close
2-DG Research

The wonders of 2-deoxy-d-glucose

print
by Haibin Xi, Metin Kurtoglu, and Theodore J. Lampidis

Introduction

Although results from countless positron emission tomography (PET) scans have repeatedly indicated that tumors from a variety of cancer cell types accumulate more glucose than surrounding normal tissues, it was not until molecular biological evidence was available before it became recognized that increased glucose metabolism is a hallmark of cancer. Thus, most of the major oncogenic events have now been shown to have direct or indirect mechanisms by which they control increased glycolysis in tumor cells under normal oxygen conditions (commonly known as aerobic glycolysis), an observation made by Warburg more than 80 years ago [1]. 2-Deoxy-d-glucose (2-DG) has been found to be uniquely suited to exploit this universal trait as it has been demonstrated to be a powerful agent for blocking and probing increased sugar metabolism in cancer cells [2-5] (Fig. 1). Taking advantage of the increased glucose uptake that occurs in most tumors provides a specificity window that increases the selectivity of this analog [3, 5-8].

Figure 1.

2-DG exploits a universal trait of cancer. Molecular evidence clearly implicates a majority of the oncogenic events required to convert a normal cell to a tumor cell converging on the upregulation of glucose uptake and metabolism, which in turn sustains the malignant phenotype. For example, it supports uncontrolled proliferation by providing the nucleotide building blocks via the PPS, reprograms energy metabolism by aerobic glycolysis, and maintains ER homeostasis through proper glycosylation, and so forth. By mimicking glucose, 2-DG offers a natural means of exploiting increased glucose metabolism, a universal trait of cancer. However, as under normoxia other energy sources can be used to fuel cells and sustain their viability, 2-DG’s toxicity is more limited to tumor cells under hypoxia. Although under normoxia most tumor cells are growth inhibited by 2-DG treatment, we as well as others have found that a select number of tumor cell types in this environmental condition undergo ER stress-induced cell death. Therefore, 2-DG holds much promise for treating most if not all cancer types. Oncogenic events are indicated by rectangles and metabolic enzymes by circles. Abbreviations: 2-DG, 2-deoxy-d-glucose; 2-DG-6-P, 2-deoxy-d-glucose-6-phosphate; ER, endoplasmic reticulum; F-6-P, fructose-6-phosphate; G-6-P, glucose-6-phosphate; HK, hexokinase; M-6-P, mannose-6-phosphate; PGI, phosphoglucose isomerase; PMI, phosphomannose isomerase; PPS, pentose phosphate shunt.

Because of its similarity to glucose, 2-DG inhibits glycolysis, but as chemically it is also 2-deoxymannose it is able to compete with mannose in the growing lipid-linked oligosaccharide chain during the initial steps of N-linked glycosylation. This mannose-like property of 2-DG results in misfolded proteins leading to endoplasmic reticulum (ER) stress [3]. Thus, 2-DG is a compound that not only naturally exploits increased glucose uptake in tumors but by its structure can induce energy as well as ER stress. These fortuitous attributes have enabled 2-DG to be used to selectively target most tumor cell types growing under normoxia as well as hypoxia and the results of some of the more important findings with this sugar analog as they relate to cancer will be considered in this review.

A Natural Window of Selectivity Exists for Eradicating Bacterial Diseases but not Cancer

Modern medicine is able to successfully treat bacterial diseases with chemotherapy (antibiotics) as prokaryotic cells (bacteria) are significantly different than normal eukaryotic cells. Thus, a window of selectivity exists that can be exploited with an antibiotic such as penicillin which attacks a biological target on bacteria that does not exist in eukaryotic cells. Unfortunately, when applying the same concept to cancer the window of selectivity naturally closes as normal (eukaryotic) cells give rise to malignant (eukaryotic) cells with both cell types sharing many of the same biological reactions.

Based on these realities, it is well-established that classical chemotherapy does not select so much between cancer and normal cells as it does between rapidly dividing and slow-dividing cells. Thus, the most rapidly dividing normal cells in the body, that is, hair follicles, blood cells, and stomach/intestinal cells, are those that are invariably most affected by chemotherapy. It follows that a major problem in cancer chemotherapy is that within most, if not every, solid tumor there are populations of cells residing within low oxygen (hypoxia) environments that are slow-dividing and consequently resistant to conventional treatment. Additionally, these cell populations will also give rise to re-growth of the rapidly dividing cells found in the normoxic locations which are initially killed by chemotherapy, accounting for the incomplete eradication of the tumor. Moreover, cancer cells surviving in hypoxia have been identified as those that have high metastatic potential [9].

Hypoxia opens the Window of Selectivity for 2-DG

Glycolysis has survived a billion years of evolution dating back to a time when there was no oxygen in the atmosphere and the only means by which microbial life could sustain itself was through the breakdown of sugar (glycolysis). Applying this fundamental tenet to cancer, our interests in investigating the use of the glycolytic inhibitor 2-DG as a potential clinical agent to target chemo-resistant hypoxic cancer cells began in the late 1990s. This idea was linked to previous studies in the early 1980s in which in collaboration with Lan Bo Chen and his group we had shown that combining 2-DG with the mitochondrial uncoupler rhodamine 123 was effective in killing carcinoma cells in vitro[10], and lowering tumor burden in vivo [11]. However, knowing that rhodamine 123 could potentially sensitize normal cells to glycolytic inhibitors such as 2-DG dampened the enthusiasm for developing this work for clinical application. It was the realization that the hypoxic areas in solid tumors harbored the slow-growing tumor cell populations that were resistant to radio- and chemo-therapy, and that hypoxia (without the need to use rhodamine 123) forced cells to rely on glycolysis for survival, that sparked our interest in targeting them with 2-DG [8]. At this time, most of the literature on this sugar analog was related to its use as an indicator (tritiated 2-DG) of glucose uptake. There were however reports by Jain and colleagues beginning in the late 1980s that were focused on administering 2-DG as a single high dose prior to radiation to increase the latter’s efficacy in the treatment of patients with brain cancer [12-14]. Although there had been other sporadic reports of the use of 2-DG in cancer cells as well as to treat cancer patients, the results were not impressive enough for attracting more interests in its further development [15-19].

Nonetheless, using three tumor cell models of simulated hypoxia, we were able to clearly demonstrate in vitro that blocking glycolysis with 2-DG led to tumor cell death [6, 7, 20]. In contrast, tumor cells treated similarly, but under normoxia, survived. These latter results could be explained by the fact that under normoxia cells take up less 2-DG than when in hypoxic conditions and or that under normoxia even if glucose utilization is blocked by 2-DG, other energy sources such as fats and proteins can be used to produce ATP via oxidative phosphorylation. In contrast, under hypoxia tumor cells depend on glucose as their main energy source and blocking glucose metabolism with 2-DG essentially starves them to death. The strategy of inhibiting glycolysis and thereby targeting the chemotherapy-resistant hypoxic tumor cells was further supported by in vivo results which showed increased treatment efficacy when 2-DG was combined with adriamycin and paclitaxel in human xenograft mouse models [21]. Moreover, in collaboration with Dr. Timothy Murray’s lab, in a transgenic model of retinoblastoma using the hypoxic indicator piminidazole, the first proof of principle that 2-DG kills hypoxic tumor cells in vivo was provided (Fig. 2 as reproduced and modified from ref. 22). Additionally, it was shown that although tumors were reduced by either agent alone, elimination of hypoxic areas within the tumors was only observed when animals were treated with 2-DG [22]. This latter result was in accord with the generally accepted understanding that chemotherapy is unable to effectively target hypoxic cells. Definitive proof that 2-DG can eliminate hypoxic tumor cells in vivo further strengthens the rationale for developing 2-DG clinically.

Figure 2.

Proof of principle demonstrating that 2-DG targets hypoxic tumor cells in vivo and significantly enhances tumor control. (A) Sixteen-week-old LHBETATAG mice, a well-established transgenic model of the childhood cancer retinoblastoma, were treated with saline, 2-DG alone (500 mg/kg), low dose carboplatin alone (31 µg), or 2-DG plus low dose carboplatin. The graph shows the decrease in tumor burden as expressed by the ratio of tumor area to globe area. The differences in tumor burden between the saline-treated group and any of the drug-treated groups are statistically significant (saline vs. 2-DG alone, P = 0.001; saline vs. carboplatin alone, P = 0.001; saline vs. 2-DG plus carboplatin, P = 0.007). (B) Representative H&E images of each group at ×40 magnification. Note that uninvolved retina appears normal, and no signs of retinal toxicity are evident. (C) Mice from each treatment group were perfused with pimonidazole to label hypoxic areas, and right eyes were sectioned and stained. Representative images of pimonidazole (green)- and DAPI (blue)-stained ocular sections are shown at ×200 magnification. Significant areas of hypoxia can be seen in the saline-treated group (arrows), with little to no reduction in these areas in the carboplatin mono-therapy group (arrows). No areas of hypoxia can be seen in the 2-DG monotherapy group. Marked reduction of tumor burden and hypoxic cells can be seen in the combination group. Abbreviations: 2-DG, 2-deoxy-d-glucose; Carbo, carboplatin. This figure is adapted from ref. 22 with the permission of the copyright holder, Association for Research in Vision and Ophthalmology.

2-Fluoro-Deoxyglucose More Closely Resembles Glucose Than 2-DG

Since the early 1950s, 2-DG has been recognized as an antagonist of glucose metabolism [23]. Like glucose, 2-DG was demonstrated to be taken up through the glucose transporters (GLUTs) and phosphorylated by hexokinase (HK) to form 2-DG-6-phosphate (2-DG-6-P). In contrast to glucose-6-phosphate (G-6-P) which progresses through the glycolytic pathway, 2-DG-6-P accumulates within the cell and is not metabolized further [23-25]. In studies in which isolated enzymatic studies were performed, it was shown that in the subsequent reversible glycolytic step through which G-6-P is converted to fructose-6-phosphate (F-6-P), 2-DG-6-P was found to compete with the former for phosphoglucose isomerase (PGI) [16, 25]. Thus, 2-DG was thought to block glycolysis primarily by competitively inhibiting PGI [23, 25]. A secondary metabolic effect of 2-DG was shown to be noncompetitive inhibition of HK by 2-DG-6-P [26, 27]. As the Ki for 2-DG-6-P inhibiting HK is well in excess of that for G-6-P, it was concluded by another group that 2-DG-6-P did not inhibit HK [24]. It is likely, however, that treatment of whole cells with 2-DG (in contrast to the studies done in isolated enzymes) leads to a buildup of 2-DG-6-P to concentrations capable of HK inhibition. This would explain results where it was shown that cells treated with 2-DG had lowered intracellular levels of G-6-P [28].

In our previous review comparing the biologic effects of 2-DG and 2-fluoro-deoxy-glucose (2-FDG) [3], we noted that 2-DG’s anti-glycolytic activity can be increased by replacing the hydrogen in the 2′ carbon with fluorine giving rise to 2-FDG. Because of the closer conformational/electronic similarity of a fluorine than a hydrogen atom to a hydroxyl group, 2-FDG more closely resembles glucose than 2-DG [29]. Better recognition of 2-FDG, as compared to 2-DG, by HK and most likely by PGI, leads to two- to threefold greater inhibition of glycolysis, as indicated by reduction in lactate as well as ATP corresponding with greater death in anaerobic metabolizing (hypoxic) tumor cells [29]. Although the comparison of 2-FDG to 2-DG was not directly tested in xenograft models, 2-FDG was more recently reported to be effective in selectively killing hypoxic cells in an in vivo transgenic model of retinoblastoma [30]. Overall, in vitro and in vivo results indicate that in addition to 2-DG, 2-FDG may also be a promising candidate to be developed clinically for selectively targeting hypoxic tumor cells.

2-DG is Not Solely a Glucose Mimetic

The importance of the hydroxyl group on the 2′ carbon of hexose molecules in determining how they are metabolized is evident when considering mannose, a natural isomer of glucose. Mannose has the same exact composition as glucose except it has its hydroxyl group on the 2′ carbon in the reverse position, which renders it a substrate for N-linked glycosylation of proteins rather than as an energy source. Thus, how the 2′ carbon is altered on the glucose molecule dictates its metabolic pathway and primary use within a cell.

Although 2-DG has generally been thought of as a glucose mimetic and many of its effects have been attributed to its interference with glucose metabolism, in the 1970s Schwarz and colleagues, studying various hexose analogs demonstrated that 2-DG has strong activity as an analog of mannose in interfering with N-linked glycosylation [31-34]. In the process of N-linked glycosylation, it has been shown that 2-DG competes with mannose for assembly of lipid-linked oligosaccharides (LLO) that are precursors of the sugar moiety in glycoproteins. LLOs are synthesized from stepwise transfer of mannose from its covalently linked carrier (GDP or dolichol) onto N-acetyl-glucosamine or mannose residues. It was demonstrated that 2-DG undergoes conversion to 2-DG-GDP or 2-DG-dolichol and competes fraudulently with mannose [35]. In contrast, the fluorine group in 2-FDG restricts it from resembling and therefore competing with mannose. 2-FDG can still however interfere with N-linked glycosylation by competing for glucose moieties that are added at the end of the LLO assembly [34]. This latter activity notwithstanding, 2-FDG has been demonstrated to be a significantly weaker glycosylation inhibitor as compared with 2-DG [31, 33, 34]. Thus, 2-FDG has been useful for probing whether a tumor cell is more vulnerable to glycolytic inhibition rather than interference with N-linked glycosylation. Moreover, it has helped in determining whether the effect of 2-DG is related to its anti-glucose or anti-mannose mechanism by repeating the same experiments with 2-FDG. If an assayed biologic effect is weaker with 2-FDG than with 2-DG, then this indicates that the mechanism by which 2-DG is causing its effect is through interference with N-linked glycosylation. Additionally, we have demonstrated that if the biologic effects of 2-DG in question can be reversed by low doses of mannose, additional information is provided indicating that the mechanism by which 2-DG is acting is through interference with N-linked glycosylation [3]. This strategy has now been used in recent studies by other groups to demonstrate 2-DG’s activity as an ER stressor [36-38]. In serum, glucose concentrations range between 4 and 6 mM, while mannose concentrations are usually at 100 µM or below [39]. Stoichiometrically, it follows that at 1–2 mM, 2-DG should predominantly interfere with mannose metabolism and at doses above 5–6 mM it should effectively inhibit the utilization of both sugar isomers.

Under normal oxygen conditions, 2-DG at low doses has been found to inhibit growth without causing cell death in most carcinoma cell lines. In the few select carcinoma cell lines as well as in leukemic and alveolar rhabdomyosarcoma cell lines where low dose 2-DG was found to induce cell death even in the presence of oxygen, its toxic effect was shown not to be due primarily to inhibition of glycolysis and lowering of ATP but rather, to interference with N-linked glycosylation resulting in accumulation of unfolded proteins and consequently endoplasmic reticulum (ER) stress [2, 36, 40]. The lowering of fully formed oligosaccharides required for N-linked glycosylation and the increase in ER stress markers induced by 2-DG in these cell lines were reversed by exogenous addition of low doses of mannose (<25% of 2-DG dose) which was accompanied by complete rescue from cell death [2, 36]. Moreover, in the few carcinoma cell lines in which 2-DG was toxic, they were found to be much less or completely insensitive to equivalent concentrations of 2-FDG further verifying the mechanism by which 2-DG was conferring its toxic effects was by interference with N-linked glycosylation. The ensuing unfolded protein response (UPR) to ER stress is similar to the action of p53 in response to DNA damage. Both processes act to protect the fidelity of proteins and DNA, respectively, by buying time for the cell to repair these distinct types of damage. At the same time, if these efforts are not successful apoptosis ensues via specific UPR- or p53-mediated pathways removing unsalvageable cells. Why a few select carcinoma cell lines undergo cell death under normoxic conditions even though they mount an UPR remains unknown. Investigations to understand the unusual vulnerability of these cell types to ER stress when treated with 2-DG under normoxia may lead to means by which 2-DG’s growth inhibition in most cell types can be converted to cytotoxicity which will be discussed below.

The fact that 2-DG interferes with both mannose and glucose metabolism raises the question of how it suppresses cell growth in most cell types under normoxia. Previously, it was assumed that growth inhibition is due to 2-DG’s activity as a glucose mimetic blocking glycolysis and thereby depleting intracellular ATP. Low ATP induces energy stress whereby the liver kinase B1 (LKB1)/AMP-activated protein kinase (AMPK) pathway signals mammalian target of rapamycin (mTOR) to stop cap-dependent protein synthesis which results in growth arrest [41]. However, data presented recently indicate that 2-DG-induced growth inhibition is mediated primarily by its ability to cause ER stress leading to attenuation of growth via the UPR [42] rather than by blocking glycolysis and causing energy stress.

It is known that when ER stress is sensed, it leads to activation of the UPR transducer protein kinase RNA-like endoplasmic reticulum kinase (PERK) which phosphorylates and shuts down a central protein translation factor, eukaryotic initiation factor 2 α subunit (eIF2α) [43]. Thus, 2-DG can inhibit growth of tumor cells under normoxia, simply via eIF2α phosphorylation. Recently, however, we reported that AMPK, a well known sensor of energy stress, can also be activated by ER stress induced by 2-DG or by the classical ER stressor tunicamycin [44]. Thus, mTOR, a downstream target of AMPK and another central regulator of protein translation, can also be inhibited in response to ER stress. How much activation of AMPK through ER stress contributes to 2-DG-induced growth inhibition remains to be determined. In addition to AMPK, REDD1, a negative regulator of mTOR, has also been recently identified as a mediator of ER stress-mTOR crosstalk [45]. In this report, it was shown that 2-DG-induced ER stress upregulates activating transcription factor 4 (ATF4), resulting in increased REDD1 expression and down-regulation of mTOR.

Interestingly, shutdown of protein translation via inhibition of eIF2α, thereby sparing the ER of any further processing of unfolded proteins, appears to be critical in maintaining cell survival after ER stress is induced by 2-DG. Thus, allowing eIF2α to function under conditions of ER stress should lead to an overwhelming amount of misfolded proteins resulting in signaling of apoptosis and essentially converting 2-DG-induced growth inhibition to cytotoxicity. Indeed, we recently found that when 2-DG is combined with PERK inhibitors at concentrations where both drugs display no toxicity alone, significant cell death ensues [42]. Additionally, eIF2α can also be phosphorylated by general control nonderepressible 2 (GCN2), which is activated following intracellular glutamine depletion [46]. In this regard, under conditions where PERK is inactive, inhibition of GCN2 activity increases 2-DG-mediated cell death [42]. In summary, it has been shown that 2-DG can induce cytotoxicity in a few tumor cell types growing under normoxia as a single agent or in several types of cancer cells when combined with PERK or GCN2 inhibitors, underscoring the potential of 2-DG as an anti-tumor agent even in the normoxic portion of solid tumors. The underlying mechanism in these processes appears to be due primarily to interference with N-linked glycosylation, in contrast to 2-DG’s effects in hypoxic tumor cells which is predominantly due to inhibition of glycolysis. In fact, it should be noted that most measurements of ATP depletion by 2-DG under normoxia are in the range of 20–30% of control levels [44]. However, under synthetically induced anaerobiosis or hypoxia, ATP levels drop by 80–90% of control [44], which appears to explain why most cell types die when treated with 2-DG in this latter condition.

Several reports have shown that 2-DG can increase reactive oxygen species (ROS) in tumor cells under normoxia leading to cell death [47-49]. An explanation offered for increased ROS is that 2-DG blocks the pentose phosphate shunt (PPS) lowering NADPH levels, which in turn leads to decreased levels of reduced glutathione, a major anti-oxidant [48]. However, it is also reported that 2-DG can be used as a substrate by glucose-6-phopshate dehydrogenase to synthesize NADPH [50, 51]. Thus, the ability of 2-DG to inhibit the PPS should only manifest when used at high doses (i.e., >20 mM) where it can completely shut down HK activity and essentially block the formation of G-6-P, the substrate of the PPS. At moderate doses of 2-DG, we observed a decrease rather than an increase in ROS [44]. We speculated that this may be due to 2-DG’s effects on glycolysis at the level of PGI thereby increasing the amount of G-6-P and shunting it through the PPS. This would theoretically increase the generation of PPS-mediated NADPH required for glutathione’s ability to lower ROS levels. At lower concentrations, however, 2-DG can effectively compete with mannose metabolism and to a lesser extent with glycolysis, which most likely underlies its growth inhibitory as well as toxic effects under normoxic conditions. Thus, a hierarchy of 2-DG’s activity appears to exist where at low concentrations, it interferes primarily with N-linked glycosylation, while at moderate concentrations glycolysis is also inhibited, and only at high concentrations is the PPS blocked. This hierarchy also suggests that there is a significant difference between treatment with 2-DG and glucose deprivation, which are often used interchangeably in the literature. Glucose deprivation results in simultaneous inhibition of PPS, glycolysis, and N-linked glycosylation. Therefore, in instances where 2-DG is used to mimic glucose starvation, these considerations should be kept in mind when interpreting experimental results.

Mechanisms and Function of Autophagy Induced by 2-DG versus Glucose Starvation

Autophagy refers to an evolutionarily conserved intracellular bulk degradation process which plays important roles in both cell homeostasis and pathogenesis [52]. In cancer, autophagy can act as a double-edged sword, preventing tumorigenesis by reducing chronic inflammation and maintaining genomic stability [53] but once malignant transformation occurs, it can protect tumor cells from pathophysiologic as well as therapeutic stresses [52].

2-DG has been reported to induce autophagy in different tumor cell types in vitro [54, 55] and in patients [56]. This response has been suggested to result from inhibition of glycolysis, and subsequent reduction of ATP leading to activation of AMPK [57]. However, more recent findings support a mechanism by which 2-DG’s interference with glycosylation and the induction of ER stress act as the predominant mechanism by which autophagy is induced [4].

The ER compartment is the major site of intracellular Ca2+ storage, and similar to the classical ER stressor tunicamycin, 2-DG induces increases in ER Ca2+ efflux resulting in elevated cytoplasmic Ca2+concentrations [44, 58]. This in turn activates Ca2+/calmodulin-dependent protein kinase kinase β (CaMKKβ) and its downstream target AMPK, resulting in autophagy induction [44]. From these results, it appears that in addition to the well-known function of AMPK as a sensor of energy stress, it also senses ER stress. Thus, in response to either type of stress AMPK is able to mediate activation of autophagy. Similarly, it suggests that in response to ER stress, the cellular benefit of shutting down protein synthesis by PERK-mediated phosphorylation of eIF2α may be aided by the blockage of mTOR via AMPK.

As noted above with the induction of ROS, the mechanisms by which 2-DG and glucose deprivation elicit autophagy are quite different from one another [44]. Although both 2-DG and glucose deprivation lower ATP levels as well as cause ER stress, 2-DG activates autophagy predominantly through this latter mechanism whereas glucose deprivation uses both signals. It is noteworthy that unlike 2-DG-induced autophagy, the Ca2+-CaMKKβ-AMPK pathway is not involved in glucose deprivation-induced autophagy downstream of ER stress. This suggests that during glucose deprivation the amount of Ca2+ release from the ER may not be enough to activate CaMKKβ.

Moreover, in the same report as noted above, it was shown that 2-DG lowers, while glucose deprivation increases cellular ROS levels, the latter contributing to autophagy activation in an extracellular signal-regulated kinase (ERK)-dependent manner [44]. Although Ramirez-Peinado et al. also found that 2-DG activates autophagy, they reported that in certain cell lines glucose deprivation inhibits rather than activates autophagy activity [59]. In regard to the latter result it remains unclear as to why in other reports glucose deprivation activates rather than inhibits autophagy [60-63]. Overall, these observations highlight the distinct mechanisms of autophagy regulation by 2-DG and glucose deprivation, which are often regarded to be one in the same and as noted earlier in this review should be taken into account when interpreting results using these two experimental perturbations.

In contrast to the upregulation of autophagy under normoxia, recently we reported that in three in vitro models of simulated hypoxia, autophagy activity falls below basal levels when treated with 2-DG or glucose deprivation [4, 44]. In these experiments, decreased autophagy activity was well correlated with the severely depleted ATP levels found in anaerobically metabolizing cells treated with either 2-DG or glucose deprivation. In addition, Ben Sahra et al. also reported diminished autophagy markers in cells treated with 2-DG in combination with the mitochondrial complex I inhibitor metformin [64]. These data are consistent with earlier studies where reduced cellular energy levels were correlated with decreased rather than increased activity of autophagic degradation [65-67]. Overall, results involving 2-DG and glucose deprivation suggest that there exists a critical intracellular energy threshold that allows for the successful execution of the ATP-dependent autophagy process. Moreover, they indicate that lowering of ATP should be considered as a modulator of autophagy, where moderate ATP reduction induces it while severe ATP depletion blocks it.

Interestingly, we have also found that in the anaerobic/hypoxia models mentioned above, 2-DG gradually loses its UPR-inducing activity as a function of increasing dose (ref. 4 and unpublished data). It is not yet clear whether this is due to the altered capabilities of 2-DG to interfere with N-linked glycosylation, the blockage of normal UPR signaling, or other potential mechanisms. Nevertheless, this finding strongly suggests that there are intricate interactions between energy- and ER-stress responses.

Although autophagy can play either a pro-survival or pro-death role [68], we and others have found that under 2-DG treatment it acts as an adaptive mechanism to support cell viability, as inhibiting autophagy enhances 2-DG-induced apoptosis and cell death [4, 54, 55]. Thus, as clinical development of 2-DG proceeds it will be important to further explore the possibility in vivo of using autophagy inhibitors to increase its efficacy.

Mechanism of Apoptosis Induction by 2-DG

There are several reports investigating the mechanism of cell death induced by 2-DG in various tumor cell types [2, 69-73]. Although the general conclusion from these studies is that 2-DG kills tumor cells via apoptosis, drug dose, and environmental conditions most likely play a significant role in the mechanism by which cell death proceeds. It was reported that under hypoxia or synthetic anaerobiosis, 2-DG at high dose (two-thirds the concentration of glucose in high glucose medium) induced significant cell death (50 and 90%, respectively) which by annexin V and PI measurements using flow cytometry was interpreted to be due to apoptosis [20]. However, since that publication it has become known that cells undergoing necrosis also stain positively for annexin V [74]. As noted above, under hypoxia 2-DG at moderate doses decreases autophagy which corresponds with severe drops in ATP levels [44]. Thus, under conditions where mitochondrial respiration is inhibited and at high enough doses of 2-DG to reduce ATP to similarly low enough levels that block autophagy, it would not be surprising that apoptosis would also be inhibited and cells would die predominantly by necrosis. Although it is likely that 2-DG-induced cell death could encompass both apoptosis and necrosis simultaneously, a threshold level of ATP must also exist where both apoptosis as well as autophagy are inhibited and the cell dies via necrosis. In this regard, Xi et al. determined that an ATP reduction of ∼50% in pancreatic cancer cell line 1420 appears to switch activation to inactivation of autophagy [44]. More detailed studies will be required to better understand the manner in which cell death proceeds when tumor cells are treated at various 2-DG doses under different levels of hypoxia.

The process of intrinsic, and in some instances, extrinsic apoptosis is mediated by the release of cytochrome c from the intermembrane space of mitochondria to the cytosol following mitochondrial outer membrane permeabilization (MOMP). MOMP is regulated by a group of proteins, known as Bcl-2 proteins, whose interactions determine the ultimate outcome of cell viability. There is clear evidence that the dynamic interaction between Bcl-2 proteins (pro- and anti-apoptotic) and MOMP underlies the fate of a cell [75]. The anti-apoptotic members of the Bcl-2 family of proteins such as Bcl-xL, Bcl-2, and Mcl-1 have been shown to sequester the pro-apoptotic members of this family, Bax and Bak, that are also the building blocks of MOMP. Conversely, upstream of these pro- and anti-apoptotic Bcl-2 members lies a subgroup referred to as BH3-only proteins that upon sensing various types of cellular stress activate Bax and Bak. These activators include Bim, Bad, Bik, Noxa, Puma, Hrk, Bmf, Mule, and Bcl-g as well as others purported to act similarly [76]. It remains unclear how each of these BH3-only proteins activate Bak and Bax either by direct interaction with them or by an indirect mechanism of interfering with their association with the anti-apoptotic proteins. Nonetheless, the end result of either or both mechanisms leads to apoptosis.

The prevailing view is that simultaneous inhibition of all major anti-apoptotic pathways, including Mcl-1, Bcl-2, Bcl-xL, and possibly another, Bcl-w, should result in MOMP and subsequently apoptosis. It should also be noted that although most tumor cell lines of epithelial origin (carcinomas) do not undergo cell death when treated with 2-DG under normoxia, there are others of different lineage, that is, hematologic [40] as well as mesenchymal [36] that have been reported to be killed when treated with 2-DG alone under these conditions. In regard to the latter, in alveolar rhabdomyosarcoma cells 2-DG was shown to induce apoptosis where the upstream pro-apoptotic BH3-only protein Noxa was found to have a dominant role [36]. Although 2-DG induced the expression of a similar pro-apoptotic protein, Bim, only knock-down of Noxa significantly reduced 2-DG-induced cell death. As Noxa predominantly interacts with Mcl-1 while Bim with all three major anti-apoptotics (Mcl-1, Bcl-2, and Bcl-xL), it may be expected that the main mechanism underlying 2-DG-induced apoptosis in alveolar rhabdomyosarcoma cells is the Noxa-Mcl-1 axis. However, in the same report, in addition to Mcl-1, Bcl-xL was shown to protect these cells from 2-DG-induced cell death suggesting that there is more than one group of anti-apoptotic Bcl-2 proteins involved in the induction of apoptosis by 2-DG (i.e., Mcl-1 and Bcl-xL). Furthermore, regardless of the apoptosis axis involved, addition of mannose abrogated the alterations in Bcl-2 proteins induced by 2-DG, indicating that ER stress is the underlying mechanism by which apoptosis is signaled [36]. Another report also supported the importance of Noxa in 2-DG-induced apoptosis in a number of different leukemic cell lines [77]. In separate findings, Mcl-1 was shown to be down-regulated when Myc-transduced lymphoma cells were treated with 2-DG [78]. Interestingly, this result was also observed even after knock-down of Noxa, suggesting that 2-DG can down-regulate the anti-apoptotic Mcl-1 independent of its upregulation of the pro-apoptotic Noxa, both of which will culminate into significant abrogation of Mcl-1’s function in preventing apoptosis.

When 2-DG was applied to T-cell lymphoma cells, Bim and possibly Bmf were found to be the main BH3-only mediators of apoptosis induction [38]. Thus, the BH3-only proteins involved in 2-DG-induced apoptosis may differ from one cell type to another although it should be noted that Noxa was not probed in this last report. In this study again as well as that cited above where rhadomyosarcoma cells were treated with 2-DG, ER stress was found to be the underlying pathway for the alteration of the levels of Bcl-2 proteins in T-cell lymphoma cells [36, 38]. Other Bcl-2 proteins, that is, Bad [79] and Puma [80], have also been reported to be involved in 2-DG-induced apoptosis but further analysis is required to understand the mechanism by which these activators are signaled. A more comprehensive consideration of how various Bcl-2 proteins may play a role in cell death induced by inhibition of glucose metabolism can be found in a review by El Mjiyad et al. [5].

A clear implication of all these studies is that 2-DG shifts the balance of Bcl-2 proteins from an anti-apoptotic to a pro-apoptotic state. Even though as a single agent, 2-DG may not be sufficient to induce cell death in most tumor cell lines, the induction of pro-apoptosis proteins and inhibition of anti-apoptosis proteins by 2-DG will sensitize tumor cells to further insults. In support of this view, there are several reports demonstrating the synergism between 2-DG and both chemotherapy and radiation (reviewed in ref. 5). Moreover, 2-DG has been shown to be an excellent synergizer of the Bcl-2 inhibitors ABT-737 and -263, which specifically bind and inhibit the function of Bcl-2, Bcl-xL and Bcl-w, but not Mcl-1 and are currently being investigated in clinical trials [81]. Thus, it follows that the ability of 2-DG to inhibit the Mcl-1 pathway [73], will sensitize tumor cells to agents such as ABT-737 that block the function of the other anti-apoptotic Bcl-2 proteins. Indeed it was demonstrated that the 2-DG/ABT-737 combination leads to activation of Bid, which subsequently activates Bak. Activated Bak is normally sequestered by Bcl-xL and Mcl-1 [81]. However, by ABT-737 blocking Bcl-xL and 2-DG interfering with the Mcl-1 pathway, active Bak cannot be sequestrated, resulting in apoptosis. These in vitro findings effectively translate into in vivo activity, as the combination of 2-DG and ABT showed significant activity in a highly resistant prostate cancer xenograft model. In a separate report, similar synergism was also observed in pediatric glioma cells [82]. Thus, combining ABT-737 (or ABT-263) and 2-DG seems to be a highly promising clinical approach which may be able to target a variety of tumor types while sparing normal cells due to 2-DG’s preferential accumulation in cancer cells.

Additional Anti-Tumor Effects of 2-DG

2-DG Inhibits Viral Replication

Recently, in collaboration with Dr. Enrique Mesri’s lab, we found that at clinically achievable doses, 2-DG interferes with Kaposi’s sarcoma-associated herpesvirus (KSHV) genome replication leading to diminished virion production during the lytic phase of virus infection, a prerequisite for KSHV tumorigenesis [83]. Moreover, 2-DG also reduces transcriptional expression of several virally encoded genes critical for KSHV oncogenesis. These effects of 2-DG on KSHV appear to result from the drug’s interference with N-linked glycosylation and induction of the UPR. It was shown that the activation of the PERK branch of UPR by 2-DG inhibits eIF2α-mediated global protein translation in virally infected cells. It is known that during active viral production, large quantities of glycoproteins are made in the ER compartment [84-86]. Despite the unusually high ER function in response to viral production, viruses have developed means which allow protein synthesis required for their replication to proceed. In the case of human cytomegalovirus (CMV), it has been reported that viral genes TRS1 and IRS1 are responsible for preventing eIF2α phosphorylation [87]. In herpes simplex virus 1 (HSV-1), γ134.5 indirectly dephosphorylates eIF2α [88] and glycoprotein B (gB) directly inhibits cellular PERK [89], both leading to resumption of protein synthesis, facilitating viral replication. Although it is unclear why KSHV is not able to silence the ER stress response elicited by exogenous 2-DG, one possibility is that as the ER stress induced by 2-DG (interference with N-linked glycosylation leading to misfolded proteins) is different from that by viral replication (protein overproduction), the adaptive mechanism of virus does not apply to 2-DG. An alternative but not mutually exclusive explanation would be that 2-DG by interfering with glycosylation might disrupt the function of the component(s) of the viral machinery that block PERK. A third explanation is that the already overloaded ER protein folding machinery becomes highly vulnerable to exogenous stresses resulting from 2-DG’s interference with N-linked glycosylation. Consequently, this could mount an overwhelming ER stress and UPR response to completely abolish viral gene expression and virion assembly. In theory, the inhibitory mechanism of 2-DG could also be applicable to other viruses.

The targeting of KSHV replication was achieved by low doses of 2-DG in lytic-phase cells without inducing cytotoxicity. Interestingly, it was recently reported that 2-DG at high doses preferentially kills KSHV-infected cells at latent stage, which is attributed to its inhibition of glycolysis [90]. Thus, although the underling mechanisms are complex, 2-DG could be a promising agent in treating Kaposi’s sarcoma through targeting this disease at both the viral and cellular level.

In addition to KSHV, 2-DG has also been reported to exert inhibitory effects on human papillomavirus 18 (HPV 18) [91]. Therein, the authors showed that 2-DG down-regulates the expression of the oncogenic HPV E6/E7 gene at the level of transcription initiation in a Ca2+-dependent manner in HPV 18-infected cells. It should be noted that unlike the mechanism of action of 2-DG on KSHV inhibition mentioned above, ER stress/UPR induction does not seem to account for the observed activity of 2-DG in the setting of HPV infection.

It is estimated that approximately one-tenth of all cancer incidences are viral-related [92], and it is suggested that increased glucose uptake and aerobic glycolysis might be a common trait of viral infection [90]. Therefore, regardless of its anti-viral mechanism, 2-DG could be preferentially enriched in infected cells and thus holds promise for treating viral-induced cancers, thereby significantly reducing the overall burden of these tumor types. Moreover, these effects of 2-DG might prove to also be effective in other non-cancerous diseases that are caused by viral infection.

2-DG Inhibits Angiogenesis

2-DG’s anti-tumor effects are not only limited to the cancer cells per se but also extend to the tumor microenvironment. Most solid tumors engage pathological angiogenesis to support their proliferation and survival when they grow beyond certain sizes [93]. This process depends on recruitment of endothelial cells and their subsequent growth and formation of new blood vessels. In a collaborative effort with Jaime Merchan’s lab, we have shown that endothelial cells are hypersensitive to 2-DG’s static and cytotoxic effects more than a panel of cancer or normal cell types [94]. In addition, 2-DG also blocks new capillary formation whereas sparing established endothelial tubes, suggesting it as a specific drug only targeting active angiogenesis (such as that occurs at tumor sites) without disrupting normal blood vessels. Moreover, the anti-angiogenic effect of 2-DG was shown to be mainly a consequence of ER stress and UPR induction resulting from perturbing N-linked glycosylation, rather than ATP reduction due to blocking glycolysis. In vivo, 2-DG’s angiogenic effect has been confirmed in two separate mouse models [94].

It was recently shown that endothelial cells are highly glycolytic and consume a large amount of glucose [95]. Thus, one possible explanation for why they are hypersensitive to 2-DG is that they accumulate an unusually high amount of 2-DG. However, as in the same report it was also shown that the glycolytic activity between endothelial and tumor cells is comparable, other factors must also contribute to the hypersensitivity of endothelial cells to 2-DG. Regardless of the exact underlying mechanism, the finding of 2-DG’s anti-angiogenic property suggests that this drug could target normoxic cancer cells directly as well as indirectly via disrupting their blood supply. Moreover, by interfering with tumor angiogenesis, 2-DG could convert the normoxic tumor cell population to a hypoxic one, thereby effectively killing both types of tumor cells as a single agent.

2-DG Inhibits Metastasis

It is noteworthy that not only can 2-DG target primary tumors but it is also capable of interfering with tumor metastasis. Sottnik et al. showed that 2-DG inhibits cancer cell invasion and migration in vitro, which correlated with decreased expression of the protease cathepsin L as well as altered cell morphology and levels of structural proteins [96]. Furthermore, in the same report it was also demonstrated that 2-DG treatment significantly delays tumor metastasis in vivo and prolongs animal survival. The detailed mechanism underlying 2-DG’s anti-metastatic effect is so far still unclear. However, as cancer stem cells (CSCs) have been suggested to be the most metastatic as well as glycolytic population within certain tumors [97, 98], an appealing hypothesis is that 2-DG is highly enriched in CSCs and thereby effectively targets this tumor cell population leading to reduced metastasis. In addition to interfering with tumor cells per se, the anti-metastatic property of 2-DG might also be attributed to its effects on tumor stroma (such as blocking angiogenesis and thus tumor cell spreading via blood vessels), or other processes required for metastasis. Currently, most of the anti-tumor drug development efforts are focused on targeting the primary tumor, whereas in reality it is metastasis that accounts for the majority of tumor-related death. To this end, 2-DG appears to be a valuable drug candidate as it targets both aspects of cancer.

2-DG Activates the Immune System

Glycosylation is known to play a role in antigen recognition by the immune system. Therefore, it is possible that 2-DG, by interfering with N-linked glycosylation, may modulate the antigenicity of tumor cells. In this regard, recently, it was reported that at low dose (approximately 0.5 mM), 2-DG augments etoposide-induced antitumor response to both lymphoma and colon cancer cells by enhancing the CD8 cytotoxic T cell recruitment into the tumor sites [99]. Furthermore, the authors demonstrate that the mechanism of enhanced immune response by 2-DG is related to its interference with N-linked glycosylation leading to ER stress. However, the notion that 2-DG treatment can lead to an immunogenic outcome is challenged by another recent report in which 2-DG was shown to inhibit natural killer cell-mediated immune response [100]. It is important to note that in this report high dose 2-DG (20 mM) was used. As mentioned above, 2-DG manifests different metabolic effects at various doses which may explain these two opposing results. Further research, therefore, appears to be warranted for better understanding how altering the dose of 2-DG modulates immune activation against tumor cells.

Clinical Studies with 2-DG

As mentioned above, earlier clinical studies using 2-DG did not result in significant tumor response. In these studies while there was a lack of anti-tumor effect, it was demonstrated that 2-DG was reasonably well-tolerated. In the earliest study with 2-DG, the doses used were 60–200 mg/kg given as an intravenous infusion [15]. In most of the patients, hypoglycemia-type symptoms including sweating and flushing were noted but no convulsions or behavioral changes were seen. Based on our current understanding as explained above, 2-DG as a monotherapy is expected to be efficacious only in select tumor types that are sensitive to this agent in normoxic conditions. Thus, retrospectively, lack of efficacy in earlier studies is not surprising and therefore clinical use of 2-DG was more recently revisited. The phase I trial conducted in our institute to assess the maximum tolerated dose of 2-DG when given orally in combination with infusion of a cytotoxic agent, that is, docetaxel, showed that 63 mg/kg was a safe dose [101]. No drug interactions were observed between these two agents indicating that 2-DG can be safely combined with conventional chemotherapy (at least with docetaxel) in patients (in animal studies 2-DG has been combined with cisplatin, carboplatin, doxorubicin, and others and no unusual adverse drug interactions were observed). Similar to earlier studies, hypoglycemia type symptoms were noted in patients that became intolerable at doses higher than 63 mg/kg. In addition, significant QT prolongations were seen at higher doses although no arrhythmia or cardiac events due to QT prolongation were observed. In another phase I trial in which 2-DG given by mouth was used as a monotherapy, 45 mg/kg was the recommended dose mainly due to concern with grade 1-2 QT prolongations at higher doses [56]. As both the hypoglycemic symptoms as well as QT prolongations were seen early after 2-DG administration, it can be concluded that the plasma peak concentration of 2-DG when given as a bolus by mouth is the underlying reason for these side effects. Therefore, alternative administration schedules such as continuous intravenous infusion should result in even better tolerability. Perhaps more importantly, when 2-DG is given as continuous infusion, a more significant anti-tumor effect was observed in preclinical xenograft models (unpublished data). This interesting observation is most likely due to tumor cells’ enhanced ability to uptake glucose as compared to normal tissues at sub-physiological concentrations, that is, micromolar levels, which resembles PET scans more closely. In conclusion, while the clinical studies established that 2-DG can be safely given to humans even in combination with cytotoxic therapy, further pharmacological optimization should lead to enhanced tolerability and efficacy.

Closing Summary

From Warburg’s data that began in the 1920s, to overwhelmingly similar results from countless of PET scans starting in the 1980s, to the more recent molecular biological confirmation of these observations, it has now become common knowledge that increased glucose metabolism is a universal principle of cancer cells. As shown in Fig. 1, most of the major oncogenic events that are known to be involved with driving a normal cell to a tumor cell have been shown to also be responsible for upregulating glucose metabolism. Thus, 2-DG takes advantage of this by naturally accumulating more in tumors than normal surrounding tissues as well as blocking a central point where oncogenes, or loss of tumor suppressor genes converge. This contrasts to other approaches where small molecules are used to target different aspects of tumor metabolism without benefitting from the natural selectivity that increased glucose uptake offers [102, 103]. In conclusion, 2-DG has been found to have wide-reaching effects on glucose metabolism in tumor cells under hypoxic and normoxic conditions as well as in endothelial cells and viruses which are summarized in Fig. 3.

Figure 3.

Summary of 2-DG’s varied effects at different doses and under normoxia or hypoxia. (A) Under normoxia, 2-DG exerts its effects in a hierarchical manner. At low doses, it mainly interferes with N-linked glycosylation, resulting ER stress. This in turn leads to inhibition of growth, viral replication, and angiogenesis as well as induction of apoptosis and autophagy. At medium doses, 2-DG also blocks glycolysis leading to ATP reduction, and the subsequent growth inhibition and viral replication defects. It is only at high doses that 2-DG starts to disrupt PPS, which causes growth arrest and oxidative stress. 2-DG has also been shown to interfere with metastasis, although the underling mechanisms need to be further explored. (B) Under hypoxia, cellular ATP levels are dramatically depleted when glycolysis is blocked by 2-DG. This results in increased cell death and decreased autophagy. It is not clear so far as how 2-DG affects N-linked glycosylation and PPS under this condition, and what are the biological consequences downstream of these two processes. Abbreviations: 2-DG, 2-deoxy-d-glucose; AMPK, AMP-activated protein kinase; CaMKKβ, Ca2+/calmodulin-dependent protein kinase kinase β; CHOP, C/EBP homologous protein; eIF2α, eukaryotic initiation factor 2 α subunit; ER, endoplasmic reticulum; GI, growth inhibition; mTOR, mammalian target of rapamycin; PERK, protein kinase RNA-like endoplasmic reticulum kinase; PPS, pentose phosphate shunt; ROS, reactive oxygen species; UPR, unfolded protein response.

With the biological trait of increased glucose metabolism so inherent in the make-up of a tumor cell, it appears only a matter of time when 2-DG will become a widely accepted treatment for cancer. Moreover, the fortuitous reality that 2-DG also mimics mannose not only has opened the opportunity to investigate how best to use it for limiting cancer growth but also presents the possibility of developing it as an anti-viral agent. These are just two of the many possible beneficial applications of this remarkable sugar analog.

Additionally, it should be noted that the non-patentable control of the compositional matter of 2-DG has slowed its development. Thus, philanthropic and or governmental resources will be required for 2-DG to reach its full clinical potential.

Acknowledgements

The authors are grateful to Huaping Liu for her input and helpful discussions. The research on 2-DG in the Lampidis lab has been supported by two consecutive 5-year National Cancer Institute grant awards #CA37109 and a Pap Corps award to T.J.L.