In 1956, Otto Warburg proposed that the origin of cancer cells was closely linked to a permanent respiratory defect that bypassed the Pasteur effect (i.e., the inhibition of anaerobic fermentation by oxygen). Since then, permanent defects in oxygen consumption that could explain the dependence of cancer cells on aerobic glycolysis have not been identified. Here, we show that under normoxic conditions exposure of leukemia cells to bone marrow–derived mesenchymal stromal cells (MSC) promotes accumulation of lactate in the culture medium and reduces mitochondrial membrane potential (ΔΨM) in both cell types. Notably, the consumption of glucose was not altered in cocultures, suggesting that the accumulation of lactate was the result of reduced pyruvate metabolism. Interestingly, the decrease in ΔΨM was mediated by mitochondrial uncoupling in leukemia cells and was accompanied by increased expression of uncoupling protein 2 (UCP2). HL60 cells fail to increase UCP2 expression, are not uncoupled after coculture, and do not exhibit increased aerobic glycolysis, whereas small interfering RNA–mediated suppression of UCP2 in OCI-AML3 cells reversed mitochondrial uncoupling and aerobic glycolysis elicited by MSC. Taken together, these data suggest that microenvironment activation of highly conserved mammalian UCPs may facilitate the Warburg effect in the absence of permanent respiratory impairment. [Cancer Res 2008;68(13):5198–205]
- Warburg effect
- uncoupling proteins
- mesenchymal stromal cells
In an article published in 1956, Otto Warburg presented a lucid hypothesis that stated that the respiration of cancer cells was damaged and that this resulted in their proglycolytic phenotype even in the presence of oxygen ( 1). The abrogation of the Pasteur effect (the inhibition of lactate generation in the presence of oxygen) in tumors became known as the Warburg effect, and for several decades, the search for permanent, transmissible injuries to mitochondrial respiration that could support Warburg's hypothesis has not yielded conclusive results. In hindsight, it is noteworthy that in his seminal article Warburg credited Feodor Lynen for first suggesting that uncoupling respiration to the formation of ATP could also result in the reversal of the Pasteur effect ( 1, 2), and in this scenario, respiratory damage need not be an absolute requirement. Lynen's hypothesis was based in part on the 1936 report by Ronzoni and Ehrenfest ( 3) who first showed in isolated frog muscle that the prototypical uncoupler 2,4-dinitrophenol could abrogate the Pasteur effect in a dose-dependent manner. More recent observations ( 4) have confirmed the reversal of the Pasteur effect by 2,4-dinitrophenol in human cells and have also shown that in a cell context-dependent manner, albeit this agent produces a rapid increase in oxygen consumption, chronic exposure may not result in increased demand for oxygen. This key finding suggests that mitochondrial uncoupling could indeed mimic the Warburg effect in the absence of permanent, transmissible alterations to the oxidative capacity of cells.
Interestingly, mitochondrial uncoupling occurs physiologically during cold acclimation in mammals and is mediated by uncoupling proteins (UCP; refs. 5, 6). UCPs are mitochondrial inner membrane proteins that short circuit the electrochemical gradient created by the mitochondrial respiratory chain by sustaining an inducible proton conductance ( 6). UCP1 was the first UCP identified and shown to play a role in energy dissipation as heat in mammalian brown fat ( 5, 7). Additional UCPs have been identified in humans (UCP2, UCP3, and UCP4), and although all promote partial mitochondrial uncoupling, their varied tissue distribution and regulation suggests that they serve distinct physiologic roles possibly unrelated to maintenance of core body temperature ( 6, 8). For instance, recent work has shown that UCP2 can decrease reactive oxygen species (ROS) production in mitochondria, protect against cell death, and promote microenvironmental thermogenesis in the central nervous system ( 9). Interestingly, UCP2 differs from UCP1 in that it requires an activation step to sustain proton conductance, and it may not conduct protons per se, but rather fatty acid anions, albeit the nature of the proton conductance of UCP2 remains controversial ( 10, 11).
Gatenby and Gawlinski ( 12) have recently suggested that the proglycolytic phenotype of solid tumors is a result of environmental selection and imparts tumor cells with survival advantages in the hypoxic and inflammatory conditions present in premalignant lesions. Hypoxia alone can also promote glycolysis via the activation of hypoxia-inducible factor-1α ( 13); however, recent evidence suggests that, in the absence of growth factor signaling, hematopoietic cells fail to express this transcription factor ( 14). This key finding indicates that, at least for hematopoietic cells, additional microenvironmental cues are essential for the establishment of the proglycolytic metabolism. Because the bone marrow microenvironment is a major contributor to the pathogenesis of leukemia ( 15– 17), we decided to investigate if bone marrow–derived mesenchymal stromal cells (MSC) could activate aerobic glycolysis in leukemia cells. Our investigations revealed that exposure of leukemia cells to MSC resulted in increased accumulation of lactate in the culture medium suggestive of the Warburg effect. Most notably, the increased proglycolytic phenotype was accompanied by decreased mitochondrial membrane potential (ΔΨM) in the leukemic cells, a result of mitochondrial uncoupling promoted by the activation of a highly conserved, energy dissipating protein, UCP2. These results are the first to show that MSCs promote the Warburg effect in leukemia cells and that this phenomenon is mediated in part via activation of mitochondrial UCPs.
Materials and Methods
Cell Lines, Chemicals, and Biochemicals
OCI-AML3, MOLM13, HL60, TF1-RAS, and TF1-SRC cells were maintained in RPMI 1640 supplemented with 5% FCS, 1% glutamine, and 100 units/mL penicillin in a 37°C incubator containing 5% CO2. TF1-RAS and TF1-SRC are stable clones of the granulocyte macrophage colony-stimulating factor–dependent cell line TF1 that express h-RAS or v-SRC oncogenes and are growth factor independent ( 18). All experiments unless stated otherwise were carried out at a cell density of 250,000 to 500,000/mL. The Amplex Red Glucose/Glucose Oxidase kit, 2-NBDG, and JC-1 were obtained from Molecular Probes. Total OXPHOS detection antibody kit was obtained from Mitosciences. UCP antibodies ( 1– 4) were obtained from Millipore. Oligomycin and carbonyl cyanide chlorophenylhydrazone (CCCP) were obtained from Sigma. MSCs were obtained from bone marrow samples and cultured at a density of 1 × 104 to 5 × 104/cm2 in Mesenpro medium (Invitrogen) until seeded as feeder layers at 5 × 104/1.9 cells per well in a 24-well plate in RPMI 1640 16 h before addition of 2.5 × 105 (TF1-RAS and TF1-SRC) or 5 × 105 (OCI-AML3, MOLM13, and HL60) leukemia cells in 1 mL of fresh RPMI 1640. For all experiments, the viability of cell cultures was quantitated in a Vi-Cell Cell Viability Analyzer (Beckman Coulter). Only cultures displaying >90% viability were used for further metabolic measurements.
Measurement of Lactate Generation and Oxygen Consumption
Lactate generation and polarographic measurements of oxygen consumption were carried out as previously described ( 19). Fluorometric oxygen measurements using the BD Oxygen Biosensor plates were carried out as previously described ( 20).
Measurement of Glucose Levels in the Medium and 2-NBDG Uptake
Glucose levels. Glucose in the medium was quantitated using the Amplex Red Glucose/Glucose Oxidase kit (Invitrogen) using a standard curve prepared with serial dilutions of RPMI 1640 (11 mmol/L glucose) into glucose-free RPMI 1640. Fluorescence was read using a Fluostar Optima microplate reader (BMG Labtech), and results were expressed as pmol/cell.
2-NBDG uptake. 2-NBDG (100 μmol/L) was added to cultures 30 min before harvesting and incubated at 37°C. Leukemia-MSC cocultures were then harvested by trypsinization followed by one wash in cold PBS. Cells were then resuspended in ice-cold PBS containing 1:100 anti-CD90 allophycocyanin (APC)-conjugated antibody (BD Biosciences) and incubated on ice for 10 min. Samples were washed once with ice-cold PBS and analyzed by flow cytometry in a FACSCalibur flow cytometer using a 488-nm argon and a 633-nm HeNe laser. The proportion of CD90-positive or negative cells (MSC) was quantitated as % cells, and the 2-NBDG uptake was presented as mean fluorescence intensity.
Ratiometric Measurement of Mitochondrial Membrane Potential
A solution of 5 mg/mL of JC-1 in formamide was diluted 1:100 in prewarmed medium with vigorous vortexing and further diluted 1:100 directly on cell cultures 15 min before harvesting. Leukemia-MSC cocultures were harvested by trypsinization followed by one wash in PBS. Leukemia-only cultures were similarly processed. After collection, cells were washed once in PBS and incubated for 5 to 10 min at room temperature with anti-CD90 APC-conjugated antibody in PBS. ΔΨM was then quantitated by flow cytometry as previously described ( 21).
Western Blot Analysis
After appropriate treatments, MSC-leukemia cocultures were collected by trypsinization, washed twice in PBS, and then resuspended in PBS with 1:100 of anti-CD90 APC-conjugated antibody. Leukemia-only cultures were identically processed. MSC and leukemia cells were then isolated by fluorescence-activated cell sorting (FACS) and collected in PBS in a chilled FACSAria (Becton Dickinson). Cell extracts were then generated and immunoblotted as previously described ( 19).
Small Interfering RNA Transfection
Silencing of UCP2 gene expression in leukemic cells was achieved by the small interfering RNA (siRNA) technique. ON-TARGETplus SMARTpool human UCP2 short interfering RNAs were obtained from Dharmacon. Nonspecific control pool containing four pooled nonspecific siRNA duplexes was also used as a negative control. Transfection of leukemic cells was carried out by electroporation using the Nucleofection System (Amaxa), as previously described ( 22).
Leukemia-MSC cocultures exhibit increased aerobic glycolysis. Our investigation started by monitoring the accumulation of lactate in the medium of leukemia cell lines cultured alone or on a feeder layer of primary bone marrow–derived MSC under normoxic conditions. For all of our experiments, leukemia cells (2.5 × 105 to 5.0 × 105/mL) were cultured with feeder layers of MSC (5 × 104/mL) for 48 h. Exposure of TF1-RAS, TF1-SRC, OCI-AML3, and MOLM13 leukemia cell lines to a MSC feeder layer resulted in increased accumulation of lactate in the culture medium, whereas no increase was seen in HL60-MSC cocultures ( Fig. 1A ). Although the source of the lactate cannot be determined a priori due to the presence of two cell types, the results in Fig. 1B indicate that, under our experimental conditions, MSC feeder layers generate <1 mmol/L, suggesting that a very large increase in lactate generation from these cells would be necessary to account for the observed differences. It is noteworthy that, under these experimental conditions, TF1-RAS and TF1-SRC, but not any of the other cell lines, proliferated slightly more when in coculture with MSC, whereas MSC proliferation was moderately inhibited by all leukemia cell lines tested (data not shown). Notably, monitoring glucose levels in the culture medium revealed no differences in the amount of glucose consumed by leukemia-MSC cocultures ( Fig. 1C), and likewise, monitoring the uptake of the fluorescent glucose derivative 2-NBDG by flow cytometry shows that neither OCI-AML3 nor TF1-RAS cells markedly altered their uptake of 2-NBDG while grown on MSC feeder layers ( Fig. 1D). Similar observations were made in TF1-SRC and MOLM13 cells (data not shown). In contrast, as shown in Fig. 1D, hypoxia significantly increased 2-NBDG uptake in OCI-AML3 cells cultured alone and markedly enhanced the uptake in both cell types cultured with MSC, suggesting that under our experimental conditions limiting oxygen concentrations, but not coculture, increase the entry of glucose into leukemic cells. In addition, the uptake of 2-NBDG by MSCs was moderately reduced in coculture with TF1-RAS but unaffected in OCI-AML3 coculture (data not shown). These observations suggest that the increase in lactate levels in leukemia-MSC coculture reflects a decrease in the amount of pyruvate that enters the Krebs cycle rather than an increased use of glucose.
MSCs induce mitochondrial uncoupling in leukemia cells. Because our results indicate that the mitochondrial metabolism of pyruvate may be compromised in leukemia cocultures, we decided to investigate mitochondrial membrane potential (ΔΨM) by flow cytometry using the ratiometric probe JC-1. As shown in Fig. 2A , OCI-AML3 cells show significantly decreased ΔΨM after 24 h of coculture with MSC, suggesting that their mitochondria either have decreased proton extrusion capacity or have increased their proton conductance. In contrast, HL60 cells do not markedly alter their ΔΨM when in coculture ( Fig. 2B), albeit as shown in Fig. 2C both HL60 and OCI-AML3 cells induce depolarization of MSC mitochondria. This key observation further suggests that although MSC mitochondrial function is altered by exposure to leukemia cells, it is the decreased entry of pyruvate into the Krebs cycle of leukemia cells that may account in large part for the increased generation of lactate in coculture supernatants.
To determine if a generalized defect in respiration is the underlying cause for the decrease in ΔΨM in leukemia cells in coculture, we monitored oxygen consumption fluorometrically in a 96-well plate format using the BD Oxygen Biosensor System. As shown in Fig. 3A , oxygen consumption in OCI-AML3 cocultures was nearly identical to oxygen consumption in leukemia cells alone, suggesting that the decrease in ΔΨM was not a result of reduced oxygen consumption. Similar results were obtained using polarographic oxygen consumption techniques ( Fig. 3B). As was the case for lactate generation under our experimental conditions, MSC feeder layers alone consumed negligible amounts of oxygen. Additionally, as shown in Fig. 3C, there was no difference in expression of oxidative phosphorylation complexes in OCI-AML3 cells cocultured for 48 h with MSC, suggesting that mitochondrial mass is not affected. Because the above observations suggest that leukemia cells maintain the ability to reduce oxygen, we postulated that the loss of ΔΨM could be a result of increased proton leakage (i.e., uncoupling). To test if leukemia mitochondria may be uncoupled in coculture, we examined the response of ΔΨM to oligomycin. Oligomycin binds to the F0 subunits 9 and 6 of complex V (ATP synthase), resulting in intramitochondrial ATP hydrolysis, uncoupled proton flux, and decreased ΔΨM ( 23, 24). As shown in Fig. 4A , OCI-AML3 cells cultured on MSC feeder layers are more resistant to the loss of ΔΨM in response to oligomycin treatment, suggesting that their mitochondria are already partially uncoupled. To then determine the kinetics of mitochondrial uncoupling in leukemia cocultures, the loss in ΔΨM in OCI-AML3 cells was evaluated 15, 30, 60, 120, and 180 min after exposure to MSC feeder layers. Our results show that ΔΨM is decreased in a time-dependent manner as early as 30 min after coculture ( Fig. 4B). Under the same experimental conditions, ΔΨM in HL60 cells was not decreased (data not shown). Interestingly, monitoring the use of oxygen revealed that cocultures of OCI-AML3, but not HL60, cells displayed a transient increased demand for oxygen 3 and 5 h after coculture that was no longer significant at 24 h ( Fig. 4C). Taken together, the above results suggest that OCI-AML3 mitochondria become rapidly uncoupled after exposure to MSC feeder layers, and this results in loss of ΔΨM with an accompanying transient increase in oxygen consumption.
UCP2 contributes to mitochondrial uncoupling in leukemia cells cocultured with MSC. To further understand the underlining mechanism by which MSCs induce mitochondrial uncoupling in leukemia cells, we investigated the expression of the energy dissipating proteins UCP1, UCP2, UCP3, and UCP4, all of which possess varying degrees of uncoupling function by short circuiting the mitochondrial electrochemical gradient. The results in Fig. 4D show that MSC feeder layers promoted the activation of UCP2 expression in OCI-AML3, but not of HL60, cells, and whereas expression of UCP1, UCP3, or UCP4 was not detected in any cell line investigated, TF1-RAS and TF1-SRC cells also increased the expression of UCP2 (data not shown). Most notably, the expression of UCP2 increased as early as 30 min after exposure of OCI-AML3 cells to MSC feeder layers ( Fig. 5A ), albeit maximal expression levels did not occur until 120 min in coculture. To then determine the contribution of UCP2 to the observed loss of ΔΨM, we decreased UCP2 expression by siRNA methodology. Albeit leukemia cells are notoriously difficult to transfect, using nucleofector technology, we achieved 33% decrease in the levels of UCP2 in OCI-AML3 cells in coculture, and this reduction was associated with a 22% protection from the loss of ΔΨM observed in coculture ( Fig. 5B). Similarly, we observed that lactate accumulation was diminished in OCI-AML3 cells treated with UCP2 siRNA, albeit our results failed to reach statistical significance ( Fig. 5C). Nonetheless, treatment with the protonophore CCCP (2 μmol/L) promoted the accumulation of lactate in the culture medium of OCI-AML3 cells, suggesting that uncoupling per se can indeed contribute to aerobic glycolysis ( Fig. 5D). Interestingly, inhibition of protein synthesis with cycloheximide in OCI-AML3 did not prevent the loss of ΔΨM observed after a 120-min exposure to MSC feeder layers ( Fig. 6A ), and exposure of OCI-AML3 cells to MSC feeder layers through a 0.4-μm filter also resulted in significantly decreased ΔΨM ( Fig. 6B), albeit this decrease was not as pronounced as the decrease seen in OCI-AML3 cells allowed to directly contact MSC feeder layers. Taken together, the above results suggest that UCP2 expression partly contributes to the observed loss of ΔΨM, but that this contribution may occur in the absence of de novo protein synthesis and may be initiated by paracrine effectors.
Therapeutic relevance of mitochondrial uncoupling in leukemia cells. Given the antiapoptotic and chemoprotective role of UCP2 and the other UCP family members, we questioned whether uncoupling per se could be chemoprotective. To test this hypothesis, we exposed OCI-AML3 cells to the protonophore uncoupler CCCP (1 and 2.5 μmol/L) followed by treatment with mitoxantrone (120 nmol/L) for 24 h. As presented in Fig. 6C, CCCP in a dose-dependent manner antagonized the proapoptotic effects of mitoxantrone, significantly (P < 0.01) protecting OCI-AML3 cells at 2.5 μmol/L. Similarly, exposure of OCI-AML3 cells to MSC feeder layers protected OCI-AML3 cells against the proapoptotic effects of mitoxantrone, cytosine arabinoside (AraC), and vincristine, significantly increasing the EC50 for each drug ( Fig. 6D; Supplementary Table S1). In contrast, HL60 cells grown on MSC feeder layers were not protected against the cytotoxic effects of mitoxantrone (Supplementary Fig. S1A) and similar observations were made using AraC (data not shown). Interestingly, MSC feeder layers did not prevent the decrease in the number of viable cells induced by either mitoxantrone or AraC and only mildly antagonized the growth-inhibitory effects of vincristine (Supplementary Fig. S1B), suggesting that MSC exposure does not modulate the growth-inhibitory effects of traditional chemotherapeutic drugs but rather the apoptotic response to them. Taken together, the above observations indicate that the antiapoptotic effects of stromal cocultures may be partly mediated by mitochondrial uncoupling, and support the notion that UCP2 may inhibit the activation of the mitochondrial permeability transition, and therefore the release of proapoptotic factors, by depolarizing the mitochondria leading to decrease uptake of calcium and decrease ROS generation ( 9, 25).
The precise mitochondrial events that contribute to the proglycolytic phenotype of cancer cells are not fully understood, but Warburg hypothesized that a profound injury to respiration was the underlying cause. Nonetheless, in a seminal article published in 1956, Warburg briefly discussed the possibility, first suggested by Feodor Lynen 14 years earlier, that in cancer cells the coupling of respiration to the synthesis of ATP could be broken (i.e., mitochondria in cancer cells could be uncoupled). On one hand, this hypothesis suggests that in the absence of permanent mitochondrial damage, cancer cells may short circuit the electrochemical gradient leading to intramitochondrial ATP hydrolysis and decreased flux of pyruvate through the Krebs cycle, resulting in the eventual accumulation of lactate. On the other hand, there is an accumulating wealth of evidence that suggests that microenvironmental cues (hypoxia, acidity, etc.) may promote the proglycolytic metabolism of tumors. Could the leukemia microenvironment promote the Warburg effect by activating mitochondrial uncoupling?
Our investigation suggests that the leukemia microenvironment, modeled by a feeder layer of MSC, can indeed promote aerobic glycolysis without a concomitant increase in glucose use; in contrast, the consumption of glucose was enhanced in cocultures under hypoxic conditions. In addition, it was observed that the accumulation of lactate was associated with a loss of ΔΨM in leukemia cells, suggesting that mitochondrial function was compromised. Although we expected that the decreased ΔΨM was the result of diminished oxygen consumption, our results indicated instead that leukemia cell mitochondria had become uncoupled, as evidenced by decreased sensitivity to the ΔΨM dissipating effects of oligomycin and a transient increase in oxygen consumption. Most surprisingly, uncoupling was mediated at least in part by the highly conserved energy dissipating protein UCP2. Mechanistically, uncoupling occurred rapidly after exposure to MSC feeder layers, was not dependent on de novo protein synthesis, and did not necessitate cell-to-cell contact, suggesting that paracrine effectors may activate basal levels of UCP2 and/or additional ΔΨM dissipating elements that orchestrate the short circuiting of the electrochemical gradient.
Our observation that oxygen is still being reduced, although pyruvate entry into the Krebs cycle is decreased, suggests that the MSCs promote a metabolic shift to the oxidation of other carbon sources. Recent evidence suggests that fatty acids can activate mitochondrial uncoupling, and conversely, mitochondrial uncoupling can promote fatty acid oxidation ( 26– 28). Moreover, a recent report shows that UCP2 promotes fatty acid oxidation while limiting the mitochondrial oxidation of pyruvate in mouse embryonic fibroblasts ( 29), and similarly, Harper and colleagues ( 30) suggested a link between increased fatty acid metabolism and reduced ΔΨM, suggesting that a decrease in the electrochemical gradient of the mitochondria may represent a metabolic shift to higher yield bioenergetic substrates. A similar metabolic shift to the use of fatty acids is promoted by leptin ( 31), a major cytokine regulator of UCP2 activity ( 32, 33). We have shown that leptin plays a role in stroma-mediated chemoprotection ( 34– 36) and that the leptin receptor is expressed in primary acute myeloid leukemia blasts and primary acute promyelocytic leukemia cells ( 34– 36). It is thus intriguing to propose that leptin may be one of the paracrine effectors that orchestrate mitochondrial uncoupling in a physiologic context.
Our data suggest that uncoupling per se can be chemoprotective, presumably by increasing the threshold to mitochondrial permeability transition and apoptosis, as has been reported for UCP2 ( 9), and agree with an earlier study that suggested an association of UCP2 and decreased ΔΨM with resistance to radiation-induced cytotoxicity of mouse lymphoblastoid cells ( 30). This antiapoptotic effect could be mediated by several factors, such as decreased mitochondrial calcium uptake ( 37), decreased generation of ROS ( 38), and potentially decreased accumulation of positively charged toxic lipophilic species in the more cationic environment of the depolarized mitochondria ( 39). More importantly, however, our observations suggest that targeting fatty acid metabolism and/or UCPs may offer clinical benefit in the therapeutic treatment of leukemias.
In conclusion, although the microenvironment has been shown to play a critical role in the pathogenesis of leukemia in terms of chemoprotection and maintenance of malignant cells ( 15, 40), our report is the first to show that it also plays a role in the promotion and/or maintenance of the Warburg effect in leukemias. Most importantly, these observations uncover an intriguing pathway activated by MSC that regulates the electrochemical gradient in leukemic mitochondria. This pathway may constitute a novel target for therapeutic intervention in hematologic malignancies.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: National Cancer Institute grant PO1 CA55164 and the Paul and Mary Haas Chair in Genetics (M. Andreeff).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We thank Leslie Calvert for laboratory assistance, Elena Vess for administrative assistance, and Drs. Marina Konopleva and Yoko Tabe for insightful discussions.
Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).
- Received February 13, 2008.
- Revision received April 17, 2008.
- Accepted May 6, 2008.
- ©2008 American Association for Cancer Research.