The metabolomic profile of gamma-irradiated human hepatoma and muscle cells reveals metabolic changes consistent with the Warburg effect

Introduction

There is an ever increasing threat from the terrorist use of nuclear isotopes. To date, the only intentional exposure of humans to ionizing radiation from a nuclear device occurred during the first use of the atomic bomb at Hiroshima and Nagasaki on 6 and 9 August 1945, respectively. Accidental exposures have also occurred. Ten persons were subjected to bursts of ionizing radiation at the Los Alamos Scientific Laboratory in accidents that occurred on 21 August 1945 and 21 May 1946 (Hempelmann, Lisco & Hoffman, 1952). The estimated median combined X-ray and γ-ray dose to these subjects was 2.7 Gy (range 0.27 to 18 Gy). The two highest exposed persons died while the others made a full recovery (Hempelmann, Lisco & Hoffman, 1952). These unfortunate accidents provided the first opportunity to examine both blood and urine for the footprints of high-dose radiation exposure. The blood chemistry provided no results that were either specific or of significance. In contrast, the subjects who had received the highest doses of radiation exhibited urinary concentrations of amino acids in the range 1 to 2 mg/ml, most notable of which was the excretion of taurine (Hempelmann, Lisco & Hoffman, 1952). These human findings were soon replicated rats receiving 8 to 25 Gy X-irradiation (Kay & Entenman, 1954) and confirmed by others in the rat (Aebi et al., 1955) and in four human subjects accidentally exposed to γ-irradiation and neutrons at the Argonne National Laboratory (Katz & Hasterlik, 1955).

A rat model of acute radiation sickness was studied in the former USSR. Administration of 2 or 4 μCi ⁹⁰Sr to rats caused a marked increase in the urinary excretion of thymidine (Uspenskaia & Rabinkova, 1965), with 8-fold increases reported in another study (Zharkov, Fedorova & Mikhailova, 1965). Enhanced urinary excretion of the related nucleoside pseudouridine was observed in cancer patients receiving X-ray therapy and in rats receiving whole-body X-irradiation. The authors believed that the spleen contributed largely to these findings (Portmann et al., 2013). The early literature, therefore, described the elevated urinary excretion of both amino acids and nucleosides in humans and rats exposed to large doses of ionizing radiation.

This remained the case until the problem was readdressed using high-throughput metabolomic technologies in the recent past. The pattern described four to five decades earlier has essentially remained the same, but with more related metabolites added. Interestingly, both taurine and thymidine have persisted as in vivo urinary biomarkers of ionizing radiation exposure in the mouse, rat and rhesus monkey (Johnson et al., 2011; Johnson et al., 2012; Tyburski et al., 2008, Tyburski et al., 2009).

There has been a recent concerted effort to discover biomarkers of ionizing radiation exposure that could lend themselves to high-throughput minimally-invasive population screening in the event of intentional or accidental mass exposures to radiation. Metabolomics has been part of this effort (Johnson et al., 2011; Johnson et al., 2012; Laiakis, Hyduke & Fornace, 2012; Laiakis et al., 2014; Lanz et al., 2009; Manna et al., 2013; Tyburski et al., 2008, Tyburski et al., 2009). Despite the uncovering of radiation biomarkers, little, if anything, is known about their cellular origins. Irradiation of cells in culture could help fill this knowledge gap.

Irradiation of cultured cells has a long history, beginning with chick embryo fibroblasts given X-ray doses of 1 to 10 Gy in an attempt to improve the efficacy of X-ray therapy with radiosensitizers (Mitchell & Simon-Reuss, 1952). More recently, both human fibroblast and lymphoblast cell lines have been γ-irradiated to define biomarkers of ionizing radiation that might assist with human in vivo studies (Patterson et al., 2008). While clear metabolomic phenotypes were found for doses of between 0.5 and 8 Gy, the only identified metabolites were those that were depleted by cellular irradiation; for example, GSH and AMP. No positive biomarkers of ionizing irradiation were reported (Patterson et al., 2008). In an effort to redress this deficiency and to define biomarkers of exposure to ionizing radiation in human cells in culture, we have conducted a study of γ-irradiation of the human hepatocellular carcinoma cell line HepG2 and human skeletal muscle cell line HMCL-7304 using gas chromatography-mass spectrometry based metabolomics. Cell lines from liver and skeletal muscle were chosen because these tissues represent the major stores of taurine (Awapara, 1956), the historical biomarker for ionizing radiation (see above), and therefore perhaps are tissues that are sensitive to ionizing radiation.

Materials and Methods

Results

It was first demonstrated that the radiation doses employed produced measurable cellular changes in HepG2 cells. Fluorescence IHC showed that γ-irradiated cells, but not sham irradiated cells, had multiple γH2AX foci indicative of DNA double-strand breaks (Fig. 1). In addition, flow cytometry analysis of sham and γ-irradiated HepG2 cells showed roughly equal populations of cells in G1 and G2/M stages of the cell cycle (Figs. 2A–2D). Although the changes were relatively small, the cells shifted significantly (r = 0.85; P = 0.007) from G1 to G2/M in relation to radiation dose, plateauing at ∼47% in G2/M (Fig. 2E).

GCMS was used to conduct metabolomic analyses on both HepG2 human hepatocellular carcinoma cells and HMCL-7304 human striated muscle myotubes. A total of 46 compounds were identified unequivocally in γ-irradiated HepG2 cell lysates, using both library spectra (>90% match) and a collection of more than 100 authentic standards. These HepG2 cell metabolites are listed in Table 1 and the most prominent 17 are shown in Fig. 3A. In the case of γ-irradiated HMCL-7304 myotubes, the resulting peaks (Fig. 3B) were smaller and, as Table 2 shows, it was only possible to identify 29 metabolites, all of which were also present in the HepG2 cell lysates.

The GCMS data from 20 cultures of HepG2 cells and 12 cultures of HMCL-7304 myotubes were subjected to multivariate data analysis, first by PCA, which shows the internal structure of the dataset and identifies any outliers. Figure 4A shows the PCA scores plot for γ-irradiated HepG2 cell lysates, with clustering of the sham, 1 Gy, 2 Gy and 4 Gy flasks. When a PLS-DA model was built for these data (Fig. 4B), the sham, 1 Gy, 2 Gy and 4 Gy clusters were all resolved. In order to check that the supervised analysis had not over-fitted the data, a leave-one-out cross-validation was performed with 100 permutations (Fig. 4C). Because the permuted correlation (R2) and predictability (Q2) fell below 0.3 and 0, respectively, the data were not over-fitted and the PLS-DA model was therefore valid. An OPLS-DA loadings S-plot was then employed to detect the upregulated metabolites in HepG2 cell lysates after 2 Gy γ-irradiation (Fig. 4D). D-Glucose (1), L-lactate (2), L-5-oxoproline (3), L-alanine (4), D-fructose (5), D-glucose 6-phosphate (6) and succinate (7) were all considered as candidates for upregulated metabolites and, together with citrate, were subjected to univariate statistics. Figure 5 shows that 2 Gy γ-irradiation of HepG2 cells led to a highly statistically significant (P < 0.001) increase in intracellular L-lactate, L-alanine, D-glucose, D-glucose 6-phosphate, D-fructose and L-5-oxoproline. This metabolic pattern is consistent with increased glucose mobilization/uptake in concert with elevated aerobic glycolysis (the Warburg effect), leading to increased synthesis of glucose 6-phosphate, lactate and alanine, the last two both from pyruvate. Increased fructose might represent shunting of excess glucose through the polyol pathway, which leads to fructose synthesis via sorbitol (Bjornstad et al., 2015), but fructose could equally be produced by other mechanisms, such as glucose 6-phosphate isomerase in the glycolytic pathway. In addition, the elevated 5-oxoproline is a sign of increased GSH turnover in response to oxidative stress (Emmett, 2014), but could potentially arise from other sources. In general, these effects were less pronounced after 4 Gy γ-irradiation of HepG2 cells. Changes in the TCA cycle intermediates citrate and succinate were of lower statistical significance or insignificant after HepG2 cell irradiation.

In summary, the pattern of metabolic reprogramming observed in both cell lines resembled that of enhanced aerobic glycolysis, with elevated intracellular glucose and lactate in both hepatoma and muscle cells, together with elevated intracellular glucose 6-phosphate and alanine in HepG2 cells. In addition, there was evidence from increased intracellular 5-oxoproline of increased GSH turnover in HepG2. Such a finding is consistent with enhanced ROS production due to γ-irradiation. These effects were not observed at all radiation doses employed. In HepG2 cells, 1 Gy γ-irradiation had no observable effect (Fig. 5). The maximum response was observed at a dose of 2 Gy (glucose↑ ∼50-fold, P < 0.001; glucose 6-phosphate↑ 2-fold, P < 0.001; lactate↑ 2-fold, P < 0.001; alanine↑ 2.5-fold, P < 0.001), with a lower but statistically significant response at 4 Gy (glucose↑ ∼40-fold, P < 0.001; lactate↑ 1.6-fold, P < 0.01). In HMCL-7304 myotubes, the metabolic effect of γ-irradiation was much less pronounced and with an intracellular rise in glucose (2-fold, P < 0.001) and lactate (2-fold, P < 0.001) only observed at 2 Gy, with doses of 1 and 4 Gy indistinguishable from sham irradiation (Figs. 6 and 7).

Discussion

Understanding the metabolic response of tissues to ionizing radiation is essential to the evaluation of the origin of metabolomic biomarkers for radiation exposure and to the development of novel treatment strategies for radiation sickness. To date, work has largely been conducted in laboratory animals, including mice, rats and rhesus monkeys, with almost no comprehension of the metabolomic response to radiation of cultured cells. One study has been reported in which human TK6 lymphoblastoid cells and the BJ fibroblast cell line were γ-irradiated and intracellular metabolites analyzed by ultra-performance liquid chromatography-electrospray ionization-quadrupole time-of-flight mass spectrometry (UPLC-ESI-QTOFMS) (Patterson et al., 2008). The OPLS-DA loadings S-plot revealed several metabolites upregulated by γ-irradiation, but none was identified. However, some intracellular metabolites were quenched by γ-irradiation and these were identified mainly as GSH, AMP and NAD⁺. Interestingly and contrary to the finding reported here, cellular 5-oxoproline was diminished in irradiated cells (Patterson et al., 2008). In this study, we sought to identify intracellular metabolites that were enhanced after γ-irradiation and chose GCMS-based metabolomics as a means to identify them. GCMS is better suited than simple reverse-phase UPLC-ESI-QTOFMS to the detection of small polar intermediate metabolites because these metabolites, such as amino acids and sugars, elute too early on UPLC (Patterson et al., 2011), making them difficult to quantitate.

The principal finding in irradiated HepG2 cells relative to sham-irradiated cells was a huge increase in intracellular glucose, with a concomitant elevation of glucose 6-phosphate and the two pyruvate metabolites lactate and alanine, suggesting enhancement of cytosolic aerobic glycolysis. As has been observed in hyperglycaemia, increased glycolytic flux leads to a saturation of hexokinase, followed by shunting of as much as one-third of glucose through aldose reductase in the polyol pathway, which uses one mole of NADPH per mole glucose metabolized and generates one mole of NADH per mole of fructose produced from the intermediate sorbitol (Yabe-Nishimura, 1998). GSH is a principal defense against oxidative stress in the liver and recycling of oxidized glutathione to GSH is carried out by glutathione reductase, for which NADPH is an obligatory cofactor. The reciprocal relationship between GSH and ROS has been demonstrated in HepG2 cells (Kim et al., 2010). In addition, when a transgenic mouse with a disrupted polyol pathway (Ho et al., 2000) was rendered short-term diabetic with streptozotocin-induced superoxide formation, GSH depletion and subsequent DNA damage, as seen in similarly treated wild-type mice, was not observed (Ho et al., 2006). Therefore, there is a clear relationship between shunting of excess glucose through the polyol pathway, GSH depletion and DNA damage. However, we have no evidence that the polyol pathway was activated in our irradiated cells, simply the conjecture that the excess glucose load makes it possible that some additional glucose is handled by aldose reductase and the polyol pathway.

In γ-irradiated HMCL-7304 myotubes, a simpler but similar metabolic pattern was observed, with elevated intracellular glucose and lactate. Unlike HepG2, this cell line was not derived from cancer cells, but rather from skeletal muscle of a 19 year-old healthy donor and a cell line developed by retroviral knock-in of the TERT and CDK4 genes (Rokach et al., 2013). These cells are therefore not expected to display significant aerobic glycolysis. However, after γ-irradiation, HMCL-7304 myotubes plainly demonstrated the metabolic fingerprint of upregulated aerobic glycolysis, with increased intracellular glucose and enhanced lactate production.

The potential augmentation of the Warburg effect was not observed at all radiation doses. In both cell lines, there was no statistically significant difference between the sham-irradiated and the 1 Gy dose, indicating a clear threshold for putative upregulation of aerobic glycolysis by γ-irradiation. For HepG2 cells, both 2 Gy and 4 Gy stimulated large and highly statistically significant increases in intracellular glucose, but the increases in both lactate and alanine were more pronounced at 2 Gy compared to 4 Gy. In contrast, only 2 Gy γ-irradiation stimulated assumed aerobic glycolysis in HMCL-7304 myotubes. Therefore, it would appear that the highest radiation dose reduced the magnitude of the metabolic reprogramming. This may be due to increased genotoxicity at the 4 Gy dose, since γ-irradiation of FL5.12 cells with a dose of 20 Gy, resulted in decreases in both glucose uptake and glycolysis 24 h after irradiation (Zhou, Vander Heiden & Rudin, 2002). We also observed a high proportion of cells (∼47%) in G2/M at the highest radiation dose. It has recently been reported that human cervical carcinoma SiHa cells, both wild-type and a sub-clone with impaired mitochondrial respiration, moved from 21–31% to 46–47% in G2/M 24 h after a dose of 5 Gy γ-irradiation (Bol et al., 2015). Our findings, both of the dose-response relationship with respect to a possible Warburg effect, together with cell cycle behaviour, are consistent with the reports of others.

Our contention that γ-irradiation augmented cytosolic aerobic glycolysis in two cell lines derived from liver and muscle may have implications for the prevention or amelioration of tissue toxicity after exposure to ionizing radiation. This would be especially pertinent in cases where vast numbers of persons were exposed to varying doses, for example, in the case of a terrorist incident involving either a radiation or a nuclear device. Ionizing radiation generates hydroxyl radicals through radiolysis of intracellular water and these highly-reactive and short-lived species lead to a state of oxidative stress that has been estimated to be responsible for 70% of the tissue damage that results from ionizing radiation (Anuranjani & Bala, 2014). The metabolic reprogramming we observed in HepG2 cells caused a huge rise in intracellular glucose. It is likely that this may result in a flux of glucose through aldose reductase and the polyol pathway, with a resulting decrement of NADPH and thus depleted glutathione reductase activity. This will also compromise GSH production and the defense against oxidative stress, leading to enhanced tissue damage. If confirmed, these findings provide clues to new potential therapeutic and preventative interventions for radiation sickness.

Conclusions

It is concluded from metabolomic profiles that γ-irradiation of human cell lines derived from liver and striated muscle causes a metabolic reprogramming consistent with amplified cytosolic aerobic glycolysis. Further investigation of this putative radiation-induced Warburg effect may lead to novel prevention and treatment strategies for radiation-induced tissue damage.