Deregulated Apoptosis Signaling in Core-Binding Factor Leukemia Differentiates Clinically Relevant, Molecular Marker-Independent Subgroups

Deregulated apoptosis signaling in core-binding factor leukemia differentiates clinically relevant, molecular marker-independent subgroups

SC Lu¨ck1, AC Russ1, U Botzenhardt1, P Paschka1, RF Schlenk1,HDo¨hner1, S Fulda2,KDo¨hner1 and L Bullinger1

1Department of Internal Medicine III, University Hospital Ulm, Ulm, Germany and 2Institute for Experimental Cancer Research in Pediatrics, Goethe-University Frankfurt am Main, Frankfurt am Main, Germany

Core-binding factor (CBF) leukemias, characterized by trans- patients relapse, putting the ‘favorable risk’ into perspective,8,9 it locations t(8;21) or inv(16)/t(16;16) targeting the CBF, constitute is of great interest to study potential cooperating events in CBF acute myeloid leukemia (AML) subgroups with favorable leukemias in order to improve and individualize targeted prognosis. However, about 40% of patients relapse and the 10,11 current classification system does not fully reflect this clinical therapies. In recent years, the more detailed molecular heterogeneity. Previously, gene expression profiling (GEP) characterization of the disease has provided novel insights with revealed two distinct CBF leukemia subgroups displaying respect to the impact of additional cytogenetic aberrations,9 significant outcome differences and identified apoptotic signal- molecular aberrations such as mutations of KIT,11 and recently ing, MAPKinase signaling and chemotherapy-resistance also aberrant gene expression patterns.7 mechanisms among the most significant differentially regulated Apoptosis, occurring either following triggering of the cell pathways. We now tested different inhibitors of the respective pathways in a cell line model (six cell lines reflecting the CBF surface death receptors (extrinsic pathway) or involving the subgroup-specific gene expression alterations), and found mitochondria (intrinsic pathway), is deregulated in most, if not apoptotic signaling to be differentiating between the CBF all, cancers.12,13 Here, inhibitors of apoptosis (IAP) proteins are subgroup models. In accordance, primary samples from newly known to have an important role in many types of human diagnosed CBF AML patients (n ¼ 23) also showed differential cancer, including leukemia, and are associated with chemore- sensitivity to in vitro treatment with a Smac mimetic such as sistance, disease progression and poor prognosis.14,15 Therefore, BV6, an antagonist of inhibitor of apoptosis (IAP) proteins, and ABT-737, a BCL2 inhibitor. Furthermore, GEP revealed the IAPs are interesting targets for cancer treatment, and currently BV6-resistant cases to resemble the previously identified several small molecule inhibitors of IAPs, for example the unfavorable CBF subgroup. Thus, our current findings show dimeric Smac mimetic BV6,16 are tested in clinical trials.14 deregulated IAP expression and apoptotic signaling to differ- Similarly, BCL2 family members have a crucial role in the entiate clinically relevant CBF subgroups, which were indepen- intrinsic apoptotic pathway, as the balance of their anti- and dent of known molecular markers, thereby providing a starting pro-apoptotic members regulates the loss of mitochondrial point for novel therapeutic approaches. membrane potential and release of cytochrome c, and thus the Leukemia (2011) 25, 1728–1738; doi:10.1038/leu.2011.154; 17 published online 21 June 2011 activation of downstream effectors. As most chemotherapeutic Keywords: core binding factor; acute myeloid leukemia; gene drugs converge on the intrinsic pathway, this balance of BCL2 expression profiling; apoptosis; BCL2; inhibitor of apoptosis proteins family members is critical in determining the sensitivity of cells to apoptosis induction by drug treatment. ABT-737, a BH-3 (BCL2 homology domain) mimetic, is to date the most potent described small molecule inhibitor of antiapoptotic BCL2 family Introduction members (along with its orally active analog, ABT-263), showing promising in vitro results in combination with Acute myeloid leukemia (AML) is a genetically heterogeneous 17,18 1 established chemotherapeutic drugs. Besides deregulated disease with a proposed multistep pathogenesis. Core-binding apoptosis, other pathways frequently activated in different factor (CBF) leukemias represent a subgroup of AML character- tumors include the classical (RAF/MEK/ERK) MAPKinase pathway, ized by chromosomal aberrations involving the heterodimeric which leads to cell cycle progression and tumor survival. transcription factor CBF, which are likely to be the primary hits. Pathway activation results often from activating mutations in a These are t(8;21)(q22;q22), producing the fusion gene RUNX1/ receptor tyrosine kinase, RAS, BRAF or MEK1, which can be RUNX1T1, and inv(16)(p13.1q22) or t(16;16)(p13.1;q22), re- 2 targeted by different small molecule inhibitors currently tested in sulting in CBFB/MYH11. Both rearrangements result in fusion clinical trials.19 Another possibility of activating MAPKinase proteins, which act as dominant negative forms of the transcrip- signaling is by PAK1, which promotes cell survival through tion factor complex CBF. As the CBF is essential in normal 3,4 phosphorylation of BAD and regulates microtubule dynamics by hematopoiesis, its disruption in murine models predisposes to 20 5 phosphorylation of Stathmin, among others. Alterations in leukemia, but does not lead to leukemia by itself. cyclin-dependent kinases (CDKs) often mediate cell cycle Like AML, in general, CBF leukemias exhibit also some 6 defects in cancer, and are also implicated in genomic and heterogeneity, which is reflected in their clinical behavior, as 21 7 chromosomal instability. well as in their gene expression profiles. As 40–50% of the Our previous gene expression profiling (GEP) study in CBF AML pointed towards the deregulation of apoptotic signaling, Correspondence: Dr L Bullinger, Department of Internal Medicine III, MAPKinase signaling, as well as chemotherapy-resistance University Hospital Ulm, Albert-Einstein-Allee 23, Ulm, Baden- mechanism pathways as potentially relevant biological basis Wurttemberg 89081, Germany. 7 E-mail: [email protected] for the clinical heterogeneity observed in CBF leukemia. As our Received 16 May 2011; accepted 19 May 2011; published online 21 results were, however, based on mRNA expression differences, June 2011 the importance of the different pathways on a functional, Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1729 cellular level remained still suppositional. Therefore, in the cytometry in selected samples, which were double stained current study we investigated the role of the respective pathways using Annexin V-PE (BD Pharmingen, Franklin Lakes, NJ, USA) by selective inhibition in a cell line model (comprising six cell and 7-AAD (7-amino-actinomycin D; BD Pharmingen) accord- lines that reflect the previously identified expression patterns), as ing to the manufacturer’s protocol, and measured using a well as by targeting them in primary CBF AML samples. These FACSCalibur (BD Pharmingen). FACS and CellTiter-Glo mea- analyses demonstrated that deregulated apoptosis seems to surements correlated very well (r40.95, Supplementary Figure indeed underlie the separation into distinct outcome-related S1). All treated samples were normalized to appropriate control CBF subgroups and might therefore constitute an attractive new treated samples. Cell lines were harvested after 48 h of treatment target in the respective AML cohorts. treatment. Primary CBF leukemia samples were measured after 24 h, as after 48 h the viable cell fraction was generally too small to allow a detection of small differences in treatment sensitivity. Patients and methods

Patient samples and cell lines Quantitative reverse transcriptase PCRs Samples (n ¼ 28, of n ¼ 23 distinct cases; n ¼ 10 peripheral For quantitative reverse transcriptase PCRs (qRT-PCRs), we first blood and n ¼ 18 bone marrow specimens) from adult CBF AML isolated RNA from samples with TRIzol reagent (Invitrogen, Life patients at diagnosis were provided by the German-Austrian Technologies Corporation, Carlsbad, CA, USA) following the AML Study Group (AMLSG) with patient informed consent manufacturer’s protocol. Reverse transcription was done with obtained in accordance with the Declaration of Helsinki and SuperScript III First-Strand Synthesis System for RT-PCR institutional review board approval from all participating (Invitrogen), using the random hexamers and following the centers. Patient age at the time of diagnosis ranged from 19.5 manufacturer’s protocol. Quantitative real-time RT-PCRs were to 59.9 years (median 36.9 years). Clinical characteristics at the done with the Fast SYBR Green Master Mix (Applied Biosystems, time of diagnosis were available for almost all cases as detailed Life Technologies Corporation, Carlsbad, CA, USA) according to in Supplementary Table S1. All cell lines (HEL, Kasumi-1, ME-1, the manufacturer’s protocol using a 7900HT Fast Real-Time PCR MONO-MAC-1, MV4-11 and OCI-AML2) were obtained from System (Applied Biosystems) in the fast mode. Primer (custom the DSMZ (German Collection of Microorganisms and Cell oligonucleotides, Invitrogen) sequences were as follows (all 50 to 30): Cultures, Braunschweig, Germany). NFKB1 forward TGGAGTCTGGGAAGGATTTG, reverse CGAA GCTGGACAAACACAGA; TNF forward CCCCAGGGACCTCTC TCTAA, reverse CAGCTTGAGGGTTTGCTACA; BCL2 forward Cell culture and in vitro treatment ATGTGTGTGGAGAGCGTCAA, reverse ACAGTTCCACAAAG Cell lines were cultivated according to the guidelines of the GCATCC; XIAP forward CATTCACTTGAGGAGTGTCTGG, DSMZ and were always kept within their optimal range of reverse TGAAACTGAACCCCATTCGT; BIRC3 forward CCAAG density using RPMI 1640 (Biochrom AG, Berlin, Germany) or TGGTTTCCAAGGTGT, reverse TTTTCATCTCCTGGGCTGTC; a-MEM (GIBCO, Invitrogen Corporation, Grand Island, NY, BIRC2 forward CCAAGTGGTTTCCAAGGTGT, reverse ATTGG USA) media, supplemented with 20% fetal calf serum (Sigma- TGGGTCAGCATTTTC; ACTB forward AGAGCTACGAGCTGC Aldrich, St Louis, MO, USA), L-glutamin (Biochrom AG), CTGAC, reverse AGCACTGTGTTGGCGTACAG. penicillin/streptomycin (GIBCO), sodium pyruvate (GIBCO) and NEAA (non-essential amino acids, GIBCO) as necessary. For in vitro treatment experiments, cell line aliquots were Western blot analysis stained with trypan blue (Sigma-Aldrich) and counted to obtain Total cell extracts were fractionated on 12% or 4–12% sodium a measure of both cell viability (exclusion criterion at o95% dodecyl sulfate polyacrylamide gels (NuPAGE Bis-Tris Gels; viability) and density. Cell lines were again split to the lower end Invitrogen) and electroblotted to polyvinylidene difluoride of their optimal density range to ensure optimal growth membranes (Immobilon-P; Millipore, Billerica, MA, USA). conditions and treated with the different agents and controls, Membranes were reacted with anti-Caspase-3 (#9662; Cell according to the respective protocol. The procedure for primary Signaling Technology, Danvers, MA, USA), anti-BCL2 CBF AML samples was the same; before treatment cells were (ab37899; Abcam, Cambridge, UK), anti-cIAP2 (#1040-1; stained with trypan blue, counted, and diluted to a density of Epitomics, Burlingame, CA, USA), anti-b-actin (ab8227; Abcam) 1.0 Â 106 cells/ml. Thawing of viably frozen samples followed or anti-a-tubulin (ab7291; Abcam), followed by incubation with the DSMZ guideline. Agents used to treat cell lines and primary secondary horseradish peroxidase-linked antibodies (GE Health- CBF AML sample were dimethyl sulfoxide (Sigma-Aldrich), care, Fairfield, CT, USA). Immunoreactivity was determined ara-C (cytarabine; cell pharm, Bad Vilbel, Germany), BV6 using ECL Western Blotting detection reagents (GE Healthcare). (kindly supplied by Genentech, South San Francisco, CA, USA), ABT-737 (Selleck, Houston, TX, USA), IPA-3 (1,1-disulfanediyl- dinaphthalen-2-ol; Sigma-Aldrich), PD98059 (20-amino-30- GEP methoxyflavone; GIBCO), olomoucine (2-(hydroxyethylamino)- Previously published cDNA microarrays data of 93 CBF AML 6-benzylamino-9-methylpurine; Promega, Madison, WI, USA) cases available through Gene Expression Omnibus (accession and colchicine (Sigma-Aldrich) at indicated concentrations. number GSE8653) were included in the study and data were filtered and normalized as reported.7 Furthermore, we profiled gene expression in 12 additional CBF AML samples using Viability assays GeneChip Human Genome U133 Plus 2.0 Arrays (Affymetrix, We performed an ATP-content measurement using the Santa Clara, CA, USA; GEO accession GSE29883). Cel files CellTiter-Glo Luminescent Cell Viability Assay (Promega), were normalized and filtered using BRB Array Tools (Biometric which reflects the amount of viable cells per sample. For read- Research Branch (BRB) ArrayTools software; BRB, National out, we used the GloMax 96 luminometer (Promega GmbH, Cancer Institute, Bethesda, MD, USA) by applying the JustRMA Mannheim, Germany). Furthermore, we performed flow algorithm and previously reported filtering criteria (13355 genes

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1730 passed ﬁltering).7 Exemplary technical validation of microarray- based gene expression ﬁndings was performed with qRT-PCR using SYBR Green I in previous work.10

Data analysis Microarray data were analyzed using BRB Array Tools Version 3.7.2, developed by Dr Richard Simon and Amy Peng Lam, as previously described.10 For hierarchical clustering average-linkage hierarchical clustering was applied in Cluster and results visualized using TreeView (http://rana.lbl. gov/EisenSoftware.htm).22 Groupwise comparisons of the distributions of clinical and laboratory variables were performed using Mann–Whitney U-test, unpaired t-test with Welch correction, Fisher’s exact test, Kolmogorov–Smirnov test, Wilcoxon test and Pearson’s w2-test, as appropriate. All tests were two-sided. An effect was considered significant if the P-value was 0.05 or less. Data was visualized using either Figure 1 Cell viability, measured after 48 h of ara-C treatment. Dashed lines and open symbols indicate cell lines of group 1 (inferior SigmaPlot (Systat Software, San Jose, CA, USA) or GraphPad outcome group in clustering of CBF primary samples7), whereas solid Prism 4 (GraphPad Software, La Jolla, CA, USA). lines and full symbols indicate cell lines of group 2 (superior outcome group). All measurements were normalized to control treated cells of the same experiment. Shown here are mean values of at least three Results independent experiments. The s.d. bars are not shown for reasons of clarity (Supplementary Figure S9); mean s.d. values were between 0.04 and 0.1 for the different cell lines. Cell line model for CBF leukemia subgroups Previously, GEP revealed two distinct CBF leukemia subgroups suitable model for further characterization of the functional displaying significant outcome differences, which might in part differences underlying the respective CBF subgroups. be attributed to a deregulation of different essential pathways.7 In order to further characterize these GEP-defined CBF subgroups, we were looking for cell line models reflecting the Pathway analysis and choice of agents for specific expression differences between the two subgroups and therefore intervention compared 18 leukemia cell line expression profiles with those of First, we reassessed our previous pathway comparison analysis,7 the respective CBF cohorts. Hierarchical clustering analysis, using the most recent BioCarta pathway annotations. In general, based on 8556 differentially expressed genes, suggested several results were confirmed and included the top pathways, which leukemia cell lines as potential models for the respective groups. were already previously determined as significantly differen- Based on qRT-PCR and western blot validation (Supplementary tially regulated between the two groups. We decided to focus on Figures S2 and S3), three cell lines were identified to be apoptosis-related pathways and MAPKinase-related signaling, as representative for each CBF subgroup, with both subgroup there were several connected pathways differentially regulated models including a CBF leukemia-derived cell line. The CBF between the two groups. Furthermore, we included the pathway subgroup with inferior outcome was represented by ME-1, ‘Stathmin and breast cancer resistance to antimicrotubule MONO-MAC-1, and OCI-AML2 (group 1 cell lines), whereas agents’ (abbreviated Stathmin pathway) in our screen, as it has Kasumi-1, HEL, and MV4-11 (group 2 cell lines) were been connected to MAPKinase signaling and to treatment representative for the superior outcome CBF subgroup. resistance in breast cancer.23 For each of these three pathway There exist only three CBF leukemia-derived cell lines groups, we picked two different agents for specific intervention worldwide, Kasumi-1 (t(8;21)), SKNO-1 (t(8;21) which is in the respective signaling pathways (Table 1). For intervention growth factor (GM-CSF) dependent and not available via the in the apoptosis signaling, we treated cells with BV6ref.16 or DSMZ (German Collection of Microorganisms and Cell ABT-737ref.18 (IAP inhibitor/Smac mimetic and BCL2 inhibitor/ Cultures)) and ME-1 (inv(16)). This was, together with our BH-3 mimetic, respectively). To intervene with MAPKinase hypothesis that a secondary, cooperative class I change in CBF signaling, we chose IPA-3ref.24 (allosteric PAK1 inhibitor) and leukemias might also be found in other AML subgroups, the PD98059 (selective inhibitor of MEK1), which was shown to reason why we included non-CBF cell lines into our model. inhibit cell growth and survival of AML cell lines while Treatment of the model cell lines with cytarabine (ara-C), a sensitizing them to drug-induced apoptosis.25 Intervention standard chemotherapy treatment in AML, revealed a differential agents for the Stathmin pathway were olomoucine, a purine response to the drug (Figure 1): as anticipated, the cell lines derivative inhibiting CDK1 and other cell cycle regulating resembling the inferior outcome CBF cohort (group 1) were less CDKs, and colchicine, a microtubule depolymerizing agent sensitive to ara-C than those modeling the good prognostic isolated from nature, which also interacts with Stathmin family subgroup (group 2). Interestingly, both CBF leukemia cell lines proteins.26,27 (Kasumi-1 with a t(8;21) and ME-1 with an inv(16)) did not show a complete loss of viability even at higher doses of ara-C, unlike the other cell lines. EC50 (half maximal effective concentration) values were between 2.3 and 5.5 mM ara-C for the inferior outcome group, Inhibitor screen (single treatment) in cell line models and between 0.1 and 1.4 mM ara-C for the superior outcome group. The six model cell lines ME-1, MONO-MAC-1, OCI-AML2, Thus, our model cell lines reflect the distinct GEP-defined CBF HEL, Kasumi-1 and MV4-11 were treated with the different leukemia subgroups not only in their expression profiles, but agents using a range of different concentrations, and viability also in their response to ara-C. Therefore, we considered them a was measured after 48 h (Figure 2a).

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1731 Table 1 Pathway comparison analysis in 93 CBF AML samples (subgroup 1 vs subgroup 2), with respective inhibitors listed as chosen for subsequent analyses

BioCarta pathway description Number of Number of LS KS Speciﬁc Inhibits dereg. genes in permutation permutation intervention genes pathway P-value P-value with

MAPKinase signaling pathway 47 89 o0.001 0.002 PD98059 MEK1 IPA-3 PAK1 Stathmin and breast cancer resistance to 914o0.001 0.004 Olomoucine CDK1 antimicrotubule agents Colchicine tubulin polymerization Role of mitochondria in apoptotic signaling 10 23 o0.001 0.127 BV6 IAPs ABT-737 BCL2 Caspase cascade in apoptosis 12 26 0.001 0.003 BV6 IAPs p38 MAPK signaling pathway 17 42 0.002 0.037 PD98059 MEK1 SODD/TNFR1 signaling pathway 6 10 0.004 0.072 BV6 IAPs TNFR1 signaling pathway 15 33 0.037 0.001 IPA-3 PAK1 Abbreviations: AML, acute myeloid leukemia; CBF, core-binding factor; dereg., deregulated; LS, least squares; KS, Kolmogorov–Smirnov. Pathway comparison analysis investigating 233 BioCarta pathways with a random variance model using 8556 genes for the random variance estimation (using BRB Array Tools) in 93 CBF AML cases. Selected pathways are listed according to their LS permutation test P-values, and the inhibitors used for speciﬁc intervention in these differentially regulated pathways, as well as the respective target of inhibition/mode of action are indicated.

Specific intervention in the apoptotic signaling revealed a be targeted by an inhibitor. As we observed a distinct sensitivity differential response of the two CBF subgroup model cell lines. to inhibitor treatment only for BV6 and ABT-737, we focused For the Smac mimetic BV6, sensitivity to treatment was in our further efforts on deregulated apoptosis signaling. general higher in the group 2 cell lines corresponding to the favorable outcome CBF subgroup (Figure 2a). However, in both subgroup models one cell line was non-responsive to BV6 Combination treatment with ara-C treatment, even at higher dosage (HEL and OCI-AML2). As novel therapies usually have to be used as combination Excluding these, EC50 values were at 0.48 and 3.25 mM for the therapies with standard chemotherapy, we wanted to assess the group 2 cell lines and 8 and 410 mM for the group 1 cell lines potential of BV6 to sensitize for ara-C induced apoptosis, and (modeling the inferior outcome CBF subgroup). Treatment with thus the model cell lines were simultaneously treated with a the BCL2 inhibitor ABT-737 resulted in an inverse response, subtoxic concentration of BV6 (as determined by our screening) with the group 1 cell lines exhibiting greater sensitivity to and increasing concentrations of ara-C for 48 h. All cell lines treatment (EC50 0.12–0.17 mM) than the group 2 cell lines (EC50 showed at least a positive additive effect of the two drugs, as the 1.56–2.69 mM). Notably, one cell line of each cluster behaved measured viability after combination treatment was at or below divergently (Figure 2b): MONO-MAC-1, the only group 1 cell the viability predicted for an additive effect according to Bliss’ 28,29 line with low BCL2 expression and an EC50 of 410 mM ABT-737, independence rule (Figure 3). was in the range of the other group 2 cell lines, and MV4-11, the group 2 cell line with high BCL2 expression, had an EC50 of 0.23 mM in range with the sensitive group 1 cell lines. Treatment of primary CBF AML samples For the MAPKinase signaling, neither of the two inhibitors Based on the differential response to BV6 and ABT-737 revealed a differential response of the two cell line groups. For treatment, we next treated mononuclear cells (mostly leukemic PD98059, a MEK1 inhibitor, EC50 values of 84–116 mM and of blasts) derived from newly diagnosed CBF AML patients with 87–148 mM were observed for group 1 and group 2, respectively. either ara-C, ABT-737 or BV6 in vitro for 24 h (Figure 4). FACS Treatment with the PAK1 inhibitor IPA-3 also resulted in similar measurements confirmed this loss of viable cells as apoptosis in sensitivity of both groups, with EC50 values of 8.3–11 mM for the a dose-dependent manner (Supplementary Figure S4). Treatment group 1 cell lines and of 6.1–8.5 mM for the group 2 cell lines (for with ara-C served as a standard chemotherapy control and an overview of all EC50 values, see Figure 2c). treatment response ranged from EC50 values of 3–25 mM, with Specific intervention in the Stathmin pathway resulted in a some samples not even responding at the highest dosage differential response of the cell lines, but again did not reveal a (100 mM ara-C). differential response according to the two cell line groups. In vitro treatment of primary samples with the IAP inhibitor Inhibiting tubulin polymerization with colchicine revealed BV6 revealed two distinct groups, with Bone third of the cases EC50 values of around 0.01–0.1 mM for both groups, with one (9/23; 39%) not responding well to treatment. The majority of exception (ME-1 of group 1, 0.22 mM). Treatment with the cases (14/23; 61%) exhibited a fair (25–50% viability at 10 mM CDK1 inhibitor olomoucine required EC50 doses around BV6) to strong (o25% viability at 10 mM BV6) response to BV6 53.6–132.1 mM for group 1 cell lines and around treatment. Both response groups comprised t(8;21) and inv(16) 44.2–115.8 mM for group 2 cell lines. cases, with 63% (5/8) of the t(8;21) and 27% (4/15) of the inv(16) Thus, intervention in apoptosis signaling revealed differential samples in the non-responsive group. sensitivity of the two cell line groups, both for BV6 and ABT-737 Following BCL2 inhibitor ABT-737 treatment, there was also a treatment, whereas we did not observe a differential response for differential response observed. Here, 11 of 19 cases (58%) any of the other interventions. These findings suggested that not showed a fair to strong response (less than 50% viability at 1 mM all gene expression differences between the two CBF subgroups ABT-737), and eight cases (42%) were categorized as weakly result in functional differences on a cellular level, which might responsive (more than 50% viability at 1 mM ABT-737). Again,

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1732

Figure 2 Cell viability of model cell lines after 48 h of treatment with different agents, as indicated. (a) Inhibitor screen: dashed lines and open symbols indicate group 1 cell lines (inferior outcome subgroup); solid lines and full symbols indicate group 2 cell lines (superior outcome subgroup). Shown are mean values of two or three independent experiments, each normalized to control treatment. Median s.d. values ranged from 0.04 to 0.06 for the different cell lines; error bars are not shown due to reasons of clarity (Supplementary Figure S10). (b) Viability after 48 h of ABT-737 treatment, with outliers in each group shown in bold, as labeled in the graph. The bar graph, right panel, shows the according results of the qRT-PCR for BCL2 expression, normalized to ACTB expression, for all cell lines (untreated material). (c)EC50 values of the different treatments in the model cell line groups, as seen in a.

both response groups comprised t(8;21) and inv(16) cases, with the BV6-resistant cases. Regarding the ABT-737 response 25% (2/8) of the t(8;21) and 55% (6/11) of the inv(16) samples in groups, we did not observe statistically significant differences the non-responsive group. In contrast to the cell line model in the expression levels of the measured genes in qRT-PCR results, treatment response of primary samples to BV6 or ABT- (Supplementary Figure S5). Expression levels of additional genes 737 was not generally reversed, as 11 samples showed a ‘same implicated in response to treatment were compared based on direction’ response (sensitive or resistant to both agents) and the corresponding microarray data. Here, we observed eight samples were characterized by an ‘opposite direction’ significantly lower levels of MCL1 in ABT-737 sensitive samples response (sensitive to one, resistant to the other agent) (Table 2). (Supplementary Figure S6). Furthermore, RIPK1 expression was As far as molecular aberrations (mutations of KIT, NRAS, FLT3) significantly higher in BV6-sensitive samples (Supplementary were concerned, neither BV6 nor ABT-737 response was Figure S7). associated with any of the measured mutations (Table 2).

BV6 treatment response correlates with prognostic Correlation of gene expression levels with treatment CBF leukemia subgroups response to BV6 and ABT-737 In order to see whether the gene expression profiles underlying To investigate whether the BV6 response is correlated with either BV6 or ABT-737 response were similar to our initial expression levels of genes, which might have a role in the observation, we profiled gene expression in 12 diagnostic, sensitivity of a sample to BV6, we measured BCL2, BIRC2, untreated primary CBF AML cases (n ¼ 3 t(8;21) and n ¼ 9 BIRC3, NFKB1, TNF and XIAP levels in untreated diagnostic inv(16)), for which BV6 treatment response data had been samples by qRT-PCR, and used ACTB expression levels for collected and which represented the overall distribution of BV6 normalization (Figure 5). We observed highly significant responsiveness (5/12 cases were resistant to BV6, as defined differences (Po0.001) in the expression levels of BIRC3 and above, and 7/12 were sensitive to BV6 treatment). Regarding the NFKB1 between BV6 sensitive and resistant cases, as well as ABT-737 treatment response, seven samples were sensitive significant differences (Po0.005) in the levels of BIRC2 and (64%, see definition above) and four samples were resistant XIAP. For all these genes, mean expression levels were lower in (36%), again resembling the overall ABT-737 responsiveness as

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1733

Figure 3 Combination treatment with ara-C and BV6 in cell lines. Cell viability after 48 h of either single treatment with ara-C (solid line) or BV6 (star, arbitrarily positioned at the lowest measured ara-C concentration), or after simultaneous combination treatment of increasing doses of ara-C with a ﬁxed concentration of BV6 (dashed line). Shown are mean values of two independent experiments, each normalized to control treated cells. Mean s.d. values were 0.06 for ara-C and 0.14 for combination treatment. Gray dashed lines denote a theoretical curve for an additive effect according to Bliss’ independence rule.

Figure 4 Treatment of primary CBF AML samples with ara-C, BV6 or ABT-737. Cell viability was measured after 24 h of treatment and normalized to control treatment. Solid black lines indicate t(8;21) cases whereas inv(16) cases are denoted by solid gray lines.

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1734 Table 2 Primary diagnostic CBF AML samples (n ¼ 23) and their treatment response to BV6 or ABT-737, as well as karyotype, age and molecular markers

CBF # BV6 ABT-737 Direction of response CBF karyotype Age FLT3 ITD FLT3 TKD NRAS KIT response response for BV6 and ABT-737 group (years)

7 Sensitive Sensitive Same t(8;21) 52.1 WT WT WT 21 Sensitive Sensitive Same t(8;21) 44.7 WT WT WT WT 5 Sensitive Sensitive Same inv(16) 28.1 WT WT WT 10 Sensitive Sensitive Same inv(16) 36.3 WT MUT WT WT 13 Sensitive Sensitive Same inv(16) 55.7 WT WT MUT WT 14 Sensitive Sensitive Same inv(16) 59.9 WT WT MUT WT 19 Sensitive Sensitive Same inv(16) 54.0 WT WT WT MUT 20 Sensitive Resistant Opposite t(8;21) 57.0 WT WT WT MUT 6 Sensitive Resistant Opposite inv(16) 46.6 WT WT WT 18 Sensitive Resistant Opposite inv(16) 50.3 WT WT WT MUT 1 Sensitive inv(16) 32.8 WT WT WT WT 2 Sensitive inv(16) 21.0 WT WT WT WT 3 Sensitive inv(16) 33.8 WT WT MUT WT 12 Sensitive inv(16) 48.0 WT WT MUT WT 11 Resistant Sensitive Opposite t(8;21) 36.9 WT WT WT MUT 15 Resistant Sensitive Opposite t(8;21) 36.4 WT WT WT MUT 22 Resistant Sensitive Opposite t(8;21) 19.5 WT WT WT WT 23 Resistant Sensitive Opposite t(8;21) 31.4 WT WT 17 Resistant Sensitive Opposite inv(16) 31.2 WT MUT MUT WT 16 Resistant Resistant Same t(8;21) 35.2 WT WT MUT WT 4 Resistant Resistant Same inv(16) 57.0 WT WT 8 Resistant Resistant Same inv(16) 41.2 WT WT 9 Resistant Resistant Same inv(16) 33.0 WT WT WT Abbreviations: AML, acute myeloid leukemia; CBF, core-binding factor. Listed are BV6 and ABT-737 treatment response in 23 primary CBF AML samples, as well as karyotypes and molecular markers, as available. Fisher’s exact test revealed no signiﬁcant associations of any response group with the listed molecular aberrations. To improve visualization mutations are highlighted in bold script, as are t(8;21) karyotypes.

differentially regulated between the different BV6 response groups. Here, we observed some overlap with our previous, outcome-related CBF subgroup signature including differential expression of FOXO1, AKAP13, PICALM and MAP4K4, among others (Supplementary Table S2). A pathway comparison analysis of the two BV6 response groups (sensitive vs resistant) found 43 of 267 investigated BioCarta pathways to be differentially regulated, with more genes differentially expressed among the two groups than expected by chance (Supplementary Table S3). These included those pathways on which we had initially focused our analysis, the ‘MAPKinase signaling pathway’, ‘p38 MAPK signaling pathway’ (significant in Goe- man’s global test), ‘Role of mitochondria in apoptotic signaling’ and ‘Caspase cascade in apoptosis’. The main component deregulated in the originally listed ‘Stathmin and breast cancer resistance to antimicrotubule agents’, CDK1, was again found to be differentially expressed between the BV6 response groups, as Figure 5 Gene expression levels, measured by qRT-PCR, grouped found in two cell cycle pathways (‘Cyclins and cell cycle by BV6 response of measured primary CBF AML samples (n ¼ 15; regulation’, ‘Cell cycle:G1/S check point’). qRT-PCR was done in untreated diagnostic samples of same primary To assess whether the BV6 treatment sensitivity in vitro did samples as were treated with BV6). ***indicates a statistically highly indeed separate the CBF cases into groups corresponding to the significant difference in the expression levels (Po0.001), **stands for a CBF subgroups as described previously, we clustered the newly statistically significant difference (P 0.005). Sensitive ¼ BV6 o profiled CBF GEPs with our previous AML cohort consisting of response with o50% viability at 10 mM BV6 treatment (labeled BV6 7 sens.); resistant X50% viability at 10 mM BV6 treatment (labeled 93 cDNA microarrays (Figure 6). Unsupervised average linkage BV6 res.). clustering based on the CBF subgroup defining signature revealed that all BV6-resistant cases, together with one sensitive case, clustered with the inferior outcome CBF cases of subgroup 1. The remaining six BV6 sensitive cases clustered with the measured in our CBF cohort (for one sample, ABT-737 treatment superior outcome subgroup 2 CBF AML samples. Regarding the was not possible due to limited material). ABT-737 sensitivity of the clustered cases, no differences were Comparison of BV6 sensitive versus resistant cases resulted in observed: two of four ABT-737-resistant and four of seven ABT- a distinct gene expression signature. Using the SAM (Signi- 737-sensitive cases clustered with subgroup 1. In general, the ficance Analysis of Microarrays) algorithm, which accounts for two clusters were well separated, and, as expected, all of the 35 multiple testing, 855 genes were found to be significantly original CBF subgroup 1 cases were found to cluster together.

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1735

Figure 6 Heatmap of 93 previously published7 and 12 newly profiled CBF samples. Red denotes overexpression of a gene, green lower expression, both compared to mean expression (black). Gray fields denote missing values. Average linkage clustering was based on the previously identified CBF subgroup 1 and 2 distinguishing top 230 genes. All 35 original CBF subgroup 1 cases (orange ‘1’) clustered into one group, together with the five newly profiled BV6-resistant cases (orange ‘resistant’). Most of the original subgroup 2 cases (blue ‘2’, 90% or 52/58) clustered together with six of seven BV6-sensitive cases (blue ‘sensitive’, one exception clustered with the other subgroup).

Discussion expression levels of the target. As the Smac mimetic BV6 inhibits cIAP1 (cellular inhibitor of apoptosis protein 1), cIAP2 Based on GEP, we have previously been able to define two and XIAP (X-linked inhibitor of apoptosis),16 overexpression of distinct CBF leukemia subgroups associated with outcome.7 In BIRC2 and BIRC3 (encoding cIAP1 and cIAP2, respectively) in order to investigate the functional relevance of the deregulated CBF subgroup 2 cases matched the greater BV6 sensitivity of pathways found to differ between CBF subgroups, we defined group 2 cell lines (Supplementary Figure S8). In contrast, the two here a cell line model resembling these two subgroups. As is the BV6 resistant cell lines, HEL (group 2) and OCI-AML2 (group 1), case for all model systems, this simplified model has limitations, both expressed BIRC2 and BIRC3 at levels comparable to the but generally well represented the gene expression differences other cell lines of their respective group. Although there were of the two CBF subgroups with regard to the pathways of also no differences with regard to the expression levels of other interest, and importantly, results could be translated to primary qRT-PCR-measured genes involved in apoptosis, it has been CBF AML samples. reported that some cell lines might show no BV6 sensitivity due In this cell line model, our screen of different inhibitors to a lack in TNF.14,30 However, both resistant cell lines in our showed that neither MAPKinase signaling intervention (IPA-3, study expressed TNF at a level comparable to the sensitive cell PD98059) nor Stathmin pathway intervention (olomoucine, lines (data not shown). Thus, the definite cause remains open, colchicine) revealed subgroup specific sensitivities. Thus, but possibly these two cell lines harbor additional, yet unknown although members of these pathways were found to be aberrations preventing their response to IAP inhibition. For differentially expressed among the CBF leukemia subgroups, treatment with the BCL2 inhibitor ABT-737, we observed a this did not result in a corresponding deregulation on cellular/ higher sensitivity in group 1 cell lines, in line with the fact that functional level, which could be targeted by any of the used BCL2 itself was overexpressed in the CBF subgroup 1 cases inhibitors. A possible explanation could be a complementary compared with subgroup 2 cases. Two cell lines, MONO-MAC-1 deregulation of the respective pathway in the other subgroup on (group 1) and MV4-11 (group 2), showed an ABT-737 sensitivity a posttranscriptional level, possibly by microRNAs or other characteristic for the converse group, which could be explained factors, thereby resulting in similar sensitivities to treatment. by their respective BCL2 expression levels. In contrast, intervention in apoptotic signaling revealed Given these findings, we assume that of the pathways subgroup-specific sensitivity of cell lines to treatment, demon- differentially expressed among the CBF subgroups, deregulated strating the relevance of the deregulation at mRNA level for the apoptotic signaling accounts for a substantial part of the separation of the two CBF subgroups. For both BV6 and functional differences between the two subgroups. As simulta- ABT-737, differences in treatment sensitivity correlated to neous treatment with the IAP inhibitor BV6 and the AML

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1736 standard treatment ara-C suggested additive to synergistic effects significantly higher levels in BV6-sensitive cases compared with in cell lines, this combination treatment represents a promising resistant cases. Due to the fact that IAP expression or function is approach to be further investigated for a potential clinical use. deregulated in many cancers, including hematological malig- Although cell lines are very useful model systems, they also nancies, several IAP-targeting agents are currently being tested entail drawbacks, such as their limited reflection of diverse in (early) clinical trials.14,15,48 Identification of molecular patients with the same diagnosis and their acquired aberrations markers to select patients that should be sensitive to treatment due to extensive culturing. Therefore, we extended our study to will be crucial, and GEP represents a valuable approach in this evaluate treatment effects also in primary CBF leukemia respect. samples. Consistent with our findings in cell lines, diagnostic We found that sensitivity to BV6 treatment correlated with CBF AML samples exhibited differential sensitivity to treatment gene expression profiles characteristic for the previously defined with ABT-737 and BV6. This raised the question whether target superior outcome CBF AML subgroup.7 As classification gene expression levels were correlated with sensitivity to according to the two CBF subgroups was not available a priori inhibitor treatment. for the primary CBF AML samples treated with BV6 in this study, In contrast to the cell line results, ABT-737 response was not we profiled their gene expression and clustered them with the associated with expression levels of BCL2 in the primary CBF 93 previously described GEPs. This approach allowed us to AML samples. This was also reported in studies of acute assign a CBF subgroup label to these newly profiled primary lymphoblastic leukemia31 and AML (cell lines and primary samples based on their association with either the inferior or samples of relapsed/refractory patients),32 although others did superior outcome group, consequently showing the association see a correlation of BCL2 expression levels to ABT-737 of BV6-resistant cases with the inferior CBF AML outcome treatment sensitivity in small cell lung cancer xenografts.33,34 group. In summary, all our data point towards apoptosis As reported by others for both de novo and acquired resistance signaling, and more specifically IAPs, as the underlying to ABT-737, we could correlate sensitivity to ABT-737 treatment deregulation to differentiate CBF leukemia subgroups indepen- to MCL1 expression levels.32,34–36 Using GEP to find predictive dently of their respective karyotypes and secondary mutations signatures for treatment sensitivity might be a beneficial strategy, (for example FLT3, KIT, NRAS). Therefore, therapies targeting and first results have been reported for leukemia/lymphoma cell IAPs, and also BCL2, should yield improved results for both CBF lines recently,37 which need to be validated in primary samples AML subgroups, and further studies to validate and more and in the clinical context. Taking all findings together, there is extensively test the preliminary effects observed in combination still a need to decide on an adequate biomarker for a clinical of ara-C and BV6 are indicated. treatment with ABT-737 or a related BCL2 inhibitor, and expanded studies like ours employing primary samples com- bined with GEP should help in this regard. In the past few years, Conflict of interest several BCL2 inhibitors have entered clinical trials, showing so far only limited activity.38,39 Only very recently, the first results The authors declare no conflict of interest. of a phase 1 study with navitoclax (ABT-263), the orally bioavailable analog of ABT-737, have shown promising clinical Acknowledgements responses in lymphoid malignancies, usually expressing in- 40 creased BCL2. As it has also been reported that BCL2 protein We thank Sabine Renz and Anna Siegmund for excellent technical levels in blasts are an important independent predictive factor assistance with Ficoll gradient separation of primary CBF AML for survival in newly diagnosed, intensively treated AML 41,42 samples. We are grateful for Juan Du’s help in establishing patients, this gives further incentive to use BCL2 inhibitors mutational screening assays. Furthermore, BV6 was kindly in the treatment of AML, in combination with standard cytotoxic supplied by Genentech. This study was supported in part by the treatment. Deutsche Jose´ Carreras Stiftung e.V. (DJCLS R 08/32f) and the For BV6 treatment, in accordance with our cell line results, German Research Foundation (DFG DO 704/3-1); LB sensitivity in CBF AML samples correlated with significantly was supported in part by the German Research Foundation elevated BIRC2 and BIRC3 expression levels, as well as with (Heisenberg-Stipendium BU 1339/3-1) and SF was supported in significantly elevated XIAP and NFKB1 expression levels. As part by the German Research Foundation, Deutsche Jose´ Carreras BV6 is an IAP inhibitor, this association was not surprising, and Stiftung e.V., the EU (ApopTrain, APO-SYS) and IAP16/8. fit well into the described mode of action.16 IAPs were originally described as caspase inhibitors, but meanwhile were found to also modulate nuclear factor (NF)-kB signaling and inflamma- Author contributions tion.16 Interestingly, NF-kB is reported to induce BIRC2 (cIAP1), BIRC3 (cIAP2) and XIAP expression in a positive feedback SCL designed the research, performed research, analyzed and 43 loop, while cIAP1 and cIAP2 are positive regulators of interpreted data, and wrote the manuscript. ACR designed the canonical NF-kB signaling and suppress the non-canonical research, collected, analyzed and interpreted data, and pathway. Thus, cIAPs have both pro- and anti-apoptotic reviewed the manuscript. UB, PP and RFS collected data properties, as not only overexpression, but also loss of cIAPs and reviewed the manuscript. HD and KD designed research and can lead to NF-kB activation, survival of tumor cells and reviewed the manuscript. SF designed the research, contributed 16,44 chemoresistance. Furthermore, Smac mimetic- and TNF- vital reagents, analyzed data and reviewed the manuscript. LB 14,30,45–47 induced cell death depends on RIPK1, a multifunc- designed the research, interpreted data and wrote the manuscript. tional protein that can promote either survival or cell death depending on its posttranslational modifications. For example, once ubiquitinated, RIPK1 can serve as an essential adapter References molecule for the activation of NF-kB, while it promotes the formation of a RIPK1/FADD/caspase-8 proapoptotic complex 1 Dohner K, Dohner H. Molecular characterization of acute myeloid when it is deubiquitinated. Interestingly, RIPK1 was expressed at leukemia. Haematologica 2008; 93: 976–982.

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1737 2 Paschka P. Core binding factor acute myeloid leukemia. Semin 26 Ravelli RB, Gigant B, Curmi PA, Jourdain I, Lachkar S, Sobel A Oncol 2008; 35: 410–417. et al. Insight into tubulin regulation from a complex 3 Hart SM, Foroni L. Core binding factor genes and human leukemia. with colchicine and a stathmin-like domain. Nature 2004; 428: Haematologica 2002; 87: 1307–1323. 198–202. 4 Speck NA. Core binding factor and its role in normal hemato- 27 Risinger AL, Giles FJ, Mooberry SL. Microtubule dynamics as a poietic development. Curr Opin Hematol 2001; 8: 192–196. target in oncology. Cancer Treat Rev 2009; 35: 255–261. 5 Castilla LH, Perrat P, Martinez NJ, Landrette SF, Keys R, Oikemus S 28 Bliss C. The toxicity of poisons applied jointly. Ann Appl Biol et al. Identification of genes that synergize with Cbfb-MYH11 in 1939; 26: 585–615. the pathogenesis of acute myeloid leukemia. Proc Natl Acad Sci 29 Greco WR, Bravo G, Parsons JC. The search for synergy: a critical USA 2004; 101: 4924–4929. review from a response surface perspective. Pharmacol Rev 1995; 6 Appelbaum FR, Kopecky KJ, Tallman MS, Slovak ML, Gundacker HM, 47: 331–385. Kim HT et al. The clinical spectrum of adult acute myeloid 30 Petersen SL, Wang L, Yalcin-Chin A, Li L, Peyton M, Minna J et al. leukaemia associated with core binding factor translocations. Br J Autocrine TNFalpha signaling renders human cancer cells Haematol 2006; 135: 165–173. susceptible to Smac-mimetic-induced apoptosis. Cancer Cell 7 Bullinger L, Rucker FG, Kurz S, Du J, Scholl C, Sander S et al. 2007; 12: 445–456. Gene-expression profiling identifies distinct subclasses of 31 High LM, Szymanska B, Wilczynska-Kalak U, Barber N, core binding factor acute myeloid leukemia. Blood 2007; 110: O’Brien R, Khaw SL et al. The Bcl-2 homology domain 3 mimetic 1291–1300. ABT-737 targets the apoptotic machinery in acute lymphoblastic 8 Marcucci G, Mrozek K, Ruppert AS, Maharry K, Kolitz JE, leukemia resulting in synergistic in vitro and in vivo interactions Moore JO et al. Prognostic factors and outcome of core binding with established drugs. Mol Pharmacol 2010; 77: 483–494. factor acute myeloid leukemia patients with t(8;21) differ from 32 Konopleva M, Contractor R, Tsao T, Samudio I, Ruvolo PP, those of patients with inv(16): a Cancer and Leukemia Group B Kitada S et al. Mechanisms of apoptosis sensitivity and resistance study. J Clin Oncol 2005; 23: 5705–5717. to the BH3 mimetic ABT-737 in acute myeloid leukemia. Cancer 9 Schlenk RF, Benner A, Krauter J, Buchner T, Sauerland C, Ehninger G Cell 2006; 10: 375–388. et al. Individual patient data-based meta-analysis of patients aged 33 Hann CL, Daniel VC, Sugar EA, Dobromilskaya I, Murphy SC, 16 to 60 years with core binding factor acute myeloid leukemia: a Cope L et al. Therapeutic efficacy of ABT-737, a selective survey of the German Acute Myeloid Leukemia Intergroup. J Clin inhibitor of BCL-2, in small cell lung cancer. Cancer Res 2008; Oncol 2004; 22: 3741–3750. 68: 2321–2328. 10 Luck SC, Russ AC, Du J, Gaidzik V, Schlenk RF, Pollack JR et al. 34 Tahir SK, Yang X, Anderson MG, Morgan-Lappe SE, Sarthy AV, KIT mutations confer a distinct gene expression signature in core Chen J et al. Influence of Bcl-2 family members on the cellular binding factor leukaemia. Br J Haematol 2010; 148: 925–937. response of small-cell lung cancer cell lines to ABT-737. Cancer 11 Paschka P, Marcucci G, Ruppert AS, Mrozek K, Chen H, Kittles RA Res 2007; 67: 1176–1183. et al. Adverse prognostic significance of KIT mutations in adult 35 Lin X, Morgan-Lappe S, Huang X, Li L, Zakula DM, Vernetti LA acute myeloid leukemia with inv(16) and t(8;21): a Cancer and et al. Seed analysis of off-target siRNAs reveals an essential role of Leukemia Group B Study. J Clin Oncol 2006; 24: 3904–3911. Mcl-1in resistance to the small-molecule Bcl-2/Bcl-XL inhibitor 12 Danial NN, Korsmeyer SJ. Cell death: critical control points. ABT-737. Oncogene 2007; 26: 3972–3979. Cell 2004; 116: 205–219. 36 Yecies D, Carlson NE, Deng J, Letai A. Acquired resistance to 13 Hanahan D, Weinberg RA. The hallmarks of cancer. Cell 2000; ABT-737 in lymphoma cells that up-regulate MCL-1 and BFL-1. 100: 57–70. Blood 2010; 115: 3304–3313. 14 Gyrd-Hansen M, Meier P. IAPs: from caspase inhibitors to 37 Tahir SK, Wass J, Joseph MK, Devanarayan V, Hessler P, Zhang H modulators of NF-kappaB, inflammation and cancer. Nat Rev et al. Identification of expression signatures predictive of sensitivity Cancer 2010; 10: 561–574. to the Bcl-2 family member inhibitor ABT-263 in small cell lung 15 Fulda S. Inhibitor of apoptosis proteins in hematological malig- carcinoma and leukemia/lymphoma cell lines. Mol Cancer nancies. Leukemia 2009; 23: 467–476. 16 Varfolomeev E, Blankenship JW, Wayson SM, Fedorova AV, Ther 2010; 9: 545–557. 38 Liu G, Kelly WK, Wilding G, Leopold L, Brill K, Somer B. An Kayagaki N, Garg P et al. IAP antagonists induce autoubiquitina- tion of c-IAPs, NF-kappaB activation, and TNFalpha-dependent open-label, multicenter, phase I/II study of single-agent AT-101 in apoptosis. Cell 2007; 131: 669–681. men with castrate-resistant prostate cancer. Clin Cancer Res 2009; 17 Vogler M, Dinsdale D, Dyer MJ, Cohen GM. Bcl-2 inhibitors: small 15: 3172–3176. molecules with a big impact on cancer therapy. Cell Death Differ 39 O’Brien SM, Claxton DF, Crump M, Faderl S, Kipps T, Keating MJ 2009; 16: 360–367. et al. Phase I study of obatoclax mesylate (GX15-070), a small 18 Oltersdorf T, Elmore SW, Shoemaker AR, Armstrong RC, molecule pan-Bcl-2 family antagonist, in patients with advanced Augeri DJ, Belli BA et al. An inhibitor of Bcl-2 family proteins chronic lymphocytic leukemia. Blood 2009; 113: 299–305. induces regression of solid tumours. Nature 2005; 435: 677–681. 40 Wilson WH, O’Connor OA, Czuczman MS, Lacasce AS, 19 Pratilas CA, Solit DB. Targeting the mitogen-activated protein Gerecitano JF, Leonard JP et al. Navitoclax, a targeted high-affinity kinase pathway: physiological feedback and drug response. Clin inhibitor of BCL-2, in lymphoid malignancies: a phase 1 dose- Cancer Res 2010; 16: 3329–3334. escalation study of safety, pharmacokinetics, pharmacodynamics, 20 Molli PR, Li DQ, Murray BW, Rayala SK, Kumar R. PAK signaling and antitumour activity. Lancet Oncol 2010; 11: 1149–1159. in oncogenesis. Oncogene 2009; 28: 2545–2555. 41 Campos L, Rouault JP, Sabido O, Oriol P, Roubi N, Vasselon C 21 Malumbres M, Barbacid M. Cell cycle, CDKs and cancer: a et al. High expression of bcl-2 protein in acute myeloid leukemia changing paradigm. Nat Rev Cancer 2009; 9: 153–166. cells is associated with poor response to chemotherapy. Blood 22 Eisen MB, Spellman PT, Brown PO, Botstein D. Cluster analysis 1993; 81: 3091–3096. and display of genome-wide expression patterns. Proc Natl Acad 42 Tothova E, Fricova M, Stecova N, Kafkova A, Elbertova A. High Sci USA 1998; 95: 14863–14868. expression of Bcl-2 protein in acute myeloid leukemia cells is 23 Alli E, Yang JM, Ford JM, Hait WN. Reversal of stathmin-mediated associated with poor response to chemotherapy. Neoplasma 2002; resistance to paclitaxel and vinblastine in human breast carcinoma 49: 141–144. cells. Mol Pharmacol 2007; 71: 1233–1240. 43 Jin HS, Lee DH, Kim DH, Chung JH, Lee SJ, Lee TH. cIAP1, cIAP2, 24 Viaud J, Peterson JR. An allosteric kinase inhibitor binds the and XIAP act cooperatively via nonredundant pathways to regulate p21-activated kinase autoregulatory domain covalently. Mol genotoxic stress-induced nuclear factor-kappaB activation. Cancer Cancer Ther 2009; 8: 2559–2565. Res 2009; 69: 1782–1791. 25 Milella M, Kornblau SM, Estrov Z, Carter BZ, Lapillonne H, 44 Gyrd-Hansen M, Darding M, Miasari M, Santoro MM, Zender L, Harris D et al. Therapeutic targeting of the MEK/MAPK signal Xue W et al. IAPs contain an evolutionarily conserved ubiquitin- transduction module in acute myeloid leukemia. J Clin Invest binding domain that regulates NF-kappaB as well as cell survival 2001; 108: 851–859. and oncogenesis. Nat Cell Biol 2008; 10: 1309–1317.

Leukemia Apoptosis deregulation in CBF AML subgroups SC Lu¨ck et al 1738 45 Wong WW, Gentle IE, Nachbur U, Anderton H, Vaux DL, Silke J. 47 Vanlangenakker N, Vanden Berghe T, Bogaert P, Laukens B, Zobel K, RIPK1 is not essential for TNFR1-induced activation of NF-kappaB. Deshayes K et al. cIAP1 and TAK1 protect cells from TNF-induced Cell Death Differ 2010; 17: 482–487. necrosis by preventing RIP1/RIP3-dependent reactive oxygen 46 Bertrand MJ, Milutinovic S, Dickson KM, Ho WC, Boudreault A, species production. Cell Death Differ 2010; 18: 656–665. Durkin J et al. cIAP1 and cIAP2 facilitate cancer cell survival by 48 LaCasse EC, Mahoney DJ, Cheung HH, Plenchette S, Baird S, functioning as E3 ligases that promote RIP1 ubiquitination. Mol Korneluk RG. IAP-targeted therapies for cancer. Oncogene 2008; Cell 2008; 30: 689–700. 27: 6252–6275.

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

Leukemia