Search

Human acute myelogenous leukemia (AML) is thought to arise from a rare population of malignant stem cells. Cells of this nature, herein referred to as leukemic stem cells (LSCs), have been documented for nearly all AML subtypes and appear to fulfill the criteria for stem cells in that they are self-renewing and give rise to the cells found in many leukemic populations. Because these cells are likely to be critical for the genesis and perpetuation of leukemic disease, the present studies sought to characterize unique molecular properties of the LSC population, with particular emphasis on the transcription factor, nuclear factor-κB (NF-κB). Previous experiments have shown that unstimulated human CD34+ progenitor cells do not express NF-κB. In contrast, primary AML CD34+ cells display readily detectable NF-κB activity as assessed by electrophoretic mobility shift assay and gene expression studies. Furthermore, detailed analyses of enriched AML stem cells (CD34+/CD38−/CD123+) indicate that NF-κB is also active in the LSC population. Given the expression of NF-κB in leukemic, but not normal primitive cells, the hypothesis that inhibition of NF-κB might induce leukemia-specific apoptosis was tested by treating primary cells with the proteasome inhibitor MG-132, a well-known inhibitor of NF-κB. Leukemic CD34+/CD38− cells displayed a rapid induction of cell death in response to MG-132, whereas normal CD34+/CD38− cells showed little if any effect. Taken together, these data indicate that primitive AML cells aberrantly express NF-κB and that the presence of this factor may provide unique opportunities to preferentially ablate LSCs.

We have examined the ability of recombinant adenoviral vectors to transduce human hematopoietic cells. Our findings indicate that adenovirus readily infects a large proportion of CD34+ cells. Using adenovirus vectors that transduce either a lacZ or an alkaline phosphatase reporter gene, we observed up to 45% of total CD34+ cells infected. Upon more detailed analysis, we observed comparable levels of transduction for CD34+/CD38- cells and for CD34+ cells in G(zero) phase of the cell cycle. Importantly, exposure to adenovirus resulted in negligible levels of toxicity as assayed by propidium iodide staining and colony-forming ability. Using adenovirus vectors, we also describe a model system for regulated gene expression in early hematopoietic tissues. CD34+ cells were simultaneously infected with two viruses, one carrying a TetR/VP16 transactivator (tTA) and the second carrying a tTA- dependent lacZ reporter gene. Using this approach, beta-gal expression was only observed upon coinfection with the transactivator vector. In addition, as shown previously (Gossen and Bujard, Proc Natl Acad Sci USA 89:5547, 1992), tetracycline was able to inhibit tTA mediated induction, thereby providing an effective means to regulate expression of the reporter gene. We conclude that recombinant adenovirus is an effective vehicle for transiently expressing genes in primitive human hematopoietic cells.

Abstract 206 Targeting leukemia stem cells (LSCs) is a priority for the development of improved therapeutic regimens. However, the intrinsic heterogeneity of malignant populations in acute myelogenous leukemia (AML) has made it challenging to identify biological properties appropriately conserved amongst primitive cell types. To better characterize physiological features of LSCs related to growth and survival, we previously investigated oxidative state and demonstrated that the majority of functionally-defined LSCs are characterized by relatively low levels of reactive oxygen species (termed “ROS-low”)(Lagadinou et al, abstract 639 ASH 2011). Based on these findings herein we have used primary AML specimens and flow cytometric sorting for endogenous ROS levels so as to enrich for ROS-low LSCs, and we have characterized mechanisms controlling LSC energy production and redox state. We report here that LSC-enriched ROS-low cells are metabolically dormant tumor populations characterized by low levels of oxygen–dependent mitochondrial respiration (OXPHOS), low rates of anaerobic glycolysis, and a low overall cellular ATP content. These properties are unique for LSCs, as bulk leukemic cells and non-tumorigenic ROS-high cells were found to be significantly more metabolically active with regard to both aerobic and anaerobic types of energy production. Intriguingly, we further demonstrate that in contrast to bulk leukemic cells, ROS-low subsets are deficient in their ability to utilize glycolysis when mitochondrial respiration is pharmacologically blocked, indicating a paradoxical dependence of LSCs on mitochondrial energy production. To investigate the mechanisms that underlie the distinct metabolic properties of ROS-low cells, we performed gene expression studies using RNA-seq based methods. In agreement with an important role of mitochondrial metabolism in LSCs we found several mitochondrial-related genes up-regulated in ROS-low cells. Importantly, we found that ROS-low cells express significantly higher levels of bcl-2 both at the mRNA and protein level. To determine if bcl-2 up-regulation relates to the metabolic status of ROS-low cells, we evaluated the bio-energetic profile of bulk AML cells and isolated ROS-low subsets +/− in vitro treatment with the bcl-2 pharmacologic inhibitor ABT-263 and the closely related compound ABT-737. We found that functional inhibition of bcl-2 by this class of drugs results in severe OXPHOS blockage both in total AML cells and ROS-low subsets, indicating a novel non-canonical activity of bcl-2 in promoting AML cell mitochondrial bioenergetics. In unfractionated total AML cells, the bcl-2-inhibitor initiated impairment of OXPHOS was associated with a robust induction of glycolysis and variable toxicity, indicating glycolysis as a compensatory protective response of leukemic cells in this class of drugs. In contrast, bcl-2 inhibition and OXPHOS impairment in ROS-low cells was not compensated by glycolysis, and resulted in depletion of cellular ATP levels, elimination of cellular glutathione pool, oxidation and profound toxicity in the LSC-enriched ROS-low compartment in vitro. Taken together, these studies indicated ABT-263 as an approach to eradicate LSCs by impairing fundamental aspects of LSC metabolism. To more directly investigate this issue we performed xenograft analyses. We first treated ROS-low AML populations in vitro with ABT-263 concentrations equal to the IC50 concentration of the total AML cells for each sample, and then transplanted treated vs. vehicle control cells into immune deficient NSG mice. We found that ABT-263 reduced LSC potential in all AML specimens evaluated by this functional assay. Next, we treated mice engrafted with primary human AML cells with ABT-737 in vivo (50mg/kg IP for 15d), and then performed serial transplantation analyses with the engrafted cells from treated and control mice. These functional studies showed that ABT-737 clearly reduced leukemia burden in the treated primary recipients, and also significantly reduced the capacity of engrafted leukemia cells to establish AML in secondary recipients. Based on these findings, our studies propose a model wherein the unique physiology of ROS-low LSCs provides an opportunity for selective targeting via disruption of Bcl-2-dependent oxidative phosphorylation. Disclosures: No relevant conflicts of interest to declare.

Abstract 3506 Many studies have indicated that induction of oxidative stress is one component of the mechanism by which anti-leukemia agents function. However, little is known about the differences between leukemic and normal redox homeostasis, and whether these agents achieve their selective toxicity to leukemia through specifically targeting leukemic redox homeostasis. Hence, the present study performed a detailed analysis of CD34+ cells derived from normal bone marrow (NBM) or primary acute myelogenous leukemia (AML) specimens. By evaluating the expression profiles of redox regulatory genes, we observed that glutathione peroxidase (GPX1), glutathione reductase (GSR), and several other genes are significantly dysregulated in AML cells. Consistent with these findings, we found a lower GSH/GSSG ratio as well as a lower amount of total glutathione in AML cells compared to normal. In addition, we also found that AML cells display a different spectrum of cell surface thiols. Together, these data suggest that leukemia cells have a distinct redox state maintained by a set of differentially regulated genes in comparison to normal cells. We next tested if parthenolide (PTL), a known ROS-inducer and an effective anti-leukemia agent, targets any of these differences to achieve its leukemia selectivity. We found that PTL directly depletes glutathione in a time and dose dependent manner. However, the amount of glutathione remaining after PTL treatment was less in primary CD34+ AML cells (∼1.5nmol/mg lysate) compared to CD34+ NBM cells (∼5nmol/mg). Thus, the net consequence on glutathione levels is much more severe in AML cells. Importantly, PTL, at the tested dose, selectively kills both AML bulk and stem/progenitor populations but spares normal cells. Given the central role of glutathione in redox homeostasis, we reason the selective toxicity of PTL to leukemia is at least partially because AML cells has a limited pool of glutathione compared to normal. Furthermore, the active moiety of PTL, a α-metheylene-γ-lactone group, predicts it will alkylate cysteine (Cys) residues required for the critical function and stability of several redox proteins. To identify these interactions, we used a biotinylated analogue of PTL to perform a streptavidin-based pull-down assays followed by liquid phase chromatography-mass spectrometry (LC-MS) analysis. We demonstrate that thioredoxin (TXN) and glutathione peroxidase (GPX1), key regulators of cell surface and intracellular redox balance, are direct binding targets of PTL. These binding events were found to at least partially inhibit the enzymatic function of these proteins, contributing to PTL-induced oxidative stress in AML cells. Knowing that GPX1 is already dysregulated in AML cells compared to normal, and considering the increased surface thiol level on AML cells, we hypothesize that these interactions are also important for the selectivity of PTL to AML cells. This is in addition to its ability to exploit the differential pool of glutathione in AML cells. Lastly, PTL induces ROS stress in AML cells as evidenced by redox dye staining, and robust Nrf2-mediated anti-oxidant responses in AML cells. Notably, SLC7A11, a gene that is required for importing cysteine into AML cells for glutathione synthesis, and GCLM which is directly involved in glutathione synthesis are also both up-regulated by PTL. In summary, our studies identify altered GSH/GSSG ratio, decreased glutathione levels, dysregulated expression of GPX1, GSR and other redox regulatory genes, and increased cell surface thiols as critical differences between leukemic and normal redox homeostasis. Our study of PTL indicates that leukemia-specific agents of this nature function by exploiting the differences between glutathione system in leukemic vs. normal cells, and further acting on redox proteins that are already dysregulated in AML cells. These findings have ramifications for the rational design of regimens that manipulate oxidative stress as a means to selectively kill malignant cells. Disclosures: No relevant conflicts of interest to declare.

Abstract 553 In order to increase our understanding of key biological properties governing the development of leukemia stem cells (LSCs), we employed a novel gene identification strategy based on cooperation between initiating oncogenes. Previous studies have demonstrated that genes whose expression is regulated in a synergistic manner as a consequence of two cooperating oncogenes (termed “cooperativity response genes”, or CRGs) are highly enriched for activity in tumor formation. Further, in contrast to the thousands of genes identified by differential expression analyses of normal vs. leukemic cell populations, CRGs represent a much smaller subset of targets; thereby, providing a defined set of genes to investigate. We adapted the CRG strategy to identify synergistically regulated genes in primitive leukemic cells. Using a mouse model of myeloid blast crisis leukemia induced through the cooperation of BCR-ABL and NUP98-HOXA9, we performed genome-wide transcriptional profiling comparing hematopoietic cells expressing each translocation alone or in combination. Using this system, we were able to model the genetic alterations induced as normal cells progressed towards LSC transformation, identifying 72 CRGs (50 aberrantly up-regulated and 22 down-regulated) with potential importance in leukemia development. To investigate the relevance of these CRGs in leukemia biology, an RNAi screen approach was employed. Primary leukemic progenitors were purified and transduced with a custom lentiviral RNAi library and subsequently transplanted into recipient animals to assess the engraftment potential upon perturbation of the individual CRGs. Our findings demonstrate that knock-down of expression in 35 of 50 (70%) leukemia CRGs reduced in vivo growth of primitive leukemia, a finding that was independently validated through single gene perturbation of several genes that scored in the RNAi screen (GJB3, EphA3, PMP22, Serinc2, SerpinB2, and CP). In particular, serpinB2, a gene that scored strongly in the RNAi analysis, was shown to directly effect the frequency of LSC in vivo. Given that the cooperative gene signature represented genes with many distinct cellular functions, we hypothesized that the CRG expression profile represents a key regulatory network in leukemia survival. To investigate our hypothesis we utilized the Broad Institute's Connectivity Map (CMAP) to identify pharmacological compounds with the ability to modulate multiple CRGs simultaneously. This analysis revealed that both Tyrophostin AG-825 (AG825) and 4-hydroxy-2-nonenol (4HNE) were predicted to reverse the gene expression induced as a consequence of leukemic transformation. To test the effect of these agents as selective toxicants to leukemia, we treated both normal and leukemia murine bone marrow cells with each compound. Both bulk and phenotypically primitive leukemic cells were eradicated in dose-responsive fashion upon treatment with either AG825 or 4HNE, while normal cells showed significantly reduced sensitivity. Progenitor function as measured by colony forming assays also showed a selective reduction in leukemia colony formation, suggesting that both these compounds are toxic to the majority of leukemic cell types. Interestingly, similar results were obtained when human normal and leukemic bone marrow specimens were treated with both drugs, suggesting the CRG signature represents an important class of genes with conserved function across species. To determine the level of conservation of the leukemia CRG signature between murine and human leukemia, we profiled eight normal and leukemic patient specimens for expression of the CRG signature. Of the 39 evaluable human CRG orthologs, 13 showed similar expression trends in human leukemia samples relative to normal controls. Intriguingly, both AG825 and 4HNE were predicted to inhibit this 13-gene signature by the CMAP database, suggesting that the compounds may act through these genes to influence leukemia cell death. Taken together, our findings demonstrate the importance of cooperative gene regulation in leukemogenesis and provide a novel platform for future research toward more effective therapeutic strategies to treat leukemia. Disclosures: No relevant conflicts of interest to declare.

Abstract 639 Failure to eliminate leukemia stem cells (LSC) appears to be a major limitation of standard regimens and a significant challenge in the development of improved therapies. To effectively target LSC, it is essential to unravel the unique functional properties of these cells with respect to bulk leukemic and normal cells. Reactive oxygen species (ROS) are oxygen-containing chemicals continuously generated in cells through various biochemical pathways. Normal stem cells appear to tightly control intracellular ROS at low levels so as to maintain long-term self-renewal and survival. Although cancer cells are generally considered to have increased ROS levels due to oncogene activation and acquisition of a tumor-specific metabolic profile, the redox regulation in cancer stem cells remains largely unknown. We hypothesized that, similar to normal stem cells, that redox state is important for LSC survival. By employing primary human acute myeloid leukemia (AML) samples and several redox-sensitive dyes that provide information on different aspects of intracellular oxidative state, we show that leukemic cells display a range of intracellular ROS, with functionally defined LSC being distributed across the ROS gradient. However, leukemic cells prospectively isolated by flow cytometry from the lowest end of the ROS gradient contain the highest concentration of leukemic stem cell activity, as defined by their preferentially quiescent cell cycle profile, enhanced potential in colony forming assays, increased engraftment in immune deficient NSG mice, and capacity to re-establish in the xenograft model an oxidative state heterogeneity similar to that observed in the primary tumor. Further, in serial transplantation assays, which evaluate the long-term maintenance of self-renewal capacity, the ROS-low leukemic population is preferentially able to engraft NSG mice in most cases. In contrast, leukemic cells residing in higher oxidative levels (ROS-high) are more actively cycling; show reduced in vivo self-renewal potential in NSG mice, and evidence of increased DNA damage. Upon therapeutic challenge, prospectively isolated ROS-low primary leukemic cells are less responsive to in vitro drug treatment with both conventional (daunorubicin) and experimental (parthenolide, temsirolimus) anti-leukemic agents, and are better able to resist induction towards an increased oxidative state after pro-oxidant treatment (hydrogen peroxide, BSO). To further investigate the relative sensitivity of ROS-low vs high primitive populations to anti-leukemic therapies, we treated sorted primary AML subsets with anti-leukemic drugs for 18hr and then transplanted into NSG mice. In this functional assay, only the ROS-low LSC fraction retains its self-renewal capacity after therapy, thereby indicating that ROS-low cells may be a critical target for therapy. To better understand mechanisms controlling redox heterogeneity in leukemia, we performed biochemical analyses on ROS-low vs high subsets. Our studies indicate that ROS-low cells adopt unique metabolic and redox regulation properties. Whereas ROS-high cells exhibit robust activation of pathways implicated in cell cycle promotion, metabolism, inflammation and senescence, the respective gene and protein expression in ROS-low cells is either markedly decreased or diminished, and ROS-low cells also show a dramatically reduced dependence on mitochondrial oxidative phosphorylation, as determined by bio-energetic analyses. Further, ROS-low cells seem to contain increased levels of glutathione, a critical cellular antioxidant. Taken together, these findings suggest that ROS-low LSC represent an important target for therapy, but as a consequence of their unique biological properties may be very difficult to eradicate. To this end, we have begun to characterize agents with the ability to enhance leukemia selective targeting of the ROS-low population. Intriguingly, our initial studies show that the naturally occurring compound rocaglamide is a potent inducer of cell death in the ROS-low compartment, as determined by both in vitro culture and xenograft analyses. Importantly, rocaglamide is much less toxic to normal stem and progenitor, indicating that this approach may have a significant therapeutic index. Thus, we propose that regimens designed to include compounds of this class may result in more effective elimination of human LSCs. Disclosures: No relevant conflicts of interest to declare.

Abstract 2463 Acute myeloid leukemia is malignant disease, characterized by an accumulation of immature myeloid cells. Recent studies have demonstrated that myeloid leukemia appears to arise from a population of leukemia stem cells (LSCs). LSCs typically reside in a quiescent state and therefore do not respond to standard chemotherapeutic agents, which generally target more actively dividing cells. However, LSCs do display certain unique molecular properties that can be exploited to target this relatively rare population of cells that drive disease pathogenesis. Specifically, NF-kB, a pro-survival transcription factor, is constitutively active in LSCs but not in normal hematopoietic stem cells (HSCs). Targeting this pathway by pharmaceutical approaches has been suggested as a potential strategy in the treatment of leukemia; however, inhibiting this pathway alone is not sufficient to strongly induce AML-specific cell death. Further investigation of pathways, that are unique to AML, is a key in designing more effective pharmacologic agents that specifically target the LSC. We have previously demonstrated that the naturally occurring compound parthenolide (PTL) induces apoptosis in primary AML cells, including the stem and progenitor cell. While the empirical anti-leukemic activity of PTL is clear, the underlying molecular mechanisms remain poorly understood. Here we investigate two properties associated with parthenolide-mediated cell death: i) activation of pro-apoptotic transcription factor p53, ii) inhibition of pro-survival transcription factor NF-kB. In order to evaluate the role of p53 signaling, AML cells were challenged with PTL resulting in the phosphorylation of p53 at serine-15, indicating activation p53 in response to PTL. To further investigate the role of p53 in PTL mediated responses, we generated a lentiviral vector expressing shRNAs specifically targeting p53. Leukemia cells were infected with the lentiviral vector encoding p53 shRNA or scramble control and evaluated by qPCR and western blot analysis. The data showed a significant knockdown of p53 mRNA and protein levels, as well as strong inhibition of p21 expression, indicating the specificity of p53 knockdown. Exposure of cells to PTL in which p53 has been specifically disrupted results in partial rescue from PTL mediated cell death, implicating the role of p53 in this response. Next, we performed a detailed analysis of the molecular mechanism by which PTL inhibits NF-kB pathway activity. Using a novel analog of PTL, we demonstrate that the compound directly binds to IKK-beta. Upon exposure to PTL, IKK-beta shows reduced kinase activity, indicating that binding of the drug directly impairs enzymatic function. Secondary to the inhibition of IKK-beta kinase activity, there is decreased phosphorylation of IkB-alpha at ser32/36, resulting in reduced proteosome mediated degradation. As expected, translocation of RelA/p65 to the nucleus was also impaired, resulting in decreased DNA binding activity as evidenced by electrophoretic mobility shift assay (EMSA). Interestingly, studies with a biotinylated analog also show that PTL appears to directly bind p65, we also observed a decreased phosphorylation of p65 at serine 536, an event mediating the transcriptional activity of DNA-bound p65. Inhibition of the NF-kB pathway by parthenolide also resulted in very significant inhibition of one of its well-known downstream target genes, ICAM-1 (CD54) at mRNA, protein and surface expression levels. Whether reduced ICAM-1 expression affects the biology of AML cells is as yet unknown. However, given the known role of ICAM-1 in integrin signaling, we propose that loss of ICAM-1 via NF-kB inhibition may impair the ability of AML cells to interact with their environment. Taken together, this study further elucidates the mechanisms by which PTL mediates pro-apoptotic activity in primary AML cells. PTL induces activation of p53 pathway and therefore knockdown of p53 by shRNA results in partial rescue from PTL mediated cell death. PTL also inhibits the NF-kB pathway, which includes binding of PTL to both IKK-beta and RelA/p65, which leads to decreased phosphorylation of IkB-alpha and reduced DNA binding of p65. In addition, we have discovered the ICAM-1 expression in AML cells is regulated by NF-kB, and that loss of NF-kB DNA binding activity results in loss of ICAM-1 expression. Disclosures: No relevant conflicts of interest to declare.

Abstract 2654 Poster Board II-630 Introduction: With the institution of multidrug, multiphase chemotherapy regimens, major improvements in clinical outcomes have been made in pediatric acute leukemia patients in the last thirty years. However, there remains a substantial percentage of pediatric patients who relapse and die of their disease, particularly with high risk ALL, T cell ALL and AML. It is possible that these patients' disease initiates from a leukemic stem cell such as those found in adult myeloid disease, or at the very least, harbor a chemo-resistant population. Our research has two main aims: first to evaluate the functional and phenotypic heterogeneity within standard risk (SR), high risk (HR) and relapsed (RD) pediatric leukemia. Second, to evaluate current treatment regimens for the selection of a chemo-resistant or LSC populations and then attempt to target this population with novel treatments. Methods: In vitro studies for functional heterogeneity include colony-forming assays (CFU) using methylcellulose and limiting-dilution suspension culture studies. Phenotypic heterogeneity is evaluated with multi-color flow cytometry and detection of alterations in aldehyde dehydrogenase activity. Xenograft studies in immune deficient mice are used to evaluate self-renewal capability, serial engraftment kinetics, and alterations in phenotype. Drug studies are performed by evaluating the differences in phenotype and CFU over time when treating with conventional induction chemotherapy or novel agents. Results: We have evaluated several SR and HR ALL samples in addition to some RD samples, which are paired with HR diagnostic samples. In vitro studies revealed the SR samples had little to no colony forming ability (0-1%) while the HR samples had approximately 3-5% and the RD samples 8-10% colony-forming ability. Likewise, the SR samples failed to engraft NOD-SCID mice while the HR samples, from patients with infantile ALL and the MLL translocation or T cell ALL, had robust engraftment in primary and secondary recipients. The engraftment kinetics were uniformly faster in secondary recipients. These findings suggest that HR leukemia may be the result of a leukemia-initiating cell with stem cell-like characteristics while SR ALL may arise from a more committed lymphoid progenitor. Interestingly, in the RD samples, several of the phenotypic markers are similar to that of the primary sample after treatment with induction therapy, particularly with regards to percentages of CD 34, 133-1, 133-2 and aldehyde dehydrogenase levels. Several HR samples have been exposed to induction chemotherapy (Decadron, Cytarabine, Doxorubicin and Vincristine), and the CFU potential and phenotype evaluated over a two-week time course. Notably, the majority of bulk disease is effectively killed, the CFU content actually increases two to three-fold, when an equivalent number of viable cells are analyzed. Furthermore, the phenotype reveals brighter staining with several proposed stem cell markers (CD34, 117, 133-1, 133-2, 123, and measurement of aldehyde dehyrogenase). These data indicate the selection of a chemo-resistant or LSC population. Conclusions: Our results to this point suggest important differences both functionally and phenotypically, between SR, HR and RD pediatric leukemia. These findings are consistent with what would be expected given clinical differences in each of these disease states and begins to establish a means of identifying a LSC or chemo-resistant population, which can be targeted with novel treatment regimens. Likewise, these techniques may also provide a means of evaluating for minimal residual disease (MRD) in a LSC or chemo-resistant population by identifying that population's phenotype by passaging the initial sample through serial murine engraftments or in vitro drug studies. Disclosures: No relevant conflicts of interest to declare.

Abstract 3473 Poster Board III-410 Leukemia arises when normal hematopoiesis is altered causing an over-production of immature blood cells. In most cases, mutations appear to occur early in hematopoietic development affecting the more primitive stem/progenitor cell compartment, generating a leukemia stem cell (LSC). The BCR-ABL translocation is a common abnormality linked to chronic myeloid leukemia (CML), predisposing patients to the acute form of the disease, blast crisis (bcCML). The genetic interplay between BCR-ABL and the secondary mutations required for acute leukemia is poorly understood. Therefore, to better define the transformation process, we adapted a strategy previously described by McMurray et al in which genes synergistically regulated by two cooperating oncogenes are identified. This approach was shown to identify key targets of cancer cell survival and potential targets for therapeutic intervention. To implement this strategy for bcCML we employed a mouse model in which primitive (lineage depleted) mouse bone marrow cells are engineered through retroviral transduction to express BCR-ABL alone or in combination with another oncogene associated with human leukemia; Nup98-HoxA9. We generated leukemic mice and isolated bone marrow cells using fluorescence activated cell sorting (FACS). Each oncogene was tagged with a distinct fluorescent marker (green fluorescent protein for BCR-ABL and yellow fluorescent protein for Nup98-HoxA9) allowing for isolation of purified cell populations. From six independent cohorts, 4 distinct primitive (lineage negative) cell populations were purified: 1) BCR-ABL+, 2) Nup98-HoxA9+, BCR-ABL+/Nup98/HoxA9+, and BCR-ABL−/Nup98HoxA9− (i.e endogenous normal cells). Probes from each cell population were analyzed using mouse 430 Plus microarray chips from Affymetrix. Genome wide microarray analysis was performed by first comparing gene expression between non-malignant (endogenous normal) and bcCML (both oncogenes) cells providing a list of genes differentially expressed in the disease state. This gene-set was further refined by comparison to cells expressing each oncogene independently, identifying 78 genes that are synergistically regulated by both BCR-ABL and Nup98-HoxA9 in primitive leukemic cells. We term this gene set, CRGs, for cooperation response genes, the nomenclature previously established by McMurray et al. The aberrant expression of 8 specific CRGs (serpinB2, AREG, EREG, pla2g7, csf2, selp, irf8, mycn) was verified using quantitative real-time PCR, and found to be in close agreement with the gene array data. Next, pathway analysis of the CRG gene set was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID). Pathways identified by this analysis include activation of JAK-STAT, Hedgehog, and Erb2 signaling, as well as increased xenobiotic metabolism. Interestingly, our data set suggests that activation of Erb2 signaling by AREG and EREG acts through STAT5, thereby influencing cell survival. In addition, drug metabolism through cytochrome P450 is observed to be up-regulated in primitive leukemia cells, indicating an enhanced ability to tolerate therapeutic agents. Taken together, the analyses performed thus far have identified gene targets implicated as central to the growth of primitive leukemia cells. Going forward, we will evaluate how the 78 synergistically-regulated genes are modulated in response to various forms of therapy, and determine whether perturbation of specific genes/pathways can selectively inhibit/eradicate LSC. Disclosures: No relevant conflicts of interest to declare.

Studies to date have shown that primary human leukemia stem cells (LSC) are resistant to standard chemotherapy agents and are likely to be a major cause of drug refractory disease and relapse. Therefore, elucidating the in vivo biology of LSC is critical in order to develop more effective therapeutic regimens. To this end, we report the first genetically defined model of LSC, using syngeneic murine systems in which the biological features of human LSC are recapitulated. The approach employs retroviral vectors to transduce normal murine hematopoietic stem cells with either BCR/ABL-GFP alone, or in combination with Nup98/HoxA9-YFP. Expression of BCR/ABL creates a well-described model of chronic phase CML, whereas expression of BCR/ABL in combination with Nup98/HoxA9 induces acute disease that mimics blast crisis CML. Analysis of the normal cell competent to generate LSC indicates that the BCR/ABL mutation must occur in primitive HSC in order to manifest disease, however, subsequent progression to blast crisis can occur through mutation in cells at the myeloid progenitor stage. Characterization of stem cells in these models revealed several striking features. First, chronic phase stem cells are1 phenotypically identical to normal hematopoietic stem cells (lin−, Sca-1+, c-kit+) and display cell cycle rates (percentage of cells in S or G2 phase) that are nearly double normal controls. However, the overall frequency of such cells is not elevated. In contrast, blast crisis stem cells show a distinct immunophenotype (lin−, Sca-1+, c-kit-lo, Flt3+, CD150−) and cycle rates nearly identical to normal controls, but are approximately 10-fold increased numbers. These data indicate that BCR/ABL alone functions as a stem cell mitogen, but does not enhance self-renewal, whereas added expression of Nup98/HoxA9 is sufficient to increase self-renewal, but return cell cycle regulation to normal levels. Furthermore, analysis of co-resident non-leukemic cells in each model shows that while the cycle activity of normal stem cells (HSC) was not affected, the cycle rates of normal progenitors (lin−, c-kit+) were substantially reduced. Thus, in either disease, active suppression of normal progenitors is evident and thereby increases the growth advantage of malignant populations. To test methods for modulation of normal vs. leukemic cells in vivo, we challenged blast crisis animals with ara-C (single dose, 100mg/kg) or imatinib mesylate (200mg/kg/day for 3 consecutive days) and assessed the consequences in primitive populations. The data indicate that ara-C reduced frequency and cycle rate of progenitor cells in vivo, but that the effects were identical between normal and malignant populations. Thus, at least for short-term studies there was no therapeutic index for ara-C at the level of primitive cells. In contrast, treatment with imatinib induced a 50% increase in the cycle rate and a 2–4 fold increase in numbers of progenitor cells. These findings imply a homeostatic mechanism in blast crisis leukemia, where pressure towards the malignant population may induce increased activity of stem and progenitor cells. In summary, this model provides a novel means by which the biology of LSCs may be directly characterized and the consequences of candidate treatment regimens can be assessed with regard to normal vs. leukemia stem cells in vivo.

Acute myelogenous leukemia (AML) arises from a rare population of leukemic stem cells (LSCs), which are thought to perpetuate both de novo and relapsed disease. Because LSCs are often refractory to standard chemotherapy agents, it is crucial to investigate the ability of current and novel drugs to target this rare population of cells. Previously, we demonstrated that combined treatment with proteasome inhibitors and anthracyclines induced apoptosis in LSCs while sparing the normal hematopoietic stem cells (HSCs). Using this approach, death of leukemic cells was characterized by NF-kB inhibition and p53 activation. In the present study, we have extended this work by investigating parthenolide (PTL), a sesquiterpene lactone found in the herbal drug feverfew and that has been reported to inhibit NF-kB activation. Primary AML cells treated with increasing concentrations of PTL (1-10 microM) showed a rapid (18h) and dose dependent decrease in cell viability. The PTL concentration and its effect on induction of apoptosis directly correlated with decreased NF-kB binding activity and increased phosphorylation of p53. Importantly, apoptosis induction was also evident in the primitive LSC population, defined by the immunophenotypic profile CD34+, CD38− and CD123+. In contrast, PTL had little to no effect in normal HSCs obtained from either cord blood (CB) or bone marrow (BM). Interestingly, apoptosis induction upon PTL treatment was completely abolished upon treatment with the anti-oxidant, N-acetyl-cysteine (NAC), suggesting that PTL effects involve reactive oxygen species (ROS) induction. This was further corroborated using the fluorescent dyes DCFH-DA and mBBr to measure ROS and thiol content, respectively. When PTL effects were compared to the common chemotherapeutic drug cytosine arabinoside (Ara-C), we observed that PTL selectively targeted leukemic cells whereas Ara-C did not. Functional analyses using in vitro colony assays and the NOD/SCID mouse xenotransplant model were then employed to address whether the drugs used in this study affected stem cell potential. PTL treatment did not affect engraftment of normal cells while leukemic cell engraftment was dramatically impaired. Similarly, methylcellulose cultures showed a severe decrease on colony formation in PTL treated AML cells, while normal colonies were not impaired. Together, these data demonstrates that PTL is able to selectively target LSCs while sparing normal hematopoietic cells. The mechanism of leukemia-specific cell death appears to involve inhibition of NF-kB, activation of p53 and induction of oxidative stress.