Smac mimetic induces cell death in a large proportion of primary acute myeloid leukemia samples, which correlates with defined molecular markers

Apoptosis is deregulated in most, if not all, cancers, including hematological malignancies. Smac mimetics that antagonize Inhibitor of Apoptosis (IAP) proteins have so far largely been investigated in acute myeloid leukemia (AML) cell lines; however, little is yet known on the therapeutic potential of Smac mimetics in primary AML samples. In this study, we therefore investigated the antileukemic activity of the Smac mimetic BV6 in diagnostic samples of 67 adult AML patients and correlated the response to clinical, cytogenetic and molecular markers and gene expression profiles. Treatment with cytarabine (ara-C) was used as a standard chemotherapeutic agent. Interestingly, about half (51%) of primary AML samples are sensitive to BV6 and 21% intermediate responsive, while 28% are resistant. Notably, 69% of ara-C-resistant samples show a good to fair response to BV6. Furthermore, combination treatment with ara-C and BV6 exerts additive effects in most samples. Whole-genome gene expression profiling identifies cell death, TNFR1 and NF-κB signaling among the top pathways that are activated by BV6 in BV6-sensitive, but not in BV6-resistant cases. Furthermore, sensitivity of primary AML blasts to BV6 correlates with significantly elevated expression levels of TNF and lower levels of XIAP in diagnostic samples, as well as with NPM1 mutation. In a large set of primary AML samples, these data provide novel insights into factors regulating Smac mimetic response in AML and have important implications for the development of Smac mimetic-based therapies and related diagnostics in AML.


IntroductIon
Acute myeloid leukemia (AML) is a genetically heterogeneous disease with a multi-step pathogenesis [1]. Leukemogenesis is considered to require deregulation of at least two different cellular processes that lead to (i) enhancement of proliferation and survival and (ii) impairment of differentiation [2]. In recent years, many chromosomal aberrations have been identified, which alter normal gene function or expression, thereby contributing to leukemic transformation. Furthermore, many of these cytogenetic aberrations provide important diagnostic and prognostic information [3]. In the large group of cytogenetically normal AML (CN-AML, 40-50% of all AML cases), which show no chromosomal aberrations in conventional banding analysis, currently Research Paper www.impactjournals.com/oncotarget the identification of novel gene mutations allows dissection of CN-AML into prognostic subgroups [4], and mutations affecting CEBPA and NPM1 are considered as provisional AML entities in the WHO classification [5].
As the hematological compartment is characterized by a fast turnover of cells, a tight regulation of cell survival and cell death is of special importance [6]. Therefore, too little cell death can contribute to a proliferative advantage of transformed cells. Apoptosis is one of the best characterized forms of programmed cell death, which is typically deregulated in most, if not all, cancers [7]. Apoptosis is engaged via ligation of death receptors at the cell surface (extrinsic pathway) or via mitochondria (intrinsic pathway) [8]. Since most current chemotherapeutic strategies depend on intact cell death signaling within cancer cells for their cytotoxic effects, deregulation of cell death programs can lead to treatment resistance [9].
Inhibitors of Apoptosis (IAP) proteins, a family of antiapoptotic proteins comprising e.g. x-linked IAP (XIAP), cellular IAP (cIAP)1 and cIAP2, are known to play a crucial role in many types of human cancer [10]. Also in leukemia, IAP proteins have been associated with chemoresistance, disease progression and poor prognosis [11]. Therefore, IAP proteins are considered as relevant targets for therapeutic intervention and several smallmolecule inhibitors have been designed to neutralize IAP proteins [10]. For example, second mitochondriaderived activator of caspases (Smac) mimetics mimick the mitochondrial intermembrane space protein Smac, an endogenous antagonist of IAP proteins that is released into the cytosol during apoptosis [10]. Currently, several Smac mimetics are being tested in clinical trials [12]. We previously demonstrated in acute lymphoblastic leukemia (ALL) and chronic lymphocytic leukemia (CLL) that small-molecule antagonists of IAP proteins can sensitize cells for Tumor-Necrosis-Factor-related apoptosis-inducing ligand (TRAIL)-, CD95-or chemotherapy-induced apoptosis [13][14][15][16]. In AML, we recently reported that Smac mimetics can prime cells for several cytotoxic agents that are being used in current treatment protocols, i.e. ara-C and epigenetic drugs such as demethylating agents and histone deacetylase inhibitors (HDACIs) [17][18][19].
However, these previous studies on Smac mimetics in AML largely embark on AML cell lines and little is yet known about the response of primary AML samples towards treatment with Smac mimetics. In this study, we therefore investigated whether or not primary AML samples are sensitive to the Smac mimetic BV6 that antagonizes XIAP, cIAP1 and cIAP2 [20], and if so, which molecular, cytogenetic or clinical markers correlate with treatment response.

Primary AMl samples show a differential response to the standard chemotherapeutic drug ara-c and to the smac mimetic bV6
To explore the therapeutic potential of Smac mimetics in primary AML samples, we investigated the in vitro sensitivity to the preclinical Smac mimetic BV6 that antagonize XIAP, cIAP1 and cIAP2 [20] in a large set of 67 newly diagnosed AML patients. To this end, we treated mononuclear cells (mostly leukemic blasts) derived from AML patients at diagnosis for 24 hours with BV6 or with ara-C, which served as a standard chemotherapy control, and determined cell viability ( Figure 1). Treatment response to ara-C ranged from mean EC 50 values of 3.9 µM (ara-C sensitive group) to 50 µM (ara-C intermediate, i.e. moderate response group) and > 100 µM (ara-C resistant group), with a total range of 1.2 µM to > 100 µM ( Figure 1A). By comparison, mononuclear cells of healthy donors remained largely resistant towards ara-C ( Figure 1A).
Treatment with BV6 revealed distinct response groups, which were defined by their respective cell viability at 1 µM BV6 and 10 µM BV6 ( Figure 1B, arrows). Samples with less than 75% viability at 1 µM BV6 and less than 25% viability at 10 µM BV6 were defined as BV6-"sensitive" (n = 26). Those with less than 75% viability at 1 µM BV6 and 25-50% viability at 10 µM BV6 were defined as BV6-"sensitive to low intermediate" (n = 8). All other response groups included samples with more than 75% viability at 1 µM BV6, which were further subdivided by their viability at 10 µM BV6 into BV6 "low intermediate" (n = 14) with 25-50% viability, BV6 "high intermediate" (n = 10) with 50-75% viability and BV6 "resistant" (n = 9) with more than 75% viability at 10 µM BV6. These five BV6 response subgroups were then compiled into three BV6 response groups, i.e. BV6-sensitive (n = 34/67; 51%) comprising "sensitive" and "sensitive to low intermediate" samples, BV6 "low intermediate" (n = 14/67; 21%) and BV6resistant (n = 19/67; 28%) containing both "high intermediate" and "resistant" samples ( Figure 1C). Thus, about half of the primary AML samples proved to be sensitive to treatment with BV6. Of note, about 42% of ara-C-resistant samples were BV6-sensitive and another 27% fell into the BV6 low intermediate response group, adding up to 69% of ara-C-resistant samples with a good to fair response to BV6. The group "low intermediate" was excluded from further analyses to obtain a clear differentiation of BV6-sensitive and -resistant characteristics. www.impactjournals.com/oncotarget As experimental therapies might be more effective in combination with standard chemotherapy, we next assessed the potential of BV6 to sensitize AML cells for ara-C-induced apoptosis. To this end, we treated primary AML samples simultaneously with a low concentration of BV6 (2 µM) and increasing concentrations of ara-C for 24 hours. In general, we observed an additive or more than an additive effect of both drugs, as the measured cell viability after combination treatment was at or below the cell viability predicted for an additive effect according to Bliss' independence rule in most conditions [21,22] ( Figure 2).

sensitivity of primary AMl samples to bV6 correlates with molecular markers
To find out which AML subgroup might benefit from treatment with Smac mimetics such as BV6, we correlated clinical, cytogenetic and molecular genetic markers with BV6 sensitivity (Table 1). While there was no correlation with age, sex and cytogenetics, there was a trend detectable for internal tandem duplication of the FLT3 gene (FLT3-ITDs) being more prevalent in the BV6sensitive group (Table 1, 45% versus 27%; P =.1141). Of note, NPM1 mutations were significantly more prevalent in BV6-sensitive compared to BV6-resistant cases ( Table 1, 58% versus 20%; P = .0076). There was no significant difference in the proportion of NPM1-mutated cases without FLT3-ITD to NPM1-mutated cases with FLT3-ITD between BV6-sensitive and -resistant cases (data not shown). Additionally, we found a significant association between CEBPA mutations and BV6 resistance (Table 1, 22% versus 0%; P = .0144).

GeP in diagnostic samples of bV6-sensitive versus -resistant cases suggests marker genes to predict response
The identification of molecular markers to select patients, which likely respond to a given treatment, is crucial for any novel therapeutic strategy. In this respect, GEP represents a valuable approach, as it allows an unbiased screen of a large number of candidate genes. Therefore, we profiled gene expression of six BV6sensitive AML samples, and compared the expression pattern to the profiles of six BV6-resistant AML samples. GEP was done in diagnostic (untreated) material, and samples of each group were matched with regard to age, sex, and karyotypes, as far as possible (Supplementary  Table 2). Class comparison analysis revealed 25 genes to be differentially expressed between BV6-sensitive and -resistant samples ( Table 2, P < .01). These included several candidate tumor suppressor genes, such as ABLIM1, CYFIP2, VPS13A, SERPINI1, TET2, HTRA4, UBE2L6, as well as oncogenes/genes implicated as biomarkers in other cancers, such as FAM65B and TCF7L2 (Table 2).

bV6 activates several cell death-related pathways in bV6-sensitive, but not in -resistant cases
To gain unbiased insights into BV6-stimulated signaling pathways in primary AML patient samples we performed whole-genome GEP of BV6-sensitive and -resistant samples after 24-hour-treatment with either BV6 or DMSO (Supplementary Table 3 Table 4). In total, 12 Biocarta pathways were significantly differentially regulated between BV6and DMSO-treated samples in BV6-sensitive cases ( Table 3, P < .05, Least Squares (LS)/ Kolmogorov-Smirnov (KS) test). In contrast, GEP of BV6-resistant cases revealed no differential regulation of these cell death-related pathways in our cohort. Here, CTL-mediated immune response against target cells was found to be most significantly enriched within the differentially regulated genes, notably with a lower expression of the key player FAS in BV6treated cells (Supplementary Table 4). Furthermore, PTEN-dependent cell cycle arrest and apoptosis was also listed to differ between BV6-and DMSO-treated BV6-resistant samples, with a lower expression of PTEN itself in BV6-treated samples (Supplementary Table 4). Overall, 26 Biocarta pathways were found to have more members significantly differentially expressed among BV6-and DMSO-treated samples in BV6-resistant cases (Supplementary Table 4, P < .05, LS/KS test).

Gene expression levels of selected cell deathrelated genes correlate with bV6 sensitivity
Based on these GEP studies, we further investigated whether the in vitro response of primary AML samples to BV6 correlates with expression levels of a defined set of cell death-related genes. To this end, we measured expression levels of BIRC2 (encoding cIAP1), BIRC3 (encoding cIAP2), XIAP, TNF, NF-κB1 and BCL2 in selected representative untreated diagnostic samples by qRT-PCR ( Figure 3). Interestingly, expression levels of XIAP were significantly lower in BV6-sensitive cases compared to BV6-resistant cases (Figure 3, P = .0241). In addition, BV6-sensitive samples exhibited significantly higher expression levels of TNF than BV6resistant samples (Figure 3, P = .0493). No significant difference was found for BCL2, NF-κB1 and BIRC2 www.impactjournals.com/oncotarget Combination of ara-C (increasing doses as indicated) with BV6 (single dose: 2 µM), according to the respective ara-C response group (sensitive (A, b), intermediate (c, d), resistant (e, F), see also Figure 1) and the achieved effect of the combination (more than additive (A, C, E) or additive and less than additive (B, D, F)). Mean and SD of all samples in the respective group (sample size as indicated in each graph) are shown, with the three curves representing ara-C single treatment, combination of ara-C with a constant dose of 2 µM BV6 (measured ara-C+BV6), and the theoretical viability of this combination, as defined by the Bliss independence rule (theoretical ara-C + BV6). expression between BV6-sensitive and -resistant cases ( Figure 3). Thus, constitutively low expression of XIAP and constitutively high levels of TNF correlate with the sensitivity of primary AML samples towards treatment with BV6.

dIscussIon
IAP proteins represent relevant targets for therapeutic intervention in AML, since they are expressed at high levels and contribute to evasion of cell death [11]. In the present study, we report that about half of the 67 investigated primary AML samples at diagnosis are sensitive to the Smac mimetic BV6 including about two thirds of ara-C-resistant cases, while mononuclear cells from healthy donors remain largely unaffected. In addition, BV6 enhances the antileukemic activity of the standard chemotherapeutic drug ara-C in an additive manner. Our present study is the largest evaluation so far of Smac mimetic's antileukemic activity in primary, newly diagnosed AML samples, thus highlighting its clinical relevance. By comparison, most previous studies focused on AML cell lines included only few, if any, primary samples [17][18][19][23][24][25][26][27][28][29]. We previously reported that inhibition of IAP proteins by small-molecule inhibitors does not sensitize unstimulated or phytohemoagglutininactivated peripheral blood lymphocytes for ara-C-induced cell death [16]. Also, we demonstrated that treatment with Smac mimetic has no adverse effect on colony formation of normal human CD34 + hematopoietic cells freshly isolated from healthy human donors at concentrations that are cytotoxic against leukemia cells [30]. However, cytokine release syndrome has been reported as a doselimiting toxicity in a recent phase I study [31].
Using whole-genome GEP, we identify a number of functional categories related to cell death that are significantly induced by BV6 treatment in sensitive samples only. Several categories are linked to apoptosis, consistent with the role of IAP proteins in the regulation of programmed cell death. These findings are in line with activation of cell death in sensitive AML samples via IAP inhibition by Smac mimetic. In addition, our GEP studies identify NF-κB Signaling Pathway as another pathway that is activated by BV6 treatment in sensitive samples. This is in line with the reported role of IAP proteins in NF-κB signaling. cIAP1 and cIAP2 are positive regulators of canonical NF-κB signaling and suppress the non-canonical NF-κB pathway [20,[32][33][34]. Accordingly, depletion of cIAP proteins by Smac mimetics such as BV6 has been shown to stimulate non-canonical NF-κB signaling via accumulation of NF-κB-inducing kinase (NIK) [20,34], a critical kinase upstream in the non-canonical NF-κB pathway that is constitutively degraded by cIAP1 and cIAP2. This leads to upregulation of NF-κB target genes  [20,34,35]. Consistently, TNFR1 Signaling Pathway is identified as another signaling cascade in BV6treated responsive AML samples. Also, NF-κB signaling has previously been implicated in mediating Smac mimeticinduced cell death [36,37]. Of note, we recently identified by GEP cell death and NF-κB among the top pathways regulated by BV6 in responsive primary CLL samples as well as in primary core-binding factor (CBF) AML samples [38], underscoring the general relevance of these signaling pathways for Smac mimetic-mediated antitumor activity. While that study also revealed redox signaling as another pathway in BV6-sensitive primary CLL samples as well as in primary CBF AML samples [38], we did not find redox-related signaling cascades among the top regulated pathways in the present set of AML samples, pointing also to some differences among these types of leukemia.
Furthermore, in the present study we identify low constitutive levels of XIAP and high constitutive expression of TNF as parameters that correlate with sensitivity to BV6 in AML samples. Since low XIAP expression favors the induction of cell death, this finding is in line with the observed response to Smac mimetics. It is interesting to note that constitutive TNFα production has previously been linked to sensitivity of human cancer cell lines towards Smac mimetics [39]. Also, TNFα/TNFR1 autocrine/paracrine signaling has been shown to trigger cell death in the presence of Smac mimetics that facilitate TNFα-mediated cell death by depleting cIAP proteins [20,34,39]. We previously reported that the TNFαblocking antibody Enbrel significantly reduces BV6/ara-C-induced cell death in AML cell lines [17], consistent with an autocrine/paracrine TNFα loop mediating cell death. However, there is also evidence showing that Smac mimetics can trigger cell death via death receptor 5 independently of TNFα/TNFR1 signaling [36]. For some probe sets, no gene symbol is annotated to date (NA). www.impactjournals.com/oncotarget Moreover, by whole-genome GEP we identify a set of 25 genes that are differentially expressed among untreated BV6-sensitive and -resistant cases, which might serve as biomarkers to identify BV6 sensitivity prior to treatment. In this regard, SERPINI1 is of special interest, as it was found to be a biomarker for hepatocellular carcinoma [40], and expression of SERPINI1 has been shown to be regulated by c-Myc [41]. Furthermore, TET2 and HTRA4 were expressed at lower levels in BV6-resistant samples and might also be at least partly involved in treatment resistance. TET2 has been implicated in leukemia and associated with decreased overall survival in AML, as its loss of function was found to impair differentiation and favor myeloid tumorigenesis [42][43][44][45]. HTRA4 is thought to be involved in the modulation of apoptosis and chemotherapyinduced cytotoxicity with a tumor-suppressive role [46]. In contrast to TET2 and HTRA4, the transcription factor TCF7L2 was less expressed in BV6-sensitive samples. TCF7L2 has been implicated in AML [47,48], regulates MYC expression and has an influence on survival and proliferation [49][50][51]. While the predictive power of these gene expression differences for BV6 treatment sensitivity requires further validation, the candidates revealed by our analysis provide a starting point for subsequent studies.
Of the tested molecular markers, we identify CEBPA and NPM1 mutations to be differentially distributed among BV6-sensitive and -resistant AML samples. Here, NPM1 mutations are more prevalent and CEBPA mutations less prevalent in BV6-sensitive cases. As far as NPM1 mutations are concerned, the association with BV6 sensitivity proved to be significant even when taking into account FLT3-ITD. While these data indicated that NPM1-mutated AML cases might be an AML subgroup that responds to BV6 treatment, additional markers may be necessary to prospectively define sensitive AML cases, as IDH mutations have been reported to confer adverse prognosis in AML with NPM1 mutations [52]. Beyond CEBPA and NPM1 mutations, our previous work in CBF AML identified a higher sensitivity to the Smac mimetic BV6 in the subgroup with superior outcome [23]. As several Smac mimetics are currently being tested in early clinical trials [12], the identification of molecular markers to select patients which likely respond to Smac mimetics becomes more and more important. Molecular markers identified by GEP or by genomic analysis represent valuable approaches in this respect. Our results point towards Smac mimetics as a novel therapeutic option in AML, especially in patients with NPM1 mutations, low XIAP expression or high TNF expression. These findings are expected to have important implications for the design of Smac mimetic-based protocols in the treatment of AML.  In total, 60 Biocarta pathways were differentially regulated specifically in BV6-sensitive cases (BV6 vs. DMSO treatment).

Primary AMl patient samples
Here we show an excerpt of 12 Biocarta pathways (P <.05 in LS/KS test), plus two others that were especially relevant.

Viability assays
We performed an ATP-content measurement using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, Madison, WI, USA), which reflects the amount of viable cells per sample. For read-out, we used the GloMax 96 luminometer (Promega GmbH, Mannheim, Germany). Furthermore, we performed flow cytometry in selected samples, which were double-stained using Annexin V-PE (BD Pharmingen, Franklin Lakes, NJ, USA) and 7-AAD (7-Amino-Actinomycin D; BD Pharmingen) according to the manufacturer's protocol, and measured using a FACSCalibur (BD). All treated samples were measured after 24 hours of treatment and normalized to appropriate control-treated samples.

data analysis
Microarray data were analyzed using BRB Array Tools Version 3.7.2 as previously described [54]. 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, and Pearson chi-squared test, as appropriate. An effect was considered significant, if the P value was 0.05 or less. Data were visualized using either SigmaPlot (Systat Software, San Jose, CA, USA) or GraphPad Prism 4 (GraphPad Software, La Jolla, CA, USA).