Radiation induces aerobic glycolysis through reactive oxygen species

Abstract

Background and purpose

Although radiation induced reoxygenation has been thought to increase radiosensitivity, we have shown that its associated oxidative stress can have radioprotective effects, including stabilization of the transcription factor hypoxia inducible factor 1 (HIF-1). HIF-1 is known to regulate many of the glycolytic enzymes, thereby promoting aerobic glycolysis, which is known to promote treatment resistance. Thus, we hypothesized that reoxygenation after radiation would increase glycolysis. We previously showed that blockade of oxidative stress using a superoxide dismutase (SOD) mimic during reoxygenation can downregulate HIF-1 activity. Here we tested whether concurrent use of this drug with radiotherapy would reduce the switch to a glycolytic phenotype.

Materials and methods

40 mice with skin fold window chambers implanted with 4T1 mammary carcinomas were randomized into (1) no treatment, (2) radiation alone, (3) SOD mimic alone, and (4) SOD mimic with concurrent radiation. All mice were imaged on the ninth day following tumor implantation (30 h following radiation treatment) following injection of a fluorescent glucose analog, 2-[N-(7-nitrobenz-2-oxa-1,3-diaxol-4-yl)amino]-2-deoxyglucose (2-NBDG). Hemoglobin saturation was measured by using hyperspectral imaging to quantify oxygenation state.

Results

Mice treated with radiation showed significantly higher 2-NBDG fluorescence compared to controls (p = 0.007). Hemoglobin saturation analysis demonstrated reoxygenation following radiation, coinciding with the observed increase in glycolysis. The concurrent use of the SOD mimic with radiation demonstrated a significant reduction in 2-NBDG fluorescence compared to effects seen after radiation alone, while having no effect on reoxygenation.

Conclusions

Radiation induces an increase in tumor glucose demand approximately 30 h following therapy during reoxygenation. The use of an SOD mimic can prevent the increase in aerobic glycolysis when used concurrently with radiation, without preventing reoxygenation.

Ionizing radiation is used to treat cancer in over 500,000 patients annually in the United States. The efficacy of radiation therapy for solid tumors is limited by normal tissue toxicity and a variety of tumor characteristics that increase resistance to killing, such as hypoxia, autophagy and senescence. We have shown that radiation-induced reoxygenation also causes phenotypic changes that promote tumor and endothelial cell survival [1,2]. Importantly, we observed a post-radiation increase in the activity of the HIF-1 transcription factor, which regulates a plethora of genes involved in angiogenesis, metabolism, invasion and protection against oxidative stress [3]. Hypoxia upregulates the activity of HIF-1, but paradoxically, HIF-1 activity has also been shown to be induced by oxidative stress that occurs during reoxygenation after radiation therapy [1,4]. Since HIF-1 regulates the majority of glycolytic enzymes, including lactate dehydrogenase, its upregulation may stimulate cells toward glycolysis [3].

Substantial data have suggested that glycolysis under aerobic conditions, also known as the Warburg effect, provides a growth advantage for tumor cells and can lead to malignant progression [5,6]. More recent reports have identified the Warburg effect to be implicated in resistance to cytotoxic stress, including ionizing radiation as well as chemotherapy [7,8]. Therefore, treatment methods which block or reduce glycolytic metabolism after radiotherapy may increase tumor cell sensitivity to radiation and chemotherapeutic killing. In this work, we examined the hypothesis that radiation therapy promotes glycolytic metabolism post-radiation treatment, during the period of tumor reoxygenation.

Additionally, we examined whether reduction of oxidative stress after radiation reduces the glycolytic phenotype. We previously demonstrated that a manganese porphyrin-based superoxide dismutase (SOD) mimic could reduce oxidative stress after radiation therapy [1] as well as decrease tumor growth rate following treatment [9]. These radiosensitization effects are linked to inhibition of hypoxia-reoxygenation injury, which upregulates HIF-1 transcriptional activity following radiation. Here, we tested a newer, optimized SOD mimic (MnTnBuOE-2-PyP⁺⁵) to investigate whether it can prevent or reduce upregulation of the glycolytic phenotype following fractionated radiation therapy. These results may have important clinical implications for trials aimed at preventing post-radiation metabolic changes and increasing the therapeutic index of radiation therapy.

Methods

Tumor treatments

40 athymic nu/nu female mice with skin fold window chambers containing 4T1 mammary carcinomas were randomized into four treatment groups: (1) no treatment, (2) radiation alone, (3) SOD mimic alone, and (4) SOD mimic with radiation treatment. Titanium window chambers were implanted into the skin-fold of the mice as previously described by Palmer et al. [10]. A 25 μL suspension (2.5 × 10⁵ cells) of 4T1-RFP cells (4T1 mouse mammary tumor cells that constitutively express red fluorescent protein (RFP)) was injected into the dorsal skin fold and closed with a glass coverslip (12 mm diameter, No. 2, Erie Scientific, Portsmouth, New Hampshire). Radiation treatment began six days following tumor implantation, delivered in daily 5 Gy fractions ×3 days (15 Gy total), using a Mark IV cesium irradiator (dose = 702 cGy/min, JL Shepherd, San Fernando, CA). 11 cm lead blocks were used to isolate the skin flap for irradiation, while shielding the rest of the body. The radiation fractionation scheme was identical to that used previously by our group, when we evaluated an earlier analog of this drug with the 4T1 tumor line [1,11]. We rationalized that it would be best to keep with the same dose fractionation scheme to be consistent with the prior work.

The SOD mimic is a manganese-based porphyrin, MnTnBuOE-2-PyP5⁺ [12]. This drug has been extensively characterized, and is known to effectively and catalytically inactivate several species associated with oxidative and nitrosative stress, including O⁻₂, NO⁻, ONOO⁻ and OH⁻ [12,13]. Preliminary data have shown this compound to be distributed in the cell in a 3.2–1 mitochondrial to cytosolic ratio. Groups treated with the SOD mimic received subcutaneous MnTnBuOE-2-PyP5⁺ injections starting four days after tumor cell implantation at a dose of 5 mg/kg, twice per day (12 h apart), for five days. In the treatment group receiving both SOD mimic injections and radiation, mice received their first drug injection two days before the start of radiation therapy. MnTnBuOE-2-PyP5⁺ administration began 2 days prior to irradiation in order for drug levels to reach steady state concentration in tissues. Pharmacokinetic analyses have shown that steady state levels are reached after two days of treatment, using the regimen shown here (data not shown). The goal was to have adequate tissue levels of drug before radiation therapy began, to maximize the effect of the drug in reducing oxidative stress associated with reoxygenation. It should be noted that although this drug is a catalytic inactivator of reactive oxygen species, it has no function as a radio-protector in tumors. It accumulates in mitochondria [14], and its mechanism of action occurs in the period of reoxygenation after radiation, not during radiation. Prior studies have shown that at pharmacologically active doses, drugs of this class have no effect on clonogenic survival of the 4T1 tumor line [11]. We have recently verified that MnTnBuOE-2-PyP5⁺ has no effect on clonogenic survival of several other tumor lines (data not shown). On radiation treatment days, mice received their first daily dose approximately 2 h prior to irradiation.

All mice were imaged on the ninth day following tumor implantation, which corresponds to approximately 30 h following the final radiation dose in the irradiated animals. Mice were imaged immediately after tail vein injection of a fluorescent glucose analog, 2-[N-(7-nitrobenz-2-oxa-1,3-diaxol-4-yl)amino]-2-deoxyglucose (2-NBDG). This fluorescent glucose analog demonstrates rapid uptake into cells by GLUT receptors, is retained intracellularly without further metabolism, and mimics the pharmacodynamics of fluoro-deoxyglucose (FDG) used in PET [15–17]. Animals were housed in an on-site housing facility with access to food and water with standard 12 h light/dark cycles. All in vivo experiments were conducted according to a protocol approved by Duke University Institutional Animal Care and Use Committee.

In vivo microscopy

Mice were fasted (NPO) 6 h prior to imaging to control blood glucose concentration prior to 2-NBDG injections. Mice were anesthetized using isoflurane, 2% for induction and 1% for maintenance. Hyperspectral images of white light transmittance (520–620 nm) through the window were recorded for 15–30 min prior to 2-NBDG injection, in order to calculate vascular hemoglobin saturation, as described by Sorg et al. [18]. Animals were administered a 100 μL tail-vein injection of 2-NBDG (6 mM dissolved in sterile saline, Small Molecule Synthesis Facility, Duke University) and 2-NBDG fluorescence (525 nm) was recorded for 85 min at 1 (first 8 min), 30 (8–45min) and 180 (45–85 min) second intervals.

All imaging was conducted using a Zeiss Axioskop 2 microscope. A 100 W halogen lamp was used for trans-illumination while a 100 W mercury lamp was used for epi-illumination fluorescence imaging of 2-NBDG and RFP expression. Optical filtering for the purpose of hyperspectral imaging was accomplished using a liquid crystal tunable filter (LCTF). Trans-illumination images were acquired from 520 to 620 nm in 10 nm increments. Epi-illumination 2-NBDG fluorescence images were acquired from 510–550 nm in 5 nm increments with a 470 nm bandpass excitation filter (40 nm bandwidth) and a 510 nm longpass dichroic beam splitter. RFP fluorescence was recorded from 610 to 690 nm using a 560 nm bandpass excitation filter (30 nm bandwidth) and a 600 nm longpass dichroic beam splitter. All images were collected with a 2.5 × objective (NA = 0.075 and field of view = 10 mm) and a DVC 1412 CCD camera (DVC Company).

Statistical analysis and modeling

All calculations were conducted using MATLAB 2011b. Unless otherwise specified, data are reported as mean and standard error. Statistical significance was determined using Student’s t-test or ANOVA where appropriate. The Kolmogorov–Smirnov test was used to test the significance of the signal enhancement ratio (SER) values and hemoglobin saturation. Only p values less than 0.05 were considered significant.

Results

Fasting affects 2-NBDG uptake

A quantification method for analyzing 2-NBDG fluorescence intensity over an 85 min interval that would allow for normalization for variations in cell density and background autofluorescence was required for this study. Signal enhancement ratio (SER) patterns have been previously used to quantitate the kinetics of contrast enhancement using dynamic contrast-enhanced MRI via three time points (Fig. 1) to characterize tumors in vivo [19–21]. Varying levels of SER have been shown to correlate with overall treatment outcome in these studies [20,21]. This method was used to compare the 2-NBDG fluorescence intensity for a tumor in a mouse which has been fasted for 6 h prior to 2-NBDG injection and the same tumor when the mouse was not fasted prior to 2-NBDG injection. It is expected that a tumor in a mouse fasted prior to imaging would have a higher glucose demand than when 2-NBDG is competitively inhibited by endogenous d-glucose in the non-fasted case [15]. Accordingly, the fasted condition shows higher 2-NBDG fluorescence intensity as well as corresponding SER values than when the same mouse is imaged without fasting (one of three representative experiments shown in Fig. 2). These results demonstrate that glucose demand, regulated by fasting, can be sufficiently monitored by 2-NBDG fluorescence and that higher corresponding SER levels indicate greater glucose demand. Based on this preliminary data, we included a 6 hr fast prior to imaging for all mice in the experimental protocol in order to provide a consistent blood glucose level at the time of imaging.

Fig. 1.

Graph depicting typical 2-NBDG fluorescence pattern in tumor and non-tumor areas. Signal enhancement ratio (SER) is calculated as (S₁–S₀)/(S₂–S₁), where S₀ is defined as the baseline fluorescence, S₁ is the peak intensity, and S₂ is the late post-contrast intensity.

Fig. 2.

2-NBDG fluorescence intensity over time (A) for fasting and non-fasted tumor and corresponding SER images (B) of control mouse NPO 6 h prior to imaging with 2-NBDG and the same mouse with no fasting prior to imaging. Representative images of n = 3 mice.

Radiation increases glucose demand while SOD mimic reduces glucose demand in tumors post-radiation

It is has been previously shown that HIF-1 activity and consequent anaerobic metabolism increased approximately 12–24 h after radiation in the 4T1-RFP tumor line. The increase in HIF-1 activity was associated with reoxygenation induced reactive oxygen species [1,2]. Since HIF-1 is known to regulate most enzymes involved in glycolysis, we sought to determine whether a reduction in reactive species via the use of the SOD mimic, MnTnBuOE-2-PyP5⁺, can reduce glucose demand. Mice were randomized to receive SOD mimic treatment +/− radiation, or no treatment.

Mice were injected with 4T1 tumor cells that were stably transfected with red fluorescence protein [22] under control of a constitutive promoter (RFP), allowing tumors to be visualized using intravital fluorescence microscopy in skinfold window chambers (Fig. 3B). Tumor areas were measured by identifying pixels that were positive for RFP. There was not a significant difference in tumor area between the treatment groups (p > 0.05). After IV injection of 2-NBDG, the average fluorescence of 2-NBDG for tumor (defined as RFP (+) pixels) was recorded for 85 min (Fig. 3A). SER for each tumor was also calculated for all mice (Fig. 3B). Tumors in the irradiated group had higher peak 2-NBDG fluorescence compared with the no treatment control, at 3 and 6 min (p < 0.05). Higher SER values spatially corresponded with the location of tumor cells (Fig. 3B). The distributions of SER values on a pixel by pixel basis for each animal were used to generate cumulative frequency histograms. These values were compiled for each group. This method of analysis facilitates comparisons between treatment groups (Fig. 3C) and has been used previously by Hardee et al. [23]. SER values obtained from the irradiated group were significantly higher (p = 0.007) than the other experimental groups, which were indistinguishable from each other (Fig. 3C).

Fig. 3.

2-NBDG fluorescence over time (A) for all treatment groups (n = 10) for each group. Error bars depict standard error of the mean. At three and six minutes, the fluorescence of the radiation group is significantly greater than that of the control group (p = 0.020, 0.031, respectively). Error bars for no treatment and radiation groups are depicted for visual clarity; error bars for two remaining groups are equal in magnitude. Representative SER image (B) for all treatment groups: (a) no treatment, (b) radiation alone, (c) SOD mimic and radiation, (d) SOD mimic alone. Cumulative distribution functions of SER values for all treatment groups (C). SER values for radiation only group are significantly higher (p < 0.007) than the other groups, which are not statistically different. Error bars depict standard error of the mean.

Radiation induced metabolic changes coincide with reoxygenation, suggesting aerobic glycolysis

Given that radiation appears to alter tumor glucose demand, we sought to determine whether these changes could be correlated temporally with reoxygenation. Hyperspectral images captured from the tumors were used to simultaneously calculate hemoglobin saturation of the tumor vasculature and 2-NBDG fluorescence (Fig. 4A). The tumors in the irradiated treatment groups displayed significantly higher hemoglobin saturation than the no treatment controls (p = 0.009). These findings strongly suggest that the radiation treatments increased tumor oxygenation. The SOD mimic only group appeared to display improved oxygenation as well, consistent with previous findings elucidating its anti-HIF/VEGF effects [9]. Altogether, these results are consistent with previous views regarding post-radiation tumor reoxygenation [1,24,25]. The finding that irradiated tumors increase glucose consumption in the presence of reoxygenation strongly suggests initiation of glycolysis rather than oxidative phosphorylation. Additionally while the treatment group receiving radiation concurrently with the SOD mimic demonstrates elevated hemoglobin saturation consistent with reoxygenation, the tumors demonstrate lower glucose demand. Consequently, these data additionally suggest that post-radiation aerobic glycolysis was prevented by scavenging of oxidative reactive species and downstream HIF-1 upregulation, without compromising the reoxygenation process.

Fig. 4.

(A) Representative hemoglobin saturation image from each treatment group: (a) no treatment, (b) radiation alone, (c) SOD mimic and radiation, (d) SOD mimic alone. (B) histogram displaying tumor vascular hemoglobin saturation in each treatment group. The y-axis represents the frequency of the pixels determined to be vasculature that displayed the corresponding hemoglobin saturation on the x-axis. The no treatment control group appeared to have a significantly lower hemoglobin saturation compared to the other three groups (p = 0.009), which were not statistically different from each other. (C) Vascular length density of tumors for all treatment groups measured 30 h following final radiation treatment. No significant difference between these groups (p = 0.09). Error bars represent the standard deviation of each group.

Increase in 2-NBDG SER is not due to increased vascular density

Since radiation treatment appeared to have increased tumor 2-NBDG fluorescence and SER values, these findings may be accounted for by either an increase in glucose consumption or increased vascularity allowing increased glucose delivery to the tumors. There was no significant difference in vascular length density among these treatment groups (Fig. 4C), suggesting that changes in 2-NBDG are most likely due to differences in glucose consumption rather than increases in marker delivery. These findings support that the increase in glucose consumption following tumor radiation may be reversed by concurrent treatment using the SOD mimic MnTnBuOE-2-PyP5⁺, through the reduction in post-radiation reactive oxygen species. The prevention of a switch to a glycolytic phenotype is expected to prevent any increase in radioresistance brought on by a switch from aerobic to anaerobic metabolism.

Discussion

Our findings demonstrate an increase in tumor glucose demand following radiation therapy, despite increased oxygen availability; these data are strongly indicative of a switch to aerobic glycolysis. We previously reported that HIF-1 is increased during reoxygenation, using this same tumor model [1]. Taken together, these data suggest that the changes in glucose demand are most likely associated with stabilization of HIF-1 during reoxygenation, as demonstrated by Moeller et al. [1]. Additionally, these findings demonstrate that a SOD mimic, given concurrently with radiation therapy, prevents the observed glycolytic changes associated with radiation alone. We also show that while free radical scavenging inhibits HIF-1 mediated increase in glycolysis, the concurrent use of the SOD mimic does not interfere with reoxygenation, as demonstrated by improvements in hemoglobin saturation (Fig. 5).

Fig. 5.

Schematic demonstrating proposed mechanism of aerobic glycolysis induction resulting from radiation therapy.

An important question is whether these tumors, which are only a few mm in diameter, exhibit hypoxia. Hypoxia-reoxygenation injury could not happen if the tissue was not hypoxic to begin with. We have studied the oxygenation state of window chamber tumors extensively. We have demonstrated that they exhibit vascular hypoxia and oxygen gradients between vessels that dip well below 10 mmHg, using oxygen microelectrodes [24,26,27]. We have verified the presence of hypoxia using pimonidazole staining and phosphorescence lifetime imaging, in collaboration with David Wilson’s group [28,29]. Using optics to measure hemoglobin saturation, we have frequently found vascular hypoxia in the 4T1 tumor model [18,30]. Most recently, we have verified correlation between hemoglobin saturation and presence of hypoxia, using an oxygen sensitive nanoparticle, which has a phosphorescence lifetime signal that is oxygen sensitive [31–33]. In this paper we used hemoglobin saturation as a surrogate for the oxygenation state of the tumor, because the optical properties of the nanoparticle are in the same wavelength range as the glucose analog. Given the extensive validation and cross-correlation between these two methods, this is a reasonable approach. The extent of reoxygenation seen after radiation in these experiments is consistent with what would be expected and is sufficient to cause an increase in reactive oxygen species, as we previously reported [1].

Radiation therapy acts to kill cancer cells primarily through free radical induced double stranded DNA breaks, but in the past decade research has shown this mechanism acts as a double-edged sword. Moeller et al. demonstrated that radiation induces HIF-1 mediated VEGF secretion, which is capable of inhibiting endothelial cell apoptosis [1]. HIF-1 has also been demonstrated to regulate a wide range of processes in tumor cells, including ATP metabolism, cellular proliferation, and p53 activation, recently reviewed by Semenza [3]. Consequently, significant research interest has stemmed from the discovery that radiation prompts the development of radioresistant phenotypes through HIF-1 stabilization. Since then, several groups have employed various HIF-1 inhibitors with the goal of radiosensitization [1,34–36]. Lu et al. most recently demonstrated radiosensitization in head and neck squamous cell carcinoma xenografts using cetuximab, an epidermal growth factor receptor (EFGR) monoclonal antibody, as a HIF-1 inhibitor [37].

These previous studies employed direct HIF-1 inhibitors. Our study aims to prevent HIF-1 stabilization through reduction of its upstream activators. It has been previously demonstrated that HIF-1 increases as 4T1 tumors are reoxygenated following radiation therapy, starting at 12 h and peaking at 48 h following treatment [2]. The hemoglobin saturation data we have presented show that post-radiation reoxygenation is present during the time period in which we are imaging using 2-NBDG, approximately 30 h following the final radiation fraction. If our current understanding of HIF-1 activation through tumor reoxygenation associated free radicals is correct, the inactivation of reactive oxygen species using a SOD mimic should demonstrate HIF-1 inhibition during this time frame. Indeed, we have previously demonstrated this phenomenon in the same tumor model used in this study. We did not use the HIF-1 reporter gene in this study because its fluorescence spectrum overlies that of 2-NBDG. However, our findings were consistent with our previous hypothesis and strongly suggest that catalytic inactivation of ROS inhibits HIF-1 associated glycolytic changes to a significant degree. Our claim that this tumor undergoes a switch to anaerobic metabolism upon reoxygenation and HIF-1 activation is based on prior work with this tumor model. Specifically, when we knocked down HIF-1 levels in this tumor, it was unable to maintain ATP levels and exhibited extensive central necrosis [2]. This finding demonstrates that it can exhibit the Pasteur effect, where anaerobic metabolism is regulated by hypoxia and HIF-1 expression. Therefore, the tumor relies on HIF-1 expression to help regulate ATP levels under hypoxic conditions. Aerobic glycolysis and increased oxidative stress have also been reported to be regulated by oncogene overexpression, such as Ras [38,39]. Since the 4T1 tumor line overexpresses Ras [40], we cannot rule out the possibility that Ras is also involved in the upregulation of glucose uptake after radiation exposure.

HIF-1 regulates many glycolytic enzymes, but this report is the first to show that radiation induced reoxygenation promotes a glycolytic switch. We were able to successfully measure glucose delivery and uptake using 2-NBDG as well as confirm the presence of reoxygenation by measuring hemoglobin saturation. In doing so, we have demonstrated that radiation induced reoxygenation causes a significant increase in peak glucose uptake by tumor. The increased 2-NBDG uptake can be attributed to either increased vascular density or an increase in glucose consumption or demand. Given that the densities of the tumor vasculature in these irradiated tumors are equal to those of the non-irradiated tumors, it is reasonable to conclude that increases in glucose uptake are thus attributed to increased glucose demand as opposed to delivery.

Since tumor cells appeared to require increased amounts of glucose in the presence of oxygen, it is more probable that an inefficient metabolic process such as glycolysis is occurring rather than oxidative phosphorylation. Further, we observed that this increase in 2-NBDG uptake occurred simultaneously with tumor reoxygenation. Consequently, our study suggests that aerobic glycolysis, also known as the Warburg effect [41], which is known to provide cancer cells an advantage in the tumor microenviroment, takes place following radiation therapy. Our findings are also consistent with previous literature suggesting a significant role of ROS in the regulation of the Warburg effect. Another possibility for the observed increase in glucose utilization includes metabolism involved in cell repopulation. However, given that the 4T1 cell line has a doubling time of approximately 12 h [42], our observation window of 30 h post-radiation would at most allow 1–2 replication cycles, buffered against a substantial amount of cell killing from the radiation exposure. Thus, we do not believe that repopulation would be a significant factor for the changes we observed.

In addition to demonstrating that radiation causes an increase in tumor glucose metabolism, we sought to test whether this increase could be inhibited through the catalytic inactivation of reactive species which are thought to activate HIF-1. Our data suggest that the concurrent use of the SOD mimic with radiation therapy does indeed prevent the increases in glucose demand seen in with radiation alone. Consequently, the importance of HIF-1’s (and potentially that of ras) role in regulating glucose consumption following radiation was verified by demonstrating significant reduction in glucose demand when radiation was combined with a SOD mimic. Tumors that were treated with the SOD mimic displayed the same levels of glucose demand as non-irradiated tumors. These data support the conclusion that post-radiation HIF-1 may be playing a major role in increasing tumor glucose metabolism and that this increase can be prevented through the scavenging of reoxygenation associated reactive species. The potential clinical significance of this study is clear: the use of a relatively non-toxic potent scavenger of reactive species may prevent deleterious post-radiation glycolytic changes without preventing reoxygenation. This drug could potentially be useful when used concurrently with radiation therapy.