and malignant immature T cells. (a) FLT3 transcripts are expressed during thymic T-cell maturation are the highest at the earliest stages and progressively decrease with maturation. T-ALL cases show a continuous spectrum of FLT3 transcript expression but a significant number of cases (22.2%) expressed FLT3 at a level significantly higher than whole thymus (fivefold) or even the most immature thymic subpopulations (P ¼ 8 Â 10 À 4). ISP, immature single positive (CD4); DP TCR NEG, double positive (CD4/8) TCR negative; DP TCR LOW/HIGH, double positive (CD4/8) TCR positive low or high. (b, c) FLT3 expression levels according to the TCR and genotypic subsets, respectively. ***Po0.0001. have been reported to be limited to a CD117 þ subset of adult FLT3 expression in these cases was at least twofold higher than T-ALLs12 and CD4 þ /CD8 À phenotype in pediatric T-ALLs.13 In the mean value observed in immature CD34 þ thymocytes our overall series, 6 of the 10 FLT3-mutated cases (2/4 adult cases) (P ¼ 0.001). As expected, all of the 10 FLT3Mut cases were classified were CD117 À and only 4/48 CD117 þ T-ALLs (2/25 adults cases) as FLT3High T-ALLs. The FLT3High cases demonstrated a higher rate demonstrated FLT3 mutation compared with 6/294 CD117 À of CD34, CD117 and myeloid antigen expression, as compared T-ALLs. Moreover, although a CD4 þ /CD8 À phenotype was more with low FLT3-expresser T-ALLs (FLT3Low) (Table 2). Similar levels common in FLT3 ITD/D835Y than FLT3 WT (FLT3WT) (Table 1), the of FLT3 expression between FLT3Low T-ALLs and normal thymic majority were CD4/8 double negative and one was CD4 þ /8 þ . All subpopulations suggest that these cases may express physiologi- the FLT3-mutated cases were CD34 þ and 9/10 expressed myeloid cal rather than oncogenic FLT3. Within FLT3WT cases, IM and markers, compared with 40% and 33% of the FLT3WT group, TCRgd þ cases showed higher FLT3 transcript levels compared respectively. Interestingly, FLT3 ITD and D835Y (FLT3Mut) were with ab-lineage cases (Figure 1b and Table 2). In keeping with this, observed in 8/120 (7%) IM T-ALL, 2/57 (3.5%) TCRgd þ T-ALL, but FLT3 transcript expression was higher in oncogenetic groups none of the 180 ab-lineage T-ALL (Po0.01). These data strongly associated with a TCRgd lineage orientation, such as CALM-AF103 suggest that FLT3 ITD/D835Y mutations are restricted to the and TLX328 and lower in oncogenetic groups with an ab-lineage immature/TCRgd lineage as are MLL and CALM-AF10 fusion orientation such as SIL-TAL129 (Figure 1c). transcripts.3,27 As expected, all the FLT3Mut cases demonstrated These data were then confirmed at the protein level when 54% high FLT3 transcript expression compared with FLT3-unmutated FLT3WT/High T-ALLs harbored a receptor detectable at the cell cases (Table 1). However FLT3-unmutated T-ALLs harbored a surface (Table 2 and Figure 2a). No difference in FLT3 transcript continuous pattern of FLT3 transcript expression, including high- levels was observed between CD135/FLT3-positive and -negative level expression (see below). cases within the FLT3WT/High group (173.0 (93.1–1970.0) versus 174.1 (100.0–1265.0), respectively, P ¼ 0.98). In contrast, most of the FLT3Low T-ALLs were CD135/FLT3 negative, and only 5% of the FLT3 expression in normal and malignant T cells latter showed FLT3 expression. FLT3 protein expression was not FLT3 transcripts were then assessed using quantitative reverse found in normal human thymus, even in immature subpopula- transcription PCR in 301 T-ALL and compared with whole-human tions (Supplementary Figure 4A). Overall, FLT3 protein expression thymus and sorted thymic subpopulations. During normal appeared ectopic and overrepresented in the FLT3High subset of thymopoiesis, the highest FLT3 expression was observed in the T-ALLs. As such, these FLT3High T-ALLs could be considered to most immature CD34 þ thymic population, with a progressive express deregulated FLT3 and suggests that FLT3 expression in decrease during maturation (Figure 1a). FLT3 expression in T-ALLs immature and TCRgd T-ALLs is probably oncogenic. was a continuous spectrum, including cases expressing high levels of FLT3 transcripts compared with normal thymocytes. In all, 67 High (22.2%) out of the 301 evaluated cases were classified as FLT3 high Oncogenic FLT3 activation in FLT3 T-ALLs expresser (FLT3High) on the basis of FLT3 transcript expression In order to further explore the potential oncogenic effect of FLT3 levels that exceeded the mean value in normal thymocytes by overexpression, we then checked whether FLT3 overexpression more than fivefold (threshold cut defined at 88.3%) (Figure 1a). and/or mutation was associated with constitutive activation of the

& 2013 Macmillan Publishers Limited Leukemia (2013) 305 – 314 A rationale for multitarget kinase inhibition L Lhermitte et al 308 Table 2. Characteristics of T-ALL patients according to FLT3 transcript levels

WT Overall WT Low WT High P-value

Patient characteristics N 234/291 (80%) 57/291 (20%) NA Age 20 (1–66) 19 (1–66) 23 (4–48) nsa Sex ratio (M/F) 229/62 (3.7) 183/51 (3.6) 46/11 (4.2) nsa Children/adults 124/167 (74%) 105/129 (81%) 19/38 (50%) nsa

Immunogenetic stage of arrest Immature 98/291 (34%) 61/291 (21%) 37/57 (65%) o0.0001a TCRgd 46/291 (16%) 36/291 (12%) 10/57 (18%) nsa ab-lineage 147/291 (50%) 137/291 (47%) 10/57 (18%) o0.0001a

Immunophenotypic features CD34 114/289 (39%) 73/232 (31%) 41/57 (72%) o0.0001a CD117/KIT 41/280 (15%) 26/225 (11%) 15/55 (27%) 0.005a CD135/FLT3 26/203 (13%) 9/168 (5%) 19/35 (54%) o0.0001a CD4 þ /8 À 38/285 (13%) 33/229 (14%) 5/56 (9%) nsa Myeloid antigen expression 103/290 (35%) 63/233 (27%) 40/57 (70%) o0.0001a

Transcript expression Median 25% 16.5% 171% o0.0001b Range 0.4–1970% 0.4–83% 93–1970% The data reported refer to 291 FLT3WT cases for which FLT3 transcript quantitation was available. Myeloid antigen expression refers to CD13, CD33 or CD123 expression. All immunophetnoypic markers herewith reported were assessed on blast cells recognized based on CyCD3 and CD45 pattern of expression. P-value refers to comparisons of FLT3High versus FLT3Low subgroups. aFisher’s exact test. bStudent’s t-test.

receptor. As expected, FLT3Low T-ALLs did not show constitutive was reported to have weak FLT3 specificity, we then evaluated a receptor phosphorylation. In contrast, FLT3High T-ALLs demon- more selective FLT3 inhibitor, AC220.36,37 Both CEP701-sensitive strated phosphorylation of the receptor, irrespective of their FLT3High T-ALLs remained significantly sensitive to AC220 inhibition mutational status, in line with similar data in AML18 (Figure 2b). (Figure 3c). In addition, three CEP701-resistant FLT3WT/High cases (UPN We then investigated the phosphorylation status of ERK and 539, 401 and 85) were evaluated using AC220; all remained resistant STAT5 signaling molecules as downstream effectors of the FLT3 to FLT3 inhibition (data not shown). Of note, resistance was also not pathway.30–32 We opted for a flow-cytometric approach in order reversed on culture of primary blast cells in the absence of FL to optimize the limited material availability and focus on blast (Figure 3d) and reversion of the phosphorylation state of FLT3 was T cells. As expected, FLT3Mut/High cases harbored constitutive evidenced for both CEP701 and AC220 (Figure 3e), whether in the phosphorylation of FLT3, and downstream ERK and STAT5 presence or absence of FL (Figure 3f), including in cases harboring molecules, with a significant increase in response to FL resistance to FLT3 inhibition. To decipher the biological relevance of (Figure 2c). As an internal control, no FLT3 phosphorylation was this heterogeneous sensitivity to FLT3 inhibition, we hypothesized detected in FLT3Low cases. The latter could display constitutive that FLT3 may not be deregulated in isolation but rather in concert activation of ERK and STAT5 molecules, possibly dependent on the with other signaling RKs, thus bypassing the inhibition of FLT3 in alternative signaling pathways, but this was not influenced by FL resistant cases. adjunction. Finally, FLT3WT/High T-ALLs demonstrated constitutive activation of FLT3, STAT5 and ERK molecules, which was Concerted transcriptional regulation of RK in FLT3High T-ALLs dramatically enhanced when FL was added to the culture regardless of the FLT3 mutational status medium. These data favored ectopic, functional and oncogenic To screen for potential co-deregulated RK, we quantified transcripts FLT3 signaling in this subset of T-ALL. In keeping with this, Pim-1 from 65 human RK38 (Website TGO: http://www.geneontology.org/) transcript levels, considered as a functional readout for FLT3 by TLDA Human Protein Kinase Array (see Material and Methods oncogenic activation,33–35 were significantly higher in the FLT3High Low section). Thymic subpopulations and 31 T-ALLs, including 7 group as compared with FLT3 cases (P ¼ 0.02; Figure 2d). FLT3Mut/High, 13 FLT3WT/High and 11 FLT3Low, were screened. Altogether, our results suggest that upregulation of FLT3 is a As an internal control, results of FLT3 transcript quantification common abnormality in T-ALL and is associated with activation of 18 High obtained by TLDA and classical real-time quantitative PCR were the receptor, as in AML. As such, FLT3 T-ALLs, which are found to be highly concordant (Supplementary Figure S2A). mostly immature, could be eligible for tyrosine kinase inhibitor- Moreover, intra- and inter-experimental controls demonstrated targeted therapy. high reproducibility of the measurements, which allowed precise comparisons of receptor tyrosine kinase profiles between individuals Heterogeneous sensitivity of FLT3 high-expressing leukemic T cells (Supplementary Figure S2B). to FLT3 inhibition Of the 65 RK, 16 were not expressed at significant levels in To determine whether FLT3 might be a target for therapy in T-ALL, either normal thymus or T-ALLs. Sixteen RKs were not expressed in we then assessed the efficiency of FLT3 inhibition on primary the thymus but were aberrantly expressed in at least one case of cultured T-ALL blast cells using the CEP701 small molecule. For T-ALL. In particular, RKs, such as TIE1, NTRK1, TYRO3 and MERTK, this purpose, we tested survival of FLT3Mut/High, FLT3WT/High and were overexpressed in a significant number of T-ALLs (9–11). FLT3Low primary blast cells exposed to incremental doses of Several other RKs were expressed in the thymus and demon- CEP701. As expected, all of the six FLT3Low T-ALLs tested were strated deviant expression patterns in T-ALL. TGFBRI, TGFBRII, resistant to CEP701. Only 1/2 FLT3Mut/High tested and 1/4 FLT3WT/High BMPR2 and DDR1 were underexpressed in most T-ALLs. showed in vitro sensitivity to CEP701 (Figures 3a and b). As CEP701 In contrast, EPHB4, FGFRL1, KIT, TIE1, NTRK1 and MERTK were

Leukemia (2013) 305 – 314 & 2013 Macmillan Publishers Limited A rationale for multitarget kinase inhibition L Lhermitte et al 309

FLT3WT/High_UPN527 FLT3Low_UPN407

RFI 2.92 RFI 0.89 cell lines T-ALL

Mut/High Low Low Mut/High WT/High FLT3 FLT3 MV4.11 DND41 UPN_153 FLT3 UPN_380 FLT3WT/High_UPN97 FLT3Low_UPN380 UPN_527 p-FLT3 160 kDa RFI 2.20 RFI 1.01 FLT3 160 kDa

Actin Count FLT3 / CD135

Mut/High WT/High WT/Low

FLT3 FLT3 FLT3

FLT3Mut/High FLT3WT/High FLT3WT/Low UPN_364 UPN_7 UPN_17 UPN_221 UPN_364 UPN_7 UPN_539 UPN_17

Control

FL- Control p-FLT3 FL+ FL- FL+

Control

FL- Control

p-STAT5 FL- FL+ FL+

Control FL- Control p-ERK FL- FL+ FL+

Figure 2. FLT3 expression and activation in T-ALL. (a) FLT3 expression was confirmed at the protein level by flow cytometry in FLT3High T-ALLs, whereas it was undetectable in FLT3Low T-ALLs. (b) Western blot analysis demonstrates that high transcriptional expression of FLT3 is associated with autophosphorylation of the receptor, both in FLT3-mutated and -unmutated cases. (c) Signaling properties of T-ALL depending on their FLT3 status evaluated by flow cytometry, in the presence or absence of FL stimulation (5 ng/ml). FLT3, STAT5 and ERK signaling molecules were monitored 30 min, 1 h and 1 h after FL exposure, respectively. Changes in phosphorylation of signaling proteins were calculated using the log2 ratio of the median fluorescence intensity of cells stained with anti-phosphoepitope antibody versus isotype control, and represented as a color scale. Results are plotted in heatmaps (left) and exemplified for representative cases in overlay histograms (right). (d) Quantitative reverse transcription PCR shows increased Pim-1 expression in FLT3High T-ALL as compared with FLT3Low cases, favoring oncogenic activation of the FLT3 pathway. *Refers to statistical significance Po0.05. overexpressed selectively in FLT3High T-ALLs. Overall, FLT3Low RK multitargeting could partially reverse leukemic T-cell T-ALLs demonstrated an expression profile similar to the resistance physiological levels seen in the whole-thymus samples. Interest- CEP701-sensitive and -resistant FLT3High cases did not split apart ingly, overexpressed RKs were all virtually found within the and intermingled within the FLT3High cluster. We then hypothesized FLT3High subgroup, with the exception of FGFR1, which was that, despite sharing a common biology, these cases may thus expressed only in the FLT3Low subset and FLT4 and LTBP1, which display variable dependence on individual RK identified within the were overexpressed in both groups (Figure 4 and Supplementary FLT3High-associated RK signature. In order to reverse resistance to Figure S5). FLT3 inhibition, we compared individual receptor kinomic specifi- Unsupervised analysis of the receptor kinome revealed two cities of CEP701-resistant and -sensitive cases. Interestingly, among distinct clusters, which spontaneously split FLT3Low and FLT3High the five FLT3High expressers explored by TLDA, the two cases that T-ALL cases, regardless of the FLT3 mutational status and as were sensitive to FLT3 inhibition expressed low level KIT transcripts, such identified a ‘FLT3High RKinome signature’. (Figure 4 and whereas the three resistant cases exhibited high level of KIT Supplementary Figure S5) Interestingly, RK profiling did not (Figure 4). Expression was confirmed at the protein level (Figure 5a). enable discrimination of FLT3 mutants from other FLT3 high The levels of KIT expression (Figure 5b) were, as for FLT3, higher than expressers, suggesting that FLT3High cases share similar underlying those seen in thymic subsets and, as such, were considered RK biology, irrespective of their mutational status. supraphysiological (Figure 5b and Supplementary Figure S4B).

& 2013 Macmillan Publishers Limited Leukemia (2013) 305 – 314 A rationale for multitarget kinase inhibition L Lhermitte et al 310

FLT3Mut/High FLT3WT/High FLT3WT/Low UPN_221 UPN_7 UPN_17

DMSO control CEP701 20nM

Control FL- FL+ FL+ CEP701+ FL+ AC220+

FLT3Mut/High CEP701 50 nM CEP701 100nM

UPN_221 UPN_647 UPN_364

Control FL-

Propidium Iodide FL+ Annexin V FL+ AC220+ FL- AC220+

Figure 3. (a) Sensitivity experiments of primary T-ALL cells upon RK inhibition. Results are expressed as the relative number of surviving cells as compared with an internal control (dimethyl sulfoxide). (b) Flow diagrams show the sensitivity to CEP701 of a representative case (UPN_221). (c) Sensitivity experiments of primary T-ALL cells to alternative FLT3 inhibitors harboring more potent and selective properties, with regard to FLT3. Cases sensitive to CEP701 remain sensitive to AC220. (d) Exposure of cells to RK inhibitors with or without FL does not result in any significant difference in sensitivity (ns, not significant). (e) Dephosphorylation of FLT3 evaluated by flow cytometry in response to FLT3- inhibitor exposure. (f) Dephosphorylation of FLT3 was not impacted by the presence of FL in FLT3-inhibitor-sensitive and -resistant T-ALLs.

Pharmacological inhibition of KIT using STI571 or AB101039 could FLT3High T-ALLs could provide the basis for rational multitargeted be tested in three resistant FLT3High cases. One case remained RK therapy in immature T-ALLs. resistant either to FLT3, KIT or to coinhibition of both of these RKs. Interestingly, leukemic cell survival was impaired in response to KIT inhibition for the two other cases and either a potentiation or DISCUSSION synergistic effect was observed when combining FLT3 and KIT We demonstrate here that FLT3 expression is common in inhibition for these patients (Figures 5c and d), supporting a immature and gd-lineage T-ALL and that FLT3 mutations are reversion of resistance by targeting several converging RKs. In line restricted to these subsets. Response to CEP701, one of the FLT3 with this, these cases showed concomitant protein expression of inhibitors,40 was restricted to FLT3High T-ALLs but was sporadic both FLT3 and KIT receptors (Figure 5e), further excluding a and independent of mutational status, probably because FLT3 mosaic expression and a simple additive effect of RK coinhibition. expression was frequently accompanied by deregulation of other At the biochemical level, we confirmed significant inhibition of KIT RK, including KIT. The synergic cytotoxic effect of FLT3 and KIT upon exposure to AB1010, but not AC220, using drug doses inhibitors provide proof of a rationale for multi-RK targeting in necessary for synergistic impairment of survival by RK coinhibition immature T-ALL. (Figure 5f). Taken together, these data favor co-deregulation of RK FLT3 mutations (D835Y and ITD) were infrequent (3%) in T-ALL networks in T-ALL and suggest that pattern of RK expression in and were restricted to cases with an immature/TCRgd phenotype

Leukemia (2013) 305 – 314 & 2013 Macmillan Publishers Limited A rationale for multitarget kinase inhibition L Lhermitte et al 311

FLT3 category

FLT3Mut/High FLT3WT/High FLT3Low

CEP-sensitivity Sensitive T-ALL 131 T-ALL 242 T-ALL 247 T-ALL 279 T-ALL 250 T-ALL 305 T-ALL 299 T-ALL 300 T-ALL 17 T-ALL 407 CD34+ 3-4+8+ DP T-ALL 363 ISP 3+4+ SP 3+8+ SP Thymus Thymus T-ALL 148 T-ALL 221 T-ALL 58 T-ALL 364 T-ALL 174 T-ALL 396 T-ALL 44 T-ALL 473 T-ALL 401 T-ALL 16 T-ALL 88 T-ALL 375 T-ALL 353 T-ALL 405 T-ALL 7 T-ALL 30 T-ALL 539 T-ALL 406 T-ALL 85 T-ALL 97 Code FLT3 Relative RK gene Resistant expression level CEP-701 sensitivity Immunogenetic Immunogenetic status Oncogenetic - t -----SSS- C --TMM ------CC - C - Notch1 Low High Immature T-ALL FBXW7 Mean TCRγδ T-ALL Pre-αβ T-ALL TCRαβ T-ALL Recurrent oncogenetic abnormalities C CALM-AF10 M MLL S SIL-TAL T TLX3 overexpression t TLX1 overexpression - none of above

Mutational status Gene mutation No gene mutation Not available or not applicable Figure 4. RK profiles of human T-ALL samples and thymic subpopulations. Hierarchical classification based on RK transcript levels shows that FLT3High/Mut and FLT3High/WT share similar pattern of RK expression, which is distinguishable from both FLT3Low cases and normal thymocytes. The heat map represents the relative and differential expression of each RK gene and is normalized relative to whole-thymic expression, shown here in gray as the reference. Over- or underexpression as compared with this reference is highlighted in red and blue, respectively, with depth of color reflecting deviation from the normal thymocytes reference. Each color change reflects a 1-delta cycle threshold step change in . Transcripts used to determine the RK signature were selected in an unsupervised manner by removing gene transcripts that had s.d. of observed values over the whole series o1.5. Complete classification based on the entire set of RK transcripts without RK selection provides similar results and is shown in Supplementary figure S4.

but were not restricted to CD117 þ or CD4 þ /8 À T-ALLs, as Our in vitro experiments demonstrated sensitivity and induction previously reported.12,13 It is likely that high-level FLT3 expression, of apoptosis in primary T-ALL blast upon exposure to CEP701. currently idiopathic, has oncogenic significance because the levels Response was restricted to FLT3High-expressing cases, in keeping were higher than those found in the thymus and its with an oncogenic role for the FLT3 pathway in these cases, but subpopulations, including the most immature (CD34 þ ) fraction, was sporadic (two out of six FLT3High T-ALLs tested) and inde- and also segregated with the same immature/TCRgd phenotype. pendent of mutational status. Primary resistance to FLT3 inhibitors These data are coherent with transcriptional profiling identifica- in AML, observed in approximately 30% of FLT3-mutated cases,41 tion of FLT3 expression as a characteristic contributor to the has been correlated with the type of mutation42–44 or considered signature of immature T-ALL.5 FLT3 overexpression in T-ALL to reflect dependence on the FLT3 molecule but not on FLT3 was associated with constitutive activation of the receptor, as kinase activity.45 Two recent studies also suggested that secon- evidenced by phosphorylation of the receptor. Few cases of T-ALL dary resistance was associated in a significant number of cases could be tested for FLT3 phosphorylation, given the limited with increases in circulating FL concentrations in vivo.46,47 Failure availability of material suitable for phosphoepitope analysis by to reverse resistance to CEP701 in the absence of FL did not western blot. T-ALL blasts usually demonstrate a short life confirm this hypothesis in our in vitro T-ALL model. expectancy ex-vivo that may alter phophoproteomic profiles and Inherent resistance has also been proposed to be linked to the no FLT3High T-ALL cell line has been identified (see Supplementary compensatory activation pathways, rendering leukemic cells Figure S6). Our observation favors oncogenic activation of the independent of FLT3 activation.48 Our data are in keeping with FLT3 pathway in FLT3Mut/High and at least a subset of FLT3WT/High this. We used a simple assay to define the receptor kinome T-ALL. The high level of PIM-1 transcripts in FLT3High T-ALLs is in signature of individual T-ALLs and screen RKs eligible for keeping with this, because PIM-1 is induced by FLT3 oncogenic multitargeting. This demonstrated that FLT3 deregulation is activation.33–35 This observation is also consistent with previous virtually always associated with deregulation of a RK network studies showing that FLT3 overexpression in AML is associated and that virtually all T-ALLs display a receptor kinome, which is with constitutive phosphorylation of the receptor, a dismal distinct from normal thymocytes. The exact oncogenic basis for prognosis, and sensitivity to kinase inhibition.18 Immature T-ALL this kinome signature remain to be further investigated. is also related to a poor response to chemotherapy, in both adults As RK signatures in FLT3-expressing immature/TCRgd T-ALLs are and children,6–9 and may therefore benefit from FLT3 inhibition. reproducible, it is probable that selective inhibition will not lead to

& 2013 Macmillan Publishers Limited Leukemia (2013) 305 – 314 A rationale for multitarget kinase inhibition L Lhermitte et al 312

CEP701 AC220 AB1010 CEP+AB AC+AB

FLT3Low_UPN407

FLT3WT/High_UPN7

FLT3WT/High_UPN539 Leukemic Cell survival Count Kit / CD117

2.0 ED50ED75 ED90 0 0 0.5 1.0 1.5 2.0 150 -2.0 UPN_97 UPN_647 100 KIT/CD117 AC (nM) Lof(fa/fu) -4.0 50 FLT3/CD135 -6.0 AC+AB AC AB Log(D) 123 UPN_647 AB (uM)

Combination IndexED50 ED75 ED90 Alpha-155D816V No drug AB1010 AC220 p-KIT UPN_647 0.79 0.71 0.65 Actin

Figure 5. RK coinhibition in T-ALL: (a) KIT protein expression by flow cytometry. KIT was detectable neither in FLT3Low T-ALL cases (left) nor in CEP701-sensitive FLT3High T-ALLs (middle). In contrast, CEP701-resistant FLT3High cases, that also expressed high levels of KIT transcript, demonstrated KIT protein expression (right) (b) Differential transcriptional expression level between classes of T-ALL as compared with normal thymocyte development. (c) In vitro sensitivity experiments of FLT3, KIT and dual FLT3/KIT inhibition to reverse ITK resistance. (d) Median-effect plot (left) and isobologram curve (right) for the effect of AC220 and AB1010 on induction of apoptosis. The latter was measured by annexin V/ propidium iodide-based assay on primary blast cells from UPN_647, and results were calculated using CalcuSyn software (see Materials and Methods). The experimental points fall to the left of the predicted line of additive effect, indicating a synergistic interaction of AC220 and AB1010. This is reflected by combination index (CI) values reported in the table below the curves, reflecting an even more synergistic effect as the CI value falls o1. D, drug dose; Fa, fraction affected; fu, fraction unaffected. (e) Flow-cytometric analysis of FLT3 and KIT protein expression. Results indicate coexpression of both RK receptors by T-ALL blast cells. (f) Western blot analysis of KIT phosphorylation upon drug exposure (AB1010 10 mM and AC220 50 nM) showed significant C-Kit phosphorylation decrease after AB1010 but not AC220 exposure.

cell death, due to unmodified activity of unaffected RK. Different Taken together, we have identified concerted receptor kinome RK may convergently affect distinct signaling nodes,49 with a deregulation in FLT3High IM/TCRgd T-ALLs, which may explain why significant functional redundancy. Several recent studies used single-agent inhibition is less appropriate than multitarget RK multitarget kinase inhibitors and demonstrated synergistic inhibition. This RK screening procedure, applied here to human cytotoxicity against blast cells, thereby suggesting that T-ALL samples, may be extended to other hematological or solid simultaneous inhibition of multiple key target kinases could be malignancies to design and test new multitargeted therapies. an efficient therapeutic strategy.50,51 One elegant study in glioblastoma highlighted coactivation of multiple RK, leading to redundant inputs and limiting efficacy of targeting single RK.52 CONFLICT OF INTEREST Noteworthy, resistance to single agents in this study could be The authors declare no conflict of interest. reversed using multiple RK-targeting agents. In keeping with this, simultaneous FLT3 and KIT inhibition by CEP701 and AB1010 was ACKNOWLEDGEMENTS synergic in one T-ALL, providing proof of concept for RK High We thank all the biologists and clinicians of the LALA/GRAALL adult and FRALLE multitargetting in immature, FLT3 T-ALLs. The identification pediatric French ALL groups for providing T-ALL samples and results. This work was of coexpressed RKs, such as EPHB4, NTRK1, and the previously supported by grants from the Association Cent Pour Sang La Vie and the association reported MERTK, also represent candidate targets for RK-inhibition Laurette Fugain. We thank the Necker-Enfants-Malades department of pediatric 53 therapy. cardiac surgery for providing thymic material.

Leukemia (2013) 305 – 314 & 2013 Macmillan Publishers Limited A rationale for multitarget kinase inhibition L Lhermitte et al 313 AUTHOR CONTRIBUTIONS 23 Chou TC, Talalay P. Quantitative analysis of dose-effect relationships: the com- LL and RBA performed research, data analysis and wrote the paper. PV, ML, bined effects of multiple drugs or inhibitors. Adv Enzyme Regul 1984; 22: NB and ASB performed research. VG performed statistical data analysis. HD and 27–55. OH performed data analysis. EM and VA oversaw conceptual development of 24 Tse KF, Novelli E, Civin CI, Bohmer FD, Small D. Inhibition of FLT3-mediated the project and wrote the paper. transformation by use of a tyrosine kinase inhibitor. Leukemia 2001; 15: 1001–1010. 25 Georgin-Lavialle S, Lhermitte L, Suarez F, Yang Y, Letard S, Hanssens K et al. Mast cell leukemia: identification of a new c-Kit mutation, dup(501–502), and REFERENCES response to , a c-Kit tyrosine kinase inhibitor. Eur J Haematol 2012; 1 Look AT. Molecular pathways in T-cell acute lymphoblastic leukemia: ramifications 89: 47–52. for therapy. Clin Adv Hematol Oncol 2004; 2: 779–780. 26 Kotecha N, Krutzik PO, Irish JM. Web-based analysis and publication of flow 2 Asnafi V, Beldjord K, Boulanger E, Comba B, Le Tutour P, Estienne MH et al. cytometry experiments. Curr Protoc Cytom. Chapter 10: Unit10 17. Analysis of TCR, pT alpha, and RAG-1 in T-acute lymphoblastic leukemias 27 Ferrando AA, Armstrong SA, Neuberg DS, Sallan SE, Silverman LB, Korsmeyer SJ improves understanding of early human T-lymphoid lineage commitment. Blood et al. Gene expression signatures in MLL-rearranged T-lineage and B-precursor 2003; 101: 2693–2703. acute leukemias: dominance of HOX dysregulation. Blood 2003; 102: 262–268. 3 Asnafi V, Radford-Weiss I, Dastugue N, Bayle C, Leboeuf D, Charrin C et al. CALM- 28 Asnafi V, Beldjord K, Libura M, Villarese P, Millien C, Ballerini P et al. Age-related AF10 is a common fusion transcript in T-ALL and is specific to the TCRgamma- phenotypic and oncogenic differences in T-cell acute lymphoblastic leukemias delta lineage. Blood 2003; 102: 1000–1006. may reflect thymic atrophy. Blood 2004; 104: 4173–4180. 4 Ferrando AA, Neuberg DS, Staunton J, Loh ML, Huard C, Raimondi SC et al. Gene 29 Macintyre EA, Smit L, Ritz J, Kirsch IR, Strominger JL. Disruption of the SCL locus in expression signatures define novel oncogenic pathways in acute lympho- T-lymphoid malignancies correlates with commitment to the T-cell receptor alpha blastic leukemia. Cancer Cell 2002; 1: 75–87. beta lineage. Blood 1992; 80: 1511–1520. 5 Soulier J, Clappier E, Cayuela JM, Regnault A, Garcia-Peydro M, Dombret H et al. 30 Hayakawa F, Towatari M, Kiyoi H, Tanimoto M, Kitamura T, Saito H et al. HOXA genes are included in genetic and biologic networks defining human acute Tandem-duplicated Flt3 constitutively activates STAT5 and MAP kinase and T-cell leukemia (T-ALL). Blood 2005; 106: 274–286. introduces autonomous cell growth in IL-3-dependent cell lines. 2000; 6 Rubnitz JE, Onciu M, Pounds S, Shurtleff S, Cao X, Raimondi SC et al. Acute mixed 19: 624–631. lineage leukemia in children: the experience of St Jude Children’s Research 31 Masson K, Ronnstrand L. Oncogenic signaling from the hematopoietic growth Hospital. Blood 2009; 113: 5083–5089. factor receptors c-Kit and Flt3. Cell Signal 2009; 21: 1717–1726. 7 Rubio MT, Dhedin N, Boucheix C, Bourhis JH, Reman O, Boiron JM et al. Adult 32 Mizuki M, Fenski R, Halfter H, Matsumura I, Schmidt R, Muller C et al. Flt3 muta- T-biphenotypic acute leukaemia: clinical and biological features and outcome. tions from patients with acute myeloid leukemia induce transformation of 32D Br J Haematol 2003; 123: 842–849. cells mediated by the Ras and STAT5 pathways. Blood 2000; 96: 3907–3914. 8 Coustan-Smith E, Mullighan CG, Onciu M, Behm FG, Raimondi SC, Pei D et al. Early 33 Kim KT, Baird K, Ahn JY, Meltzer P, Lilly M, Levis M et al. Pim-1 is upregulated by T-cell precursor leukaemia: a subtype of very high-risk acute lymphoblastic leu- constitutively activated FLT3 and plays a role in FLT3-mediated cell survival. Blood kaemia. Lancet Oncol 2009; 10: 147–1456. 2005; 105: 1759–1767. 9 Asnafi V, Buzyn A, Thomas X, Huguet F, Vey N, Boiron JM et al. Impact of TCR 34 Kim KT, Levis M, Small D. Constitutively activated FLT3 phosphorylates BAD status and genotype on outcome in adult T acute lymphoblastic leukemia: partially through pim-1. Br J Haematol 2006; 134: 500–509. a LALA-94 study. Blood 2005; 105: 3072–3078. 35 Vu HA, Xinh PT, Kano Y, Tokunaga K, Sato Y. The juxtamembrane domain in ETV6/ 10 Brown P, Meshinchi S, Levis M, Alonzo TA, Gerbing R, Lange B et al. Pediatric FLT3 is critical for PIM-1 up-regulation and cell proliferation. Biochem Biophys Res AML primary samples with FLT3/ITD mutations are preferentially killed by FLT3 Commun 2009; 383: 308–313. inhibition. Blood 2004; 104: 1841–1849. 36 Pratz KW, Sato T, Murphy KM, Stine A, Rajkhowa T, Levis M. FLT3-mutant allelic 11 Taketani T, Taki T, Sugita K, Furuichi Y, Ishii E, Hanada R et al. FLT3 mutations in burden and clinical status are predictive of response to FLT3 inhibitors in AML. the activation loop of tyrosine kinase domain are frequently found in infant ALL Blood 2010; 115: 1425–1432. with MLL rearrangements and pediatric ALL with hyperdiploidy. Blood 2004; 103: 37 Zarrinkar PP, Gunawardane RN, Cramer MD, Gardner MF, Brigham D, Belli B et al. 1085–1088. AC220 is a uniquely potent and selective inhibitor of FLT3 for the treatment of 12 Paietta E, Ferrando AA, Neuberg D, Bennett JM, Racevskis J, Lazarus H et al. acute myeloid leukemia (AML). Blood 2009; 114: 2984–2992. Activating FLT3 mutations in CD117/KIT( þ ) T-cell acute lymphoblastic leukemias. 38 Manning G, Whyte DB, Martinez R, Hunter T, Sudarsanam S. The protein kinase Blood 2004; 104: 558–560. complement of the . Science 2002; 298: 1912–1934. 13 Van Vlierberghe P, Meijerink JP, Stam RW, van der Smissen W, van Wering ER, 39 Dubreuil P, Letard S, Ciufolini M, Gros L, Humbert M, Casteran N et al. Masitinib Beverloo HB et al. Activating FLT3 mutations in CD4 þ /CD8 À pediatric T-cell (AB1010), a potent and selective tyrosine kinase inhibitor targeting KIT. PLoS One acute lymphoblastic leukemias. Blood 2005; 106: 4414–4415. 2009; 4:e7258. 14 Scharnhorst V, Wals J, Beverloo HB, Langerak AW, van der Velden VH. Mutation of 40 Levis M, Allebach J, Tse KF, Zheng R, Baldwin BR, Smith BD et al. A FLT3-targeted FLT3 is not a general phenomenon in CD117-positive T-ALL. Leuk Res 2006; 30: tyrosine kinase inhibitor is cytotoxic to leukemia cells in vitro and in vivo. Blood 245–246. 2002; 99: 3885–3891. 15 Birg F, Courcoul M, Rosnet O, Bardin F, Pebusque MJ, Marchetto S et al. Expression 41 Stone RM, DeAngelo DJ, Klimek V, Galinsky I, Estey E, Nimer SD et al. Patients with of the FMS/KIT-like gene FLT3 in human acute leukemias of the myeloid and acute myeloid leukemia and an activating mutation in FLT3 respond to a small- lymphoid lineages. Blood 1992; 80: 2584–2593. molecule FLT3 tyrosine kinase inhibitor, PKC412. Blood 2005; 105: 54–60. 16 Carow CE, Levenstein M, Kaufmann SH, Chen J, Amin S, Rockwell P et al. 42 Bagrintseva K, Schwab R, Kohl TM, Schnittger S, Eichenlaub S, Ellwart JW et al. Expression of the hematopoietic receptor FLT3 (STK-1/Flk2) in Mutations in the tyrosine kinase domain of FLT3 define a new molecular human leukemias. Blood 1996; 87: 1089–1096. mechanism of acquired drug resistance to PTK inhibitors in FLT3-ITD-transformed 17 Rosnet O, Buhring HJ, Marchetto S, Rappold I, Lavagna C, Sainty D et al. Human hematopoietic cells. Blood 2004; 103: 2266–2275. FLT3/FLK2 receptor tyrosine kinase is expressed at the surface of normal and 43 Clark JJ, Cools J, Curley DP, Yu JC, Lokker NA, Giese NA et al. Variable sensitivity of malignant hematopoietic cells. Leukemia 1996; 10: 238–248. FLT3 activation loop mutations to the small molecule tyrosine kinase inhibitor 18 Ozeki K, Kiyoi H, Hirose Y, Iwai M, Ninomiya M, Kodera Y et al. Biologic and clinical MLN518. Blood 2004; 104: 2867–2872. significance of the FLT3 transcript level in acute myeloid leukemia. Blood 2004; 44 Grundler R, Thiede C, Miething C, Steudel C, Peschel C, Duyster J. Sensitivity 103: 1901–1908. toward tyrosine kinase inhibitors varies between different activating mutations of 19 Swerdlow S, Campo E, Harris N, Jaffe E, Pileri S, Stein H et al. WHO Classification the FLT3 receptor. Blood 2003; 102: 646–651. of Tumors of Haematopoietic and Lymphoid Tissues. 4th edn. IARC Press: 45 Breitenbuecher F, Markova B, Kasper S, Carius B, Stauder T, Bohmer FD et al. Lyon, 2008. A novel molecular mechanism of primary resistance to FLT3-kinase inhibitors in 20 Libura M, Asnafi V, Tu A, Delabesse E, Tigaud I, Cymbalista F et al. FLT3 and MLL AML. Blood 2009; 113: 4063–4073. intragenic abnormalities in AML reflect a common category of genotoxic stress. 46 Levis M, Ravandi F, Wang ES, Baer MR, Perl A, Coutre S et al. Results from a Blood 2003; 102: 2198–2204. randomized trial of salvage chemotherapy followed by for patients 21 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis and display of with FLT3 mutant AML in first relapse. Blood 2009; 117: 3294–3301. genome-wide expression patterns. Proc Natl Acad Sci USA 1998; 95: 14863–14868. 47 Sato T, Yang X, Knapper S, White P, Smith BD, Galkin S et al. FLT3 ligand impedes 22 Chou TC. Theoretical basis, experimental design, and computerized simulation of the efficacy of FLT3 inhibitors in vitro and in vivo. Blood 2011; 117: 3286–3293. synergism and antagonism in drug combination studies. Pharmacol Rev 2006; 58: 48 Knapper S, Mills KI, Gilkes AF, Austin SJ, Walsh V, Burnett AK. The effects of 621–681. lestaurtinib (CEP701) and PKC412 on primary AML blasts: the induction of

& 2013 Macmillan Publishers Limited Leukemia (2013) 305 – 314 A rationale for multitarget kinase inhibition L Lhermitte et al 314 cytotoxicity varies with dependence on FLT3 signaling in both FLT3-mutated and treatment of leukemias caused by oncogenic PTKs. Proc Natl Acad Sci USA 2004; wild-type cases. Blood 2006; 108: 3494–3503. 101: 3130–3135. 49 Chalandon Y, Schwaller J. Targeting mutated protein tyrosine kinases and their 52 Stommel JM, Kimmelman AC, Ying H, Nabioullin R, Ponugoti AH, Wiedemeyer R signaling pathways in hematologic malignancies. Haematologica 2005; 90: 949–968. et al. Coactivation of receptor tyrosine kinases affects the response of tumor cells 50 Jiang BH, Liu LZ. Role of mTOR in anticancer drug resistance: perspectives for to targeted therapies. Science 2007; 318: 287–290. improved drug treatment. Drug Resist Updat 2008; 11: 63–76. 53 Linger RM, DeRyckere D, Brandao L, Sawczyn KK, Jacobsen KM, Liang X et al. 51 Mohi MG, Boulton C, Gu TL, Sternberg DW, Neuberg D, Griffin JD et al. Mer receptor tyrosine kinase is a novel therapeutic target in pediatric B-cell Combination of rapamycin and protein tyrosine kinase (PTK) inhibitors for the acute lymphoblastic leukemia. Blood 2009; 114: 2678–2687.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia (2013) 305 – 314 & 2013 Macmillan Publishers Limited