|Home | About | Journals | Submit | Contact Us | Français|
Most forms of chemotherapy employ mechanisms involving induction of oxidative stress, a strategy that can be effective due to the elevated oxidative state commonly observed in cancer cells. However, recent studies have shown that relative redox levels in primary tumors can be heterogeneous, suggesting that regimens dependent on differential oxidative state may not be uniformly effective. To investigate this issue in hematological malignancies, we evaluated mechanisms controlling oxidative state in primary specimens derived from acute myelogenous leukemia (AML) patients. Our studies demonstrate three striking findings. First, the majority of functionally-defined leukemia stem cells (LSCs) are characterized by relatively low levels of reactive oxygen species (termed “ROS-low”). Second, ROS-low LSCs aberrantly over-express BCL-2. Third, BCL-2 inhibition reduced oxidative phosphorylation and selectively eradicated quiescent LSCs. Based on these findings, we 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.
Biological analyses of leukemia stem cells (LSCs) has been the focus of numerous studies over the past decade, with reports describing molecular, cellular, biochemical, and genetic properties in both human and mouse systems. It has been assumed to at least some extent that the biology of LSCs mirrors that of normal tissues, where malignant stem cells reside at the apex of a hierarchical developmental structure. Indeed, key stem cell properties such as self-renewal are readily evident in stem cells from multiple tumor types (Magee et al., 2012). In addition, quiescent cell cycle status, a very well described feature of normal hematopoietic stem cells (HSCs), also appears to be a central property of functionally-defined acute myeloid leukemia (AML) stem cells (Saito et al., 2010). Further, several molecular mechanisms that regulate normal stem cell properties are also frequently active (and/or aberrant) in cancer cell types as well (Magee et al., 2012). These observations have led investigators to attempt targeting of LSC populations based on properties that are thought to be shared among normal and tumor primitive cells and are conserved from patient to patient.
Despite the similarities noted above, it has become increasingly clear that primitive cancer cell types possess important differences from normal tissues, some of which run counter to conventional views of stem cell biology. For example, recent studies have shown that cell surface immunophenotype can vary widely amongst specimens derived from primary AML patients, and that commonly used stem cell antigens such as CD34 and CD38 do not show consistent levels of expression on LSCs (Eppert et al., 2011; Sarry et al., 2011). Thus, unlike normal hematopoietic tissues, which show substantial inter-specimen similarities, some of the biological properties of independent LSCs may be highly disparate and/or unstable.
Given the observations outlined above, we sought to better define fundamental aspects of LSC biology, which are conserved irrespective of inter- or intra-patient heterogeneity. To this end, we investigated energy metabolism, a basic requirement of any cell type. Cancer cells appear to commonly acquire aberrancies in energy metabolism and cellular redox state (Cairns et al., 2011). In contrast to their normal counterparts, cancer cells tend in the presence of oxygen to employ glycolysis instead of aerobic mitochondrial respiration to generate energy(Warburg, 1956). Notably, in at least some instances, cancer cells generate increased levels of reactive oxygen species (ROS) (Trachootham et al., 2008). This notion has provided the impetus to explore redox modulation as a therapeutic target in cancer, with the rationale that pro-oxidants may eradicate tumor cells by pushing them beyond the limits of viability (Trachootham et al., 2009). However, while such tumor-related metabolic adaptations characterize bulk tumor tissues, it is unclear whether cancer stem cells (CSCs) also exhibit an increased oxidative state. Indeed, previous studies by Diehn et al demonstrated that breast cancer cells enriched for the breast CSC - related CD24−/CD44+ phenotypic profile showed lower levels of ROS than the “non CD24−/CD44+“ cells (Diehn et al., 2009). Gaining further insights into the metabolic properties of CSCs may reveal physiological dependencies that can be targeted for therapy.
In the present study we used primary human AML specimens and a stringent xenograft assay to investigate the oxidative state and the bio-energetic properties of LSCs. We show that in contrast to the non-tumorigenic cells, the majority of self-renewing and chemotherapy resistant primitive leukemic populations are quiescent tumor subsets characterized by a low rate of energy metabolism and a low cellular oxidative status (termed “ROS-low”). Surprisingly, ROS-low cells are unable to utilize glycolysis when mitochondrial respiration is inhibited. Thus, we show that the maintenance of mitochondrial function is essential for LSC survival. Finally, we demonstrate a unique role for BCL-2 in ROS-low cell mitochondrial respiration, and show that small molecule BCL-2 inhibitors can effectively target chemotherapy resistant LSCs by impairing their energy generation capacity and redox control. Our findings indicate it is feasible to eradicate therapy-resistant LSC populations by targeting their unique metabolic dependencies.
To characterize leukemia cells with differing oxidative state we employed a method previously reported by Jang and Sharkis (Jang and Sharkis, 2007). In this study, the authors used cell sorting based on the relative fluorescence of redox sensitive probes to isolate HSC with differing endogenous ROS levels. Since a number of recent studies have highlighted that LSC may display disparate phenotypic patterns besides CD34+/CD38−/CD123+ (Eppert et al., 2011; Sarry et al., 2011), in performing our initial analyses of ROS levels in primary AML samples, we did not rely on standard phenotypic profiles, but rather examined the entire leukemic population. By employing several redox sensitive probes commonly used to measure intracellular ROS (CM-H2DCFDA, CellROX, MitoSox, Mitotracker, DHE) we consistently observed substantial heterogeneity in the redox staining profile of AML specimens (Figure 1A and Figure S1A–B). Since the fluorescent dyes CM-H2DCFDA and CellROX provide a broader measure of intracellular redox levels, we used those two dyes to sort AML specimens to subsets with low and high endogenous ROS (defined as the 15% dimmest and brightest dye fluorescence distribution as shown in Figure 1A) and then analyzed the two populations for stem cell properties.
A critical feature of both normal and leukemia primitive populations is quiescence, as increased proliferation can lead to exhaustion of stem cell function (Ito et al., 2006; Tothova et al., 2007). We hypothesized that quiescence may reduce LSC metabolic activity and thus lower ROS production, so slowly proliferating LSC should be enriched in the redox low compartment. To investigate this issue, we co-labeled sorted ROS-low and ROS-high fractions with anti-Ki67 and 7AAD to evaluate cell cycle status. As shown in Figures 1B and 1C, leukemic ROS-low cells are preferentially enriched for G0 quiescent cells. This finding was corroborated by analyses of cell cycle components, which demonstrated reduced expression of CDK1, cyclins A2, B2, D3 and E2, and increased expression of the cell cycle inhibitor p27 in ROS-low cells (Figures S1C–D). ROS-low cells also showed reduced expression of the MAPK kinase phospho-p38, which is associated with redox state regulation and cell cycle activity in normal hematopoiesis (Figure S1D) (Ito et al., 2006). Interestingly, ROS-low subsets isolated from the M9-ENL1 AML cell line and from normal marrow CD34+ cells were also more quiescent as compared to corresponding ROS-high cells (Figure S1E), indicating that the association between lower ROS and quiescence is conserved in both cell lines and normal HSCs.
To assess whether AML cells with lower endogenous ROS are enriched for primitive leukemia cells, we first performed in vitro colony assays (CFU). In four evaluable AML samples, the leukemic cells with lower ROS gave growth to more CFU in comparison to ROS-high or total AML cells (Figure 1D), suggesting enrichment of primitive cells in the ROS-low compartment. To functionally evaluate LSC potential, xenograft studies were performed using transplantation of primary human cells into immune deficient NOD/SCID-IL2Rγ−/− (NSG) mice. As shown in Figure 1E, LSC activity (i.e. successful engraftment of NSG mice) is readily evident in all ROS-low AML subsets evaluated (5 of 5 independent specimens). Notably, in 2 of 5 specimens we also detected some LSC activity (albeit at lower levels) with ROS-high AML cells, suggesting that functionally defined primitive leukemia populations can exist across a relatively broad ROS gradient in at least some cases. However, the leukemic subsets isolated from the lowest end of the ROS range were consistently more enriched for LSC activity as compared to AML cells containing high levels of ROS (Figure 1E, S1F). Indeed, in transplantation experiments (Figure 1E), despite the fact that AML ROS-low cells were injected in NSG mice at a 5-fold lower dose than unfractionated total AML cells (Table S2), the ROS-low population engrafted at levels comparable to the total AML cells (AML ROS-low=44.6±38 vs Total AML=51±38, p=0.122>0.05, n=5). Further, to address whether ROS-low cells fulfill the stem cell criterion of long-term self-renewal, we performed serial transplantation studies in secondary mouse recipients. ROS-low leukemic cells re-established AML in the secondary recipients in all cases, thus confirming their leukemic stem cell potential (Table S3). Interestingly, in 2 out of three AML specimens where ROS-high AML cells engrafted at levels that allowed comparative serial transplantation analyses, we observed long-term self-renewal potential in secondary recipients suggesting that LSCs are significantly fewer but evident within the redox high leukemic compartment (Table S3). Since normal HSCs have been suggested to reside in a reduced oxidized state we next tested if the increased engraftment potential of the ROS-low AML subset resulted from contributing HSCs (Tothova et al., 2007). To this end we analyzed the human cells from mice engrafted with ROS-low vs high AML cells (patient sample #1) for expression of mutant nucleophosmin (NPM), which is a frequently observed leukemia marker (Falini et al., 2007). Engrafted AML cells from both leukemic subsets were positive for this marker (Figure S1G), thus confirming the leukemic origin of engrafted cells in both cases.
We next asked if AML cells containing lower ROS are enriched for the CD34+/CD38− phenotypic profile, which is commonly associated with stem cells in both normal and leukemic hematopoiesis. We found an increased percentage of CD34+/CD38− cells in the ROS-low subset as compared to ROS-high, but no significant enrichment of this phenotypic profile when the ROS-low cells where compared to the total AML population (Figure S1H). Further, analysis of AML specimens with an additional collection of antibodies associated with primitive cell types (CD123, CD33, CD117, CD90, CD44) did not reveal any clear association between surface expression profile and a lower oxidative state (data not shown).
Taken together, these findings indicate that endogenous ROS levels can be used as a tool to prospectively isolate LSC-enriched populations. To identify properties of LSCs that can be targeted for therapy we therefore used endogenous ROS sorting to enrich for LSCs, and studied mechanisms controlling energy metabolism in LSC enriched populations.
To gain insights in the metabolic regulation of LSC, we performed bio-energetic analyses in ROS-low and high primary AML specimens. We first asked what is the relative dependence of the ROS-low AML cells to the two major energy-generating pathways: mitochondrial oxidative phosphorylation (OXPHOS) and glycolysis. To this end, we used the Seahorse XF24 extracellular flux analyzer to measure the oxygen consumption rate (OCR) indicative of OXPHOS, and the lactate production indicative of glycolysis (extracellular acidification rate; ECAR). As shown in Figure 2A, LSC enriched ROS-low cells showed a significantly lower OCR rate indicative of lower oxidative metabolism as compared to the total and the ROS-high AML cells. While oxidative processes occur mainly in mitochondria, oxidative activity can also be found in the cytoplasmic membrane due to the presence of the NADPH oxidases (Hole et al., 2010). Therefore, we evaluated the protein expression of NADPH oxidase subunit NOX2 (gp91phox) by flow cytometry. As shown in Figure 2B, NOX2 levels were found to be quite low in ROS-low cells, indicating mitochondria as the primary site of oxidative metabolism in LSC enriched populations. AML ROS-low cells isolated based on the fluorescence of the redox dye CM-H2DCFDA also showed reduced labeling with the mitochondrial-specific redox probe MitoSox-Red, indicating a lower content of mitochondrial-derived ROS (Figure S2). We next examined relative rates of glycolysis and observed that ROS-low cells are substantially less active with respect to this anaerobic form of energy production as well (Figure 2C). While the basal rate of glycolysis in ROS-low cells is low, it is possible that under certain conditions (i.e. in hypoxic niches in vivo) LSCs might compensate for low OXPHOS by up-regulating glycolysis. To investigate this issue we measured lactate generation indicative of the cellular glycolytic rate after treatment of cells with the mitochondrial respiration inhibitors oligomycin (OLI) and FCCP. OLI and FCCP inhibit mitochondrial respiration and force cells to the glycolytic pathway to maintain energy supply, so treatment with these agents is indicative of the reserved cellular glycolytic potential. Surprisingly, LSC enriched ROS-low AML cells showed a significantly decreased potential to up-regulate the glycolytic machinery and thus sustain energy generation when mitochondrial energy production was inhibited (Figures 2D–E). In agreement with our bio-energetic data, ROS-low cells showed significantly decreased ATP levels in all samples evaluated (Figure 2F).
To determine whether the metabolic status of ROS-low AML cells represents an aberrant condition, we performed comparative bio-energetic analyses of normal bone marrow CD34+ progenitors isolated from healthy volunteers. A recent study showed that murine hematopoietic stem populations are predominantly glycolytic (Simsek et al., 2010), but to our knowledge the bio-energetic profile of human marrow progenitors has not been previously investigated. We found that, in contrast to AML ROS-low cells, normal CD34+ cells are metabolically active cells with baseline glycolysis and OXPHOS rates similar to CD34− (Figures 2G and I). Further, normal CD34+ progenitors preferentially (in comparison to CD34− cells) up-regulate glycolysis when mitochondrial OXPHOS is blocked (Figures 2G–H).
Taken together, our metabolic analyses indicate that LSC enriched subsets are metabolically dormant tumor populations that reside in a substantially decreased energetic state. Importantly, our findings further suggest that in contrast to normal progenitor cells, LSC enriched subsets may paradoxically depend more on mitochondrial respiration rather than glycolysis to meet their energy demands and maintain their survival.
To investigate mechanisms that control oxidative state and metabolic processes in ROS-low and high populations, we performed gene expression studies using RNA-seq based methods. Since our findings indicated LSC enriched populations contained lower ROS, we investigated if up-regulation of antioxidant genes represents a prominent feature of ROS-low AML cells. We found only a small number of genes related to antioxidant defenses to be significantly altered in ROS-low vs high AML cells, none of which were consistently up-regulated in the ROS-low population (Figure S3A). We next investigated the relative expression of genes related to mitochondria and energy metabolism. In agreement with an important role of mitochondrial metabolism in LSCs, we found several mitochondrial-related genes up-regulated in ROS-low cells. These included the SLC25A42 and SLC25A38 mitochondrial carriers that transport metabolites and amino acids across the mitochondrial membrane, and the gene FIS1 which is critical for mitochondrial dynamics and fission (Figure 3A) (Westermann, 2010). Intriguingly, we found that the LSC enriched ROS-low population expressed significantly higher levels of BCL-2 (Figure 3A). We validated this initial observation using quantitative PCR in a larger AML sample cohort (Figure 3B). BCL-2 mRNA expression correlated with protein expression, as ROS-low AML cells expressed higher levels of BCL-2 protein (Figure 3C), whereas they showed no significant up-regulation of other anti-apoptotic BCL-2 family members as BCL-XL and MCL-1 (Figures S3B–C). Parallel analyses of BCL-2 gene expression in normal marrow specimens showed, in agreement with previous observations (Delia et al., 1992), that normal CD34+ progenitors express significantly higher levels of BCL-2 as compared to more differentiated CD34− cells (Figures S3D–F). To investigate the extent to which normal CD34+ progenitors share the BCL-2 expression profile observed in leukemic cells we further examined BCL-2 levels in normal CD34+ DCF-low vs high populations. To this end, we enriched for CD34+ progenitors using an immunomagnetic affinity column and then isolated CD34+ DCF-low and high populations by flow cytometry. CD34+ DCF-low cells showed a modest increase in BCL-2 mRNA expression as compared to CD34+ DCF-high, but we found no difference in the BCL-2 protein levels among the two populations in three independent experiments (Figures S3G–H). Based on these data, we conclude there is no differential activity of BCL-2 in normal ROS-low vs. high compartments.
The up-regulation of BCL-2 in the LSC is potentially important, since a) BCL-2 has an established role as an inhibitor of the mitochondrial-initiated pro-apoptotic pathway and thus can represent an important contributor in the chemo-resistance properties of LSC (Del Poeta et al., 2003) and b) recent evidence point to a non-canonical activity of BCL-2 in regulating oxidative state and mitochondrial metabolism (Chen and Pervaiz, 2007). Thus, elevated BCL-2 could contribute to key functional properties of LSCs.
To investigate the role of BCL-2 in the metabolic homeostasis of primitive leukemia cells, we determined the bio-energetic properties of primary AML cells treated with BCL-2 pharmacologic inhibitors. We found that within minutes of treatment the BCL-2 inhibitor ABT-263 induces severe impairment of OXPHOS in primary unfractionated AML cells (Figure 4A and S4A). Similarly, a second chemically distinct BCL-2 inhibitor, obatoclax, also inhibited oxidative respiration in leukemia cells (Figure S4B), albeit to a lesser extent than ABT-263. This effect is accompanied by a robust glycolytic response (Figure 4D), indicating glycolysis is a compensatory mechanism activated upon inhibition of OXPHOS by BCL-2 inhibitors. ABT-263 similarly impaired OXPHOS in the LSC–enriched ROS-low cells (Figure 4B). However, in agreement with our previous findings showing that AML ROS-low cells have reduced reserved glycolytic capacity (Figures 2D–E), ROS-low cells were not able to induce glycolysis (Figure 4D). These findings suggested that ABT-263 should selectively inhibit energy generation in the ROS-low population. Indeed, ABT-263 rapidly depleted cellular ATP in the ROS-low AML subset (Figure 4E, left panel), whereas it had no effect in the ATP content of the ROS-high population (Figure 4E, middle panel).
To investigate the role of BCL-2 in normal populations, we tested the effect of ABT-263 treatment in CD34+ cells from healthy donors. As shown in Figure 4C, OCR was impaired, albeit in a lesser extent than observed in leukemic cells. In agreement with our data showing that normal CD34+ cells possess reserve glycolytic capacity (Figure 2H), we detected a clear up-regulation of glycolysis in normal CD34+ subsets in response to ABT-263 induced OXPHOS inhibition (Figure 4D). Importantly, evaluation of the effect of ABT-263 on the ATP content of normal CD34+ progenitors revealed no significant impact, unless ABT-263 was combined with the glycolysis inhibitor 2-deoxyglucose (2-DG) (Figure 4E, right panel). The combination of ABT-263 and 2-DG induced a decrease in the ATP content of normal CD34+ cells at levels similar to that observed in AML ROS-low cells treated with ABT-263 alone. These data indicate glycolytic capacity represents an important biological property that distinguishes AML ROS-low from normal CD34+ cells.
To more directly address the role of BCL-2 in the oxidative respiration of leukemia cells, we employed an shRNA strategy to genetically reduce gene expression. Using lentiviral-mediated gene transfer, we expressed BCL-2 shRNAs in primary AML cells as well as the U937 leukemia cell line (Figure S4C). Primary leukemic and U937 cells transduced with BCL-2 shRNA showed a significantly reduced basal oxygen consumption rate as compared to cells transduced with a control vector (Figure 4F and S4D), indicating BCL-2 is a positive regulator of OXPHOS in leukemia cells. Further, the capacity of ABT-263 to suppress oxidative respiration was markedly decreased in BCL-2 shRNA transduced primary AML cells as compared to vector control cells (Figures 4G and S4E). Knocking down of BCL-XL (which is another recognized target of ABT-263)(van Delft et al., 2006) by specific shRNA did not have a significant effect on the OXPHOS of primary leukemic cells (Figures S4F). Taken together, these findings strongly indicate that inhibition of oxidative respiration induced by treatment with ABT-263 is mediated by inhibition of BCL-2.
Based on our finding that BCL-2 inhibition induced mitochondrial dysfunction and energy depletion in ROS-low cells, we next asked if oxidation was promoted. To this end, we measured mitochondrial ROS levels using the mitochondrial superoxide detector MitoSOX-Red. We found that ABT-263 induces a clear up-regulation of mitochondrial ROS selectively in LSC-enriched ROS-low cells (Figure 5A and S5). As BCL-2 inhibition increased the oxidative state of ROS-low cells, we investigated whether changes in anti-oxidant defenses were evident. We found that ABT-263 treatment resulted in decreased levels of reduced glutathione (GSH) in the ROS-low subset (Figure 5C). Kinetic studies determined that mitochondrial oxidation and glutathione depletion are detectable after 4 to 6hr of in vitro exposure of ROS-low cells to BCL-2 pharmacologic inhibitors, and these events occur subsequent to the effects of BCL-2 inhibitors on OXPHOS and the cellular ATP content (not shown). Taken together, our redox and bio-energetic studies showed that BCL-2 inhibition severely impaired LSC redox and energy control. To investigate BCL-2 inhibition as a means for eradicating primitive AML cells, we evaluated the in vitro sensitivity of ROS-low AML to BCL-2 inhibitors. We first used ABT-737, which is a closely related analog of ABT-263 with similar structural features, but which, in contrast to ABT-263, is not orally bioavailable. Those studies showed a dramatic targeting of ROS-low cells with much less effect in the ROS-high compartment (Figure 5B). Additional studies with ABT-263 show consistently strong toxicity to the LSC enriched ROS-low population with generally less activity towards ROS-high cells indicating a selective role of BCL-2 in LSC survival (Figure 5D).
To evaluate the relative contribution of ABT-263-induced oxidation in LSC cell death, we investigated whether ROS-low cells would be rescued from BCL-2 inhibitor toxicity by the anti-oxidants NAC and GSH-ester. NAC and the GSH-ester did not significantly reduced ABT-263 toxicity in ROS-low cells (Figure 5E), indicating the effects of ABT-263 on LSC oxidative state as a late component of the LSC cell death process rather than initiators of cell death. The AML ROS-low cell oxidation following OXPHOS inhibition and cell death induction in response to ABT-263 is not comparable to our findings in steady state conditions, where lower overall and mitochondrial ROS associated with a relatively lower OCR rate (Figures 2A, S2). Further, although ROS-low cells treated with ABT-263 became oxidized, they remained in a quiescent state (Supplementary Table 4), excluding an increase in LSC cell cycle activity as a mechanism of cell death in response to BCL-2 inhibition. Taken together, our findings suggest that BCL-2 inhibitors target LSC by selectively impairing LSC energy homeostasis.
To further investigate BCL-2 inhibition as an approach to target primitive AML cells, we first examined the colony forming potential of primary AML cells transduced with BCL-2 shRNA vs control cells. As shown in Figure 6A, BCL-2 knockdown dramatically reduced the colony forming capacity of primary leukemic blasts. Next, we performed xenograft analyses. We treated LSC enriched ROS-low AML populations in vitro with concentrations of the BCL-2 inhibitor ABT-263 equal to the IC50 concentration of the total AML cells for each sample, and then transplanted treated vs. vehicle control cells into NSG mice. As shown in Figure 6B, ABT-263 reduced LSC potential in all AML specimens evaluated by this functional assay. Importantly, parallel studies performed with the standard chemotherapy agent daunorubicin (DNR), showed almost no activity towards ROS-low LSC engraftment (Figure S6). Next, we also tested the anti-leukemic effect of pharmacologic BCL-2 inhibitors in vivo by treating mice engrafted with primary human AML cells with ABT-737 (50mg/kg IP for 15d). As shown in Figure 6C, ABT-737 clearly reduced leukemia burden in the treated mice. Human leukemic cells isolated from mice that received treatment with ABT-737 were transplanted into secondary recipient mice and demonstrated reduced engraftment compared to leukemic cells from control mice (mean±SEM: 43.2±5.7 vs 89.2±2.5, P<0.05)(Figure 6D). These findings indicate that in addition to reducing the bulk tumor, ABT-737 targets the LSC compartment. As previously shown (Konopleva et al., 2006), we also found that ABT-263 did not impact the in vitro survival of normal bone marrow CD34+ cells, or the clonogenic potential of normal marrow progenitors (Figure 6E–F). Further, functional analysis of total bone marrow mononuclear cells treated overnight in vitro with 250nM ABT-263 and then transplanted in NSG mice, showed no significant effect of ABT-263 on the engraftment potential of normal HSC (Figure 6G). Taken together, these results indicate BCL-2 inhibition may be an effective means to target quiescent ROS-low LSCs.
Developing effective strategies to target human LSCs has proven to be a challenging task. While a number of strategies have been proposed based on studies in preclinical models, the clinical utility of such methods remains unclear. An added difficulty that has come to light in recent years is the likely complexity and heterogeneity of the LSC population(Sarry et al., 2011). Evidence from human leukemia studies, as well as related efforts in other cancers, have suggested that the properties defining “stemness” may vary amongst individual AML patients, as well as within the same patients during the course of disease pathogenesis (Eppert et al., 2011; Sarry et al., 2011). Consequently, it has become increasingly important to determine whether any biological properties are relatively consistent in LSC populations, irrespective of disease stage or the nuances of any individual tumor population. Indeed, fundamental LSC-specific characteristics remain a highly desirable focal point for the development of improved therapies.
In the present study, we examined the properties of LSCs with respect to oxidative status and energy production. In agreement with previous reports on normal stem cells and breast CSCs (Diehn et al., 2009; Jang and Sharkis, 2007; Smith et al., 2000), we observe substantial heterogeneity in LSC oxidative state, an underlying measure of energy metabolism. However, we show that similar to normal stem cell populations (Jang and Sharkis, 2007; Smith et al., 2000), functionally defined LSCs preferentially reside in a reduced state (ROS-low). Our findings in LSC-enriched populations, and previous studies in breast CSCs indicate a low oxidative state as a potentially frequent CSC property (Diehn et al., 2009). Further, we show here that LSC-enriched populations are characterized by quiescent cell cycle status and low energy production. This relatively dormant condition likely enables LSCs to persist, even under restrictive conditions such as low nutrient or oxygen levels. Intriguingly though, in contrast to our findings in normal CD34+ and previous findings in murine HSCs, both showing that normal stem cells efficiently utilize glycolysis for energy homeostasis (Simsek et al., 2010), ROS-low LSCs appear to be deficient in their ability to employ glycolysis, and are highly reliant on oxidative phosphorylation. Glioma stem cells were also shown to be more reliant on oxidative respiration for energy generation, suggesting that lack of glycolytic activity in stem cells may be a broader characteristic of cancers (Vlashi et al., 2011). If true, then therapeutic strategies based on the assumption that tumors are preferentially reliant on glycolysis may need to be reconsidered.
Why ROS-low self-renewing LSC populations, in contrast to the bulk tumor and normal HSC (Simsek et al., 2010) are less efficient in employing glycolysis represents an important avenue of future investigation. We consider it unlikely that the LSC metabolic profile identified in our study is an artifact of ex vivo cell manipulation, since if this were the case we would expect all leukemic cells to exhibit this phenotype, instead of just the ROS-low subset. Intriguingly, it is recently proposed that cancer cell subsets with different dependencies in energy generating pathways co-exist within tumors in a symbiotic manner(Nakajima and Van Houten, 2012). In particular, cancer populations that depend on glycolysis for energy generation secrete lactate as a byproduct that is subsequently imported and used as a fuel for oxidative respiration by cancer subsets dependent on oxygen metabolism. Evaluating if this symbiotic model has relevance to the CSC model represents an intriguing line of future investigation. Further, it is possible that the paradoxical OXPHOS dependence of the ROS-low LSC-enriched subset is part of a wider metabolic adaptation of these cells, as they successfully maintain survival despite a dramatically reduced overall metabolic rate. Since oxidative respiration is a slower but significantly more efficient mechanism of energy generation, quiescent LSC subsets may benefit in the long term by using OXPHOS in the context of the tumor microenvironment, as they can more efficiently utilize the limited available nutrients.
By performing gene expression analyses and molecular studies, we show that BCL-2 is preferentially elevated in LSC enriched ROS-low cells. Although BCL-2 is commonly up-regulated in cancer, a differential role for BCL-2 in cancer stem cells has not been widely considered. Notably though, one study by Konopleva et al previously demonstrated the activity of BCL-2 inhibitor ABT-737 towards leukemia cells types, including primary AML CD34+/CD38− cells(Konopleva et al., 2006). These initial studies first established the concept of using BCL-2 inhibition as a strategy for targeting primitive leukemia cells. Our studies extend the work by Konopleva et al and provide novel insights on the role of BCL-2 in promoting LSC metabolic homeostasis. Further, by employing xenograft analyses we demonstrate here that both ABT-737 and its optimized orally bioavailable analogue ABT-263, not only decrease bulk leukemia burden, but also target functionally defined LSCs.
From a mechanistic standpoint the high levels of BCL-2 expression we observe in ROS-low cells may promote LSC survival in response to stressful stimuli such as chemotherapy and irradiation by inhibiting the mitochondrial pro-apoptotic pathway (Del Poeta et al., 2003). In addition, we demonstrate that BCL-2 mediates oxidative respiration, which our findings indicate as essential for LSC energy homeostasis (Figure 7). The newly identified role of BCL-2 in promoting mitochondrial bioenergetics offers a unique opportunity for drug targeting in the LSC-specific context, as LSC not only express high levels of BCL-2 but also have a selective dependency on oxidative respiration. In this vein, we demonstrate that the BCL-2 inhibitors ABT-737 and ABT-263 inhibit LSC oxidative respiration, which is accompanied by selective eradication of ROS-low cells. Importantly, we, and others have shown that this class of drugs does not affect normal HSCs(Konopleva et al., 2006). The identification of BCL-2 inhibitors as LSC targeting agents together with the absence of significant toxicity to normal cells is intriguing, since it provides a basis for clinical investigation of those drugs at disease stages where targeting residual LSCs is essential, as for instance in consolidation therapy or maintenance treatment during remission.
In conclusion, the concept of targeting cancer cell metabolism has emerged as an intriguing approach to the development of improved therapeutic regimens. We demonstrate here that it is feasible to eradicate resistant LSC populations by targeting their unique metabolic dependencies. Understanding further the metabolic regulation of leukemia stem cells may yield improved therapeutic strategies to eradicate this highly drug resistant tumor population.
Normal and leukemic human bone marrow samples were obtained after informed consent from volunteer donors at the University of Rochester Medical Center. Patient clinical characteristics are detailed in Table S1. Total bone marrow mononuclear cells were isolated by standard Ficoll procedures (GE Healthcare, Piscataway, NJ) and cryopreserved in freezing medium consisting of Cryostor CS10 (BioLife Solutions). The viability of leukemic cells after thawing was 50 – 90%. Normal bone marrow total mononuclear cells were further enriched for CD34 positive cells using MACS CD34 enrichment kit (Milltenyi Biotec, Auburn, CA).
Endogenous ROS levels of primary AML specimens were detected by labeling 1×106 leukemic cells for 15min at 37°C with the redox sensitive probes CM-H2DCFDA (5µM), CellRox (5µM), MitoSox (5µM), Dihydroethidium (2µM) or Mitotracker (1µM)(Life Technologies). Labeled cells were re-suspended in 0.2ml of 10uM DAPI (Invitrogen) solution and analyzed using BD LSRII flow cytometer (BD Biosciences) and FlowJo software (Treestar).
For cell seperation based on endogenous ROS levels, 1×108 cells were stained with 10µM CM-H2DCFDA or 8µM CellRox for 15min at 37°C, washed with PBS with 0.5% FBS and re-suspended in 10uM DAPI solution. Viable AML cells containing lower ROS (bottom 15% of dye distribution) or higher ROS (top 15% of dye distribution) were collected after gating in the DAPI− population by using a BD ARIAII cell sorter (San Jose, CA).
Sub-lethally irradiated NSG (NOD.Cg-Prkdcscid Il2rgtm1Wjl/SzJ) mice (Jackson Laboratories, Bar Harbor, ME) were injected with human leukemic cells via tail vein in a final volume of 0.2 mL of PBS with 0.5% FBS. Total AML: sorted ROS-low: sorted ROS-high cells were transplanted at a constant cell ratio of 5:1:1. Based on the isolation of ROS-low (and respective ROS-high) AML subset as ~15% of the bulk AML cells, this adjustment of the injected cell ratio enables transplantation of approximately equal absolute numbers of ROS-low cells in mice injected with total leukemia cells and in mice injected with purified ROS-low, so as to provide a relative estimate of the contribution of the AML ROS-low cells to the engraftment of the total AML cells.
For secondary transplantation studies of AML specimens, engrafted human cells from mice injected with total, ROS-low and ROS-high AML cells were adjusted to equal cell numbers and injected to irradiated NSG secondary recipients. In primary and secondary transplantation experiments, animals were killed after 6 to 8 weeks and bone marrow was analyzed for the presence of human cells by evaluation of human CD45 and human CD33 by flow cytometry.
02 consumption and lactate generation were measured using the Seahorse XF24 extracellular Flux analyzer as previously described(Varum et al., 2011). Briefly, 5 replicate wells of 5×105 leukemic or normal cells per well were seeded overnight in 24-well XF24 well plates coated with BD Cell-Tak (BD Biosciences) in serum-free culture medium. One hour prior to analysis the medium was replaced by un-buffered DMEM and the well plates were incubated to 37°C for pH stabilization. Analyses were performed both at basal condition, as well as after injection of Oligomycin (1µg/ml), FCCP (1µM), Antimycin A (5µM), ABT-263 (250nM–1µM) and obatoclax (10µM). ATP levels were determined according to manufacturer instructions by the ATP Bioluminescence Assay Kit HS II (Roche). The protein expression of the NOX2 (gp91phox) subunit of the NADPH oxidase was detected by flow cytometry after labeling cells with anti-NOX2 antibody (bs-3889R, Bios USA) as previously described(Hole et al., 2010).
Irradiated NSG mice were transplanted with human leukemia cells (3×106/mouse) via tail vein in a final volume of 0.2 mL of PBS with 0.5% FBS. After two weeks, animals were size matched to treatment and control groups (10 animals per group) and received either ABT-737 (50mg/kg, IP for 14 daily doses) or vehicle. ABT-737 was formulated in 30% propylene glycol, 5% Tween 80 and 65% dextrose and prepared freshly every day. Following treatment, mice were sacrificed and human engrafted cells were evaluated by flow cytometry.
The authors gratefully acknowledge critical review and comments by Drs. Fay Young, H. Leighton Grimes, Martin Carroll and Helene McMurray, and technical assistance by the Functional Genomics Center of the University of Rochester. Craig Jordan is supported by an endowed chair generously provided by Philip and Marilyn Wehrheim. Research studies described in this manuscript were supported by grants from the Leukemia and Lymphoma Society (6230-11 and 6133-12, CTJ), NYSTEM (CO24964, CTJ), and the US National Institutes of Health (HL-071158, PSB).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.